encyclopedia of chemical processing and design: 69

Encyclopedia of Chemical Processing and Design: 69 Supplement 1

Marcel Dekker, Inc.

Rayford G. AnthonyJohn J. McKetta

Encyclopedia ofChemical Processingand DesignEDITOR Rayford G. Anthony

SENIOR ADVISORY EDITOR John J. McKetta

69 Supplement 1

Library of Congress Cataloging in Publication DataMain entry under title:

Encyclopedia of chemical processing and design.

Includes bibliographic references.1. Chemical engineering—Dictionaries 2. Chemistry,

Technical—Dictionaries. I. McKetta, John J.II. Cunningham, William Aaron.Tp9.E66 660.2′8′003 75-40646ISBN: 0-8247-2621-9

HeadquartersMarcel Dekker, Inc.270 Madison Avenue, New York, NY 10016tel: 212-696-9000; fax: 212-685-4540

Eastern Hemisphere DistributionMarcel Dekker AGHutgasse 4, Postfach 812, CH-4001 Basel, Switzerlandtel: 41-61-261-8482; fax: 41-61-261-8896

World Wide Webhttp:/ /www.dekker.com

COPYRIGHT 2002 by MARCEL DEKKER, INC. ALL RIGHTS RESERVED.

Neither this book nor any part may be reproduced or transmitted in any form or by any means, electronicor mechanical, including photocopying, microfilming, and recording, or by any information storageand retrieval system, without permission in writing from the publisher.

MARCEL DEKKER, INC.270 Madison Avenue, New York, New York, 10016

Current printing (last digit):10 9 8 7 6 5 4 3 2 1

PRINTED IN THE UNITED STATES OF AMERICA

Contributors to Volume 69

Steve Chum, Ph.D. Research Fellow, The Dow Chemical Company, PolyolefinsResearch, Freeport, Texas: Structure, Properties, and Applications of PolyolefinsProduced by Single-Site Catalyst Technology

Ray A. Cocco, Ph.D. Senior Specialist, The Dow Chemical Company, Midland,Michigan: Circulating Fluidized Bed Reactors: Basic Concepts and Hydrody-namics

David M. Fishbach, P.E. Senior Consulting Engineer, Starfire Electronic Devel-opment & Marketing, Ltd., Bloomfield Hills, Michigan: Nanophase Materials inChemical Process

Avery N. Goldstein, Ph.D. Research Director, Starfire Electronics Develop-ment & Marketing, Ltd., Bloomfield Hills, Michigan: Nanophase Materials inChemical Process

Manfred Grove Senior Partner, Intermacom A.G., Technology Consultants,Auerich, Switzerland: Introduction to the Selective Catalytic Reduction Tech-nology

Dennis Hendershot Rohm and Haas Company, Bristol, Pennsylvania: Funda-mentals of Process Safety and Risk Management

Trevor A. Kletz Process Safety Consultant, Cheshire, United Kingdom: Funda-mentals of Process Safety and Risk Management

Fu-Ming Lee, Ph.D. Director of Technology, GTC Technology Corporation,Houston, Texas: Recent Development of Extractive Distillation: A Distillation Al-ternative

Jim Makris Director, Chemical Emergency Preparedness and Prevention Office,U.S. Environmental Protection Agency, Washington, D.C.: Process Safety andRisk Management Regulations: Impact on Process Safety

M. Sam Mannan, Ph.D., P.E. Associate Professor of Chemical Engineering,Mary Kay O’Connor Process Safety Center, Texas A&M University, College Sta-tion, Texas: Fundamentals of Process Safety and Risk Management; Process Safetyand Risk Management Regulations: Impact on Process Industry

Rajen M. Patel, Ph.D. Technical Leader, The Dow Chemical Company, Polyo-lefins Research, Freeport, Texas: Structures, Properties, and Applications of Polyo-lefins Produced by Single-Site Catalyst Technology

H. James Overman Dow Chemical Company, Freeport, Texas: Process Safetyand Risk Management Regulations: Impact on Process Industry

iii

iv Contributors to Volume 69

Michael V. Pishko, Ph.D. Assistant Professor, Department of Chemical Engi-neering, Texas A&M University, College Station, Texas: Recent Advances in Bio-materials

Alan W. Weimer, Ph.D., P.E. Professor of Chemical Engineering, Universityof Colorado, Boulder, Colorado: Effect of Pressure and Temperature in BubblingFluidized Beds

CONTENTS OF VOLUME 69

Contributors to Volume 69 iii

Conversion to SI Units vii

Bringing Costs up to Date ix

Circulating Fluidized Bed Reactors: Basic Concepts andHydrodynamicsRay A. Cocco 1

Effect of Pressure and Temperature in Bubbling Fluidized BedsAlan W. Weimer 35

Fundamentals of Process Safety and Risk ManagementM. Sam Mannan, Dennis Hendershot, and Trevor A. Kletz 49

Introduction to the Selective Catalytic Reduction TechnologyManfred Grove 94

Nanophase Materials in Chemical ProcessAvery N. Goldstein and David M. Fishbach 150

Process Safety and Risk Management Regulations: Impact onProcess IndustryM. Sam Mannan, Jim Makris, and H. James Overman 168

Recent Advances in BiomaterialsMichael V. Pishko 194

Recent Development of Extractive Distillation: A DistillationAlternativeFu-Ming Lee 207

Structure, Properties, and Applications of Polyolefins Produced bySingle-Site Catalyst TechnologyRajen M. Patel and Steve Chum 231

v

Conversion to SI Units

To convert from To Multiply by

acre square meter (m2) 4.046 � 103

angstrom meter (m) 1.0 � 10�10

are square meter (m2) 1.0 � 102

atmosphere newton/square meter (N/m2) 1.013 � 105

bar newton/square meter (N/m2) 1.0 � 105

barrel (42 gallon) cubic meter (m3) 0.159Btu (International Steam Table) joule (J) 1.055 � 103

Btu (mean) joule (J) 1.056 � 103

Btu (thermochemical) joule (J) 1.054 � 103

bushel cubic meter (m3) 3.52 � 10�2

calorie (International Steam Table) joule (J) 4.187calorie (mean) joule (J) 4.190calorie (thermochemical) joule (J) 4.184centimeter of mercury newton/square meter (N/m2) 1.333 � 103

centimeter of water newton/square meter (N/m2) 98.06cubit meter (m) 0.457degree (angle) radian (rad) 1.745 � 10�2

denier (international) kilogram/meter (kg/m) 1.0 � 10�7

dram (avoirdupois) kilogram (kg) 1.772 � 10�3

dram (troy) kilogram (kg) 3.888 � 10�3

dram (U.S. fluid) cubic meter (m3) 3.697 � 10�6

dyne newton (N) 1.0 � 10�5

electron volt joule (J) 1.60 � 10�19

erg joule (J) 1.0 � 10�7

fluid ounce (U.S.) cubic meter (m3) 2.96 � 10�5

foot meter (m) 0.305furlong meter (m) 2.01 � 102

gallon (U.S. dry) cubic meter (m3) 4.404 � 10�3

gallon (U.S. liquid) cubic meter (m3) 3.785 � 10�3

gill (U.S.) cubic meter (m3) 1.183 � 10�4

grain kilogram (kg) 6.48 � 10�5

gram kilogram (kg) 1.0 � 10�3

horsepower watt (W) 7.457 � 102

horsepower (boiler) watt (W) 9.81 � 103

horsepower (electric) watt (W) 7.46 � 102

hundred weight (long) kilogram (kg) 50.80hundred weight (short) kilogram (kg) 45.36inch meter (m) 2.54 � 10�2

inch mercury newton/square meter (N/m2) 3.386 � 103

inch water newton/square meter (N/m2) 2.49 � 102

kilogram force newton (N) 9.806kip newton (N) 4.45 � 103

knot (international) meter/second (m/s) 0.5144league (British nautical) meter (M) 5.559 � 103

league (statute) meter (m) 4.83 � 103

vii

viii Conversion to SI Units

To convert from To Multiply by

light year meter (m) 9.46 � 1015

liter cubic meter (m3) 0.001micron meter (m) 1.0 � 10�6

mil meter (m) 2.54 � 10�6

mile (U.S. nautical) meter (m) 1.852 � 103

mile (U.S. statute) meter (m) 1.609 � 103

millibar newton/square meter (N/m2) 100.0millimeter mercury newton/square meter (N/m2) 1.333 � 102

oersted ampere/meter (A/m) 79.58ounce force (avoirdupois) newton (N) 0.278ounce mass (avoirdupois) kilogram (kg) 2.835 � 10�2

ounce mass (troy) kilogram (kg) 3.11 � 10�2

ounce (U.S. fluid) cubic meter (m3) 2.96 � 10�5

pascal newton/square meter (N/m2) 1.0peck (U.S.) cubic meter (m3) 8.81 � 10�3

pennyweight kilogram (kg) 1.555 � 10�3

pint (U.S. dry) cubic meter (M3) 5.506 � 10�4

pint (U.S. liquid) cubic meter (m3) 4.732 � 10�4

poise newton second/square meter (N ⋅ s/m2) 0.10pound force (avoirdupois) newton (N) 4.448pound mass (avoirdupois) kilogram (kg) 0.4536pound mass (troy) kilogram (kg) 0.373poundal newton (N) 0.138quart (U.S. dry) cubic meter (m3) 1.10 � 10�3

quart (U.S. liquid) cubic meter (m3) 9.46 � 10�4

rod meter (m) 5.03roentgen coulomb/kilogram (c/kg) 2.579 � 10�4

second (angle) radian (rad) 4.85 � 10�6

section square meter (m2) 2.59 � 106

slug kilogram (kg) 14.59span meter (m) 0.229stoke square meter/second (m2/s) 1.0 � 10�4

ton (long) kilogram (kg) 1.016 � 103

ton (metric) kilogram (kg) 1.0 � 103

ton (short, 2000 pounds) kilogram (kg) 9.072 � 102

torr newton/square meter (N/m2) 1.333 � 102

yard meter (m) 0.914

Bringing Costs up to Date

Cost escalation via inflation bears critically on estimates of plant costs. Historicalcosts of process plants are updated by means of an escalation factor. Several pub-lished cost indexes are widely used in the chemical process industries:

Nelson Cost Indexes (Oil and Gas J.), quarterlyMarshall and Swift (M&S) Equipment Cost Index, updated monthlyCE Plant Cost Index (Chemical Engineering), updated monthlyENR Construction Cost Index (Engineering News-Record), updated weeklyVatavuk Air Pollution Control Cost Indexes (VAPCCI) (Chemical Engineering),

updated quarterly

All of these indexes were developed with various elements such as materialavailability and labor productivity taken into account. However, the proportionallotted to each element differs with each index. The differences in overall resultsof each index are due to uneven price changes for each element. In other words,

TABLE 1 Chemical Engineering and Marshall and Swift Plant and Equipment CostIndexes since 1950

Year CE Index M&S Index Year CE Index M&S Index

1950 73.9 167.9 1973 144.1 344.11951 80.4 180.3 1974 165.4 398.41952 81.3 180.5 1975 182.4 444.31953 84.7 182.5 1976 192.1 472.11954 86.1 184.6 1977 204.1 505.41955 88.3 190.6 1978 218.8 545.31956 93.9 208.8 1979 238.7 599.41957 98.5 225.1 1980 261.2 659.61958 99.7 229.2 1981 297.0 721.31959 101.8 234.5 1982 314.0 745.61960 102.0 237.7 1983 316.9 760.81961 101.5 237.2 1984 322.7 780.41962 102.0 238.5 1985 325.3 789.61963 102.4 239.2 1986 318.4 797.61964 103.3 241.8 1987 323.8 813.61965 104.2 244.9 1988 342.5 852.01966 107.2 252.5 1989 355.4 895.11967 109.7 262.9 1990 357.6 915.11968 113.6 273.1 1991 361.3 930.61969 119.0 285.0 1992 358.2 943.11970 125.7 303.3 1993 359.2 964.21971 132.3 321.3 1994 368.1 993.41972 137.2 332.0 1995 381.1 1027.5

1996 381.7 1039.21997 386.5 1056.8

ix

x Bringing Costs up to Date

TABLE 2 Nelson-Farrar Inflation Petroleum Refinery Construction Indexes since 1946(1946 � 100)

NelsonMaterials Labor Miscellaneous Inflation

Date Component Component Equipment Index

1946 100.0 100.0 100.0 100.01947 122.4 113.5 114.2 117.01948 139.5 128.0 122.1 132.51949 143.6 137.1 121.6 139.71950 149.5 144.0 126.2 146.21951 164.0 152.5 145.0 157.21952 164.3 163.1 153.1 163.61953 172.4 174.2 158.8 173.51954 174.6 183.3 160.7 179.81955 176.1 189.6 161.5 184.21956 190.4 198.2 180.5 195.31957 201.9 208.6 192.1 205.91958 204.1 220.4 192.4 213.91959 207.8 231.6 196.1 222.11960 207.6 241.9 200.0 228.11961 207.7 249.4 199.5 232.71962 205.9 258.8 198.8 237.61963 206.3 268.4 201.4 243.61964 209.6 280.5 206.8 252.11965 212.0 294.4 211.6 261.41966 216.2 310.9 220.9 273.01967 219.7 331.3 226.1 286.71968 224.1 357.4 228.8 304.11969 234.9 391.8 239.3 329.01970 250.5 441.1 254.3 364.91971 265.2 499.9 268.7 406.01972 277.8 545.6 278.0 438.51973 292.3 585.2 291.4 468.01974 373.3 623.6 361.8 522.71975 421.0 678.5 415.9 575.51976 445.2 729.4 423.8 615.71977 471.3 774.1 438.2 653.01978 516.7 824.1 474.1 701.11979 573.1 879.0 515.4 756.61980 629.2 951.9 578.1 822.81981 693.2 1044.2 647.9 903.81982 707.6 1154.2 622.8 976.91983 712.4 1234.8 656.8 1025.81984 735.3 1278.1 665.6 1061.01985 739.6 1297.6 673.4 1074.41986 730.0 1330.0 684.4 1089.91987 748.9 1370.0 703.1 1121.51988 802.8 1405.6 732.5 1164.51989 829.2 1440.4 769.9 1195.91990 832.8 1487.7 797.5 1225.71991 832.3 1533.3 827.5 1252.9

Bringing Costs up to Date xi

TABLE 2 Continued

NelsonMaterials Labor Miscellaneous Inflation

Date Component Component Equipment Index

1992 824.6 1579.2 837.6 1277.31993 846.7 1620.2 842.8 1310.81994 877.2 1664.7 851.1 1349.71995 918.0 1708.1 879.5 1392.11996 917.1 1753.5 903.5 1418.91997 923.9 1799.5 910.5 1449.2

the total escalation derived by each index will vary because different bases areused. The engineer should become familiar with each index and its limitationsbefore using it.

Table 1 compares the CE Plant Index with the M&S Equipment Cost Index.Table 2 shows the Nelson-Farrar Inflation Petroleum Refinery Construction In-dexes since 1946. It is recommended that the CE Index be used for updating totalplant costs and the M&S Index or Nelson-Farrar Index for updating equipmentcosts. The Nelson-Farrar Indexes are better suited for petroleum refinery materials,labor, equipment, and general refinery inflation.

Since

CB � CA(B/A)n (1)

Here, A � the size of units for which the cost is known, expressed in terms ofcapacity, throughput, or volume; B � the size of unit for which a cost is required,expressed in the units of A; n � 0.6 (i.e., the six-tenths exponent); CA � actualcost of unit A; and CB � the cost of B being sought for the same time period ascost CA.

To approximate a current cost, multiply the old cost by the ratio of the currentindex value to the index at the date of the old cost:

CB � CA IB/IA (2)

Here, CA � old cost; IB � current index value; and IA � index value at the dateof old cost.

Combining Eqs. (1) and (2),

CB � CA(B/A)n (IB/IA) (3)

For example, if the total investment cost of plant A was $25,000,000 for 200-million-lb/yr capacity in 1974, find the cost of plant B at a throughput of 300million lb/yr on the same basis for 1986. Let the sizing exponent, n, be equal to0.6.

From Table 1, the CE Index for 1986 was 318.4, and for 1974 it was 165.4.Via Eq. (3),

xii Bringing Costs up to Date

TABLE 3 Vatavuk Air Pollution Control Cost Indexes (VAPCCI). First Quarter 1994� 100.0 (index values have been rounded to the nearest tenth).

1994 1995 1996Control Device (Avg.) (Avg.) (Avg.)

Carbon adsorbers 101.2 110.7 106.4Catalytic incinerators 102.0 107.1 107.0Electrostatic precipitators 102.8 108.2 108.0Fabric filters 100.5 102.7 104.5Flares 100.5 107.5 104.9Gas absorbers 100.8 105.6 107.8Mechanical collectors 100.3 103.0 103.3Refrigeration systems 100.5 103.0 104.4Regenerative thermal oxidizers 101.4 104.4 106.3Thermal incinerators 101.3 105.9 108.2Wet scrubbers 101.3 112.5 119.8

CB � CA(B/A)n (IB/IA)

� 25.0(300/200)0.6 (318.4/165.4)

� $61,200,000

Table 3 shows the Vatavuk Air Pollution Control Cost Indexes (VAPCCI) since1994. For details, see the Vatavuk Air Pollution Control Cost Indexes article involume 61.

Editor’s note: For a more thorough explanation of updating costs, see the arti-cle, ‘‘Tower Cost Updating’’ in volume 58.

john j. mcketta

Encyclopedia ofChemical Processingand Design

69

Circulating Fluidized Bed Reactors:Basic Concepts and Hydrodynamics

Introduction

Circulating fluidized beds (CFBs) consist of two basic designs, as shown in Fig. 1.One design involves a fast-fluidized bed where high gas velocities convey a sub-stantial amount of solids to one or more cyclones. The separated particles are fedback to the fluidized bed using a standpipe. The second basic design uses a riserto convey solids to one or more cyclones. The separated particles are fed to anoptional fluidized bed and then back to the riser. Solids flow rates can be controlledusing nonmechanical L- and J-valves or using a mechanical slide valve.

The large-scale commercial realization of CFBs occurred in the early 1940s,although some coal gasification was done in a fluidized bed as early as 1926 [1].With the increased demand for gasoline during World War II, major efforts wereunderway to develop reactors to crack petroleum feedstocks into usable fuels moreproductively than the moving bed or snake reactors (i.e., the Houndry Process)used at that time. The result was a fluidized catalyst cracker (FCC), where highcatalyst circulation rates allowed a balance between the exothermic burning ofcoke on the catalyst in the regenerator and the endothermic hydrocracking of petro-leum in the reactor. The continuous circulation or regeneration of catalyst providedfresh catalyst for petroleum cracking and thereby resulted in high sustainable pro-ductivities. With the addition of a stripping section after the reactor, even higheryields were obtained. The addition of steam, CO2, or other inerts would removethe product from and around catalyst particles flowing toward the regenerator.Today, the evolution of the FCC unit has results into several basic designs, asshown in Fig. 2.

In 1960, circulating fluidized beds contributed to another breakthrough processfor the petroleum and chemical industry. Standard Oil of Ohio (SOHIO) developeda fluidized-bed reactor for the ammoxidation of propene to acrylonitrile. Previoustechnology was done in tube-and-shell fixed-bed reactors. However, the high heatof reaction of 160 kcal/mol limited the economic feasibility of those units. Thehigh heat transfer characteristic of fluidized-bed reactors made them ideal for theproduction of acrylonitrile. Today, nearly all large-scale acrylonitrile plants arebased on the SOHIO design, with capacities up to 180,000 tons per year [4].

The greatest challenge in developing the SOHIO process was in the manage-ment of backmixing. The inherent hydrodynamics of fluidized beds, where solidsand, to a lesser extent, gas circulate from the top of the bed to the bottom, thento the top again, would have a deleterious effect on acrylonitrile selectivity. Toovercome backmixing, SOHIO developed sieve trays to compartmentalize the gasflow in the fluidized-bed reactor to resemble a more plug-flow characteristic [5].In 1979, SOHIO redesigned the acrylonitrile reactor to more of a ‘‘tube-and-shell’’fluidized-bed unit [6], as shown in Fig. 3.

1

2 Circulating Fluidized Bed Reactors

FIG. 1 Basic design of circulating fluidized beds.

FIG. 2 Typical FCC units based on the designs of (a) Standard Oil Development, (b) UOP, (c) Kel-logg, and (d) Exxon. (Adapted from Refs. 2 and 3.)

Circulating Fluidized Bed Reactors 3

FIG. 3 Two-dimensional schematic of the SOHIO acrylonitrile processes. (Adapted from Ref. 5.)

During the late 1970s and early 1980s oil crisis, circulating fluidized beds foundapplications in coal combustion. The high-heat-transfer capabilities of these re-actors resulted in lower operating temperatures, thereby reducing NOx and SO2

emissions. In addition, the high gas velocities resulted in significant turbulence,which provided uniform temperatures in the combustor. With the surplus of oilstarting in the late 1980s, fluidized-bed combustors became economically less at-tractive. As of the early 1990s, only Dynergy (via the Destec process) and Lurgiand Ahlstrom are practicing this technology [7].

Today, circulating fluidized beds are used in a wide array of chemical pro-cesses, as shown in Table 1. With fluidized beds having the unique distinction ofexcellent heat transfer and continuous in situ regeneration, the economic attrac-tiveness of processing thermally sensitive chemicals or using catalysts that require

TABLE 1 Some Fluidized and Circulating Fluidized Bed Reactor Processes

Product Process Developer

Acrolynitrile Propene ammoxidation SOHIOAniline Nitrobenzene hydrogenation BASF, Cyanamid, LonzaChloromethanes Cat. oxidaton of methane Asahi GlassGoal gasification Oxidation WinklerHydrocyanic acid Ammono-dehydrogenation ShawiniganMaleic anhydride Butane oxidation Alusuisse & Lummus (Alma Process)Maleic anhydride Butane oxidation DuPontMaleic anhydride Butene oxidation MitsubishiPerchlorethlyene Chlorination Diamond ShamrockPhthalic anhydride Naphthalene oxidation Badger/Sherwin-WilliamsPolyethylene Ethlyene polymerization Union Carbide (Unipol)Synthesis gas Fisher–Tropsch SASOL, KelloggVinyl chloride Ethylene chlorination, Ethyl, Hoechst, Mitsui, Toatsu, Mon-

oxychlorination santo


frequent regeneration are more realized. Once the obstacles of backmixing, masstransfer, and attrition have been addressed, these reactors often set the standardsin reactor design.

Basic Concepts

As the gas velocity through a bed of solids increases, the bed undergoes severalregimes, as shown in Fig. 4. At first, the gas velocity is insufficient to fluidizethe particles and the bed remains fixed. With increasing velocity and under idealconditions, the fixed bed expands smoothly and uniformly. Particles move in alimited fluidlike fashion and the bed pressure drop becomes constant. At this point,the bed is commonly referred to as undergoing minimum fluidization.

Further increases in gas velocity results in further bed expansion and particlesappear to freely move throughout the bed. The gas permeates through the bedwithout the formation of bubbles. This regime is referred to as smooth fluidizationand is only observed for Geldart Group A powders (see Appendix A). These pow-ders require noticeably higher gas velocities to promote the formation of gas bub-bles after minimum fluidization. In contrast, Group B powders begin bubblingshortly after minimum fluidization. Group C powders, being cohesive, may evenshow signs of bubbling prior to minimum fluidization; however, this is usually theresult of channeling.

The onset of bubbles in the fluidized bed is commonly referred to as bubblingfluidization. Here, gas bubbles form at or near the distributor and grow to a maxi-mum bubble size as they propagate through the bed. The top of the fluidized bedis still well defined, as it was in the minimum and smooth fluidization regimes.The pressure drop across the bed is still constant, on average, but starts exhibitinglarge, but regular, fluctuations with time.

As the gas velocity continues to increase, the top of the bed becomes lessdefined. Large amounts of particles are ejected into the freeboard region abovethe bed. Concurrently, sizeable regions of voidage and particle clusters are seen

FIG. 4 Various fluidization regimes with increasing superficial gas velocity.


in the bed itself. For Group A and B powders, this transition from the bubblingfluidized-bed regime is called the onset of turbulent fluidization. Group C and Dpowders may show a slugging behavior prior to the turbulent fluidization regime.

During fast fluidization, the gas velocity is sufficient enough that the surfaceof the bed can no longer be discerned. Particle density is still higher at the bottomof the unit compared to the top, suggesting that some sort of bed exists. Particleclusters and streamers are readily observed and, in some cases, a core–annulusradial variation in particle density begins to take shape. Particle entrainment is highand the total disengagement height may be well beyond the physical dimensions ofthe fluidized-bed unit. To overcome the losses of particles due to entrainment,cyclones may be used to capture entrained particles and recirculate them back intothe bed.

At very high gas velocities, nearly all the particles are entrained from the bed.This regime is commonly referred to as pneumatic conveying. In this regime, axialvariation in particle density is no longer observed, except maybe in entrance andexit regions. Radial variation in particle density can vary dramatically and rangefrom a core–annulus profile to a uniform profile. For dense systems, clusters andstreamers are readily observed.

Thus, for gas–solid systems, increases in the gas velocity results in dramaticand sometimes sharp transitions in the hydrodynamics. In the design of fluidizedbeds, it is crucial that one knows the fluidized regime that will exist at operatingconditions. The simple transition from one regime to another can have significantimpacts on reaction, heat transfer, attrition, and entrainment rates. For circulatingfluidized beds, where several regimes may exist in a single unit (i.e., from con-veying in the riser to a bubbling fluidized-bed regime in the regenerator), knowl-edge of the fluidization regimes is paramount.

In order to gain better understanding of these regimes, the methodology usedto determine the onset of each fluidization regime is discussed in the followingsections. Keep in mind that most of the correlations are empirical and may notfully represent every system. With the cost of these units running in the tens ofmillions of dollars for large-scale plants, experimental validation of the expectedregimes is critical when designing these processes.

Minimum and Smooth Fluidization

As a gas permeates through a fixed or packed bed, the pressure drop can be de-scribed by the Ergun equation [8]:

∆P

Lgc � 150

(1 � ε bp)2µug

ε 3bp(Φdp)2

� 1.75(1 � ε bp)ρgu 2

g

ε 3bp(Φdp)2

(1)

With increasing gas velocity, the bed reaches a point where the drag force exceedsthe force of gravity on the particles. The particles become mobile and the bedbecomes fluidized. The gas velocity at the onset of this type of fluidization isreferred to as the minimum fluidization velocity or umf. Assuming that the weight


of the particles in the fluidized bed corresponds to ∆P/L, the Ergun equation canbe written as

1.75ε3

mf Φ�dpumf ρg

µ �2

�150(1 � εmf)

ε3mf Φ 2 �dpumf ρg

µ � �d 3

p gρg(ρs � ρg)µ2

(2)

or

1.75ε3

mf ΦRe2

p,mf �150(1 � εmf)

ε3mf Φ 2

Re p,mf � Ar (3)

where the Archimedes number, Ar, is defined as

Ar �d 3

pgρg(ρs � ρg)µ2

(4)

and Rep is the particle Reynolds number having the expression

Re p,mf �dpumf ρg

µ(5)

Equation (3) can be written as a quadratic with the coefficients K1 and K2 havingthe form

K1 Re2p,mf � K2 Rep,mf � Ar (6)

By solving for Rep,mf, Eq. (6) can be rewritten as

Rep,mf � �� K1

2K2�

2

� �ArK1��

0.5

� � K1

2K2� (7)

where the particle Reynolds number at minimum fluidization is a simple functionof the Archimedes number and two constants (K1/2K2 and 1/K1).

Many correlations for the minimum fluidization velocity are based on Eq. (7)


TABLE 2 K1 and K2 Values for Eq. (7)

Reference K2/2K1 1/K1 Comments

Wen and Yu [9] 33.7 0.0408 For fine particlesRichardson [10] 25.7 0.0365Saxena and Vogel [11] 25.3 0.0571 Dolomite at high temperature and pressureBubu et al. [12] 25.3 0.0651Grace [13] 27.2 0.0408Chitester et al. [14] 28.7 0.0494 For large particles

Source: Adapted from Ref. 7.

for the constants K1/2K2 and 1/K1. These constants are presented in Table 2 fora wide range of studies. For typical Geldart Group A powder, the constants ofWen and Yu are most often used. However, these correlations are specific to agroup of particles with common characteristics and may not represent a less-than-ideal particle morphology and texture.

The minimum fluidization velocity can be experimentally determined by mea-suring the pressure drop across a bed of particles with increasing superficial gasvelocity. For smooth, round, and noncohesive particles, the pressure drop increaseslinearly with gas velocity until the minimum fluidization velocity is reached. Withfurther increases in the gas velocity, the pressure drop remains constant. Hence,the minimum fluidization velocity is the intersection of the linearly increasing linewith the constant-pressure-drop line.

Figure 5–8 demonstrate the results of such an experiment. Figure 5 is thepressure-drop curve for alumina particles with a mean particle diameter of 60 µmin a 4.5-in.-inner diameter fluidized bed unit. The minimum fluidization velocityfor these particles was determined to be 6.5 cm/min. When measuring the mini-mum fluidization velocity, less scatter in the data is obtained from larger or higherbeds. The scatter in Fig. 5 suggests that perhaps a higher bed should have beenused. The diameter of the fluidized bed used in this type of experiment is also

FIG. 5 Minimum fluidization curve for smooth and round alumina particles, dp,ave � 60 µm.


FIG. 6 Minimum fluidization curve for smooth and round alumina particles, dp,ave � 60 µm, withincreasing and decreasing gas velocities.

critical to obtaining accurate data. For Geldart Group A powders, bed diametersless than 3 in. can result in experimental data that are influenced by frictionaleffects at the wall. For the coarser Group B powders, the minimum diameter ismuch larger.

Figure 6 shows two pressure drop versus superficial gas velocity curves forthe same particles used in Fig. 5. The black data points are the pressure drop withincreasing gas velocity and the gray data points are the subsequent pressure-dropmeasurements with decreasing gas velocities. For round, smooth, and noncohesiveparticles, the two curves should overlap each other, as shown in Fig. 6. However,for irregular, rough, or cohesive particles, a hysterisis effect is typically observed.This is obvious in Fig. 7 for rough and irregularly shaped alumina particles with

FIG. 7 Minimum fluidization curve for rough alumina particles, dp,ave � 92 µm.


FIG. 8 Minimum fluidization curve for cohesive iron catalyst particles, dp,ave � 68 µm.

a mean particle diameter of 92 µm. This resulted in higher solid shear forces duringfluidization such that the pressure drop is dependent on previous conditions or ispath dependent. Figure 8 shows the pressure-drop curve for a catalyst supportedon alumina with a mean particle diameter of 68 µm. Here, it is almost impossibleto detect the minimum fluidization velocity. High cohesive forces result in a fluid-ized bed prone to channeling. Each peak or spike in Fig. 8 is the result of anotherchannel achieving fluidization while the majority of the bed remains fixed. Thisbehavior is typically of Group C powders.

Bubbling Fluidization

As discussed earlier, beds with Group A powders pass from minimum fluidizationto smooth fluidization to bubbling fluidization with increasing gas velocity. GroupB and D powders exhibit bubbling fluidization at the onset of minimum fluidiza-tion. Oddly enough, gas bubbles in all fluidized beds behave similarly to gas bub-bles in low-viscosity liquids [7]. Large gas bubbles are typically spherical on topand flattened or even inverted on the bottom; smaller bubbles tend to be completelyspherical. As in liquid systems, gas bubbles in fluidized beds can coalesce intolarger bubbles or split into smaller bubbles, depending on bed conditions. Also,as gas bubbles approach the top of a fluidized bed, they collapse such that solidsare propagated into the freeboard region. Higher pressures or temperatures resultin a decrease in the maximum bubble size (due to changes in the gas physicalproperties) and tend to make fluidization smoother [1].

The minimum velocity for bubble formation is referred to as the minimumbubbling velocity or Umb. For Geldart Group A and C powders, Abrahamsen andGeldart [15] proposed that the minimum bubbling velocity can be calculated from


umb

umf

�2300ρ0.13

g u0.52g

d 0.8p,ave(ρs � ρs)0.93

exp(0.72P45µm) in SI units (8)

This expression is based on observations that the minimum bubble velocity isstrongly dependent on P45 µm, the probability of finding a particle with a diameterless than 45 µm. For calculations where less information regarding the system isknown, Eq. (8) can be simplified into the form

umb � 33dp�ρg

µg�

0.1

in SI units (9)

For Geldart Group B and D powders, the minimum bubbling velocity is near theminimum fluidization velocity. Thus, the onset of fluidization and the formationof gas bubbles nearly coincide and

umb � umf (10)

As gas bubbles rise through the fluidized bed, the bubble size increases until amaximum or equilibrium size is achieved, provided that the bed is high enough.For Group A and B powders, Davidson and Harrison [16] proposed that the maxi-mum stable bubble size can be determined from

db,max �2ut

g(11)

where ut is the terminal free-fall velocity of the particle (see Appendix B). Geldart[1] found that Eq. (11) provided a better fit to experimental data if an effectivediameter, d ′p, was used to calculate the terminal velocity [i.e., ut � f(d ′p)]. Theeffective diameter is defined as

d ′p � 2.7dp (12)

For Group D powders, the maximum stable bubble size is so large that in realisticfluidized-bed applications the bubbles size is restricted by the bed diameter.

Turbulent Fluidization

Further increases in the gas velocity result in a slugging or turbulent fluidized bed.If the bed diameter is small and the bed is sufficiently high, the fluidized bed willslug before entering the turbulent fluidization regime. For Group A, B, and Dpowders, slugging is basically the result of a bubble diameter that exceeds abouttwo-thirds the bed diameter. The wall stabilizes the bubble such that almost theentire bed is translated up to the top of the bed or even higher. Group C powdersmay also exhibit slugging behavior even in large-diameter beds due to the cohesiveforces. Thus, the larger and more cohesive the particles or the smaller the beddiameter, the higher the probability of a bed exhibiting slugging. For these cases,


the minimum slugging velocity, Ums, can be estimated using the expression ofStewart and Davidson [17], where

ums � umf � 0.07(gD)0.5 (13)

For Group A or B powders in large-diameter beds, where the maximum bubblediameter is smaller than two-thirds the bed diameter, higher gas velocities resultin the onset of turbulence fluidization directly from the bubbling bed regime. Theonset of turbulent fluidization is defined as the point where the standard deviationof the pressure fluctuations reaches a maximum [18]. Grace and Bi [19] notedthat fluidization at this point reflects the balance between bubble coalescence andbreakup. Higher velocities cause this balance to shift toward a dominating bubblebreakup mechanism. The minimum velocity for the onset of turbulence fluidiza-tion, umt, can be calculated using the expression of Bi and Grace [20]:

umt � 1.24 Ar 0.45� µρg dp

� (14)

where 2 � Ar � 1 � 108 or, in terms of dimensionless numbers,

Remt � 1.24 Ar 0.45 (15)

where 2 � Ar � 1 � 108.

Fast-Fluidization Regime

Detecting the point of transition between the turbulent fluidized bed and the fast-fluidized bed regimes has been under debate for many years. Yerushalmi [21]proposed that this transition occurs when a significant number of particles becomeentrained from the fluidized bed. This transition can be observed by detecting asudden change of pressure drop with respect to entrainment rate for increasingsuperficial velocities. The superficial gas velocity corresponding to this point isreferred to as the minimum transport velocity or umr. Schnitzlein and Weinstein[22], however, were unable to determine umr using Yerushalmi’s method. Theirobservations suggested that the value of umr was strongly dependent on the locationand the distance separating the two pressure taps used to measure the pressuredrop.

Bi and Grace [20] measured the entrainment rate versus superficial velocity fora wide range of fluidized-bed systems. They noted that the onset of fast fluidizationcorrelated to the point where significant particle entrainment was observed. ForGroups A and B powders, this minimum transport velocity can be estimated withthe expression

umr � 1.53 Ar 0.45� µρgdp

� (16)


where 2 � Ar � 4 � 106. For the larger Group D powders, Eq. 16 may underpredictumr compared to using the terminal velocity as the minimum transport velocity.Under these conditions, the minimum transport velocity should be set to equal theterminal velocity [19].

Pneumatic Conveying

As the velocity continues to increase in a fast-fluidized bed, the axial transitionbetween the dense and lean regions disappears. This transition marks the onset ofthe pneumatic conveying regime and the superficial velocity corresponding to thispoint is called the minimum conveying velocity or umc. The onset of pneumaticconveying is readily measured by starting with a high superficial gas velocity anddecreasing its value while holding the entrainment rate constant. The superficialgas velocity where the suspension collapse (i.e., choking) is observed correspondsto the minimum conveying velocity [19]. Yang [23] proposed that this velocitycan be estimated with the equation

2gD(ε�4.7mc � 1)

umc/εmc � ut

� 681,000�ρg

ρs�

2.2

(17)

Regime Map

The bed hydrodynamics associated with each powder classification used in fluid-ized beds are summarized in Fig. 9. Increasing the gas velocity through a powderbed results in the transition of several regimes, which are dependent on the particleproperties. For Group A powders, increasing the gas velocity results in bed hydro-dynamics that start in the fixed-bed regime and continue through smooth fluidiza-tion, bubbling fluidization, turbulent fluidization, fast fluidization, and, finally,pneumatic transport. Group B powders exhibit the same types of transition, withthe exception that smooth fluidization is typically not observed; the onset of fluid-

FIG. 9 Flow regime map for various powders. Slugging for Group A and B powders depend onvessel diameter. Group C powders tend to slug and Group D powders almost always exhibitslugging.


ization corresponds with the onset of bubble formation. For Group C powders,smooth and bubbling fluidization are often replaced by channeling due to largesolid stresses or large cohesive forces on these particles. Group C powders alsohave a higher tendency to exhibit slugging prior to turbulent fluidization. GroupD powders often exhibit slugging during the onset of fluidization. The large particlesizes and/or high densities of Group D powder result in the formation of largemaximum bubble diameters that often exceed the diameters of commercial-scalefluidized beds. In contrast, slugging for Group A, B, and, to a lesser extent, Cpowders may not be observed in large commercial units where the effective beddiameter often exceeds the maximum stable bubble size.

In an effort to develop a design guide for fluidized unit operations, variousattempts have been made to quantify the flow regimes for gas–solid flow. Reh[24] first made the attempt for a unified flow regime map by comparing the parti-cle’s Reynolds number, Rep, to the inverse of the drag coefficient, 1/CD. Furthermodification to Reh’s map were made by Werther [25]. Li and Kwauk [26] andAvidan and Yerushalmi [27] took a different approach to developing a unified flowregime map by comparing the relationship of the superficial gas velocity to solidsvoidage. This approach was further modified by Rhodes [28]. Yet, another ap-proach was taken by Leung [29], Klinzing [30], and Yang [23], who comparedthe superficial gas velocity to the solid flux.

In each of these cases, however, the transition from the fast-fluidized bed toa pneumatic transport regime was not well defined. Grace [31] resolved this prob-lem by generating a unified flow regime map based on the dimensionless variablesd*p and u* of Zenz and Othmer [32]. By comparing the dimensionless particlediameter, d*p � Ar 1/3, with the dimensionless gas velocity, u* � Rep/Ar 1/3, Gracewas able to discern the transition to pneumatic conveying. Kunii and Levenspiel[7] further modified Grace’s work by including the observations of van Deemter[33], Horio et al. [34], and Catipovic et al. [35]. Figure 10 shows the result of thiseffort. It is interesting to note that Kunii and Levenspiel’s [7] demarcation fromfast fluidization to pneumatic conveying is not identical to that first proposed byGrace. Kunii and Levenspiel describe this transition as not being well defined.

With the aid of Fig. 10 and the relatively simple calculation of d*p and u*, theflow behavior of most gas–solid systems can be readily obtained. Keep in mind,however, that the boundary for each regime shown in Fig. 10 should not be consid-ered as sharp transitions, but more of a gray area between two adjacent regimes.

Pressure Balance in a CFB

In designing a circulating fluidized bed (CFB), special attention needs to be givento the pressure-loop calculations. This is especially true in the proper design ofstandpipes and mechanical or nonmechanical valves (i.e., L-valves, J-valves, slidevalves). The bed height in a standpipe counterbalances the dynamic pressure dropoccurring in the riser or fast-fluidized-bed section. Mechanical or nonmechanicalvalves provide better response for this counterbalancing. For instance, an increasein the gas flow rate in a riser would require a constriction in a slide valve or lessaeration in a L- or J-valve to prevent backflow. This correction results in a higherstandpipe bed height and pressure drop. The higher pressure in the standpipe coun-


FIG. 10 Flow regime diagram proposed by Kunii and Levenspeil. (From Ref. 7.)

terbalances the increased pressure in the riser. Counterbalancing can also happenwithout the aid of mechanical or nonmechanical valves, but the response time isslow and the probability for backflow is much greater.

Figures 11–13 give qualitative examples of the pressure-loop behavior in theCFBs shown in Fig. 1. The pressure drop in the riser or fluidized-bed section mustequal the pressure drop in the cyclone, fluidized bed, standpipe, and control valvesections. For example, in Fig. 11, the pressure loop is defined as

[(P4 � P3)]fluid bed � [(P5 � P4)]freeboard

� [(P6 � P7) � (P5 � P6)]cyclone (18)

� [(P8 � P9) � (P7 � P8)]standpipe

or

∆P|fluid bed � ∆P|freeboard � ∆P|cyclone � ∆P|standpipe (19)

Assuming that particle frictional (both particle–particle and particle–wall) and par-ticle acceleration effects are negligible, the pressure drop in the fluidized bed canbe approximated by

∆P|fluid bed � Lbρsg(1 � ε) (20)

where Lb is the height of the fluidized bed. Similarly, the freeboard can be described


FIG. 11 Pressure loop for a CFB of Design I in Fig. 1.

using

∆P|freeboard � ρsg�Lu

Lb

[1 � ε(z)] dz � ρsg(1 � ε)(Lu � Lb) (21)

where Lu is the height of the unit. Because the voidage profile, ε(z), in the freeboardregion is rarely known, Eq. (21) can be approximated by using the average voidagein the freeboard ε.

For the right-hand side of Eq. (18), the pressure drop in the cyclone can bedetermined using the expression described by Muschelknautz and Greif [36],where

∆P|cyclone � ∆P|wall friction � ∆P|vortex losses (22)

The first term of this expression corresponds to the friction losses at the wall dueto gas. The second terms accounts for momentum losses in the vortex. Under theassumption that the flow over the wall of the cyclone is similar to flow over a flatplate, the pressure drop due to wall friction in a cyclone can be calculated using

∆P|wall friction � λpρg Asurface

1.8Qg

(u0ub)0.67 in SI units (23)

The pressure drop in the vortex depends on the average velocity in the vortex tubeand the tangential velocity in the vortex tube or exit. This pressure drop can be


calculated from

∆P|vortex losses � ρgv 20�2 � 3 �u0

v0�

1.33

�u0

v0�

2

� in SI units (24)

where the average velocity in the vortex tube is

v0 �Vg

πr 20

(25)

For the standpipe, the pressure drop is determined in a method similar to that usedfor the fluidized bed under the same assumptions. The pressure drop in the leanphase of the standpipe can be estimated as

∆P|standpipe, lean phase � ρsg �Ls ,0

Ls,b

[(1 � ε)(z)] dz � ρsg(1 � ε)(Ls ,0 � Ls,b) (26)

where Ls,0 and Ls,b are the height of the standpipe and the dense-phase bed in thestandpipe, respectively. For the dense-phase region of the standpipe, the pressuredrop is approximated as

∆P|standpipe, dense phase � ρsg(1 � ε)Ls,b (27)

For the CFB in Fig. 12, the pressure loop has the expression

FIG. 12 Pressure loop for CFB with a fluidized bed as shown in Design II of Fig. 1.


FIG. 13 Pressure loop for CFB with no fluidized bed as shown in Design III of Fig. 1.

(P4 � P3)riser � [(P5 � P4) � (P6 � P5)]cyclone

� [(P7 � P6) � (P8 � P7)]1st Standpipe (28)

� [(P9 � P8) ]2nd standpipe � [(P10 � P9) � (P3 � P10)]L-valve

or

∆P|riser � ∆P|cyclone � ∆P|1st standpipe � ∆P|2nd standpipe � ∆P|L-valve (29)

Similarly, the pressure loop for the CFB in Fig. 13 can be described as

(P4 � P3)riser � [(P5 � P4) � (P6 � P5)]cyclone

� [(P7 � P6) � (P8 � P7)]standpipe (30)

� [(P9 � P8) � (P3 � P9)]L-valve

or

∆P|riser � ∆P|cyclone � ∆P|standpipe � ∆P|L-valve (31)

A similar set of expressions can be obtained for the CFB in Fig. 13. Comparedto the pressure-loop calculations used for the system in Fig. 11, the only newexpressions needed to complete the pressure-loop calculations in Figs. 12 and 13are the pressure drop across the L-valve. Yang and Knowlton [37] noted that the


pressure drop across an L-valve is similar to that used by Jones and Davidson [38]for a mechanical valve where

∆P|mechanical �1

2ρs(1 � εmf) � Ws

CDA0�

2

(32)

For mechanical valves, the valve opening area, Ao, may vary during operation butis always known. For an L-valve, however, the opening area, Ao, depends on theamount of aeration and is not known. Yang and Knowlton [37] found an empiricalexpression for this opening area of the L-valve based on 158 experimental datapoints. Their results showed that the opening area can be expressed as

Ao � 1.41Qaeration

ut

� 0.0623πD 2

L-valve L L-valve

ut

in Foot Pounds Seconds (FPS) units

(33)

Thus, pressure drops through a CFB can be estimated and can provide the basicfoundation for design. However, as these units become more complicated or areoperated at higher and higher solid circulation rates, this procedure may not beenough. Indeed, the pressure-drop calculations presented assume that frictional andacceleration effects are negligible. Although this may be a safe assumption forrough estimates of the pressure loop, detailed design calculations cannot neglectthese effects. In addition, particle properties such as roughness, sphericity, andparticle size distribution can also have a significant effect on the pressure drop.Indeed, the addition of fines to a Group A powder can result in reduced pressuredrops for fluidized beds and risers. Using computation fluid dynamics, Sinclairand Mallo [39] demonstrated that this may be due to the fact that smaller particlesdampen the wakes generated by larger particles or bubbles. Hence, a clear under-standing of the hydrodynamics is needed to fully describe pressure loss in CFBs.Fortunately, this has been the focus of intensive research in the last few years.

Gas–Solid Hydrodynamics

In general, the hydrodynamics of a riser can be divided into macro-scale and meso-scale flow behavior. Macro-scale behavior is mostly concern with solids concentra-tion profiles and solids velocity on a large scale. Risers typically exhibit a wideand diverse range of axial and radial solids profiles that are highly dependent onoperating and design conditions. For instance, the design of the entrance and exitregions of a riser can influence the solid profile throughout the riser. Hence, it isimportant to understand this macro-scale behavior in order to provide the correctriser design for the desired solids concentration and velocity profiles.

To make matters even more complicated, fast-fluidized beds and risers haveshown evidence of dynamic meso-scale flow behavior in the form of particle ag-glomeration called clusters and streams. The size and frequency of these clusters


and streams are also highly dependent on operation and design conditions. Boththe cluster size and frequency have a substantial impact on both catalytic reactionrates and heat transfer.

Fortunately, a clearer understanding of these macro-scale and meso-scale be-haviors are under intensive investigations. Today, fast-fluidized beds and riserscan be designed with the solids concentration and radial profile in mind. However,this can only be achieved if one has an understanding of gas–solid hydrodynamics.

Macro-Scale Behavior: Solids Profile

Axial Profile of Solids in a Fast-Fluidized Bed and Riser

In 1971, Reh noted that there exists an axial gradient of solids concentration in ariser similar to that observed in a fast-fluidized bed. Figure 14 illustrates thesesubtitle differences in the axial solids concentration profile commonly observedthroughout a fast-fluidized bed and riser. As the gas velocity increases through afluidized bed, the boundary between the dense fluidized-bed region and the leanfreeboard region becomes indistinguishable. Indeed, having a distinguishable bedheight is one of the indicators for fast fluidization (see the subsection Fast-Fluid-ized Regime). However, it was surprising that this axial gradient also exists in ariser (i.e., pneumatic conveying region) where higher gas velocities are used.

This behavior was later quantified by Li and Kwauk [26], Weinstein et al. [40],Hartge et al. [41], and Rhodes et al. [42], who found that the axial gradient of thesolids concentration exhibited an S-shaped curve. Horio [43] suggested that thisS-shaped curve is restricted to units with a large L/D, riser length to diameter ratio.Large-scale units, such as atmospheric fluidized-bed combustors, may not exhibitthis axial profile. Unfortunately, data on large units are limited and the solids con-centration profiles in these units are still a subject of debate.

FIG. 14 Axial solids concentration profile in (a) a fast-fluidized bed and (b) a riser.


FIG. 15 Length-normalized pressure drop in a riser with increasing solids flux.

Recent studies have shown that the design of the entrance and exit region ina riser can have a substantial effect on the resulting performance. As illustratedin Fig. 15, the pressure drop, normalized by the distance between pressure taps,versus the riser length to diameter ratio, L/D, can be affected by the entrance designfor 20 to 30 L/D’s. Furthermore, this effect becomes more pronounced with highersolids fluxes. Using an x-ray absorption technique, Kostazos et al. [44] were ableto further substantiate this effect by examining the radial profile in a riser at variedfeed ports. Their results showed that an asymmetric position of the feed manifesteditself in the asymmetry of the radial profile at an axial position of up to 30 L/D’s.

Radial Profile of Solids in a Fast-Fluidized Bed and Riser

The radial profile of solids that exists in fast-fluidized beds and risers is even moresurprising. At some point beyond the entrance region of the fast-fluidized bed orriser, particles segregate toward the wall to form a core–annulus profile, as illus-trated in Fig. 16. Studies using kinetic sampling probes, a γ-ray densitometer, andfiber optic probes were able to resolve this core–annulus profile [1,45–48]. Theirresults showed that the core consists of a lean concentration of solids moving upthe riser, whereas the annulus consists of a dense concentration of particles. Atmoderate solids fluxes, particles in the annulus region actually exhibit a downwardvelocity, as shown in Fig. 16 [46,48,49]. Karri and Knowlton were able to quantifythis downflow as a function of radial profile by measuring the solid mass fluxesin a 20-cm-diameter by 14-m-high riser [49]. Figure 17 presents their results wheredownflow in the annulus regions was observed for solid mass fluxes of 49 and 93


FIG. 16 Representation of the core–annulus profile in a riser where downflow is observed near thewalls. (Adapted from Ref. 48.)

kg/m2 s. Miller and Gidaspow [50] showed that the largest magnitude of annulardownflow flux at and near the wall was near the bottom of the riser. At less than2 m from the inlet of Miller’s 7.5-cm-diameter riser, downward fluxes were severaltimes the average feed flux [51].

The implications of this behavior can be substantial. For many catalytic reac-tions, backmixing near the feed region and, to a lesser extent, up throughout theriser can have a significant impact on productivity. Fortunately, many of thesereactions require very high solids fluxes where downflow may be less of an issue.For example, Fig. 17 shows that operating Karri and Knowlton’s riser [49] at orabove solid fluxes of 195 kg/m2 s, results in a core–annulus profile where particlesat the wall move in the same direction as those in the core region (positive solidsmass flux of �1 kg/m2 s). In this case, backmixing was limited. Similar findings

FIG. 17 The effects of solids mass flux on the radial net solids mass flux profile. (Adapted fromRef. 49.)


FIG. 18 Representation of the core–annulus profile in a riser where upflow is observed near thewalls. (Adapted from Ref. 49.)

have also been reported by Issangya et al. [52]. A representation of this behavioris presented in Fig. 18.

Particle segregation appears also to be influenced by the core–annulus profilein a riser. Karri and Knowlton [49] observed that in the presence of downflow,the particle size distribution in the annulus region was larger than that found inthe core. In contrast, this effect appears to only occur for downflow operations.Particle segregation was not observed for core–annulus upflow profiles for eithervery high or low solid mass fluxes [49]. Jones et al. [53] examined this phenomenausing, the Laser Doppler Velocimetry (LDV) of particle-laden jets. Their resultsshowed that eddies or recirculation zones were responsible for this particle segrega-tion. Hence, the high shear and resulting recirculation zones generated from thesolids downflow near the wall may be responsible for the segregation effect ob-served by Karri and Knowlton [49]. With upflow at the wall, the low shear maynot generate strong enough eddies to effect the particle size distribution across theriser diameter.

There are also design features that can reduce backmixing in risers. Bafflescan induce wakes and turbulence, which limit the core–annulus profile. Of course,the added attrition caused by baffles needs to be factored into the design process.Another option is to use secondary feeds to produce a higher plug flow or uniformsolids velocity profile at the entrance region. A core–annulus profile may stilldevelop further up the riser, but backmixing is less severe in this region.

As with the axial profile, the design of the entrance and exit region can havea substantial effect on the solid radial concentration profile. Rhodes et al. [42] useda nonisokinetic sampling probe to examine the radial solids loading in a 0.09-m-inner diameter by 7.2-m-high riser. Their results showed that a side solids feedresulted in a nonuniform radial distribution of solids beyond 40 L/D’s, as depictedin Fig. 19. In addition, Rhodes et al. noted that the asymmetries in solids radialdistribution were more noticeable in the interphase between the dense and diluteregions. Thus, depending on the design of the feed region, a nonuniform radialprofile may exist throughout many industrial risers.

In a similar fashion, the exit configuration of a riser can have an impact onthe solids profile for several L/D’s below the exit region. Brereton and Grace [54]observed this effect for smooth and abrupt riser exits. As shown in Fig. 20, using


FIG. 19 Illustration of the nonuniform solids radial profile in a riser due to solids feed on the sideof the riser (not drawn to scale). (Adapted from Ref. 28.)

FIG. 20 Effect of exit configuration on solids volume fraction for a 0.15-m-diameter by 9.3-m-highriser with a superficial gas velocity of 7.1 m/s, initial solids flux of 73 kg/m2 s, and 148-µm sand particles. (From Ref. 54).


FIG. 21 Solids flux ratio with respect to radial position for a riser with a smooth, rounded exit at asuperficial gas velocity of 4.2 m/s, solids flux of 50 kg/m2 s, and 80-µm sand particles.(From Ref. 55.)

a smooth, wide-radius bend to terminate the riser resulted in little deviation in theaxial solids concentration profile. However, an abrupt bend, such as a square bendor tee, resulted in backmixing, which affected the overall riser solids volume frac-tion profile up to 20 L/D’s below the exit region.

Similar effects for solids fluxes are reflected in the data of Kruse and Werther[55] who compared normalized solid fluxes to radial solids loadings for a 0.4-m-diameter by 15.1-m-high riser, as shown in Figs. 21 and 22. For smooth bends,substantially less downflow is observed compared to the abrupt exit configurations.In addition, the region of downflow for the abrupt exit configuration was overtwice the size of that observed for the smooth configuration.

These results provide a good example of the importance of riser design forchemical production. For combustors, where backmixing is tolerable and some-times even desired, asymmetric feed designs and abrupt exits are less critical. How-

FIG. 22 Solids flux ratio with respect to radial position for a riser with an abrupt, squared exit at asuperficial gas velocity of 4.2 m/s, solids flux of 71 kg/m2 s, and 80-µm sand particles.(From Ref. 55.)


FIG. 23 Illustration of riser entrance region for a more uniform solids loading profile.

ever, in chemical production such as in oxidation and chlorination, asymmetricsolids profiles and backmixing can seriously reduce selectivity and activity. Fortu-nately, both the entrance and exit can be designed such that minimal asymmetricsolids profiles and backmixing ensues. For the entrance region, care needs to betaken such that entering solids are well mixed with the entraining gas. One suchdesign is shown in Fig. 23. A fluidizing gas is used to distribute incoming solids,and one or more jets are used to entrain catalyst into the riser. Similarly, the exitregion should have either a long radius bend or a disengagement section. Typically,industrial risers have a stripper section at the top of the riser to not only strip gasbut also minimize exit effects on the riser, as shown in Fig. 24.

Meso-Scale Behavior: Clusters and Streamers

Riser sections in circulating fluidized beds exhibit a core–annulus profile withdownward flow resulting in the formation of clusters and streamers of particles.

FIG. 24 Illustration of riser exit region (i.e., stripper) for a more uniform solids loading profile.


This phenomenon was proposed by Squires et al. [56] and Yerushalmi and Avidan[21,57]. Under the assumption that the pressure drop equals the weight of the solidsin suspension, the resulting slip velocity, calculated as

vslip � vg �Gs

ρs εs

(34)

was found to be several times larger than the terminal velocity [51]. Because theslip velocity cannot exceed the terminal velocity, it was postulated that the particlesmust be forming clusters that effectively act as larger particles. Today, cluster andstreams are frequently observed. High-speed movies [24], laser sheet [58], infraredimagining [59], and fiber-optic probes [60–62] all reveal the presence of wavepacket of particles near the riser wall moving with the downward annulus flow incontinuous but dynamic and unstable sheets. These sheets of particles (commonlycalled clusters, streamers, swarms, strings, or strands) are represented in Fig. 25.

Laser sheet and fiber-optic studies of Horio [43] and Rudnick and Werther[61] have further demonstrated that these clusters are three dimensional in natureand can be found in the annulus and core regions. In general, it was observed thatclusters move in a direction parallel to the flow of the suspension phase. In otherwords, in a core–annulus profile with downward flow at the walls, clusters in theannulus region flow downward and clusters in the core region flow upward.

Soong et al. [63] experimentally measured the cluster length and time-averagedlocal solid volume fraction in riser flow. Their results were in agreement withYerushalmi and Avidan’s [57] earlier empirical correlation of

dc � dp � εs(0.027 � 10dp) � 32ε 6s (35)

However, Soong et al. replaced the average local solids volume fraction of parti-cles, εs, by the solids volume fraction of a cluster, εcl, as

dc � dp � εcl(0.027 � 10dp) � 32ε 6cl (36)

FIG. 25 An illustration of the cluster or streamers often observed in risers.


Gu and Chen [64] further correlated Soong’s data such that the solids volumefraction of a cluster can be related to the local solids volume fraction of particlesusing the expression

εcl � εs,max�1 � � εs

εs,max�

3.4

� (37)

where the maximum local solids volume fraction, εs,max, is 0.57. Tsuo and Gidaspow[65] observed that, for Group A powders, the clusters were 2–3 cm in lengthin the down-flowing annulus region. Furthermore, cluster density increased withincreasing solids flux, decreasing gas velocity, or decreasing pipe diameter.

The mechanism for the formation and degeneration of clusters in a riser is stillunder some dispute. Tsuo and Gidaspow [65] proposed that clusters are the resultof partially inelastic collision with the walls. Particles hit the wall, lose energy,and fall, to collide with another particle. This process continues until a cluster isformed.

Senior and Grace [66] proposed that inelastic wall collisions cannot accountfor all the energy loss needed to form clusters. For this to happen, wall collisionswould need a coefficient of restitution of less than 0.1, which is unlikely. Perfectlyelastic collisions have a coefficient of restitution of 1. Instead, Senior and Graceproposed that the balance between shear-induced lift and drag forces on a particleact to momentarily detain particles at the wall region of the riser. After a particlecollides with the wall, it has insufficient momentum to counteract the lift force.As a result, the particle continues to hit the wall, each time losing more lateralvelocity. When the particle-to-wall friction slows the particle below the local gasvelocity, lift forces acting in the opposite direction move the particle away fromthe wall to the near-wall regions. Clusters form when many particles undergo thislateral-velocity-reduction process. The downward motion of a cluster is caused bythe net lift forces on a cluster being less than the sum of the forces on each individ-ual particle.

Particle migration to the wall is not only dependent on axial and lateral veloci-ties but also on the particle diameter. Lee and Durst [67] found that 100- and 200-µm-diameter glass beads readily accumulated at the wall, whereas larger particles,with a 400–800-µm diameter, did not. The larger particles were traveling at sig-nificantly lower axial velocities and were less influenced by lift forces directed tothe wall. Tsuji et al. [68] was also able to measure this crossover of particles tothe wall for smaller particles. Senior and Grace [66] were able to model thesetrajectories and found similar conclusions. For particles larger than 500 µm indiameter, no range of lateral velocities was found that slowed the particle signifi-cantly enough for crossover to the wall. Yet, for 230-µm particles, a significantconcentration of particles was predicted to accumulate near the wall for initiallateral velocities of 4.5–5.5 cm/s. For 40-µm particles, concentrations were foundto be an order of magnitude larger at the wall than that found for 230-µm particleswith lateral velocities ranging from 3 to 22 cm/s.

What is interesting here is that if particle collisions with the wall help createthe formation of clusters, how do cluster form in the core region as observed byHorio [43] and Rudnick and Werther [61]. Furthermore, Karri and Knowlton [49]


observed that particles in the annulus region with a downward flow had a largerparticle size distribution than that in the core. In a core–annulus profile where bothregions have an upflow profile, no particle size distribution effects were observed.Both of these results contradict the above postulate of Senior and Grace in whicha wall is needed for cluster formation and larger particles prefer the core region.Most likely, the magnitude of the solids flux or solids concentration may have asignificant impact on the locality of cluster formation and particle-size-segregationeffects. These macroscopic properties may also need to be considered.

Horio [43] proposed that another mechanism may be responsible for clusterformation. A particle in flight can have either attractive or repulsive forces withnearby particles. Two particles traveling perpendicular to the gas flow tend to re-peal each other, whereas two particles aligned parallel with the flow tend to attracteach other due to the nearest-neighbor effect on lift and drag. Yet, in riser flow,particle alignment is not stable, as particles undergo collisions, bumping, tumbling,and other nonelastic processes. This may be the very mechanism to ensure cluster-ing. The combination of nonelastic processes and the parallel-aligned particle flowmay provide the attractive force needed to promote clustering.

Furthermore, Horio noted that buoyancy forces dominate in lean-phase regions,whereas gravitational forces dominate in dense-phase regions. This results in ashear between clusters and the lean-phase region. The interaction of these forcescontrols the development of particle groups to form steady but turbulent structuresof a certain characteristic length. Two-dimensional simulations of particles showthis very event, where a homogeneous suspension evolved into clusters [69].

Closing Remarks

To date, a complete description of the CFB hydrodynamics is not possible evenafter decades of intensive research. Hence, predicting accurate pressure profiles,reaction productivities, heat transfer, and gas and solid residence times is still moreart than science. Yet, circulating-fluidized-bed reactors have been in productiveuse for more than 50 years. Indeed, a broad knowledge base exists for the designof fluidized catalyst crackers using empirical-based models. Even though a funda-mental understanding of the physics behind gas–solid flow is limited, the design,construction and operation of these units can be done in a relatively short periodof time.

For the design, construction, and operation of CFBs in the chemical industry,this is not the case. Unlike the petroleum industry where FCC catalysts have similarproperties, the physical properties of the catalyst commonly used in the chemicalindustry are diverse. Thus, it is more important to understand and apply at leastthe known physics in gas–solid flow. This is not a daunting task; just careful con-sideration of each design aspect of the CFBs needs to be addressed.


Nomenclature

Ar Archimedes’ numberAsurface Surface areadb Bubble diameterdb,max Equilibrium bubble diameterdc Cluster diameterdp Particle diameterdp′ Effective particle diameterdp,ave Average particle diameterd*p Dimensionless particle diameterD Vessle diameterDL-valve Diameter of an L-valveg Acceleration of gravitygc Newton’s law proportionality factor, 9.8 m/s2 or 32.2 ft/s2

Gs Solids flux rateL LengthLb Height of fluidized bedL L-valve Height of L-valve (distance between L-valve feed and horizontal leg into

reactor)Lu Height of fluidized-bed vesselP45µm Proportion of particles with a diameter less than 45 µmQaeration Volumetric flow rate of the gas for L- or J-valveQg Volumetric flow rate of the gasr RadiusR Vessel radiusRep Particle Reynolds numberRepmf Particle Reynolds number at minimum fluidizationug Superficial gas velocityumc Superficial gas velocity at the onset of pneumatic conveying in a fluid-

ized bedumb Superficial gas velocity at the onset of bubbling in a fluidized bedumf Superficial gas velocity at minimum fluidizationumr Superficial gas velocity at the onset of fast fluidization in a fluidized bedums Superficial gas velocity at the onset of slugging in a fluidized bedumt Superficial gas velocity at the onset of the turbulence regime in a fluid-

ized bedui Inlet velocity of a cycloneu0 Tangential velocity at the wall of a cycloneu* Dimensionless gas velocityut Particle terminal velocityVg Volumetric gas flow ratevslip Slip velocityz Distanceε Solids void fractionεbp Solids void fraction for a packed bedεcl Average solids void fraction of a cluster


εms Solids void fraction at the onset of pneumatic conveying in a fluidizedbed

εmf Solids void fraction for a bed at minimum fluidizationεs Average solids void fraction∆P Pressure dropλp Dimensionless solids friction coefficientµ Gas viscosityρg Gas densityρs Solids densityΦ Sphericity

Appendix A: Geldart Powder Classification

Simpson and Rodger [70], Jackson [71], and Verloop and Heertjes [72] suggestedthat the fluidization of particles can be classified into two categories: particulateand aggregative. The particulate category pertains to powders that fluidize in aliquidlike fashion. As the superficial gas velocity increases, these particles movefurther apart in an independent fashion. In contrast, particles in the aggregatescatagory would exhibit bubble formation with increasing gas velocities. For theaggregate particles, the resulting dense phase remains unchanged in terms of solidsconcentrations after initial fluidization. Most particles adhere to the aggregativebehavior.

Unfortunately, this two-category methodology falls short in adequately describ-ing fluidization behavior. The particulate category is limited to a small group ofparticles leaving the aggregative category to describe everything else. To addressthis shortcoming, Geldart and Rhodes [1,73] demonstrated that particles could beclassified into four distinct categories or groups. Today, these groups are referredto as Geldart Group A through D.

Figure A1 illustrates how the fluidization behavior of a particular powder canbe predicted using the Geldart Group classification. This graph is the basic founda-tion of modern-day fluidization engineering. By comparing the particle density(less the gas density) with the mean particle diameter, the ‘‘type’’ of fluidizationcan be determined. In general, Group A powders undergoing fluidization behavesignificantly different than the other groups. Thus, the design and operation of afluidized-bed unit containing a Group A powder would not be the same as thatused for a Group B powder. To better clarify these differences, a description ofthe fluidization behavior is presented in the order of increasing particle size.

Geldart Group C powders are typically less than 50 µm in diameter and arethe most difficult to fluidize. These particles are considered cohesive and almostalways exhibit significant channeling during fluidization. To limit this effect,Group C powders are usually fluidized with the aid of baffles and/or mechanicalvibration. Sometimes, larger particles, such as Group B powders, are added to thebed to promote smoother fluidization.

Geldart Group A powders are the most common type of powder used in fluid-ization. For example, most FCC units are designed for Group A powders, which


FIG. A1 Geldart powder classification at ambient conditions. (Adapted from Ref. 7.)

typically have a mean particle diameter of 75 µm and a particle density of 1.0 g/cm3. At low gas velocities, Group A powders tend to exhibit significant bed expan-sion without the formation of bubbles (i.e., smooth fluidization). At higher gasvelocities (i.e., greater than umb), bubbles appear and rise more rapidly than thegas in the rest of the bed. Gas in the dense or emulsion phase tend to percolatethrough the bed compared to the residence time of the gas in the bubbles. Typically,Group A powders do not promote maximum bubble size greater than 10 cm [7].

Geldart Group B powders have particle diameters typically ranging from 200to 800 µm. Unlike Group A powders, where smooth fluidization is observed, thesepowders exhibit the formation of bubbles at the onset of fluidization. The bubblesize in Group B powders can be large, on the order of feet in some cases. GroupB powders fluidize easily and are used in a wide range of fluidization unit opera-tions with little difficulties. Care should be taken that slugging does not occur insmaller fluidized beds.

Geldart Group D powders have the largest particle diameters of all other Geld-art groups. As a result, gas requirements for fluidization are large. These powdersare typically processed in spouting beds where gas requirements are less than thatneeded in standard fluidized beds. During fluidization, Group D powders haveenormous bubble diameters and slugging is commonly observed even in large flu-idized beds.

It should be noted that Figure A1 was developed for particles at ambient condi-tions. Under high pressures, a Group B powder may behave as a Group A powderFurthermore, the transition from one group to another is not well defined. In somecases, the behavior of a powder can be classified under more than one group. Forexample, some powders fluidize well as a Group A powder, but become perma-nently defluidized as a Group C powder in nonaerated horizontal sections of aCFB. These powders are sometimes referred to as Group AC powders.


Appendix B: Terminal Velocity

The velocity of a free-falling particle will increase according to the accelerationof gravity. At some point, the drag on the particle will reduce the acceleration tozero and the particle reaches its terminal velocity. Haider and Levenspiel [74]defined the terminal velocity of a single particle as

u*t � � 18(d*p )2

�2.335 � 1.744Φ

(d*p )0.5 ��1

(B1)

where

u*t � ut � ρ2g

µ(ρs � ρg)g�1/3

Thus, to avoid significant carryover in a fluidized bed, the superficial gas velocityshould be less than the terminal velocity, ut.

References

1. D. Geldart, ‘‘Introduction,’’ in Gas Fluidization Technology (D. Geldart, ed.), JohnWiley & Sons, Chichester, 1986, p. 5.

2. R. H. Perry and C. H. Chilton, Chemical Engineers’ Handbook, 5th ed., McGraw-Hill, New York, 1973, p. 20–71.

3. M. Sittig, Chem. Eng., 60, 219–231 (1953).4. K. Weissermel and H.-J. Arpe, Industrial Organic Chemistry, 2nd ed., VCH, Wein-

heim, 1993, p. 302.5. J. L. Callahan, E. C. Milberger, and R. K. Grasselli, ‘‘Process for Preparing Olefin-

ically Unsaturated Aldehydes and Nitriles,’’ U.S. Patent No. 3,4827,343 (February11, 1969).

6. J. L. Callahan, H. F. Hardman, and E. C. Milberger, ‘‘Reactor for Contacting Gasesand a Particulate Solid,’’ U.S. Patent No. 4,152,393 (May 1, 1979).

7. D. Kunni and O. Levenspiel, Fluidization Engineering, 2nd ed., Butterworth–Heine-mann, Boston, 1991, p. 44.

8. W. L. McCabe and J. C. Smith, Unit Operations of Chemical Engineering, 3rd ed.,McGraw-Hill, New York, 1976, p. 149.

9. C. Y. Wen and Y. H. Yu, AIChE J., 12, 610–612 (1966).10. J. F. Richarson, ‘‘Incipient Fluidization and Particle Systems,’’ in Fluidization (J. F.

Davidson and H. Harrison, eds.), Academic Press, New York, 1971, p. 26–64.11. S. C. Saxena and G. J. Vogel, Trans. Inst. Chem. Eng., 55, 184–189 (1977).12. S. P. Babu, B. Shah, and A. Talwalkar, AIChE Symp. Ser. 176, 176–186 (1978).13. J. R. Grace, ‘‘Fluidized Bed Hydrodynamics,’’ In Handbook of Multiphase Systems

(G. Hetsroni, ed.), Hemisphere, Washington, DC, 1982, p. 8–1.14. D. C. Chitester, R. M. Kornosky, L.-S. Fan, and J. P. Danko, Chem. Eng. Sci., 39,

253–261 (1984).


15. A. R. Abrahamsen and D. Geldart, Powder Technol., 26, 47–55 (1980).16. J. F. Davidson, and D. Harrison, Fluidized Particles, Cambridge University Press,

Cambridge, 1963.17. P. S. B. Stewart and J. F. Davidson, Powder Technol., 1, 61–80 (1967).18. J. Yerushalmi and N. T. Cankurt, Powder Technol., 24, 187–205 (1979).19. J. R. Grace and H. Bi, ‘‘Introduction to Circulating Fluidized Beds,’’ in Circulating

Fluidized Beds (J. R. Grace, A. A. Avidan, and T. M. Knowlton, eds.), London:Blackie Academic & Professional, London, 1997, pp. 1–20.

20. H. T. Bi and J. R. Grace, Chem. Eng. J., 57, 261–271 (1995).21. J. Yerushalmi, ‘‘High Velocity Fluidized Beds,’’ in Gas Fluidization Technology (D.

Geldart, ed.), John Wiley & Sons, New York, 1986, pp. 127–196.22. M. G. Schnitzlein and H. Weinstein, Chem. Eng. Sci., 43, 2605–2614 (1988).23. W. C. Yang, Powder Technol., 35, 143–150 (1983).24. L. Reh, Chem. Eng. Prog., 67, 58–63 (1971).25. J. Werther, Chem. Eng. Sci., 35, 372–379 (1980).26. Y. Li and M. Kwauk, ‘‘The Dynamics of Fast Fluidization,’’ in Fluidization (J. R.

Grace and J. M. Matsen, eds.), Plenum, New York, 1980, pp. 537–544.27. A. A. Avidan and J. Yerushalmi, Powder Technol., 32, 223–232 (1982).28. M. J. Rhodes, Chem. Eng. Res. Des., 67, 30–37 (1989).29. L. S. Leung, Powder Technol., 25, 185–190 (1980).30. G. E. Klinzing, Gas–Solid Transport, McGraw-Hill, New York, 1981.31. J. R. Grace, Can. J. Chem. Eng., 64, 353–363 (1986).32. F. A. Zenz and D. F. Othmer, Fluidization and Fluid Particle Systems, Van Nostrand

Reinhold, New York, 1960.33. J. J. van Deemter, in Fluidization, (J. R. Grace and J. M. Matsen, eds.), Plenum, New

York, 1980, pp. 69–89.34. M. Horio, N. M. Hoshiba, K. Morishita, Y. Kobukai, J. Naito, O. Tachibana, K. Wata-

nabe, and N. Yoshida, ‘‘Coal Combustion in a Transparent Circulating FluidizedBed,’’ in Circulating Fluidized Bed Technology (P. Basued, ed.), Pergamon, NewYork, 1986, pp. 225–262.

35. N. M. Catipovic, G. N. Jovanovic, and T. J. Fitzgerald, AIChE J., 24, 543–547 (1978).36. E. Muschelknautz and V. Grief, ‘‘Cyclones and Other Gas–Solids Separators,’’ in

Circulating Fluidized Beds (J. R. Grace, A. A. Avidan, and T. M. Knowlton, eds.)Blackie Academic & Professional, London, 1997, pp. 181–213.

37. W. C. Yang and T. M. Knowlton, Powder Technol., 77, 49–54 (1993).38. D. R. Jones and J. F. Davidson, Rheol. Acta, 4, 180–192 (1965).39. J. Sinclair and T. Mallo., FED (Am. Soc. Mech. Eng.), 245, 12/58–12/63 (1998).40. H. Weinstein, R. F. Graff, M. Meller, and M. J. Shao, ‘‘The Influence of the Imposed

Pressure Drop Across a Fast Fluidized Bed,’’ in Fluidization IV (K. Kunii and R.Toei, eds.), Engineering Foundation, New York, 1983, pp. 299–306.

41. E. U. Hartge, D. Rensner, and J. Werther, ‘‘Solid Concentration and Velocity Patternsin Circulating Fluidized Beds,’’ in Circulation Fluidized Bed Technology II (P. Basuand J. F. Large, eds.), Pergamon Press, Oxford, 1988, pp. 165–180.

42. M. J. Rhodes, M. Sollaart, and X. S. Wang, ‘‘Structure of the Dense–Dilute Interfacein Fast Fluidization,’’ in Fluidization IX (L-S. Fan and T. M. Knowlton, eds.), Engi-neering Foundation, New York, 1998, pp. 141–148.

43. M. Horio, ‘‘Hydrodynamics,’’ in Circulating Fluidized Beds (J. R. Grace, A. A. Avi-dan, and T. M. Knowlton, eds.), Blackie Academic & Professional, London, 1997,pp. 21–85.

44. A.E.Kostazos,H.Weinstein, andR.A.Graff, ‘‘TheEffectof theLocationofGas Injectionon the Distribution of Gas and Catalyst in a Riser,’’ in Fluidization IX (L-S. Fan and T.M. Knowlton, eds.), Engineering Foundation, New York, 1998, pp. 221–228.


45. H. Weinstein, M. Shao, M. Schnitzlein, and R. A. Graff, ‘‘Radial Variation in VoidFraction in a Fast Fluidized Bed,’’ in Fluidization V (K. Ostergarrds and A. Sorensen,eds.), AIChE Proceedings on the 5th Engineering Foundation Conference on Fluidiza-tion, New York, 1986, p. 329.

46. R. Bader, J. Findlay, and T. M. Knowlton, ‘‘Gas–Solid Flow Pattern in a 30.5cm Diameter Circulating Fluidized Bed,’’ in Circulating Fluidized Bed TechnologyII (P. Basu and J. F. Large, eds.), Pergamon Press, New York, 1988, pp. 123–127.

47. P. A. Galtier, R. J. Pointer, and T. E. Patureaux, ‘‘Near Full-Scale Cold Flow Modelfor the R2R Catalytic Cracking Process,’’ in Fluidization VI (J. R. Grace, L. W.Shemilt, and M. A. Bergougnou, eds.), Engineering Foundation, New York, 1989,pp. 17–24.

48. A. Miller, ‘‘Dense, Vertical Gas-Solid Flow in a Pipe,’’ Ph.D. dissertation, IllinoisInstitute of Technology, Chicago (1991).

49. R. S. B. Karri and T. M. Knowlton, ‘‘Flow Direction and Size Segregation of AnnulusSolids in a Riser,’’ in Fluidization IX (L-S. Fan and T. M. Knowlton, eds.), Engi-neering Foundation, New York, 1998, pp. 189–196.

50. A. Miller and D. Gidaspow, AIChE J., 38, 1801–1815, (1992).51. D. Gidaspow, Multiphase Flow and Fluidization: Continuum and Kinetic Theory De-

scriptions, Academic Press, Boston, 1994, p. 209.52. A. S. Issangya, D. Bai, and J. R. Grace, ‘‘Solids Flux Profiles in a High Density

Circulation Fluidized Bed Riser,’’ in Fluidization IX (L-S. Fan and T. M. Knowlton,eds.), Engineering Foundation, New York, 1998, pp. 196–204.

53. N. E. Jones, C. U. Yurteri, and J. L. Sinclair, ‘‘Effects of Solids Loading on ParticleMotion in Gas–Solid Flow,’’ AIChE Annual Meeting Preprint ‘‘Fluidization andFluid–Particle Systems,’’ in Dallas, TX (L. R. Glicksman, ed.), 1999, p. 65.

54. C. Brereton and J. R. Grace, ‘‘End Effects in Circulation Fluidized Bed Hydrodynam-ics,’’ in Circulating Fluidized Bed Technology IV (A. A. Avidan, ed.), AIChE, NewYork, 1990, pp. 137–144.

55. M. Kruse and J. Werther, J. Chem. Eng. Process., 34, 185–203 (1995).56. A. M. Squires, M. Kwauk, and A. A. Avidan, Science, 230, 1329–1337 (1985).57. J. Yerushalmi and A. A. Avidan, Powder Tech., 32, 223–232 (1982).58. M. Horio and H. Kuroki, Chem. Eng. Sci., 49, 2413–2421 (1994).59. P. D. Noymer and L. R. Glicksman, ‘‘Temperatures of Clusters at the Wall of a Circu-

lating Fluidized Bed,’’ in Fluidization IX (L-S. Fan and T. M. Knowlton, eds.), Engi-neering Foundation, New York, 1998, pp. 429–436.

60. H. T. Ishii, T. Nakajima, and M. Horio, J. Chem. Eng. Japan, 22, 484–490 (1989).61. C. Rudnick and J. Werther, ‘‘The Discrimination of Cluster Characteristics from

Fiber-Optical Probe Signals in Circulating Fluidized Beds,’’ in Fluidization IX (L-S.Fan and T. M. Knowlton, eds.), Engineering Foundation, New York, 1998, pp. 572–580.

62. S. Krol and H. de Lasa, ‘‘CREC–GS–Optiprobe for Particle Cluster Characteriza-tion,’’ in Fluidization IX, (L-S. Fan and T. M. Knowlton, eds.), Engineering Founda-tion, New York, 1998, pp. 565–572.

63. C. H. Soong, K. Tuzla and J. C. Chen, ‘‘Experimental Determination of Cluster Sizeand Velocity in Circulation Fluidized Bed,’’ in Fluidization VIII, (J. F. Large and L.Claude, eds.) Engineering Foundation, New York, 1995, pp. 219–227.

64. W. K. Gu and J. C. Chen, ‘‘A Model for Solid Concentration in Circulating FluidizedBeds,’’ in Fluidization IX (L-S. Fan and T. M. Knowlton, eds.), Engineering Founda-tion, New York, 1998, pp. 501–508.

65. Y. P. Tsuo and D. Gidaspow, AIChE J. 36, 885–896 (1990).66. R. C. Senior and J. R. Grace, ‘‘Formation of Particle Streamers in the Wall of Circulat-

Pressure and Temperature in Fluidized Beds 35

ing Fluidized Bed Risers,’’ in Fluidization IX (L-S. Fan and T. M. Knowlton, eds.),Engineering Foundation, New York, 1998, pp. 165–172.

67. S. L. Lee and F. Durst, Int. J. Multiphase Flow, 8, 125–146 (1982).68. Y. Tsuji, Y. Morikawa, and H. J. Shiomi, J. Fluid Mech., 139, 417–434 (1984).69. S. Yonemura, T. Tanaka, and Y. Tsuji, ‘‘Cluster Formation in Dispersed Gas–Solid

Glow: Effects of Physical Properties of Particles,’’ Proceedings to 2nd InternationalConference on Multiphase Flow, Kyoto, 1995, vol. 3, pp. PT4-25–PT4-30.

70. H. C. Simpson and B. W. Rodger, Chem. Eng. Sci., 16, 153–180 (1961).71. R. Jackson, Trans. Inst. Chem. Eng., 41, 13–28 (1963).72. J. Verloop and P. M. Heertjes, Chem. Eng. Sci., 25, 825 (1970).73. D. Geldart and M. J. Rhodes, ‘‘From Minimum Fluidization to Pneumatic Transport:

A Critical Review of the Hydrodynamics,’’ in Circulation Fluidized Bed Technology(P. Basu, ed.), Pergamon Press, New York, 1986, pp. 21–31.

74. A. Haider and O. Levenspiel, Powder Technol. 58, 63–70 (1989).

RAY A. COCCO

Effect of Pressure and Temperature in BubblingFluidized Beds

Introduction

Typical commercial gas fluidized-bed processes operate at up to 7.1 � 106 Pa(�70 atm) pressure and 1273 K (1000°C). Hence, an understanding of the effectof pressure and temperature on hydrodynamics is critical for propert fluidized-bedsystem design.

Temperature and pressure affect fluidized-bed systems because they affect thegas density and gas viscosity, two critical parameters impacting bed operation.These effects cannot be considered independently of particle size, as particle sizedramatically impacts fluid–particle interactions. In this section, we will summarizethe impact of pressure and temperature on the critical hydrodynamic parametersof minimum fluidization velocity, umf, bed voidage at minimum fluidization veloc-ity, εmf, minimum bubbling velocity, umb, bubble diameter, db, bubble frequency,f, dense phase voidage, εD, and particle entrainment and elutriation rates, K *i∞. Wewill also summarize the reported effect of pressure on chemical reaction conversionbecause reactor performance is the result of all combined effects the pressure andtemperature may have on hydrodynamics as well as on the reaction. The readeris referred to two in-depth reviews of the subject [1,2].

36 Pressure and Temperature in Fluidized Beds

Two-Phase Flow Model

A bubbling fluidized bed can be considered as consisting of two phases (Fig. 1):(1) a dense (or emulsion) phase of solid particles with a surrounding interstitialdense (or emulsion) phase gas and (2) a dilute phase consisting of gas jets whichprotrude up through a perforated gas distributor or series of nozzles, and gas bub-bles that form at the top of the jets. The gas that enters a fluidized bed at thevolumetric flow rate qG(in) enters in the form of gas jets. At the top of the jets, thisgas is distributed between bubbles that form (qB) and the interstitial dense phasegas (qD) that flows through the solid particles. Mass and heat transfer (interchange)occur between the jets and the dense phase gas, between the bubbles and the densephase gas, and between the dense phase gas and solids. Heat is also transferredby the dense phase to internal heat-exchange surfaces.

A critical factor affecting the overall performance of a fluidized bed is theinterchange between the dilute and dense phases. Interchange rates can be in-creased by (1) increasing the ratio of the gas jet volume to the bubble volume inthe bed (i.e., VJ/VB), (2) increasing the ratio of the gas flowing through the densephase to the gas flowing as bubbles (i.e., qD/qB), and (3) maintaining a small bubblesize throughout the bed. Small bubbles provide higher mass-transfer rates and in-creased residence time in the bed. It is imperative to note that although we wantbubbles to be small, bubbles must not disappear. Bubbles must be present in orderto achieve rapid mixing in the reactor, contributing to high heat-transfer rates tointernal surfaces.

FIG. 1 Transport, phases, and pressure in fluidized beds.


Pressure and temperature can have a significant effect on the distribution ofgas between the bubble and dense phases and the mass- and heat-transfer rates.

Minimum Fluidization Conditions

The minimum fluidization velocity, umf, is a fundamental characteristic of a fluid-ized bed. Its accurate prediction is important for the successful design and opera-tion of a fluidized-bed process. At minimum fluidization conditions, the drag forceby upward-moving gas equals the weight of particles lifted in the fluidized bed.The following relationship can be shown to hold (see the Nomenclature sectionfor terms):

Ar � 1501 � εmf

φ2ε3mf

Remf �1.75φε3

mf

Re2mf (1)

or

Ar �2C1

C2

Remf �1

C2

Re2mf (2)

where

Ar �d 3

p ρg(ρp � ρg)g

µ2g

, Remf �dpumf ρg

µg

C1 �42.86(1 � εmf)

φ(3)

C2 �φε3

mf

1.75(4)

Working on the assumption that the particles constituting the bed can be approxi-mated by a constant value of φ and that εmf remains constant over the entire rangeof operating conditions of temperature and pressure, various investigators haveproposed values of C1 and C2 on the basis of experimental results to be in therange 18.75–33.7 and 0.0313–0.0651, respectively [3].

The first terms on the right-hand sides of Eqs. (1) and (2) are important iflaminar, or viscous, flow predominates in the system, whereas the second termsare important if turbulent, or inertial, flow predominates.

For small particles (Remf � 20), the simplified form of Eqs. (1) and (2) is

umf � � C2

2C1� �d 2

p (ρp � ρg)g

µg� (5)


For large particles (Remf � 1000), the simplified form becomes

u2mf � C2 �d p (ρp � ρg)g

ρg� (6)

With gas viscosity, µg, almost independent of pressure and because ρp �� ρg formost materials, Eq. (5) indicates that umf is relatively unaffected by changes inpressure for fine particles. On the other hand, according to Eq. (6), umf for largerparticles will vary with (1/ρg)0.5, indicating that umf decreases for increasing pres-sure.

For changes in temperature, Eq. (5) indicates that umf varies with 1/µg. Becausethe gas viscosity, µg, increases with temperature, umf decreases with an increasein temperature for fine-particle systems in which viscous forces dominate. Forlarge particles, Eq. (6) indicates that umf will increase with temperature becausean increased temperature results in a decreased ρg.

For systems in the intermediate regime (20 � Remf � 1000), Eq. (2) can berearranged and used:

umf � � µg

dpρg� [(C 2

1 � C2 Ar)0.5 � C1] (7)

The trends that umf is relatively unaffected by changes in pressure and decreasesfor increases in temperature for fine-particle systems and that umf decreases forincreases in pressure and increases for increases in temperature for large-particlesystems are consistent with those reported experimentally. However, the absolutevalues of the predictions may be incorrect because of the difficulty in determininga representative value for dp and φ, or estimating εmf.

In the absence of reliable data, use C1 � 33.7 and C2 � 0.0408 [4] for fine-particle systems and C1 � 28.7 and C2 � 0.0494 [5] for coarse-particle systems.

A recommended method for improving the accuracy of umf is to first determineumf experimentally at ambient conditions and to back-calculate an effective particlediameter (i.e., deff) from Eq. (5), (6), or (7), using fixed C1 and C2. Using thiseffective particle diameter, calculate umf at the desired conditions of temperatureand pressure. This method substitutes an effective value for dp and φ, which areindependent of temperature and pressure; however, it does not account for anychanges in εmf which might occur for changes in temperature and pressure.

The effect of temperature and pressure on εmf has been studied by a numberof investigators. Several studies [5,6] have indicated that pressure has essentiallyno effect on εmf for fine particles and a slight increasing effect on εmf for largerparticles. The effect of temperature on εmf has been reported to be much moresignificant [7–10] and to affect fine-particle systems more than coarse-particle sys-tems. The temperature effect appears to be the result of interparticle forces whichaffect packing properties [8,10]. The dependence of εmf on temperature can beexpressed in linear form as [8]

εmf � εmf(amb) � k(T � T(amb)) (8)

where the parameter k is a function of particle properties.


Although substantial data have been reported in the literature, no reliable correla-tions are available for predicting εmf for a given system. However, it is clear thatincorporating an accurate value of εmf with an effective particle size, deff, in Eq. (1)will allow an accurate calculation of umf. Without experimental data, the use of Eq.(7) with the recommended values for C1 and C2 is suggested for calculating umf.

Minimum Bubbling Velocity, umb and Dense PhaseVoidage, εD

As the gas velocity is increased above that required to incipiently fluidize the bedof particles, gas bubbles eventually form (at u0 � umb) and rise through the densephase. In fine-particle systems at high pressures, it is possible to observe a particu-lately (or homogeneously) fluidized bed without bubbles for intermediate gas ve-locities between umf and minimum bubbling, umb (i.e., for umf � u0 � umb). Theminimum gas velocity at which bubbles appear, umb, has been found to equal umf

for coarse-particle systems [11]. However, for fine-particle systems, there is a rangeof velocities between umf and umb over which the bed expands uniformly. Thevelocity range over which this ‘‘delayed bubbling’’ occurs can be extended withan increase in operating pressure.

Abrahamsen and Geldart [12] observed that umb and, hence, the region of partic-ulate expansion is increased by adding fines (increased F, weight fraction of parti-cles �45 µm) to a fluidized bed of fine Group A powders:

umb

umf

�2300ρ0.126

g µ 0.523g exp(0.716F)umf

d 0.8p g0.934(ρp � ρg)0.934

(9)

Equation (9) indicates that umb will increase for an increase in gas temperature andpressure (via increased ug and ρg) and that the sensitivity to viscosity is moresignificant than the sensitivity to pressure. This equation has been shown to bevalid over a wide range of pressures for fine-particle systems [13].

For fluidized beds of fine Group A powders that particulately fluidize (i.e.,umb � umf), the voidage of the dense phase, εD, exceeds εmf. This voidage hasbeen shown to be adequately described for high-pressure systems of fine powders[14] by the empirical correlation of Kmiec [15]:

εD �(18Re � 2.7Re1.687)0.209

Ar 0.209(10)

It is important to note that an increase in εD results in a decrease in dense phaseviscosity, µD.

The effect of umb � umf and εD � εmf is to increase the interchange betweenfeed gases and solid particles because more feed gas flows directly through thedense phase (i.e., � qD/qB). Hence, overall reaction rates for catalytic and gas–solid fluidized-bed reaction processes involving fine particles will increase with


increased pressure due to the effect of pressure on fluidization hydrodynamicsalone.

Jetting Region

As gas flows through a perforated or multinozzle distributor plate, it enters a fluid-ized bed in the form of gas jets. This jetting region (Refs. 16 and 17, among others)has high gas–solid contacting efficiencies relative to the bubbling region above.Several investigators [18,19] have shown that the penetration of the gas jets intothe fluidized bed increases substantially with an increase in operating pressure.The effect of gas viscosity was not found to be significant [20]. The contributionof pressure and temperature impacting the jet length, h j, is through their effect ongas density via the equation [19]:

h j

do

� 21.2Fr 0.37Re0.05 �ρg

ρp�

0.68

�dp

do�

0.24

(11)

Bubbling Region

Bubble size is one of the most important parameters of conventional gas–solidfluidized beds. Overall reaction rates depend on the size of bubbles because bubblesize is a primary contributor to bubble rise velocity, gas rate of exchange betweenphases, heat transfer, and fine-particle elutriation.

Fine-Particle Systems

For fine-particle Group A systems, increased pressures yield smaller-sized bubbles.Studies have indicated [21] that the smaller bubbles result from a decrease in thestability of bubbles leading to their breakup into smaller voids. X-rays [22] of gasbubbles in fluidized beds have shown that an initial disturbance, indenting theupper surface of a bubble, grows to split it from above. It has been suggested thatthis splitting results from a Taylor instability [23] wherein a heavy fluid overliesa lighter one [24].

In applying the Taylor theory to bubbles in fluidized beds, the dense phaseover the bubble is the heavy fluid and the gas in the bubble is the lighter fluid.An initial, random, small disturbance with amplitude η0 is assumed to perturb theessentially horizontal upper surface of the bubble and to grow according to

η � η0 exp(nt) (12)


where n is a parameter depending on the physical properties of the system and onthe wavelength of the disturbance. In gas fluidized beds, n may be obtained from [25]

R 4 � 2R 2 � 4R � 1 �g

(k3ν2p)

� 0 (13)

where

k �2πλ

(14)

is the wave number of the disturbance and

R � 1 � n(νpk 2) (15)

A numerical solution for Eq. (13) has been presented elsewhere [25]. For each νp,there is a specific wavelength called the ‘‘most dangerous’’ wavelength, λmax,which exhibits the maximum growth rate.

Clift et al. [25] have noted that disturbances initiated on the roof of a bubbleare swept around the periphery. In practice, a bubble does not split unless thedisturbance has grown sufficiently before the tip of the growing spike reaches theside of the bubble. The likelihood of splitting can be estimated by comparing thetime constant for the growth of a disturbance (i.e., te � 1/n) with the maximumtime available for growth [25]:

tam � �R0

g �0.5

�ln1

tan(θ/2)� (16)

If the time available for growth, tam, is greater than the required growth time, te,the bubble is liable to split. Otherwise, the disturbance grows so slowly that itdoes not achieve an amplitude large enough to cause splitting before it is sweptaround the bubble equator.

Observations of splitting bubbles suggest that disturbances usually develop ina regular pattern on either side of the node. Assuming a node is located λ/4 fromthe bubble nose so that the node is an antinode in the initial disturbance,

θ �λ2θ

�π

kR0

(17)

The relationship between the stable bubble frontal diameter, df , and the correspond-ing νp is shown in Fig. 2. It is clear from Eq. (13) and from Fig. 2 that νp is thedominant factor determining the growth of the instabilities, the most dangerouswavelength, λmax, and, hence, the maximum stable bubble size, dbmax. Thus, the pre-diction of the effect of system properties on the bubble stability depends on theprediction of the effect of system properties on νp. Smaller particles and higherpressures are known to lower νp [26] and thus, according to the Taylor instabilitytheory, result in smaller df . The major difficulty in testing the ability of using the


FIG. 2 Effects of kinematic viscosity on stable bubble frontal diameters.

Taylor instability theory to explain observed limited bubble growth in high-pressurefluidized beds is the lack of data, correlations, and theory relating the kinematicviscosity of the dense phase, νp, to system parameters. The relationship betweenthe stable bubble frontal diameter, df , and the corresponding νp is shown in Fig. 2.

The actual bubble volume is approximately 75% of that of a sphere of diameterequivalent to df [22]. Hence, df and db are related:

df � 1.1db (18)

For Taylor instabilities governing the maximum stable bubble size in fluidizedbeds, dbmax is dependent on νp. Because νp decreases with an increase in εD andεD increases with an increase in pressure and temperature, particularly for fine-particle systems, increased pressure and temperature result in a decrease in dbmax.However, the effect of particle size must be taken into consideration because theeffect of pressure and temperature on dbmax is minimal for coarse-particle systemscompared to fine-particle systems.

A suitable relationship for estimating µD from variations in εD for gas fluidizedbeds is needed in order to predict dbmax a priori. Although such a relationship isnot currently available, calculations using equations developed for systems of solidspheres in liquids have been shown to be consistent with µD needed to limit bubblegrowth in high-pressure systems [14]. Hence, no suitable relationship exists todayto allow the calculation of dbmax for fine-particle systems as impacted by tempera-ture and pressure. Nonetheless, experimentally determined values for db indicate


that higher pressures and temperatures yield smaller-diameter bubbles in gas fluid-ized beds, particularly for fine-particle systems.

It has generally been reported that the bubble frequency, f, increases with in-creases in temperature [27–29] and pressure [30]. These results are consistent withdecreases in bubble size for increases in temperature and pressure.

Coarse-Particle Systems

A generalized correlation for bubble size has been developed by Cai et al. [31]for pressurized fluidized-bed combustors (PFBC) on the basis of a comprehensiveanalysis of previous work. The correlation takes into account the different flowregimes at different pressures and gas velocities, as well as the special variationof bubble size within the lower-pressure range of the bubbling regime.

For PFBC (P in Pa),

db � 0.1905h0.8P0.06(u0 � umf)0.42

(19)� {exp[�1.4 � 10�14P2 � 0.25(u0 � umf)2 � 1 � 10�6P(u0 � umf)]}

The equivalent bubble size of the entire bed is

dbe � 0.1002L 0.8f P0.06(u0 � umf)0.42

(20)� {exp[�1.4 � 10�14P2 � 0.25(u0 � umf)2 � 1 � 10�6P(u0 � umf)]}

Particle Entrainment and Carryover

Entrainment occurs when gas bubbles break at the top of the fluidized bed andthrow particles up into the freeboard region above the bed surface. At low gasvelocities, these particles quickly fall back into the bed and are retained, but asthe fluidizing velocity is increased, more and more particles are transported togreater heights above the bed surface and there exists a particle density gradientextending some distance above the surface. For sufficiently tall freeboards, therewill be a certain height at which the density gradient eventually falls to zero, andabove this height, the entrainment flux will be constant. This height is called thetransport disengagement height (TDH). If the bed solids have a wide size distribu-tion and the gas velocity in the freeboard exceeds the terminal fall velocity of thesmaller ones, then these will be carried out of the system or elutriated.

Entrainment from fluidized beds is affected by changing pressure and tempera-ture. An increase in the operating pressure increases the carrying capacity of the gasand increases the amount of solids carried over [32,33]. Chan and Knowlton [32]also found that the TDH increased linearly with pressure in the range 1–30 atm.

Findlay and Knowlton [20] studied the effect of increasing temperature onentrainment. They investigated temperatures in the range of ambient to 1033 K


while maintaining the system pressure constant; thus, the dominant effect was theincrease in gas viscosity. The increased gas viscosity decreased the terminal fallvelocity and increased the rate of entrainment for a given fluidizing velocity.

For the discharge point exceeding TDH, Geldart [11] recommended variouselutriation rate constants for particle size fraction dpi. The Zenz–Weil [34] equationis recommended for particles of diameter dp � 100 µm and superficial gas veloci-ties u0 � 1.2 m/s when the entire bed is potentially entrainable (even the largestparticles have a terminal velocity ut � u0). Some of the applicable materials includecracking catalyst, coal char, and other low-density solids:

K*i∞(µgu0)

� A� u 20

gdpiρ2p�

B

(21)

where

A � 1.26 � 107 and B � 1.88 when [(u 20/(gdpiρ2

ρ) � 3 � 10�4

A � 4.31 � 104 and B � 1.18 when [(u 20/(gdpiρ2

ρ) � 3 � 10�4

Geldart [11] proposed that

K*i∞(ρgu0)

� A′ exp��B′ut

u0� (22)

where

A′ � 23.7 and B′ � 5.4 for higher velocities and coarser particlesA′ � 31.4 and B′ � 4.27 for beds consisting largely of 1-mm solidsA′ � 49.1 and B′ � 4 for 2.5-mm coarse particles in the bed

It is clear that the elutriation rate constants will increase for increases in gas viscos-ity and gas density according to Eqs. (21) and (22).

Also, the elutriation rate will increase for a decrease in the particle terminalvelocity, ut, according to Eq. (22). Increasing the temperature will cause a reductionin the terminal velocity for particles in the Stokes region (small particles) andan increase for those in Newton region (large particles). Terminal velocities areunaffected by the gas density in the Stokes region (small particles), slightly in thetransition region, and in the Newton region (large particles) according to (1/ρg)0.5.Hence, as the gas pressure is increased at constant gas velocity, the size of theparticles carried over will increase slightly.

Effect of Hydrodynamics on Reaction Conversion

Because the conversion of fluidized-bed chemical reactors depends on the contactof solids (either catalyst or reacting) with reactant gases, it is advantageous to be


able to improve the amount of feed gas contacting the solids. This contacting canbe improved by forcing more feed gas to flow directly into the dense phase (i.e.,increasing qD relative to qB) and to keep the bubbles, which do form, small (im-proved mass transfer between dilute and dense phases). Because higher pressuresincrease gas–solid contacting for fine-particle systems, one would expect thathigher pressures would result in improved reactor conversions.

Weimer et al. [21] presented the only comparative results ever reported for apilot-scale tubular fixed-bed (3.8 cm inner diameter � 22 cm long) and fluidized-bed (15 cm inner diameter � 2.7 m long) catalytic process operating at comparativegas hourly space velocities (GHSV) and temperatures. They investigated the exo-thermic Fischer–Tropsch synthesis of syngas at pressures of P � 3549 and 6996kPa and temperatures between T � 613 and 663 K. The fluidizable particle sizewas measured to be around dp � 100 µm.

Their results showed that for nominal pressures of P � 3549 kPa, the conver-sion of carbon monoxide (XCO) was noticeably higher for the fixed-tube reactor.For GHSV �1000 h�1 and T � 623 K, XCO � 0.7 in the pilot fixed bed versusXCO � 0.55 for the pilot fluidized bed. At nominal P � 6996 kPa, the fixed-bedconversions were only slightly higher than those in the fluidized bed were. ForGHSV �1600 h�1 and T � 621 K, XCO � 0.74 in the fixed bed versus XCO � 0.71in the fluidized bed.

The better agreement between results at nominal P � 6996 kPa versus P �3549 kPa was believed due to smaller bubbles at the higher pressures. Bubbleswere reported to be db �5 cm in diameter for the P � 3549 kPa operation and tobe db � 2 cm in diameter for the P � 6996 kPa operation. The number of gasinterchanges for bubbles traversing the bed (i.e., the number of times the bubblegas was exchanged with the dense phase gas as bubbles rise in the bed) was calcu-lated to be between 5 and 6 for the smaller bubbles at higher pressures versus 1for the larger bubbles at lower pressures. The investigators believed that even betterresults could have been achieved if finer particles were being fluidized (�100µm).

These results indicate that the smaller bubbles resulting from higher pressuresin fine-particle systems contribute to substantial improvement in mass interchange,allowing fixed-bed-type conversions. The scale-up of high-pressure, fine-particlefluidized beds from laboratory data is simplified because bubble growth is limitedby hydrodynamics. Normally, fluidized-bed scale-up is complicated by the factthat the wall for small-particle systems limits bubble growth and that bubbles growas the bed gets larger. Results reported in the literature indicate that the high-pressure, fluidized-bed bubble size can be controlled by proper selection of fine-particle size.

Nomenclature

A Constant in Eq. (21)A′ Constant in Eq. (22)Ar Archimedes number � d 3

pρg(ρp � ρg)g/µ 2g


B Constant in Eq. (21)B′ Constant in Eq. (22)db Bubble diameter (m)C1 C1 � 42.86(1 � εmf)/φC2 C2 � φε 3

mf/1.75dbmax Maximum bubble diameter (m)df Frontal bubble diameter (m)do Jet nozzle or distributor orifice diameter (m)dbe Equivalent bubble diameter, averaged over bed height (m)dp Particle diameter (m)deff Effective particle diameter based on experimental umf calculation (m)dpi Diameter of particle size fraction I (m)F wt fraction of fine particles �45 µm in diameterFr Froude number � u 2

0/gdp

f Bubble frequency (1/s)g Gravitational force (m/s2)h Distance up the bed axially (m)hj Distance the jets protrude up the bed (m)K*i∞ Elutriation rate constant for size fraction dpi(kg/m2 s)k Linear slope for rate of change of εmf with temperature (l/K)L f Expanded bed height (m)n Parameter used in Eq. (12)P Pressure (Pa)qB Volumetric flow rate of gas as bubbles (m3/s)qD Volumetric flow rate of dense phase gas (m3/s)qG(in) Volumetric flow rate of gas (m3/s)R Defined by Eq. (15)R0 Frontal radius of a bubble (m)Remf Particle Reynolds number � dpumf ρg/µg

T Temperature (K)T(amb) Ambient temperature (K)tam Time available for growth (s)te Required growth time (s)u0 Superficial gas velocity (m/s)umf Minimum fluidization velocity (m/s)ut Terminal velocity (m/s)VB Bubble phase volume (m3)Vj Jet phase volume (m3)XCO Conversion of carbon monoxide

Greek

εD Dense phase voidageεmb Voidage at minimum bubblingεmf Voidage at minimum fluidizationεmf(amb) Voidage at minimum fluidization at ambient temperature


η Amplitude of disturbance for Taylor theory [Eq. (12)]η0 Initial amplitude of disturbance for Taylor theory [Eq. (12)]φ Particle sphericityλ Wavelength of disturbance for Taylor theory [Eq. (12)]λmax Maximum wavelength of disturbanceµg Gas viscosity (kg/m s)νp Kinematic dense phase viscosityθ Angular position where disturbance originatesρg Gas density (kg/m3)ρp Particle density (kg/m3)

References

1. T. M. Knowlton, ‘‘Pressure and Temperature Effects in Fluid-Particle Systems,’’ inFluidization VII (O. E. Potter and D. J. Nicklin, eds.), Engineering Foundation, NewYork, 1992, pp. 27–46.

2. J. G. Yates, ‘‘Effects of Temperature and Pressure on Gas–Solid Fluidization,’’ Chem.Eng. Sci., 51(2), 167–205 (1996).

3. A. Mathur, S. C. Saxena, and Z. F. Zhang, ‘‘Hydrodynamic Characteristics of Gas–Solid Fluidized Beds over a Broad Temperature Range,’’ Powder Technol., 47, 247–256 (1986).

4. C. Y. Wen and Y. H. Yu, ‘‘Mechanics of Fluidization,’’ Chem. Eng. Progr. Symp.Ser., 62, 100–111 (1966).

5. D. C. Chichester, R. M. Kornosky, L.-S. Fan, and J. P. Danko, ‘‘Characteristics ofFluidization at High Pressure,’’ Chem. Eng. Sci., 39(2), 253–261 (1984).

6. P. A. Olowson and A. E. Almstedt, ‘‘Influence of Pressure on the Minimum Fluidiza-tion Velocity,’’ Chem. Eng. Sci., 46(2), 637–640 (1991).

7. J. S. M. Botterill, Y. Toman, and K. R. Yuregir, ‘‘The Effect of Operating Temperatureon the Velocity of Minimum Fluidization Bed Voidage and General Behaviour,’’ Pow-der Technol., 31, 101–110 (1982).

8. B. Formisani, R. Girimonte, and L. Mancuso, ‘‘Analysis of the Fluidization Processof Particle Beds at High Temperature,’’ Chem. Eng. Sci., 53(5), 951–961 (1998).

9. A. Lucas, J. Arnaldos, J. Casal, and L. Puigjaner, ‘‘High Temperature Incipient Fluid-ization in Mono and Polydisperse Systems,’’ Chem. Eng. Commun., 41, 121–132(1986).

10. G. Raso, M. D’Amore, B. Formisani, and P. G. Lignola, ‘‘The Influence of Tempera-ture on the Properties of the Particulate Phase at Incipient Fluidization,’’ Powder Tech-nol., 72, 71–76 (1992).

11. D. Geldart, ‘‘Particle Entrainment and Carryover,’’ in Gas Fluidization Technology(D. Geldart, ed.), Wiley, Chichester, 1986, pp. 123–154.

12. A. R. Abrahamsen and D. Geldart, ‘‘Behavior of Gas Fluidized Beds of Fine PowdersI. Homogeneous Expansion,’’ Powder Technol., 26, 35–46 (1980).

13. K. V. Jacob and A. W. Weimer, ‘‘High-Pressure Particulate Expansion and MinimumBubbling of Fine Carbon Powders,’’ AIChE J., 33(10), 1698–1707 (1987).

14. A. W. Weimer and G. J. Quarderer, ‘‘On Dense Phase Voidage and Bubble Sizein High Pressure Fluidized Beds of Fine Powders,’’ AIChE J., 31(6), 1019–1028(1985).


15. A. Kmiec, ‘‘Equilibrium of Forces in a Fluidized Bed—Experimental Verification,’’Chem. Eng. J., 23, 133 (1982).

16. L. A. Behie and P. Kehoe, ‘‘The Grid Region in a Fluidized Bed Reactor,’’ AIChEJ., 19, 1070 (1973).

17. L. A. Behie, M. A. Bergougnou, and C. G. J. Baker, ‘‘Mass Transfer from a Grid Jetin a Large Gas Fluidized Bed,’’ in Fluidization Technology, Vol. 1, D. L. Keairns,ed.), Hemisphere, Washington, DC, 1976, p. 261.

18. I. Hirsan, C. Shishtla, and T. M. Knowlton, ‘‘The Effect of Bed and Jet Parameters onVertical Jet Penetration Length in Gas Fluidized Beds,’’ 73rd Annual AIChE Meeting,Chicago, IL, 1980.

19. J. G. Yates, V. Bejcek, and D. J. Cheesman, ‘‘Jet Penetration into Fluidized Beds atElevated Pressures, Fluidization 5 (K. Ostergaard and S. Sorenson, eds.), EngineeringFoundation, New York, 1986, pp. 79–86.

20. J. G. Findlay and T. M. Knowlton, Final Report for Department of Energy, ProjectDE-AC21-83MC20314 (1985).

21. A. W. Weimer, G. J. Quarderer, G. A. Cochran, and M. M. Conway, ‘‘Design andPerformance of a High-Pressure Fischer-Tropsch Fluidized Bed Reactor,’’ in Fluidiza-tion VI (J. R. Grace, L. W. Shemilt, and M. A. Bergougnou, eds.), Engineering Foun-dation, New York, 1989.

22. P. N. Rowe and B. A. Partridge, ‘‘An X-ray Study of Bubbles in Fluidized Beds,’’Trans. Inst. Chem. Eng., 43, 157 (1965).

23. G. I. Taylor, ‘‘The Instability of Liquid Surfaces When Accelerated in a DirectionPerpendicular to Their Planes,’’ Proc. Roy. Soc., A201, 192 (1950).

24. R. Clift and J. R. Grace, ‘‘The Mechanism of Bubble Break-up in Fluidized Beds,’’Chem. Eng. Sci., 27, 2309 (1972).

25. R. Clift, J. R. Grace, and M. E. Weber, ‘‘Stability of Bubbles in Fluidized Beds,’’Ind. Chem. Fundam., 13, 45 (1974).

26. D. F. King, F. R. G. Mitchell, and D. Harrison, ‘‘Dense Phase Viscosities of FluidizedBeds at Elevated Pressures,’’ Powder Technol., 28, 55 (1981).

27. T. Mii, K. Yoshida, and D. Kunii, ‘‘Temperature Effects on the Characteristics ofFluidized Beds,’’ J. Chem. Eng. Japan, 6, 100–102 (1973).

28. T. Otake, S. Tone, M. Kawashima, and T. Shibata, ‘‘Behaviour of Rising Bubblesin a Gas-Fluidized Bed at Elevated Temperature,’’ J. Chem. Eng. Japan, 8, 388–392.

29. K. Yoshida, T. Ueno, and D. Kunii, ‘‘Mechanism of Bed to Wall Heat Transfer in aFluidized Bed at High Temperatures,’’ Chem. Eng. Sci., 29, 77–82 (1974).

30. I. H. Chan, C. Shishtla, and T. M. Knowlton, ‘‘The Effect of Pressure on BubblingGas Fluidized Beds,’’ Powder Technol., 53, 217–235 (1987).

31. P. Cai, M. Schiavetti, G. De Michele, G. C. Grazzini, and M. Miccii, ‘‘QuantitativeEstimation of Bubble Size in PFBC,’’ Powder Technol., 80, 99–109 (1994).

32. I. H. Chan and T. M. Knowlton, ‘‘The Effect of Pressure on Entrainment from Bub-bling Gas Fluidized Beds,’’ in Fluidization (D. Kunii and R. Toei, eds.), EngineeringFoundation, New York, 1984, pp. 283–290.

33. S. T. Pemberton and J. F. Davidson, ‘‘Elutriation of Fine Particles from BubblingFluidized Beds,’’ in Fluidization (D. Kunii and R. Toei, eds.), Engineering Founda-tion, New York, 1983, pp. 275–282.

34. F. A. Zenz and N. A. Weil, AIChE J., 4, 472 (1958).

ALAN W. WEIMER

Process Safety and Risk Management 49

Fundamentals of Process Safetyand Risk Management

Introduction

The growth of industry and, as a result, the economy are dependent on technologyadvances and innovations. However, these same activities often lead to more com-plex processes, especially in the chemical industry, which is using comparativelysevere operating conditions (temperature, pressure, flow rate, etc.), more reactivechemicals, and exotic chemistry. These more complex processes require in-depthanalysis and knowledge of process chemistry and hazards. It is even more impor-tant now to design the process and equipment to precise standards based on acomplete understanding of the underlying hazards, process chemistry, and the im-pact of operating conditions. Recently, much attention has been paid to humanfactors and its impact on chemical plant incidents. However, one can also say thatprocess knowledge and understanding is the most human factor. This is based onthe concept that inadequate knowledge, information, and understanding of the pro-cess hazards, chemistry, and impact of operating conditions are the root cause ofmany process plant incidents.

Managing safety is no easy task, but it makes bottom-line sense. There is adirect payoff in savings on a company’s workers’ compensation insurance, whosepremiums are partly based on the number of claims paid for job injuries [1]. Theindirect benefits are far larger, for safe plants tend to be well run in general andmore productive. The recipe for safety is remarkably consistent from industry toindustry. It starts with sustained support of top management followed by imple-mentation of appropriate programs and practices that institutionalize safety as aculture as compared to add-on procedures. The ingraining of safety as second na-ture in day-to-day activities requires a paradigm shift and can only be accomplishedwhen safety is viewed as an integral and comprehensive part of any activity ascompared to being a stand-alone or add-on activity.

Accident Process and Multiple Barrier Concept

Most chemical plant accidents follow a typical pattern. It is important to studythese patterns in order to be able to develop management systems to prevent theseaccidents. Also, many accidents occur as a result of the failure of multiple systemsor ‘‘barriers.’’ In fact, it can be argued that many of these accidents may not haveoccurred, had at least one of the ‘‘barriers’’ not failed. Thus, it is important to studythe concept of multiple barriers and its role in preventing process plant accidents.

50 Process Safety and Risk Management

The Accident Process

Most chemical accidents follow a three-step process, as described by Crowl andLouvar [2]:

• Initiation: the event, which starts the accident process• Propagation: the event, series of events, or condition which allows the accident

process to continue, or which expands the magnitude of the accident• Termination: the event or events, which stop the accident

The following is an example of the process:

• A seal on a sulfuric acid pump leaked, requiring replacement (initiating event).• The pump was drained and washed, but some time passed before maintenance

began (propagating event).• An isolation valve between the pump and the sulfuric acid supply was leaking

(propagating event).• The mechanic wore most of the required personal protective equipment, but

failed to wear rubber boots (propagating event).• When the mechanic began to work on the pump, he was splashed on the foot

when a small amount of sulfuric acid was released, resulting in an acid burn(terminating event—all of the acid in the pump was released).

To prevent accidents, we must modify this accident process. This can be doneby eliminating or reducing the likelihood of initiating events or propagating events,reducing the ability of propagating events to increase the magnitude of the accident,or by providing terminating events to interrupt the accident sequence before unac-ceptable consequences can occur. For the example described, some corrective ac-tions might include the following:

• Using a pump with an improved design, which would require less frequent sealrepair (reducing the likelihood of the initiating event)

• Providing a double block between the sulfuric acid supply and the pump, andimproving procedures and training to ensure timely washing of equipment anduse of protective equipment (reducing likelihood of propagating events)

• Training the mechanic to assume the pump contains sulfuric acid and to drainit to a safe place before he begins his work (provide a safe terminating eventby safely removing the acid)

Multiple Barrier Concept (Layers of Protection)

Chemical processes traditionally rely on multiple layers of protections, or barri-ers, between a hazardous agent and the people, environment, and property whichmight be adversely impacted by an incident. This concept is illustrated in Fig. 1


FIG. 1 Typical layers of protection for a chemical process. (Based on Fig. 2.2 of Ref. 3.)

[3]. The layers of protection might include the basic process design, basic processcontrols and operating procedures, critical alarms and process shutdown proce-dures, safety interlocks, emergency equipment such as rupture disks and pressurerelief valves, physical containment systems such as catch tanks and spill con-tainment dikes, emergency response equipment and services such as sprinkler sys-


tems and fire-fighting equipment and personnel, and personnel evacuation proce-dures.

Multiple barriers are generally required because no barrier will be perfect—all are subject to potential failure. An inherently safer process (discussed elsewherein this article) will reduce or eliminate the hazard and will require fewer or lessrobust layers of protection—and, if the hazard is sufficiently small, there may beno need for additional protective layers at all. This is highly desirable because thelayers of protection may require significant initial capital investment and ongoingoperating costs to ensure their continued effectiveness. Also, although the layersof protection may be highly reliable and the risk of an accident may be small, itcan never be zero—there is always a possibility that all of the layers of protectionwill fail simultaneously and the accident will occur.

The number and required reliability of the barriers or layers of protection mustbe established through the use of the various hazard and risk analysis techniquesdescribed in the following sections. This requires a complete understanding of thehazards of the process and plant-hazard identification, and an understanding of themechanisms or scenarios by which those hazards might result in harm to people,the environment, or property—hazard analysis or hazard evaluation.

Regulations

During the past 15 years, a number of chemical or related incidents in the petro-chemical industry have adversely affected surrounding communities. A few ofthese incidents, such as the vapor cloud explosion in Flixborough in 1974, theliquefied petroleum gas explosion in Mexico City in 1984, the toxic material re-lease in Bhopal in 1984, and the fire and radiation release in Chernobyl, werereported worldwide. Both governmental agencies and trade organizations re-sponded by developing standards and regulations to improve process safety. TheAmerican Petroleum Institute (API) and the American Chemistry Council (ACC)started to work with their members to develop organizational guidelines. The U.S.Department of Labor directed the Occupational Safety and Hazard Administration(OSHA) to develop federal standards for managing process safety.

A consensus started to emerge in 1990. Although the language, application,and extent of each document differed, the contents and objectives were almost thesame. The API published Recommended Practice 750: Management of ProcessHazards [4] in January 1990. OSHA published the proposed federal process safetyrule [5] in July 1990. In October 1990, the ACC published its Resource Guide forImplementing the Process Safety Management Code of Practices [6]. In addition,the Clean Air Act Amendments of 1990 directed OSHA and the EnvironmentalProtection Agency (EPA) to develop process safety management regulations toprotect workers and the environment. The final OSHA rule on Process Safety Man-agement of Hazardous Chemicals (29 CFR 1910.119) was published in the FederalRegister [7] on February 24, 1992. A matrix showing the relevance of OSHAProcess Safety Management (PSM) elements to the Center for Chemical Process


TABLE 1 Summary Comparison of OSHA Elements with CCPS Elements

CCPS 12 elements of chemical processsafety management Relevant paragraphs of OSHAs PSM rule

1. Accountability: Objectives andGoals

2. Process Knowledge and Documenta- Process Safety Information § 1910.119 (d)tion

3. Capital Project Review and Design Pre-Startup Safety Review § 1910.119 (i)Procedures (for new and existingplants, expansions, and acquisi-tions)

Mechanical Integrity § 1910.119 (j)4. Process Risk Management Process Hazard Analysis § 1910.119 (e)

Pre-Startup Safety Review § 1910.119 (i)5. Management of Change Management of Change § 1910.119 (l)6. Process and Equipment Integrity Process Hazard Analysis § 1910.119 (e)

Operating Procedures § 1910.119 (f)Mechanical Integrity § 1910.119 (j)

7. Human Factors Process Hazard Analysis § 1910.119 (e)Operating Procedures § 1910.119 (f)

8. Training and Performance Operating Procedures § 1910.119 (f)Training § 1910.119 (g)Pre-Startup Safety Review § 1910.119 (i)Emergency Planning and Response

§ 1910.119 (n)9. Incident Investigation Incident Investigation § 1910.119 (m)

10. Standards, Codes, and Laws11. Audits and Corrective Actions Compliance Audits § 1910.119 (o)12. Enhancement of Process Safety

Knowledge

Safety’s (CCPS) chemical process safety management elements is given in Table 1.EPA published the Risk Management Program in June 1996.

The international chemical and petroleum community has also been addressingprocess safety management through regulations and recommended practices. TheNorwegian Petroleum Directorate issued rules [8] in 1981 requiring quantita-tive hazard analyses for offshore petroleum operations. In response to the 1976chemical dioxin release in Seveso, Italy, a European Directive [9] (commonlycalled the Seveso Directive) on process safety management was issued in 1982.More recently, the British government has issued process safety managementregulations [10] for North Sea petroleum operations, following the recommenda-tions of the widely distributed Cullen Report, which investigated the 1985 PiperAlpha offshore platform tragedy. Outside of Europe, the World Bank [11] hasprovided process safety management guidance for third-world projects. Similarly,the International Labor Office in Geneva has issued hazard analysis recommenda-tions [12].


The Process Safety Management Program

The 14 elements of the OSHA Process Safety Management (PSM) regulation (29CFR 1910.119) were published in the Federal Register on February 24, 1992 [7].The objective of the regulation is to prevent or minimize the consequences ofcatastrophic releases of toxic, reactive, flammable, or explosive chemicals. Theregulation requires a comprehensive management program: a holistic approach thatintegrates technologies, procedures, and management practices.

The process safety management regulation applies to processes that involvecertain specified chemicals at or above threshold quantities, processes that involveflammable liquids or gases on-site in one location, in quantities of 10,000 lbs, ormore (subject to few exceptions), and processes that involve the manufacture ofexplosives and pyrotechnics. Hydrocarbon fuels, which may be excluded if usedsolely as a fuel, are included if the fuel is part of a process covered by this regula-tion. In addition, the regulation does not apply to retail facilities, oil or gas welldrilling or servicing operations, or normally unoccupied remote facilities.

The process safety management regulation requires a systems approach formanaging safety. Segments of the hazardous chemicals industry have for sometimepracticed some or all of the required programs. The promulgation of the regulationformalized the requirements and established a minimum criterion. This is bothgood and bad. The regulation now requires everyone to establish the managementsystems and apply the technologies needed to comply with the regulation. How-ever, because of the same reason, there is a tendency to look for ‘‘paper compli-ance’’ as compared to making real improvements in safety programs and technolo-gies.

The Risk Management Program

In 1996, the EPA promulgated the regulation for Risk Management Programs forChemical Accident Release Prevention (40 CFR 68). This federal regulation wasmandated by section 112(r) of the Clean Air Act Amendments of 1990. The reg-ulation requires regulated facilities to develop and implement appropriate riskmanagement programs to minimize the frequency and severity of chemical plantaccidents. In keeping with regulatory trends, EPA required a performance-basedapproach toward compliance with the risk management program regulation.

The EPA regulation also requires regulated facilities to develop a Risk Manage-ment Plan (RMP). The RMP includes a description of the hazard assessment, pre-vention program, and the emergency response program. Facilities submit the RMPto the EPA and, subsequently, it is made available to governmental agencies, thestate emergency response commission, the local emergency planning committees,and communicated to the public.

The risk management program regulation defines the worst-case release as therelease of the largest quantity of a regulated substance from a vessel or processline failure, including administrative controls and passive mitigation that limit thetotal quantity involved or release rate. For gases, the worst-case release scenarioassumes the quantity is released in 10 min. For liquids, the scenario assumes an


instantaneous spill and that the release rate to the air is the volatilization rate froma pool 1 cm deep unless passive mitigation systems contain the substance in asmaller area. For flammables, the scenario assumes an instantaneous release anda vapor cloud explosion using a 10% yield factor. For alternative scenarios (note:EPA used the term alternative scenario as compared to the term more-likely sce-nario used earlier in the proposed regulation), facilities may take credit for bothpassive and active mitigation systems.

Appendix A of the final regulation lists endpoints for toxic substances to beused in worst-case and alternative scenario assessment. The toxic endpoints arebased on ERPG-2 (Emergency Response Planning Guidelines—Level 2) or levelof concern data compiled by the EPA. The flammable endpoints represent vaporcloud explosion distances based on overpressure of 1 psi or radiant heat distancesbased on exposure to 5 kW/m2 for 40 s.

Hazard and Risk

Hazard

A hazard is a physical or chemical characteristic of a material or process whichhas the potential to cause harm to people, the environment, or property. A hazardcan be a characteristic property of a material, it can be a result of the conditionsof use of the material, or it can be the result of an interaction among two or morematerials or sources of energy. Some examples of hazards include the following:

• Chlorine is a toxic gas.• Gasoline is a flammable liquid.• Sulfuric acid is corrosive.• A cylinder containing compressed air contains significant potential energy from

the pressurized gas.• A mixture of a vinyl monomer and a peroxide initiator has significant potential

chemical energy of reaction.• A 600 psig steam pipe is at elevated temperature and also contains a lot of

energy from the pressure of the steam.

These hazards cannot be changed; they are intrinsic to the material or its conditionsof use. The only way to eliminate or reduce hazards is to change the material orconditions of use. Although it is generally preferable to eliminate or reduce hazards(see later discussion on inherent safety), this is not always possible. The propertiesof a material or system, which create a hazard, may be the same as the properties,which make the material or system useful. A highly reactive monomer, when poly-merized under controlled conditions, will produce a valuable product. However,if the polymerization is not controlled, the result could be overpressurization of areactor and a possible explosion. Therefore, it is often necessary to manage andcontrol the hazards of a process and plant. To do this, you must first identify andunderstand the hazards—hazard identification. Then, you must understand how


the harm to people, the environment, or property can be realized from the hazard-ous material or condition—hazard evaluation or hazard analysis.

Risk

Risk is a measure of human injury, environmental damage, or property loss ex-pressed in terms of both the likelihood of the incident and the magnitude of theinjury, damage, or loss. Risk can be considered to be a function of the potentialincident occurrence, incident consequence, and incident likelihood:

Risk � f (incident, consequence, likelihood)

There can be many different kinds of risk associated with a chemical process orplant; for example, safety risk to plant workers, health risk to workers, health riskto neighbors, risk of various kinds of environmental damage, risk of damage tothe plant or other property, risk of producing product which does not meet specifi-cations and cannot be sold, risk of loss of business due to a plant outage, businessrisk that the product cannot be sold, and others. All of these risks must be under-stood and managed to successfully operate a profitable plant and business over thelong term.

There are many different measures for each of the risks associated with a chem-ical manufacturing facility. For example, some measures of risk to employees ina plant include the following:

• Average risk of fatality from a process accident to an employee in the plant• Maximum risk of fatality from a process accident to the employee at greatest

risk• Average risk to a specific employee in a plant over the course of a normal

working day as he does various specific jobs• An estimate of the distribution of likelihood of accidents of various size (im-

pacting one employee, two employees, three employees, etc.)• Average risk of injury to an employee

The CCPS Guidelines for Chemical Process Quantitative Risk Analysis [13] de-scribes many different measures of safety risk which might be used in understand-ing the risk of a chemical plant and also provides quantitative methodologies forcalculation.

Hazard Identification and Hazard Evaluation

Objective of Hazard Identification and Evaluation

The objective of hazard identification is to fully understand the hazards of a chemi-cal process, including the hazards associated with the following:


• Materials; for example, toxicity, reactivity• Process conditions; for example, high temperature, high pressure• Potential interactions among one or more materials; for example, chemical reac-

tion, decomposition, corrosion

A full understanding of all hazards of a process is the essential first step in eliminat-ing, minimizing, or managing those hazards.

The objective of hazard evaluation is the identification of specific mechanismsby which the potential harm associated with the process hazards can be realized.Hazard evaluation techniques can also include the identification of protective mea-sures which have been incorporated into the process design to manage the hazards,qualitative assessment of the risk of the specific incident scenarios identified, andevaluation of the adequacy of existing protective features or recommendation foradditional safeguards.

Hazard Identification

Hazard identification is based on a complete knowledge of the properties of thematerials being handled and the chemical and physical processes used. Hazardsof materials can be identified from literature searches, publications such as Sax’sDangerous Properties of Industrial Materials [14], and libraries of Material SafetyData Sheets. The best source of hazard information for raw materials is often thematerial supplier. Bretherick’s Handbook of Reactive Chemical Hazards [15] pro-vides a comprehensive summary of reactive chemical hazard literature. Checklistsare a good mechanism for identifying process hazards (checklists as a hazard evalu-ation technique will be discussed later). Two specific tools for hazard identification,which are particularly useful in understanding chemical reaction hazards, are dis-cussed in more detail in the following sections.

Interaction Matrix

The interaction matrix (Fig. 2) is intended to identify chemical reaction hazardsamong materials and energy sources in a chemical process. This tool is particularlyuseful early in the development of a new chemical process. To create an interactionmatrix, list all of the materials, materials of construction, likely contaminants, po-tential sources of energy, process utilities (such as steam, water, nitrogen, com-pressed air, ethylene glycol coolant, and heat transfer oil), and other relevant pa-rameters along each axis of the matrix. It is a good idea to also include ‘‘people’’on one of the axes, to prompt questions about toxicity and other adverse impactsof materials on people. Then, ask what happens for each interaction where thematrix columns and rows intersect. The matrix should go beyond a simple yes–no answer, but rather should provide some detailed information on the nature ofthe interactions identified. Often an interaction matrix will generate more questionsthan answers, particularly early in development. In this case, it may be appropriateto recommend a literature search or laboratory experiments to understand potentialinteractions.


FIG. 2 Example interaction matrix.

Chemistry Hazard Analysis

The Chemistry Hazard Analysis (CHA) is derived from the Hazard and Operability(HAZOP) study methodology. The thought process of a HAZOP study can beapplied at any stage in process development. The CHA is a HAZOP applied to achemical reaction, without the detailed plant design information required for atraditional HAZOP study. For the CHA, the chemist or engineer usually assumesthat the deviation identified by the application of the guide word to the chemicalreaction does occur for some reason, not developing specific causes, and investi-gates the consequences. If the consequences are known, the designer should deter-mine if they represent a hazard which must be understood and managed as a partof the process development, and document this information for future action orreference. In many cases, early in process development, the consequences may notbe known, and additional research or experiments may be needed.

Hazard Evaluation: Selection of Procedure

A number of different hazard evaluation techniques are in common use in thechemical industry, as listed in Table 2. Some techniques, particularly those basedon logic models, require more detailed plant design information and it may notbe possible to apply them early in process development. Table 3 provides some


TABLE 2 Process Hazard Analysis Tools Commonly Used in the Chemical Process Industries

Category Process Hazard Analysis Tool Description and Comments

Brainstorming tech- Safety review Relatively unstructured brainstormingniques Preliminary hazard analysis techniques to identify hazards and po-

What If tential accident scenarios.Hazard and Operability Study (HAZOP) A structured analysis procedure that fo-Failure Mode and Effect Analysis cuses brainstorming activities, includ-(FMEA) ing use of a specific set of guide words

or knowledge and checklists of knownequipment failure modes.

Checklist techniques Checklist Predefined checklists based on previ-What If/checklist ous experience compare a design to

specific standards or good practice.When combined with What If analysis,the checklists are used to prompt brain-storming activities.

Risk-ranking techniques Relative ranking A general category that includes alarge number of quantitative and semi-quantitative techniques which usechecklists or equations based on mate-rial properties, quantities, and handlingconditions to numerically rank risk. Ex-amples include the Dow Fire and Ex-plosion Index and the Dow ChemicalExposure Index.

Logic model techniques Fault-tree analysis Logic models which identify specificEvent-tree analysis causes combinations of events which

lead to a potential accident scenario.These techniques require much detaileddesign information and usually focuson analyzing a few specific accidentscenarios in detail. These techniquescan be quantified and are importanttools in quantitative risk analysis

guidance on how the various techniques have commonly been applied through thelife cycle of a chemical process.

What-If Analysis

What-If Analysis is a brainstorming technique in which a team with expertise onthe process asks ‘‘what-if ’’ questions about the process to identify potential haz-ards or incident scenarios. The Preliminary Hazard Analysis and Safety Reviewtechniques are forms of What-If Analysis. In a What-If Analysis, a team of expertson the process and plant meet in a free brainstorming session to ask ‘‘what-if ’’questions to identify what can go wrong. The technique is very flexible and can


TA

BL

E3

Proc

ess

Haz

ard

Eva

luat

ion

Tec

hniq

ues

Proc

ess

Haz

ard

Ana

lysi

sT

echn

ique Fa

ilure

Haz

ard

Mod

ean

dan

dPr

elim

inar

yE

ffec

tO

pera

bilit

yFa

ult-

Eve

nt-

Che

mic

alPr

oces

sSa

fety

Rel

ativ

eH

azar

dW

hat-

If/

Ana

lysi

sA

naly

sis

Tre

eT

ree

Lif

e-C

ycle

Stag

eR

evie

wC

heck

list

Ran

king

Ana

lysi

sW

hat-

IfC

heck

list

(FM

EA

)(H

AZ

OP)

Ana

lysi

sA

naly

sis

Res

earc

han

dde

velo

pmen

t�

��

��

��

��

�

Con

cept

ual

desi

gn�

��

��

��

��

�

Det

aile

den

gine

erin

g�

��

��

��

��

�

Con

stru

ctio

n�

��

��

��

��

�

Star

t-up

��

��

��

��

��

Rou

tine

oper

atio

n,m

odifi

-�

��

��

��

��

�

catio

ns,

and

expa

nsio

nsD

ecom

mis

sion

ing

��

��

��

��

��

Dem

oliti

on�

��

��

��

��

�

Not

e.—

�:

PHA

tech

niqu

eco

mm

only

used

;�

:PH

Ate

chni

que

rare

lyus

edor

not

appr

opri

ate.

Sour

ce:

Bas

edon

Ref

.5.


TABLE 4 Example of the Results of a What-If Analysis for a Batch Process

Hazard orWhat-If...? consequence Safeguards Recommendations

1. Reactant feed 1. Heat generation 1. High reactant 1. Establish preven-rate is too high? rate exceeds heat flow rate inter- tive maintenance

removal capabil- lock shuts down and testing pro-ity, increased tem- feed. High reactor gram for flowperature, poten- temperature in- and temperaturetial runaway terlock shuts interlocks.reaction. down feed. Rup-

ture disk sized ad-equately to pro-tect reactor formaximum feedrate with no cool-ing.

2. Reactor tempera- 2. Reaction may 2. Low reactor tem- 2. Evaluate ruptureture is too low? stall, resulting in perature alarm disk size relative

a buildup of re- warns operator. to potential pool-actants. This pool ing of unreactedof unreacted mate- material. Basedrial has signifi- on result, deter-cant potential en- mine if additionalergy, possible protection isrunaway reaction needed.if temperature issubsequently in-creased.

be applied at any stage in the process life cycle. The unstructured nature of What-If Analysis can be both an advantage and a disadvantage. With an experienced andknowledgeable team, the technique can be powerful. The discussion and interactionamong team members in the meeting can enhance the identification of hazards.However, the unstructured nature may result in an incomplete analysis by an inex-perienced team. Table 4 is an example of the results of a What-If Analysis. What-If is often combined with a checklist to ensure that a minimum set of ‘‘What-If’’questions are covered by the review team.

Checklist Analysis

A checklist is a list of items used to verify that a plant or process is designed andoperated consistently with a predetermined set of good practices defined by thechecklist. A checklist is often used to confirm that a plant complies with codes,standards, or regulations. Checklists vary from general lists of questions describingcommon chemical hazards and processing concerns to detailed lists of specificrequirements of a standard. Many checklists are simple, requiring only ‘‘yes, no,


not applicable’’ answers. The use of checklist analysis depends on the availabilityof suitable checklists for the plant being reviewed. Good checklists are most likelyto be available for common types of installation, such as flammable-solvent storagefacilities, and are unlikely to be available for unique process operations. Checklistcompleteness depends on the experience of the checklist authors. The output of achecklist analysis is a list of responses to the checklist questions, with areas ofnoncompliance highlighted and recommendations for bringing the facility intocompliance.

Combining a What-If Analysis with a checklist can be a very effective hazardidentification and evaluation technique. The What-If allows creative brainstormingof a team to identify hazards, and the checklist ensures that the team considers aspecified list of hazards based on prior experience, addressing the concern aboutcompleteness of the What-If analysis.

Hazard and Operability Study

A Hazard and Operability Study (HAZOP) is a guide word hazard evaluation tech-nique normally done by a review team. HAZOP begins with the premise that theprocess is safe if operated, as intended, and the team must agree that this is true.Incidents are assumed to result from deviation from intended operation. Guidewords are used in conjunction with the process operating parameters to identifypotential deviations, and the review team determines the consequences of thosedeviations.

A Hazard and Operability Study is best applied when specific process and plantinformation is available (e.g., a detailed plant design or an operating plant). Todo a HAZOP, the process is first divided into sections, or nodes, which are analyzedindividually. A node might be a transfer line from one vessel to another, a pieceof process equipment such as a reactor or heat exchanger, or a step in a batchprocess. The team states the intended operation of each process node, includingvalues of the process parameters—the process intention. The team then appliesguide words to the parameters to identify potential deviations from intended opera-tion. The following are basic guide words:

• No• More• Less• Part Of• Reverse• As Well As• Other

As an example, the guide word ‘‘less’’ can be combined with a specified reactortemperature intention to arrive at the deviation ‘‘less (lower) reactor temperature.’’Once a deviation has been identified, the team determines as many potential causesof the deviation as possible. For the deviation ‘‘less reactor temperature,’’ causesmight include a cooling water control valve stuck open, incorrect temperature setpoint, and others. The team determines the consequences of each deviation, cause


combination, and existing safeguards. It then qualitatively judges the effectivenessof the safeguards to determine if they are adequate. A semiquantitative risk-rankingsystem is often used to aid the team in evaluating the significance of the hazardsidentified. If the existing safeguards are judged to be inadequate, the team shouldrecommend appropriate action to mitigate the potential hazard. The team continuesto apply the guide words to each node until no additional deviations can be identi-fied. These steps are repeated for each process node, until the entire process hasbeen reviewed. Table 5 shows a part of the output of a typical HAZOP study.

Fault-Tree Analysis

A fault tree is a logic model which identifies the multiple ways in which equipmentand human failures in a system can combine to cause an undesired event (the ‘‘TopEvent’’). The analyst begins with a specific undesired event and develops a modelusing Boolean ‘‘AND’’ and ‘‘OR’’ logic gates to identify the immediate causes.These immediate causes can then be further developed to identify their causes,also using Boolean logic gates. The development of the tree continues until itreaches a level of resolution judged to be adequate for understanding potentialincident sequences and identifying system improvements. These events are notfurther developed and are called basic events. Normally, basic events are at thelevel of failure of individual plant components (e.g., a shutoff valve stuck in theopen position, a pump failure to run, a pressure sensor failing to detect high pres-sure). Human error can be incorporated into a fault tree by including specific opera-tor actions (e.g., opening the wrong manual valve) or errors (e.g., failure to act inresponse to a high temperature alarm). Figure 3 is an example of the top levelsof a fault tree for a fire.

A fault tree can be solved using Boolean algebra techniques to identify specificcombinations of individual equipment failures and human errors, which can causethe undesired top event. These combinations of failures and errors describe specificpotential incident scenarios and are called minimal cut sets. The cut sets can beused to identify areas where the system can be improved. Fault trees can also bequantified by assigning failure rate and probability data to the basic events. Thesedata can be mathematically manipulated to estimate the likelihood of the top eventand to understand the relative contribution of individual basic events and cut setsto the total probability of failure.

Event-Tree Analysis

An event tree is a graphical logic model which shows the possible outcomesresulting from an initiating event. An event tree describes the response of a sys-tem to a disturbance created by the initiating event. For example, Fig. 4 shows anevent tree for the sulfuric acid splash incident used to describe the accident pro-cess. An event tree describes a number of potential outcomes from a single initiat-ing event. These outcomes may vary in severity, and the event tree is useful inunderstanding the full range of possible outcomes that can result from a singlesystem failure. Event trees are very useful in understanding the effectiveness of


TA

BL

E5

Exa

mpl

eof

Part

ial

Res

ults

ofa

HA

ZO

P

Dev

iatio

nC

ause

sC

onse

quen

ces

Safe

guar

dsR

ecom

men

datio

ns

1.M

OR

Eth

an16

0lb

s/h

A.

Wro

ngse

tpo

int

(too

A.

Rea

ctor

R-3

10te

mpe

ra-

A.

1.H

igh

flow

alar

mon

A.

Con

firm

that

oper

ator

Raw

Mat

eria

lA

flow

high

)on

FIC

-301

-01

ture

incr

ease

s,he

atba

l-FI

C-3

01-0

1w

arns

oper

a-tr

aini

ngin

clud

espr

oper

B.

FIC

-301

-01

cont

rol

valu

ean

cein

dica

tes

ther

eis

noto

rs.

resp

onse

toFI

C-3

01-0

1st

uck

open

pote

ntia

lfo

rru

naw

ayre

-2.

Hig

hte

mpe

ratu

rean

dT

AH

-310

-05

high

C.

FIC

-301

-01

flow

sens

orac

tion,

even

for

max

i-al

arm

TA

H-3

10-0

5on

Re-

alar

ms.

mis

calib

rate

d—re

ads

low

mum

poss

ible

flow

rate

;ac

tor

R-3

10w

arns

oper

-pr

oduc

tw

illno

tm

eet

ator

.sp

ecifi

catio

nsif

R-3

10B

.Sa

me

asA

.te

mpe

ratu

rein

crea

ses

C.

A-2

only

appl

ies.

A-1

abov

e90

°C.

will

not

prov

ide

anal

arm

B.

Sam

eas

A.

beca

use

the

flow

sens

oris

C.

Sam

eas

A.

mis

calib

rate

d.Pl

ant

expe

-ri

ence

indi

cate

sth

isis

not

likel

y.

Not

e.—

Inte

ntio

n:Fe

ed15

0–16

0lb

s./h

of42

–44%

aque

ous

Raw

Mat

eria

lA

solu

tion

from

Feed

Tan

kF-

101

usin

gPu

mp

P-11

0th

roug

hlin

e10

436

and

flow

cont

rolle

rFI

C-

301-

01to

Rea

ctor

R-3

10.

Raw

Mat

eria

lA

solu

tion

tem

pera

ture

is20

–35°

C.

Whi

lefe

edin

g,R

eact

orR

-310

agita

tor

isru

nnin

gat

50rp

m.


FIG. 3 Top levels of a fault tree for a fire.

the multiple layers of protection, which are often present in a chemical process.Each independent layer of protection is a branch point in the event tree, with thebranches corresponding to the success or failure of the layer of protection. Anevent tree, like a fault tree, can also be quantified by estimating a frequencyfor the initiating event and the probabilities of success and failure at each branchpoint in the event tree. Quantitative event tree analysis is often combined withfault-tree analysis: Fault trees are used to quantify the frequency of the initiatingevent and the probability of failure of the protective systems at the event-treebranch points.

Failure Mode and Effect Analysis

A Failure Mode and Effect Analysis (FMEA) lists the known failure modes ofspecific pieces of equipment in a plant and determines the impact of those failureson the plant. FMEA and HAZOP are similar; the main difference is the startingpoint for identifying potential hazardous incident scenarios. HAZOP starts by pos-tulating a deviation in the value of a process parameter (e.g., more flow) and askingwhat kind of equipment failures or operating errors might have caused that devia-


FIG. 4 Example event tree.

tion and what the process impact will be. FMEA starts by postulating a knownequipment failure mode (e.g., control valve stuck open) and asks what impact thisfailure will have on the operation of the process.

The FMEA starts with a functional description of each piece of process equip-ment and identifies ways in which that piece of equipment might fail to performas designed. A good understanding of the equipment and potential failure modesis required. The FMEA determines how the process will respond to the potentialequipment failure, determines if a potentially hazardous incident will result, identi-fies existing safeguards, evaluates their effectiveness, and develops recommenda-tions for action where appropriate. These steps are very similar to the correspond-ing step in a HAZOP study. Table 6 shows a part of the output from a typicalFMEA study.

Risk-Ranking Techniques

Risk-ranking techniques such as the Dow Fire and Explosion Index [16] and theDow Chemical Exposure Index [17] develop a numerical risk-ranking index basedon various process characteristics such as material properties, chemical reactions,unit operations, operating conditions, and other factors. These risk indices providea relative ranking of specific types of process hazards (e.g., fire and explosionhazards) and are useful for comparing alternate process or plant designs (includinglocation and siting), understanding the parts of a plant or process which are themajor sources of risk, and prioritizing other hazard evaluation and risk managementactivities.


TA

BL

E6

Exa

mpl

eof

Part

ial

Res

ults

ofa

Failu

reM

ode

and

Eff

ects

Ana

lysi

s

Item

and

Des

crip

tion

Failu

rem

ode

Eff

ects

Safe

guar

dsR

ecom

men

datio

ns

1.C

oolin

gW

ater

Con

trol

A.

Val

vest

uck

incl

osed

A.

Una

ble

toin

crea

seflo

wA

.H

igh

tem

pera

ture

inte

r-A

.C

onfir

mth

atth

ere

isno

Val

veFI

C-3

01-1

0.Pn

eu-

posi

tion

ofco

olin

gw

ater

tore

ac-

lock

TA

H-3

10-0

8on

Re-

build

upof

unre

acte

dm

a-m

atic

ally

oper

ated

valv

e,to

rto

cont

rol

tem

pera

-ac

tor

R-3

10cl

oses

raw

teri

alan

dst

oppi

ngfe

edfa

ilscl

osed

onai

rsu

pply

ture

ofba

tch

duri

ngex

o-m

ater

ial

feed

valv

e.w

illst

opre

actio

n.R

e-fa

ilure

ther

mic

reac

tion.

view

mai

nten

ance

proc

e-Po

ssib

leun

cont

rolle

dre

-du

res

and

mak

esu

reac

tion

gene

ratin

ghi

ghV

alve

FIC

-301

-10

isre

g-te

mpe

ratu

rean

dpr

essu

re.

ular

lyin

spec

ted

and

test

ed.

B.

Air

failu

reto

FIC

-301

-B

.V

alve

will

open

,pu

tting

B.

Full

cool

ing

issu

ffici

ent

10fu

llco

olin

gon

the

reac

-to

cont

rol

tem

pera

ture

atto

r.m

axim

umra

wm

ater

ial

feed

rate

.


Summary of Hazard Identification and Evaluation

This section has briefly described a number of commonly used hazard identificationand evaluation tools. Most of these tools are best used in a multidisciplinary teamenvironment, providing a wide variety of plant and process experience and interac-tive discussion to understand the process and identify potential hazards and inci-dent scenarios. CCPS [18] and Wells [19] provide more information on the applica-tion of these and other hazard evaluation tools in the chemical industry.

Consequence Analysis

As mentioned earlier, risk is defined as a function of incident occurrence, fre-quency, and consequence. Consequence analysis is the quantitative estimation ofthe consequence of a chemical process incident—an estimate of the magnitude ofthe potential harm to people, the environment, or property. Because there is awide range of potentially harmful impacts of chemical process incidents, there isa number of different tools which may be useful in analyzing these impacts. Inthis discussion, the consequence analysis tools described will be limited to thosecommonly used to estimate the potential for injury or fatality to people as an imme-diate result of exposure to harmful materials or energy. However, it is recognizedthat there is a wide variety of other potential consequences of incidents and acorrespondingly wide variety of tools used to understand these consequences.

Consequence models can be quite complex and can only be described in generalterms in this discussion. A number of publications by the Center for ChemicalProcess Safety [20–24] describe specific types of consequence analysis models indetail. Les [25] also provides a detailed description of incident consequence mod-els. There are also a number of public domain and proprietary commercial com-puter-modeling systems available for chemical release consequence analysis.

Source Models

Source models quantitatively estimate the magnitude, rate, duration, physical state(solid, liquid, gas, or a combination), and temperature or other physical conditionof a chemical release based on the physical and chemical parameters associatedwith a particular release scenario.

Most source models are well developed in chemical engineering theory andare essentially the same as the models used for similar material flow scenariosused to design plant equipment. These include single-phase and multiphase flowmodels for flow-through holes, orifices, and pipes, which are readily adapted todescribe flow from a leaking pipe or vessel. Two-phase flashing flow models arebased on technology developed by the Design Institute for Emergency Relief Sys-tems (DIERS) [26]. Two-phase or flashing jet release models must also considerthe formation of fine aerosols in the discharge and the potential for the small drops


to remain suspended in the atmosphere rather than ‘‘raining out’’ into an evaporat-ing pool on the ground. For a discharge of material from a reactive system, themodels required to fully understand the system may be quite complex and datafrom reaction calorimetry tests may be required.

If a release is wholly or partially in the form of a liquid, it will form a poolon the ground. The evaporation of vapor from this pool is another potential sourceterm for atmospheric dispersion models, which estimate the downwind concentra-tion of the vapor. The first step in estimating evaporation from a liquid pool is toestimate the size of the pool. Pool size models consider the momentum of theliquid stream entering the pool, gravity spreading resulting from the depth of thepool, and the liquid physical properties (e.g., viscosity, surface tension, and surfacewetting properties). Physical constraints such as dikes and containment systemsmay also determine the size of a pool of spilled liquid.

There are three major pool evaporation situations, which are typically modeled:boiling liquid pools, volatile liquid pools, and relatively nonvolatile liquid pools.

• Boiling liquid pools occur when the pool liquid boils at a temperature belowthat of its surroundings (the ground and atmosphere). In this case, vapor genera-tion is controlled by heat transfer into the liquid pool, both from the groundand from the surrounding atmosphere. The vapor release rate is determined froman estimate of the total heat transfer into the pool and the heat of vaporizationof the liquid.

• Volatile liquid pools exert a significant vapor pressure but are at a temperaturebelow the liquid boiling point. Evaporation models for volatile liquid poolsconsider both heat transfer into the pool and mass transfer rates into the atmo-sphere from the pool surface.

• The evaporation of relatively nonvolatile liquid pools is primarily determinedby mass transfer at the surface of the pool. Because the evaporation rate is low,the pool temperature will be essentially the same as the temperature of thesurroundings after any initial temperature differences equilibrate. Evaporationmodels are based on standard methodologies for estimating convective masstransfer from a liquid into a gas.

Vapor Cloud Dispersion

Vapor cloud dispersion models estimate the area covered by the vapor cloud froma chemical release as it disperses in the atmosphere, and they estimate the vaporconcentrations at specific locations in the cloud. Some of the data required for avapor cloud dispersion model include the following:

• Characteristics of the release, including rate, total quantity released, location ofthe release

• Characteristics of the release (phase, direction, velocity, composition, tempera-ture, pressure)


• Atmospheric conditions, including wind speed, atmospheric stability, tempera-ture, pressure

• Characteristics of the surface, including surface roughness

Some of the complex vapor cloud dispersion models may require additional infor-mation to characterize the release, the atmospheric conditions, and surface condi-tions.

Vapor cloud dispersion models consider three typical types of behavior:

• Neutrally buoyant gases (having a density close to the density of air)• Positively buoyant gases (having a lower density than air)• Dense or heavy (negatively buoyant) gases

Two major types of releases must also be considered:

• Instantaneous (puff releases)• Continuous releases (plumes)

The CCPS [20] describes vapor cloud models in detail, including all of themajor types of dispersion models and release types. The CCPS [22] provides amore condensed summary of some of these models. The output of these modelsdescribes the concentration of the released material in both time and space as thevapor cloud travels downwind.

Fires

Incident consequence analysis may require consideration of one or more of severaldifferent types of fire:

• Pool fire: a burning pool of a flammable or combustible liquid• Jet fires: burning of a flowing jet of flammable liquid or gas, usually from a

pipe or vessel• Flash fire: nonexplosive combustion of a flammable mixture of a combustible

vapor in air

Pool Fire

The primary mechanism of damage from a pool fire is thermal radiation from theflame. Pool fire models estimate the thermal effects based on the properties of thematerial burning in the pool, the geometry of the pool, atmospheric characteristics,and geometry of the fire relative to the receiving source. Pool fire models arewell developed. They are based on empirically determined characteristics such asburning rate, flame height, surface emissive power, and atmospheric transmissivity,all of which are well established in the literature. Pool fire models provide anestimate of the thermal radiation at locations of interest surrounding the fire.


Jet Fire

Many jet fire models are based on models used for the design of flare systems.As for pool fires, the damage from a jet fire results primarily from thermal radiationfrom the fire. Jet fire models require an understanding of the characteristics of thejet (discharge rate and velocity, material burning properties) and, like pool fires,characteristics of the atmosphere. Jet fire models are primarily empirical but arederived from much data and experience. The models produce an estimate of thethermal radiation at locations surrounding the jet fire.

Flash Fire

A flash fire results from the ignition of a cloud of flammable gas (a cloud containinga flammable material at a concentration between its lower and upper explosivelimits in air). Such a cloud can explode under the proper conditions (size anddegree of confinement), resulting in a vapor cloud explosion. If the conditionsrequired for a vapor cloud explosion are not present, the cloud may still ignite andburn. In this case, the burning cloud will not generate pressure and an explosion,but the flash fire is still capable of causing significant damage. The primary hazardis from direct contact with the flame and from thermal radiation, which is normallyfor a brief time of a few tenths of a second. Flash fires are normally modeled bydetermining the dimensions of the flammable cloud using vapor dispersion modelsand estimating the thermal radiation resulting from combustion of the cloud.

Explosions

Chemical incident consequence analysis may need to consider four types of explo-sion:

• Physical explosions: the failure of a vessel containing material under pressurewithout chemical reaction (e.g., due to a vessel defect or excess pressure in thevessel).

• Vapor cloud explosions: explosion of a cloud of flammable vapor dispersed inthe atmosphere.

• Confined explosions: explosion resulting from a rapid chemical reaction gener-ating high temperature and pressure inside a confined space such as a vesselor a building.

• Boiling liquid expanding vapor explosions (BLEVE): the catastrophic failureof a vessel containing a superheated liquid. If the liquid is flammable, ignitionmay result in a fireball.

The CCPS [20,22] describes commonly used explosion models in detail.


Physical Explosions

Physical explosion models generally estimate the amount of energy which wouldbe released by the sudden expansion of the material contained in a vessel from itsinitial temperature, pressure, and volume to atmospheric pressure. This estimatedenergy is then converted to an equivalent amount of TNT. A number of correlationsof explosion pressure as a function of distance from a TNT explosion have beenpublished, and these can be used to estimate damage. It may also be necessary toconsider the potential impact of the vessel fragments, which result from a vesselexplosion. Empirical models to estimate the number and size of fragments, theirtravel distance, and energy are available.

Vapor Cloud Explosions

If ignited, a flammable vapor cloud can burn as a flash fire, or, if the flame speedaccelerates sufficiently, it can produce significant blast pressure from a vapor cloudexplosion. A number of factors have been found to be important in determiningwhether a vapor cloud explosion occurs when a flammable vapor cloud is ignited.These include the following:

• Turbulence in the vapor cloud. This turbulence may arise from the energy fromthe release of the fuel itself (from a jet or catastrophic loss of containment) orfrom the interaction of the cloud with its surroundings during the combustionprocess.

• Partial confinement of the vapor cloud as a result of obstacles, structures, orother factors, which could cause local partial confinement. The explosive com-bustion in the locally confined cloud can propagate into the rest of the cloud.

• Mass of the cloud. Experimental studies have demonstrated that there is a mini-mum mass of flammable material required to transition to a vapor cloud explo-sion. The CCPS [21] reports studies indicating that this minimum mass is inthe range of 1 to 15 tons for typical hydrocarbons.

• Combustion properties of the fuel. Materials with a high fundamental burningvelocity such as ethylene oxide and ethylene are reported to be more readilyinclined to propagate to a vapor cloud explosion.

Vapor cloud explosions are modeled using three types of model:

• TNT Equivalency Models. The total energy available from the combustion isestimated from the mass of fuel in the cloud and the heat of combustion of thefuel. This combustion energy is then converted to an equivalent mass of TNTand reduced by an ‘‘explosion efficiency’’ factor, which is empirically esti-mated. The explosion overpressure and other characteristics can then be esti-mated as a function of distance from the cloud using readily available experi-mental data for TNT explosions. TNT equivalency models are empirical, andthe results are strongly dependent on the explosion efficiency, which may notbe known for a particular material or cloud configuration. TNT equivalencymodels also do not characterize the vapor cloud explosion well in the area close


to the cloud, where they may predict much higher pressure than typically resultfrom the combustion of a flammable cloud.

• Multienergy Method. This model is based on the assumption that the blast char-acteristics of a flammable vapor cloud depend more on the level of congestionand confinement than on the fuel. The models require dispersion models todetermine the size of the cloud. Then, areas with different confinement andcongestion characteristics are identified and considered to be sources of strongblasts. The energy from each blast source is estimated, and the potential damageis estimated from empirically derived correlations.

• Baker–Strehlow Method. This model also considers confinement as the basisfor the size of the flammable vapor cloud. It also considers burning characteris-tics and reactivity of the fuel, geometry of the confined volume, and the degreeof confinement created by the obstacles in the confined volume. Blast character-istics are then estimated using a set of correlations and charts.

These models are discussed in detail in Refs. 20 and 22.

Confined Explosions

Confined explosions result from combustion or another rapid chemical reactionin a confined vessel or building. Confined combustion reactions may occur withflammable vapor–air mixtures or from the dispersion of a cloud of combustibledust in air. The combustion or reaction products are often gases, and pressure isgenerated by the gas and also the elevated temperature resulting from the heat ofcombustion or reaction. Confined explosions are modeled by estimating the peakpressure that can be generated from the chemical reaction. The models are specificto the reaction and may require considerable reaction thermodynamic and kineticdata. The maximum pressure resulting from the reaction model is then comparedto the failure characteristics of the confining vessel or building. If the pressureexceeds the expected failure pressure of the vessel, the damage resulting fromvessel failure and the potential for damage or injury from fragments can be esti-mated using the methods for physical explosions discussed earlier.

Boiling Liquid Expanding Vapor Explosions

A BLEVE is the rapid release of a large amount of superheated liquid to the atmo-sphere. It often occurs as a result of weakening of a pressure vessel caused bydirect flame impingement on the vessel above the liquid level. This weakens themetal vessel and it can fail rapidly and catastrophically. The sudden loss of con-finement allows the superheated liquid to rapidly flash, increasing its volume sev-eral hundred times and generating a pressure wave and fragments. If the releasedliquid is flammable, it can also ignite, resulting in a fireball. BLEVE models arebased on the expansion energy of the flashing liquid. Blast effects tend to be local,and the impact of the fireball, which usually accompanies a BLEVE of a flammablematerial, is the more important source of damage. BLEVE fireball models empiri-cally estimate the fireball dimensions based on the quantity of material released.


Thermal radiation characteristics of the fireball are then modeled using a combina-tion of empirically derived relationships and fundamental models for the geometryof the fireball with respect to the receptor and atmospheric transmission of thethermal radiation. The result is an estimate of the radiant energy flux level andduration at various locations surrounding the BLEVE.

Effect Models

The result of the application of the models discussed in this section so far is anestimate of some type of physical parameter at various locations surrounding achemical release: a concentration of toxic gas in the atmosphere, the amount ofradiant energy at a specific location from a fire, and the peak pressure and impulseduration from an explosion. Effect models estimate the damage, which results fromthese physical effects. There are a wide range of possible effect models correspond-ing to the wide range of potential damage to people, the environment, and property,which can result from exposure to toxic materials, fires, and explosions.

The CCPS [22] provides a summary of effect models commonly used to esti-mate the impact of toxic vapors, fires, and explosions on people. These modelsare generally empirical and are based on experimental data and evaluation of theconsequences of past incidents. Models are available to estimate the impact of ahazardous agent using the dose-response relationship (e.g., relating probability offatality to concentration and duration of exposure by inhalation of a toxic gas,relating severity of burns to intensity and duration of exposure to thermal radiation,or estimating damage to structures based on peak overpressure and duration).

Risk Assessment

However many the resources we devote to the prevention of accidents, we cannever eliminate every risk. We have to decide our priorities: Which risks shouldwe deal with first? Which are so small compared with the other risks to which weare exposed that we should tolerate them, at least for the time being? Often, thejudgment is qualitative: Some risks are illegal; some are obviously intolerable largeor acceptably small; sometimes a generally accepted standard or code of practicetells us what to do. In other cases, the decision is not obvious and we use a numeri-cal method known as quantitative risk assessment (QRA), probabilistic risk assess-ment (PRA), or, in the chemical industry, hazard analysis (Hazan).

Stages of Risk Assessment

Before carrying out a risk assessment, we have to identify the hazards (i.e., thesubstances, objects, or situations that can give rise to injury or damage) using oneof the methods described earlier. (A risk, in contrast, is the probability that injuryor damage will occur.) There are then three questions to answer:


• How often will injury or damage occur?• What is the extent of the injury or damage?• What action should we take?

Whenever possible, the answer to the first question should be based on experi-ence, but often there is no experience, as the equipment is new or failure has neveroccurred. We then estimate a failure rate for the equipment as a whole, based on theknown failure rates of its components, as described earlier. Similarly, the answer tothe second question should be based on experience whenever possible but can beestimated as by one of the methods described earlier. The answer to the thirdquestion depends on the nature of the consequences. If damage is possible butinjury is not, then the average cost of the damage (including consequential loss)is compared with the cost of prevention.

If injury is possible, then the QRA approach is to set a target or criterion,usually based on the risk to life. Risks above a certain level should be removedor reduced as a matter of priority. Those below this level can be left alone, at leastfor the time being. Thus, QRA is a method for determining priorities. In a laterdevelopment, there are two levels of risk. Risks above an upper level are consideredintolerable; if they cannot be reduced, the plant should not be built (or should notbe operated if it is already built). The risk considered tolerable for members ofthe public is much lower than that considered tolerable for employees. Risks belowa much lower level are considered acceptable and need not be reduced. In betweenthe two levels, we reduce the risks if we can, but we tolerate them if it is impractica-ble or very expensive to do so. The pressure to reduce them is great if the risk isnear the intolerable level and reduces as we approach the acceptable level.

The extent to which this approach is used and the risk levels are made explicitdiffers from country to country. The United Kingdom has long accepted the princi-ple that we should compare the size of a risk with the cost, in money, time, andtrouble, of removing it (although the ability to pay is not a deciding factor). Ifthere is a gross disproportion between them, the risk being insignificant in relationto the cost, the risk can be tolerated. QRA was therefore accepted readily and theregulatory authority has suggested figures for the tolerable and acceptable risklevels. Other governments have been reluctant to admit that even trivial and infre-quent risks should be tolerated and this has hindered the use of QRA.

The actual risk levels suggested for the United Kingdom are as follows. Theyare similar to those used by many organizations elsewhere.

Risk of death perperson per year

Maximum tolerable risk (employees) 10�3

Maximum tolerable risk (public) 10�4

Maximum tolerable risk (public—nuclear risks) 10�5

Broadly acceptable risk (employees and public) 10�6

The maximum tolerable risk to employees seems rather high, but this risk is, infact, tolerated in some industries.


For comparison, the annual risk of death from all causes is about 10�4 forsomeone aged 20 years and about 10�3 for someone aged 60 years.

Public Attitudes

Quantitative risk assessment is difficult to explain to the public. They pick on thefact that a number of people could be killed in an industrial accident but ignorethe fact that the probability that this will occur is extremely low. The death of 10people once in 10 years is given far more publicity than the death of 1 person peryear for 10 years. As a result, public pressure often compels industry and govern-ment to reduce risks which are already low but which the public perceives as high.At its best, this is democracy in action; at its worst, it is giving the most to thosethat shout the loudest.

The public tends to oppose risks with the following traits:

• Imposed rather than accepted voluntarily• Not under the individual’s control• Of no obvious benefit to them• Man-made rather than natural• Unfamiliar• Dreaded (e.g., cancer is more dreaded than heart disease though the latter kills

far more people)• Immoral (e.g., crime is feared more than road accidents)• Associated with unpleasant events (e.g., nuclear power reminds us of atomic

bombs)

When the public cannot judge the message, they judge the messenger. Unfortu-nately, most of these concerns make the man in the street oppose the chemicalindustry: The risks are imposed, not under his control, man-made, unfamiliar, anddreaded; past experience has been unpleasant; the industry does not obviouslybenefit him; and the spokesmen for the industry are often outsiders. There is noeasy way of countering this perception. We try to explain the benefits of the indus-try and the low levels of risk, but we cannot say that accidents will never happen.

Incident Investigation

The purpose of incident investigation is to find out why the incident occurred sothat we can prevent it from happening again. The purpose is not to find out whoshould be blamed. Many people have an opportunity to prevent almost every inci-dent. Figure 5 shows by example the opportunities that are available to prevent afire or minimize the consequences of an apparently simple incident: An expansionjoint (bellows) was incorrectly installed in a pipeline so that it was distorted. Aftersome months, it leaked and a passing vehicle ignited the escaping vapor. Damage


FIG. 5 An example of an accident chain. An expansion joint (bellows) was incorrectly installed sothat it was distorted. After some months, it leaked and the escaping vapor was ignited by apassing vehicle. Damage was extensive, as the surrounding equipment had not been fire-protected to save the cost. Many people in various functions could have prevented the incidentor minimized the consequences.

was extensive, as the surrounding equipment had not been fire-protected to savethe cost.

Many people could have prevented the fire, not just the fitter who installed theexpansion joint incorrectly. The fire could have been prevented by better detaileddesign (not using expansion joints for hazardous materials), by better design meth-ods (using HAZOP, consulting experts, better design standards, better training ofdesigners), by better training of the fitter, by better inspection of workmanship,by keeping eyes open on plant visits, and by not tolerating poor workmanship inthe past.

We should investigate all incidents, including those, which, by good fortune,caused no injury or damage, but might easily have done so. Next time, they may.


Finding the Facts

• Include people with a variety of experience on the investigating panel. It shouldnot be too large; four or five people are usually sufficient.

• Do not disturb evidence that may be useful to experts who may be called inlater.

• Draw up a list of everyone who may be able to help, such as witnesses, experts,designers, and people on other shifts.

• Be patient when questioning witnesses. Valuable information may be missedif we try to take police-type statements. Do not put ideas into people’s minds.Avoid questions to which the answer is ‘‘yes’’ or ‘‘no.’’

• Make it clear that the objective of the investigation is to find out the facts, sothat we can prevent the incident happening again, not to establish blame.

• Inform any authorities who have to be notified.• Record information on damage and injuries so that others can use it for predic-

tion.

Drawing Conclusions from the Facts

Accident investigation is like peeling an onion. Beneath the immediate technicalcauses, look for ways of avoiding the hazard, such as inherently safer design. Lookalso for weaknesses in management, such as poor training or instructions or turninga blind eye to previous failures to follow instructions.

Concentrate on prevention rather than causes. Look for causes that lead toactions. Do not, for example, quote corrosion as a cause and stop there. Ask if itwas foreseen. If not, why? If it was foreseen, why did it occur? Was the rightmaterial of construction used? Were operating conditions outside the design range?Was monitoring carried out? If so, were the results followed up?

Preventing leaks is a more effective way of preventing liquid and gas fires thanremoving sources of ignition (although we should also do what we can to removeknown sources of ignition). Avoid the use of the term ‘‘human error’’ and neverrecommend someone to take more care. Instead, ask if we need better training,better instructions, or better compliance with instructions and, if so, say how thiswill be achieved. If an error was due to a slip or lapse of attention, inevitable fromtime to time, look for ways of removing opportunities for error.

It is often useful, especially when investigating fires and explosions, to askwhy it occurred when it did and not at some other time. It is also useful to askif similar incidents, perhaps with less serious results, have occurred before,what recommendations were then made, and if they were effective or allowed tolapse.

Avoid long shopping lists of possible recommendations. Ask if the cost of eachrecommendation is proportional to the size of the risk. Consider alternative solu-tions as well as the obvious ones.

For each recommendation, make it clear who will carry it out and when. Bringthe report forward at that time. Otherwise, nothing will happen except a repeat ofthe incident.

Managers should not accept reports that fall short in any of these respects.


They should look out for what is not said. For example, writers of accident reportsare naturally reluctant to draw attention to similar incidents that had occurred else-where and, if they had been followed up, could have prevented the accident.

Spreading the Message

Many companies restrict the circulation of incident reports, but this will not preventthe incident from happening again. We should circulate the essential messagesthroughout the company. There is no need to say where the incident occurred.Remember that incident reports grab people’s attention and are read, whereas ad-vice and instruction are put aside to be read when we have time (if we ever do).Having paid the high price of an accident, we can recover some of the cost byturning it into a learning experience.

Circulate reports containing new or forgotten information throughout the indus-try, so that others can learn from them. There are several reasons for doing so.

• Moral: If we have information that might prevent another accident, we have aduty to pass it on.

• Pragmatic: If we tell other organizations about our accidents, they may tell usabout theirs.

• Economic: We would like our competitors to spend as much as we do on safety.• The industry is one: Every accident affects its reputation.

Remembering the Message

Incident reports are written, acted on, and then filed and forgotten. After a fewyears, people forget the reasons for the changes that were made. Procedures lapseor the equipment falls out of use and the incident happens again, even in the plantwhere it happened before. To prevent this from happening we should do the fol-lowing:

• Include in every instruction, code, and standard a note on the reasons for it andaccounts of accidents that would not have occurred if the instruction, code, orstandard had been followed.

• Never remove equipment before you know why it was installed. Never abandona procedure before you know why it was adopted.

• Describe prior accidents as well as recent ones in safety bulletins and discussthem at safety meetings. Giving the message once is not enough.

• Follow up at regular intervals to see that the recommendations made after acci-dents are being followed, in design as well as operations.

• Remember that the first step down the road to an accident occurs when someoneturns a blind eye to a missing blind.

• Include important accidents of the past in the training of undergraduates andcompany employees.

• Keep in every control room a folder of reports on past accidents. It should beread by all new arrivals and others should browse it during quiet shifts.


• Devise better retrieval systems so that we can find, more easily than at present,details of past accidents in our own and other companies and the recommenda-tions made afterward.

The Management of Safety

Inherently Safer Design

The first step in the management of safety, after the hazards have been identified(see the section Hazard Identification and Hazard Evaluation), is to see if they canbe removed. Only when we cannot do so, should we look for ways of keepingthem under control or mitigating their consequences. When we remove a hazard,the safety is inherent in the design and cannot be lost. When we control a hazard,the protective equipment may fail, or be neglected, or the safety procedures maylapse.

Note that we refer to inherently safer, not safe, design as we can rarely, if ever,remove every hazard. The principle routes to inherently safer design are as follows:

• Intensification or minimization: Using so little hazardous material that it doesnot matter if it all leaks out. ‘‘What you don’t have, can’t leak.’’ This mayseem obvious but until the explosion at Flixborough, UK in 1974 little thoughtwas given to ways of reducing the amount of hazardous material in a plant.Engineers simply designed a plant and accepted whatever inventories the designrequired, confident that they could keep it under control. Flixborough weakenedthat confidence, and 10 years later, Bhopal almost destroyed it.

Microreactors promise much greater intensification than has been possible inthe past. Intensification, when it is practicable, is the first choice, as it brings aboutgreater reductions in cost. If less material is present, we need smaller pipes andvessels and smaller structures and foundations.

• Substitution: If intensification is not possible, then an alternative is substitution,using a safer material in place of a hazardous one. Thus, it may be possibleto replace flammable solvents by nonflammable ones and processes that usehazardous raw materials or intermediates by processes that use safer ones.

• Attenuation or moderation: Another alternative to intensification is attenua-tion—using a hazardous material under the least hazardous conditions. Thus,liquefied chlorine and ammonia can be stored as refrigerated liquids at atmo-spheric pressure instead of storing them under pressure at ambient temperature.The lower pressure results in smaller leaks through a hole of a given size leakand the lower temperature results in less evaporation.

• Limitation of effects, by changing designs or reaction conditions rather than byadding on protective equipment, which may fail or be neglected. For example,it is better to prevent overheating by using steam or oil at a safe temperaturethan by using a hotter medium and a control system.


• Simplicity: Simpler plants are safer than complex plants, as they provide feweropportunities for error and contain less equipment that develop faults. They areusually also cheaper.

Defense in Depth

The hazards that cannot be removed have to be controlled. Because we dependon equipment and people, both of which may fail, we use defense in depth. If wehandle flammable liquids or gases and an inherently safer design is not possible,we use some or all of the following:

• Prevent leaks by good design, construction, maintenance, and operation.• Install automatic detectors so that leaks are detected promptly and people not

required to deal with the leak can leave the area.• Install remotely operated emergency isolation valves in places where leaks are

most likely to occur or where a large quantity could leak.• Remove all known sources of ignition.• Minimize damage by installing fire protection. Passive equipment such as fire

insulation is usually better than active equipment such as water spray turnedon by automatic equipment. This is better than active equipment turned on bypeople.

• Provide fire-fighting equipment.

It is essential to carry out regular audits—tests and inspections to make surethat automatic equipment is in working order and those procedures have not lapsed.

Human Factors

Engineers are interested in equipment, its failures, and ways of preventing themand often less interested in people. However, all systems involve both equipmentand people. Engineers, whether they are designers, supervisors, or managers, there-fore, should understand the way people react with equipment and why they some-time fail to act in the way we instruct them or expect them to act.

• Some errors, usually called mistakes, occur because people do not know whatto do. The intention was wrong. Employers should provide adequate trainingand instructions and should not write the sort of instructions that are designedto protect the writer rather than help the reader. However, for many instructionswe write, problems will arise that are not covered by them and so people, partic-ularly operators, should be trained in flexibility (i.e., the ability to diagnose andhandle unforeseen situations). If instructions are not being followed, are theytoo complex? Can the job be simplified?

• Some errors, usually called violations or noncompliances, occur because some-one knows what to do but makes a deliberate decision not to do it. Some viola-tions occur because all people carrying out routine tasks tend to cut cornersafter a while. Many more occur because people think they know a better way


of doing the job. Note that if the instructions are wrong, noncompliance mayachieve the intention. There is a fine line between showing initiative and break-ing the rules.

To prevent or reduce violations, we should do the following

• Explain the reasons for the instructions. We do not live in a society in whichpeople will simply do as they are told. They want to know the reason why.

• If possible, simplify the job. If the correct method is difficult, an incorrectmethod will be used.

• Carry out checks from time to time to see that instructions are being followedand do not turn a blind eye if they are not.

• Some errors (mismatches) occur because the job is beyond the physical or men-tal ability of the person asked to do it, sometimes beyond anyone’s ability. Forexample, errors occur if people are overloaded, or underloaded, or asked tobreak well-established habits. We should change the plant design or method ofworking.

• The fourth category is the commonest—a momentary slip or lapse of attention.People know what to do, intend to do it, and are able to do it, but it slips theirmind. Compared with mistakes, the intention is correct but is not fulfilled. Theyhappen to everyone from time to time and cannot be prevented by telling peopleto be more careful or by telling them to keep their minds on the job. All wecan do is to change the plant design or method of working so as to removeopportunities for error (or minimize the consequences or provide opportunitiesfor recovery). We should, whenever possible, design inherently safer plantswhich can withstand errors (and equipment failures) without serious effects onsafety (and output and efficiency).

Managers and designers as well as operators make errors, but because theyusually have time to check their work, slip and lapses of attention are infrequent.Most of their errors are mistakes or violations.

Management Systems

Some management systems have been discussed in earlier sections on risk assess-ment, hazard identification, and accident investigation. The following are also im-portant:

The preparation of equipment for maintenance: Many accidents have occurredbecause equipment was not isolated correctly, was not freed from hazardousmaterials, or was not correctly identified and the wrong equipment was openedup. Sometimes, procedures were poor; sometimes, they were not followed.

The management of change: Many accidents have occurred because a change toplant, process, or organization had unforeseen effects. Before any change ismade, it should be examined by professionally qualified people using HAZOP(or a simpler technique if the change is minor) and then inspected after comple-tion to make sure that the intention has been followed and that the modification


looks right. What does not look right is often wrong and should always bechecked.

Testing and inspection of equipment: All protective equipment is liable to fail andshould be tested or inspected at regular intervals. When active equipment suchas relief valves and interlocks fails, the failure is usually hidden and regulartesting is necessary. If passive equipment such as fire insulation is missing,this is visible, but, nevertheless, it should be checked regularly. If 10% of thefire insulation on a vessel is missing, the rest is useless. The following equip-ment is often overlooked but should be tested or inspected regularly:• Drain holes in relief valve tailpipes. If they choke, rainwater will accumulate

in the tailpipe.• Drain valves in tank bunds. If they are left open, the bund is useless.• Emergency equipment such as diesel-driven firewater pumps and genera-

tors.• Earth connections, especially the moveable ones used for earthing road

tankers.• Fire and smoke detectors and fire-fighting equipment.• Flame arrestors.• Hired equipment. Who will test it, the owner or the hirer?• Labels are a sort of protective equipment. They vanish with remarkable

speed and regular checks should be made to make sure that they are stillthere.

• Mechanical protective equipment such as overspeed trips.• Nitrogen blanketing (on tanks, stacks and centrifuges).• Nonreturn valves and other backflow prevention devices, if their failure can

affect the safety of the plant.• Open vents. These are the simplest possible sort of relief device and should

be treated as relief valves.• Spare pumps, especially those fitted with auto-starts.• Steam traps.• Trace heating (steam or electrical).• Valves, remotely operated and hand-operated, which have to be used in an

emergency.• Ventilation equipment.• Water sprays and steam curtains.

All protective equipment should be designed so that it can be tested or inspected.Test results should be displayed for all to see, for example, on a board in thecontrol room.

Operators sometimes regard tests and inspections as a nuisance, interferingwith the smooth operation of the plant. Training should emphasize that protectiveequipment is there for their protection and they should ‘‘own’’ it.

Remembering the past: A most important system, discussed in the subsection Re-membering the Message, is one to ensure that the lessons learned from pastaccidents, in our own and other companies, is not forgotten and that the infor-mation can readily be retrieved.


Introducing and maintaining systems: When systems are introduced or changed,they should be discussed with those who will have to operate them and notjust sent to them through the mail. Discussions should start with descriptionsof incidents that would not have occurred if the systems had been in operationat the time. These have much more impact than mere procedures and bringout the need for the changes. Discussions will allow the manager to check thatthe message has been received and understood and he may discover that it isimpracticable or difficult to use in its present form.

All systems are subject to a form of corrosion more rapid than that whichaffects the steelwork and can vanish without trace once managers lose interest.Continuous monitoring is necessary to make sure that systems continue in use.

Limitations of systems: Some managers seem to believe that good safety manage-ment systems will ensure a safe plant. All the systems can do, however, isensure that people’s knowledge and experience are applied systematically. Ifthe staff lack knowledge and experience, then the systems are empty shells.People will go through the motions, but the output will be poor. Without asystem, people will not achieve their full potential. Without knowledge andexperience, systems will achieve nothing. This is a particular danger at timeswhen companies are reducing manpower and experienced people are leaving.Senior managers should systematically assess the levels of knowledge and ex-perience needed and ensure that they are maintained.

Audits

We need audits of equipment and procedures by outsiders because of the following:

• Those who work in a plant do not notice the hazards they see everyday.• Auditors may have specialized knowledge and thus see hazards not apparent

to others.• Auditors have more time for investigation in depth than those who work regu-

larly on a plant.

Safety auditing should not be a police activity; it is intended to help the localmanagement, who may miss hazards through familiarity, ignorance, or lack oftime.

Auditors should pay particular attention to the following:

• The quality of the training and instructions and the knowledge and experienceof employees.

• The procedures for preparing equipment for maintenance, controlling modifica-tions, and testing protective equipment and whether or not these procedures areactually followed.

• Procedures for investigating accidents, passing on the lessons learned, and en-suring that they are not forgotten.

• Process hazards as well as mechanical ones.


• Places where others do not look, behind, and underneath equipment.• Although it may be a separate exercise, process hazards should be reassessed

every few years in the light of new knowledge and new techniques.

Auditors (and managers) should visit the plant at night and at weekends, not justduring the day.

The Measurement of Safety

Whenever possible, we should provide a numerical measure of the success of eachmanagement function. Accident rates in good companies are now so low that theusual measure of safety, the lost-time accident rate, merely measures luck and thewillingness of injured people to remain at work. In any case, it never measuredprocess safety. Possible additional or alternative measures are as follows:

• An index based on audit results. Unlike many other measures of safety, thisone tries to detect falling standards before an accident occurs.

• A monthly summary of the cost of incidents.• An annual report of the progress made in reducing inventories of hazardous

substances.• The number of faulty permits-to-work found by routine inspection.• The number of faulty protective systems found by routine testing.

Mitigation

Mitigation is the cornerstone of emergency management. It is the ongoing effortto lessen the impact disasters have on people and property. Mitigation involveskeeping homes and populated areas away from industry, engineering process plantsto be inherently safer, and creating and enforcing effective engineering codes toprotect employees, the public, and the environment from potential process plantupsets and incidents.

Mitigation is defined as ‘‘sustained action that reduces or eliminates long-termrisk to people and property from hazards and their family and belongings are betterprotected from floods, earthquakes, hurricanes, and other natural hazards. Theycan be utilized to help business and industry avoid damages to their facilities andremain operational in the face of catastrophe. Mitigation technologies can be usedto strengthen hospitals, fire stations, and other critical service facilities so that theycan remain operational or reopen more quickly after an event. In addition, mitiga-tion measures can help reduce disaster losses and suffering so that there is lessdemand for money and resources in the aftermath.

In practice, mitigation can take many forms. It can involve actions such as thefollowing:

• Promoting sound land use planning based on known hazards• Buying flood insurance to protect your belongings


• Relocating or elevating structures out of the floodplains• Having hurricane straps installed to more securely attach a structure’s roof to

its walls and foundation• Developing, adopting, and enforcing effective engineering codes and standards• Engineering process plants to be inherently safer• Using fire-retardant materials in new construction• Developing and implementing a plan in your business or community to reduce

your susceptibility to hazards

In the multiple-barriers concept and development of inherently safer layers ofprotection, mitigation is a much lower-level activity and should be looked at onlywhen all other measures in the inherently safer hierarchy are exhausted. For exam-ple, mitigation should be addressed only after the following options have beenexhausted:

• Inherent safety: These include inventory reduction (i.e., less chemicals storedor less in process vessels), substitution of a less hazardous chemical for onemore hazardous, and use of lower temperatures and pressures.

• Engineering design: Examples are use of better seals or materials of construc-tion, ensuring proper operating conditions and material purity, and installingdikes and spill vessels.

• Management: Examples include consistent operating policies and procedures,training for vapor release prevention and control, audits and inspections, equip-ment testing, maintenance program, management of modification and changesto prevent new hazards, and general plant security.

Some of the common mitigation techniques employed by process plants areas follows:

• Early vapor detection and warning: Detection by sensors or personnel. De-pending on the nature and extent of the chemical hazards, some plants maychoose to employ very sophisticated sensor systems. For example, a pipelinecompany handling sour gas mixtures with very high H2S content decided toinstall an early warning H2S-sensing system. The system known, in the industryas Teledyne Geotech’s ‘‘LASP’’ [27,28] (Leak Alarm System for Pollutants)consisted of a semipermeable tubing which is laid above the sour gas pipelineunder the ground. The tubing is capable of drawing air through it, which isanalyzed for H2S contamination at regular intervals. Another common techniqueutilized to protect high-hazard pipelines is the installation of labeled warningribbons approximately 1 ft below grade over the pipeline.

• Use of engineered and management systems to impede the progress of the re-leased chemicals. Some of the engineered systems, which have been very effec-tive in mitigation, include water sprays, water curtains, steam curtains, and aircurtains. Management systems may include standing procedures to deliberatelyignite explosive clouds, procedures for forced dilution of contaminant, and pro-cedures to mitigate or suppress released chemicals by the use of foams andother suppressants.


Isolation by distance of a chemical process from on-site and off-site sur-rounding populations is generally a very effective consequence–mitigation mea-sure. Separating the process from vulnerable populations affords both attenuationof the effects and time to provide emergency response. The isolation distancesneeded to appreciably reduce blast effects and impact from toxic releases is sig-nificantly large, whereas this type of mitigation measure is not useful for the protec-tion of on-site personnel. Other hazard mitigation measures should be used forprotection of on-site personnel. For example, explosion hazards may warrant theconstruction of blast-resistant buildings or blast walls. Toxic release hazards mayrequire the availability of shelter-in-place facilities or escape respirators. Someprocess plants use consequence modeling in deciding the layout of the plant. Forexample, a hazards analysis early in the design stage may identify one particularunit as having the greatest potential for a toxic release and that unit may then belocated on the site as far as possible from off-site neighbors, perhaps consideringprevailing directions as well.

Process integrity may also be addressed in the engineering design. Processintegrity involves the chemistry of plant design and operation. Mitigation after lossof containment can also be effective and usually must be provided for in the processdesign stage. Secondary containment by double-walled piping or double-walledvessels may be needed. Dikes, curbs, and trenches leading away from storage ves-sels to strategically located impoundments can be used to reduce the rate of evapo-ration, help keep the liquid source of the vapor away from the most sensitive areasof the plant, and limit the extent of emergency response activities.

Mitigation measures such as active and passive scrubbers, stacks, flares, catchtanks for vapor–liquid separation, incinerators, absorbers, adsorbers, and condens-ers are used widely for reducing the impact after loss of containment. CCPS’sGuidelines for Vapor Release Mitigation [29] provides detailed discussions onthese mitigation techniques.

Response

Emergency response plans for process plants should be developed in accordancewith applicable governmental regulations and operating company requirements.Written emergency response procedures, accident investigation protocol and proce-dures, and repair procedures should be prepared, and the appropriate operatingpersonnel should be trained in their proper use. Results or risk assessment studiesand hazard zone calculations should be used in the formulation of emergency re-sponse actions contained in the emergency plans.

In addition to being quickly and effectively warned of a dangerous situation,personnel need to know ahead of time the best response to minimize their chancesof being affected by the vapors. Questions that should be settled in advance includeemergency shutdown criteria, incident commander designation, and the roles ofincident commander and other emergency personnel. Other issues that should besettled in advance include circumstances which would warrant shelter-in-place forthe employees and the affected public.


Communicating warnings to potentially exposed personnel is essential in anemergency. Warnings can be communicated through the use of public addresssystems, alarm systems, and sirens. Recently, many companies are also movingtoward the installation of automatic telephone dialing and alerting systems or com-municating hazard warnings to downwind personnel. Preprogrammed computerscan be used to dial thousands of preselected numbers in very short periods of time.

Every plant should establish clearly defined procedures for emergency shut-down of equipment. The procedures should clearly indicate what constitutes anemergency situation. The extent and nature of steps taken to bring the processback under control should also be clearly indicated. Finally, the conditions whichwould unambiguously require a shutdown should be spelled out. Personnel author-ized to make the decision to shutdown should be aware of their responsibilitiesand the role of other personnel.

Technology Advances

Advances in the understanding of chemical hazards have led to the developmentof new technology in the arena of process design, equipment, and risk management.A discussion of three major areas of development is given here.

Relief Valve Sizing and Overpressure Alternatives

Recent research on relief valve sizing and overpressure protection alternatives hasfocused on the development of validated engineering design procedures for theproper sizing of safety relief valves for systems, which involve two-phase flowsof viscous fluids. Systems which are being considered include single-phase viscousliquids and gas flows, ‘‘frozen’’ (e.g., air–liquid) two-phase flows of gases andviscous liquids, and flashing flows of viscous and nonviscous liquids.

Reactive Chemistry

In the Reactive Chemistry arena, calorimeters are being used increasingly forstudying the thermal behavior of reactive systems. One such calorimeter is theReactive Systems Screening Tool (RSST), which is designed for rapid measure-ment of thermal behavior of small samples (10 cm3) for temperatures up to 400°Cand pressures to 500 psia. Another apparatus is the Automatic Pressure TrackingAdiabatic Calorimeter (APTAC) for detailed analyses of thermal behavior of largersamples (up to �130 cm3) for temperatures up to 450°C and pressures up to 2000psia. In this calorimeter, closed-cell sample pressures are continuously matchedby an external pressure of nitrogen so that sample cells of low mass and thereforelow thermal inertia can be used for highly sensitive measurements of sample ther-mal behavior. Other advanced features of the APTAC include in situ additions tothe sample cell of reactants or catalysts with a high-pressure syringe pump.


During typical experiments with each calorimeter, the sample temperature ismeasured during temperature scans when the temperature is increased at a steadyrate or held isothermally (in the APTAC). Thermal energy released or absorbedby the sample (as defined by sample temperature changes of �0.04°C/min) ismeasured by the calorimeter as a nearly adiabatic excursion from the thermal scanbaseline. For each sample, the thermal peaks can be identified and measured fromambient up to 450°C (up to 400°C in the RSST). With this experimental capability,investigations of thermal behavior of wide ranges of reactive systems and systemsof questionable chemical compatibility can be performed, which, in turn, is usedto design safe processes and choose safe operating conditions.

Safety Integrity Levels

Industry is moving toward the use of high-integrity protection systems to reduceflare loading and alleviate the need to upgrade existing flare systems when expandingfacilities. In the process industry, a key safety consideration is the control and re-sponse to overpressure situations. Industry standards from the American PetroleumInstitute (API) and American Society of Mechanical Engineers (ASME) providecriteria for the design of vessels and the protection of these vessels from overpres-sure. Traditionally, pressure relief valves and flares were used to handle the relievingof vessels in the worst credible scenario. Flare loading calculations gave no creditfor operator intervention, fail-safe equipment operation, or trip systems.

In many communities and countries around the world, the belt is tighteningon the venting and combustion of gases. It is simply not acceptable to flare largevolumes of gas. In addition, the cost of designing and installing large flare systemshas continued to rise. API 521 and Case 2211 of ASME Section VIII, Division 1and 2, provide alternatives in the design of overpressure protection systems. Thesealternatives revolve around the use of an instrumented system that exceeds theprotection provided by a pressure relief valve and flare system.

ASME Code Case 2211, approved in 1996, sets the conditions under whichoverpressure protection may be provided by an instrumented system instead of apressure relief valve (PRV). The ruling is intended to enhance the overall safetyand environmental performance of a facility by utilizing the most appropriate engi-neered option for pressure protection. Although there are no specific performancecriteria in the Case Code, the substitution of the high-integrity protection systemsfor the pressure relief valve should provide a safer installation. Consequently, thesubstitution is generally intended for limited services where the PRV may notwork properly due to process condition (e.g., plugging, multiple phases, etc.). Theoverpressure protection can be provided by a safety instrumented system in lieuof a pressure-relieving device under the following conditions:

One of the most important criteria for safety instrumented system (SIS) designis the requirement that the User assign and verify the safety integrity level (SIL)for the SIS [30]. The assignment of SIL is a corporate decision based on riskmanagement philosophy and risk tolerance. The SIS should be designed to meeta safety integrity level, which is appropriate for the degree of hazard associatedwith the process upset. Safety integrity levels per draft IEC 61508 [31,32] andANSI/ISA S84.01 [33–37] are designated in Table 7.


TABLE 7 Safety Integrity Levels

Availability Probability to failSafety integrity level required on demand 1/PFD

IEC 61508:4 �99.99% 10�5 to 10�4 100,000–10,000

ISA S84:3 99.90–99.99% 10�4 to 10�3 10,000–1,0002 99.00–99.90% 10�3 to 10�2 1,000–1001 90.00–99.00% 10�2 to 10�1 100–10

Industrial Hygiene and Toxicology

As the world becomes more industrialized in an attempt to increase the quality oflife, more and more environmental problems will require the attention of engineers,managers, and planners. In the coming years, findings and advances on chemicaltoxicity will require the implementation of stringent industrial hygiene standards.Although there are many toxic effects, both acute and long term associated withchemicals, one of the most dreaded is cancer. It is quite apparent that industrialhygiene requirements in the coming years will be dictated to a large extent by theupcoming findings on carcinogenesis and mutagenesis. By one estimate, there are500 new chemicals marketed each year [38]. Thus, determining the toxicologicaleffects of these chemicals and, as a result, developing industrial hygiene programsto protect people from these effects will command significant attention and re-sources from process safety personnel.

Although workers are often exposed to contaminant mixtures, exposure regula-tions do not take into account the effects of the various types of contaminant inter-actions capable of modifying toxicity. To remedy this situation, the scientific re-search and development of databases on the toxicity of mixtures are needed. Withthis information, health and safety specialists will be able to quantify contaminantinteractions for any given situation.

Although substantial progress has been made in the United States toward im-proving worker protections since 1970 (largely a result of occupational safety andhealth research), workplace hazards continue to inflict a tremendous toll in termsof human and economic costs. Clearly, there is much work to be done.

The practical impact of the toxicological and industrial hygiene research pro-grams on the workplace largely depends on the actions of employers, employees,and partners in governmental agencies, industry, labor, academia, and communityorganizations. The stated objectives in this area include target levels of improve-ments in work-related conditions. Examples are reducing work-related deaths andinjuries, reducing lost work days and incidences of cumulative trauma and skindisorders, and increasing the number of workplaces with rehabilitation and safetyand health programs. The work of governmental agencies and research organiza-tions has had and will continue to have an impact on improving health and safetyat the workplace; therefore, it will help address many of the issues related to work-related hazards, injuries, illnesses, and deaths (such as musculoskeletal problems,


skin diseases, violence in the workplace, employee stress, and back injuries) aswell as categories of workers and prevention strategies for mine workers, farmworkers, and adolescents. In addition, surveillance efforts will assist the develop-ment of comprehensive databases, thereby helping to establish baseline and trendinformation in the occupational safety and health area.

Future Developments

Increasingly, the process safety requirements for chemical plants will become moreand more stringent. In addition, the pressure to operate safely from the point ofview of competitiveness and profitability will also keep increasing. Finally, thepublic outcry for improved safety performance also creates significant pressure onthe industry. In fact, in future processes, safety performance will quite likely bedictated by national goal setting. This would require the establishment of a baselineassessment of the status of process safety incidents. Given a baseline assessment,National Chemical Safety Goals can be established, with the identification of activ-ities necessary to accomplish the goals and the development of a measurementsystem to measure progress toward the goals.

Regulatory programs and industrial standards and practices in the United Stateshave quite often been reactive (i.e., in response to catastrophic accidents or otherevents). The pros and cons of establishing national process safety goals and evalua-tion approaches include the following:

1. Stakeholder consensus on national chemical (process) safety goals2. Identification of where we want to be and by when in relation to national

chemical safety goals3. List of activities that need to be implemented to accomplish Step 2 above4. Agreement on some common metrics for measurement of progress toward

national chemical safety goals

Summary and Conclusions

The industrial revolution brought prosperity and, along with it, the use of hazardousprocesses and complex technologies. Growing economies and global competitionhas led to more complex processes involving the use of hazardous chemicals, ex-otic chemistry, and extreme operating conditions. As a result, a fundamental under-standing of the hazards and associated risks is essential. Process safety and riskmanagement requires the application of the basic sciences and a systematic ap-proach. Recent advances, such as overpressure protection alternatives and reactivechemistry, allow safer design and operation of processes.

In the multiple-barriers concept, plants are designed with several layers, so thatan accident would require the failure of several systems. Another novel approach to


process safety and risk management is to consider various actions in a descendinghierarchical order. Inherently safer design consideration should be first in the hier-archy, followed by prevention systems, mitigation, and response. The success ofthese systems is dependent on the fundamental understanding of the process andthe associated hazards. Chronic as well as catastrophic consequences resultingfrom toxic and flammable substances can be reduced and/or eliminated throughappropriate design and operating practices.

In the end, progress toward the improvement in safety performance can bemeasured only by a reduction in occupational injuries, illnesses, and fatalities. Infact, measurable progress has been made in the period 1970 to 1995, during whichthe rate of workplace fatalities fell by 78% and the number of workplace deathshas declined by 62%. We have also seen a 25% decline in the rate of occupationalinjuries and illnesses from 1973 through 1994. These reductions are the result ofthe combined efforts of all the partners in occupational safety and health: industry,labor, academic researchers, National Institute of Occupational Safety and Health,Occupational Safety and Health Administration, Mining Safety and Health Admin-istration, state and local agencies, and others. No single partners can claim exclu-sive credit for the progress. Thus, if further progress is to be made, all of thepartners must act—from identifying the causes of disease and injury through con-trolling or eliminating the hazards or exposures at the worksite.

References

1. M. Connors, ‘‘The Battle for Industrial Safety,’’ Fortune, 116[C-P] (August 4, 1997).2. D. A. Crowl and J. F. Louvar, Chemical Process Safety: Fundamentals with Applica-

tions, Prentice-Hall, Englewood Cliffs, NJ, 1990.3. Center for Chemical Process Safety, Guidelines for Safe Automation of Chemical Pro-

cesses, American Institute of Chemical Engineers, New York, 1993.4. American Petroleum Institute, Recommended Practice 750: Management of Process

Hazards, API, Washington, DC, 1990.5. ‘‘Notice of Proposed Rule Making on Process Safety Management of Highly Hazard-

ous Chemicals’’: 29 CFR 1910.119, Federal Register, Washington, DC, July 17, 1990.6. Chemical Manufacturers Association, Resource Guide for Implementing the Process

Safety Management Code of Practices, Chemical Manufacturers Association, Wash-ington, DC, 1990.

7. ‘‘Final Rule on Process Safety Management of Highly Hazardous Chemicals’’: 29CFR 1910.119, Federal Register, Washington, DC, February 24, 1992.

8. Norwegian Petroleum Directorate, ‘‘Safety Evaluation of Platform Conceptual De-sign,’’ Stavanger, Norway, 1981.

9. European Community Directive, ‘‘On the Major Accident Hazards of Certain Indus-trial Activities,’’ 82/501/E, J. Eur. Community, L230 (June 1982).

10. Offshore Installation (Safety Case) Regulation 1992, Health and Safety Executive,London, UK, 1992.

11. ‘‘Techniques for Assessing Industrial Hazards,’’ World Bank Technical Paper #55,Washington, DC, 1988.

12. Major Hazard Control, a Practical Manual, International Labour Office, Geneva,1988.


13. Center for Chemical Process Safety, Guidelines for Chemical Process QuantitativeRisk Analysis, American Institute of Chemical Engineers, New York, 1989.

14. R. J. Lewis (ed.), Sax’s Dangerous Properties of Industrial Materials, 9th ed., JohnWiley & Sons, New York, 1996.

15. P. G. Urben (ed.), Bretherick’s Handbook of Reactive Chemical Hazards, 5th ed.,Butterworth-Heinemann Boston, 1995.

16. Dow Chemical Company, Dow’s Fire and Explosion Index Hazard ClassificationGuide, 7th ed., American Institute of Chemical Engineers, New York, 1994.

17. Dow Chemical Company, Dow’s Chemical Exposure Index Guide, American Instituteof Chemical Engineers, New York, 1994.

18. Center for Chemical Process Safety (CCPS), Guidelines for Hazard Evaluation Proce-dures, 2nd ed., with Worked Examples, American Institute of Chemical Engineers,New York, 1992.

19. G. Wells, Hazard Identification and Risk Assessment, Institution of Chemical Engi-neers, Rugby, Warwickshire, UK, 1996.

20. Center for Chemical Process Safety (CCPS), Guidelines for Use of Vapor CloudDispersion Models, 2nd ed., American Institute of Chemical Engineers, New York,1996.

21. Center for Chemical Process Safety (CCPS), Guidelines for Evaluating the Character-istics of Vapor Cloud Explosions, Flash Fires, and BLEVES, American Institute ofChemical Engineers, New York, 1994.

22. Center for Chemical Process Safety (CCPS), Guidelines for Consequence Analysis ofChemical Releases, American Institute of Chemical Engineers, New York, 1999.

23. G. E. DeVaull, J. A. King, R. J. Lantzy, and D. J. Fontaine, Understanding Atmo-spheric Dispersion of Accidental Releases, American Institute of Chemical Engineers,New York, 1995.

24. The Netherlands Organization for Applied Scientific Research (TNO), Methods forthe Calculation of Physical Effects, Part 1 and 2 CPR-14, 3rd ed., SdU Uitgevers,The Hague, 1997.

25. F. P. Lees, Loss Prevention in the Process Industries, 2nd ed., Butterworth-Heine-mann, Boston, 1996.

26. H. G. Fisher, H. S. Forrest, S. S. Grossel, J. E. Huff, A. R. Muller, J. A. Noronha, D. A.Shaw, and B. J. Tilley, Emergency Relief System Design Using DIERS Technology,American Institute of Chemical Engineers, New York, 1992.

27. M. Mannan, D. B. Pfenning, and C. D. Zinn, ‘‘Sour Gas Pipeline—1: Risk-AnalysisProcedures Ensure System Safety,’’ Oil Gas J. 83–87 (June 3, 1991).

28. M. Mannan, D. B. Pfenning, and C. D. Zinn, ‘‘Sour Gas Pipeline—Conclusion: Line,Weather Conditions Among Variables to Determine Public Risk,’’ Oil Gas J., 34–35 (June 10, 1991).

29. Center for Chemical Process Safety (CCPS), Guidelines for Vapor Release Mitigation,American Institute of Chemical Engineers, New York, 1988.

30. A. E. Summers, ‘‘Techniques for Assigning a Target Safety Integrity Level,’’ ISATrans., 37, 95–104 (1998).

31. IEC 61508, 65A/255/CDV, ‘‘Functional Safety of Electrical/Electronic/Programma-ble Electronic Safety Related Systems, Parts 1, 3, 4, and 5,’’ International Electrotech-nical Commission, Final Standard, December 1998.

32. IEC 61508, 65A/255/CDV, ‘‘Functional Safety of Electrical/Electronic/Programma-ble Electronic Safety Related Systems, Parts 2, 6, and 7,’’ International Electrotechni-cal Commission, Final Draft International Standard, January 1999.

33. ‘‘Safety Instrumented Systems (SIS)—Safety Integrity Level (SIL) Evaluation Tech-niques, Part 1: Introduction,’’ TR84.0.02, Draft, Version 4, March 1998.

34. ‘‘Safety Instrumented Systems (SIS)—Safety Integrity Level (SIL) Evaluation Tech-


niques, Part 2: Determining the SIL of a SIS via Simplified Equations,’’ TR84.0.02,Draft, Version 4, March 1998.

35. ‘‘Safety Instrumented Systems (SIS)—Safety Integrity Level (SIL) Evaluation Tech-niques, Part 3: Determining the SIL of a SIS via Fault Tree Analysis,’’ TR84.0.02,Draft, Version 3, March 1998.

36. ‘‘Safety Instrumented Systems (SIS)—Safety Integrity Level (SIL) Evaluation Tech-niques, Part 4: Determining the SIL of a SIS via Markov Analysis,’’ TR84.0.02, Draft,Version 4, March 1998.

37. ‘‘Safety Instrumented Systems (SIS)—Safety Integrity Level (SIL) Evaluation Tech-niques, Part 5: Determining the PFD of SIS Logic Solvers via Markov Analysis,’’TR84.0.02, Draft, Version 4, April 1998.

38. T. F. Yen, Environmental Chemistry: Essentials of Chemistry for Engineering Prac-tice, Prentice-Hall, Englewood Cliffs, NJ, 1999.

M. SAM MANNANDENNIS HENDERSHOT

TREVOR A. KLETZ

Introduction to the Selective CatalyticReduction Technology

Introduction

Overview

The selective catalytic reduction (SCR) process has been originally developed forreducing oxides of nitrogen (NOx). The process is being discussed within the frame-work of air-pollution control policies and practices of industrialized countries, ac-counting for about 20% of the world population and generating 80% of the globalcombustion air-pollution. The United States with approximately 5% of the worldpopulation accounts for about 28% of the world’s fossil fuel consumption (Interna-tional Energy Agency, Paris, France), more than twice the per-capita consumptionof other major industrial countries. Stringent Energy Conservation and EmissionReductions Policies have a much longer tradition in Europe and Japan than in theUnited States. In the United States, due to low-cost energy, excess fuel is oftenused to reduce NOx of IC engines, which accounts for over 50% of the UnitedStates’ total NOx inventory, by a 4-degree time-retard timing. This at the expenseof major increases in CO emissions, fuel consumption, and carcinogenic particulatematter (PM) emission. (Source: BACT best available control technology of U.S.EPA, Cal.-ARB, and SCAQMD).

The SCR process has been designed to reduce oxides of nitrogen (NOx), aprecursor of regulated low-level ozone (O3) emission. The ozone is formed primar-

Selective Catalytic Reduction 95

ily by the two precursor NOx and volatile organic compounds (VOC) in the pres-ence of sunlight through photosynthesis. However, advanced SCR developmentshave been achieving VOC/HC (hydrocarbons) and particulate matter (PM) emis-sion reductions simultaneously with the NOx reduction. Also, in recent years, Euro-pean application engineering technology developments have broadened the appli-cation of SCR technology. Originally developed in the 1970s for stationary utilitypower generation equipment, the SCR process can be used to reduce emissions ofpractically all types of combustion equipment, including mobile applications suchas heavy-duty diesel (HDD) trucks, portable generation sets, diesel locomotives,coastal and ocean-going vessels, earth-moving equipment, and so forth. Today,SCR technology is being used in retrofit and integrated emission control systems.The SCR combustion pollution control technology is able to overcome the targetconflict of IC combustion equipment, increasing the NOx emission by reducingthe fuel consumption or increasing the PM and CO emission by increasing fuelconsumption through fuel-based emission controls (see Fig. 1).

The earliest SCR applications were located primarily in oil-burning utilityboiler facilities in Japan in the 1970s. Then, in the 1980s, European SCR applica-tions for hard-coal-burning utility and oil- and gas-burning industrial boilers, andstationary gas, diesel, and dual-fuel IC engines were added. In the United States,gas turbine applications were introduced in the 1980s and many coal-burning utilitypower plants require SCR retrofit system installations by 2004. Figure 2 showshigh-dust and low-dust utility boiler applications.

However, because the number of utility plants are relatively few in comparisonto the number of other nonutility SCR applications to be considered for OEM andretrofit applications in future, accounting for over half of the total U.S. NOx pollu-tion inventory, the nonutility and the new mobile combustion sources are beingemphasized in this article.

The SCR utility applications have to compete with pollution control processes,

FIG. 1 The target conflict of IC equipment emission reduction versus fuel consumption. (Source:German Engineering Society (VDI), May 1993.)

96 Selective Catalytic Reduction

FIG. 2 Utility boiler applications. (Courtesy of Siemens.)

such as the selective noncatalytic reduction process (SNCR), featuring low capitalinvestment but high operating costs. This process is limited to a small operatingtemperature window of usually 1350–1550 F (750–900°C) and requires an ammo-nia or urea consumption of up to four times the stoichiometric requirements of theSCR process.

In the United States, natural-gas- and fuel-oil-burning industrial boilers and ICengines and turbines were equipped with SCR systems in a few regional nonattain-ment areas for ozone since the 1980s and gas- and coal-burning utility boilers sincethe 1990s.

In Europe, due to the development of a diesel SCR catalyst with an operatingtemperature window of 300–1020 F (150–550° C), air quality regulators cooper-ated with the trucking industry and the SCR equipment industry in the developmentand field testing of SCR systems for heavy-duty diesel truck engine applicationsin the 1990s. These truck field tests have lasted for some 4 years, accumulatingapproximately 3 million road and highway miles with 20 Class 8 type HDD trucks.The SCR systems will go to market in 2002. Today, the range of SCR applicationsis rather broad, as shown by Fig. 3: From (clockwise) a coal-burning utility plantto HDD truck, portable generator set, vessel, railroad, construction equipment,pipeline pump station, and gas turbine applications.

Today, the SCR technology allows the simultaneous reduction of combustionemission of 70–95% NOx, 20–50% PM, and 85–95% VOC/HC at minimum fuelconsumption and CO2 emission. It may be interesting to note that the emissionreduction of VOC include air toxins such as aldehydes and polycyclic aromatichydrocarbons (PAH), which are almost completely oxidized in the SCR process.Hence, SCR technology is considered the most promising, cost-effective technol-ogy available to reduce combustion emissions, enabling air quality regulatorsworldwide to achieve more drastic emission reductions of combustion equipmentthan thought possible just a few years ago.

Various cost-effectiveness calculations showed annualized costs of �$500 to�1500 per annual ton of NOx reduction, depending on the type and size of theSCR equipment. The capital investment cost of SCR systems are �US$ 10 to 50per BHP h, depending on custom engineering requirements and the number ofsystems of the same design fabricated (Siemens).


FIG. 3 The Range of SCR applications. (Courtesy of Siemens.)

Environmental Policies

After the United Nations summits on the environment in the 1990s in Rio, Braziland Kyoto, Japan, environmental policy-makers around the world have been focus-ing on the reduction of emission generated by combusting fossil fuel.

• Health effects of NOx, PM, and VOC/HC have been the driving force for airquality legislation in the past. The air quality regulations, however, were devel-oped in a piecemeal approach: first NOx, then VOC/HC to reduce ground-levelozone, and, finally, PM. This has been a major disadvantage for SCR technologyapplications, as no considerations were given to the capability of reducing VOCand PM simultaneously with NOx at no extra cost.

• Global warming is becoming an increasing concern domestically and interna-tionally, considering changes in weather pattern, shorelines of the ocean, andso forth. The industrialized countries must reduce their fossil-fuel combustionemission substantially to allow developing nations to generate more power fortheir growing economies. The various emission inventory data for the UnitedStates for NOx indicate that the leading sources of combustion emission are fromon-road and off/nonroad emission sources. Gasoline engines, at the expense ofhigher fuel consumption and N2O (laughing gas) secondary emission, have beenequipped with three-way catalyst systems for many years. For the most fuel-efficient device in converting fossil fuel into useful energy and power, the dieselengine, no cost-effective, commercially available solution to reduce NOx, PM,and VOC/HC emissions substantially and simultaneously existed until the early1990s. Therefore, as mentioned above, fuel is often used today to reduce emis-sions, increasing the fuel consumption of engines and other combustion equip-ment and, with it, global warming gas emissions. Increases of 10–13% in CO2


FIG. 4 Target conflict: reduction of NOx versus increase in fuel consumption, CO2, and PM. (Courtesyof Siemens.)

and up to 70% of PM emission of diesel engines were reported at public Cal.ARB meetings in 1999 and 2000.

• Combustion emission reductions through fuel savings and internal combustionequipment design modifications such as electronic engine management systems,unit electronic–hydraulic injection, intercooling, and exhaust gas recycling islimited. In Fig. 4, the target conflict of the U.S. EPA NOx emission reductionsversus exponential increases in fuel consumption, CO, VOC/HC, and PM emis-sion of Class 8 HDD truck engines, is shown by reducing NOx emissions ofengines by modifications such as U.S. EPA’s BACT mentioned earlier. There-fore, to avoid higher fuel cost and to achieve future U.S. EPA emission reduc-tion goals beyond 2002 and 2004, exhaust gas aftertreatment technologies willbe required in mobile and other applications.

Environmental Regulations

In 1970, the U.S. Ambient Air Quality Act was passed and then amended in 1990,allowing a maximum ground-level ozone concentration of 0.12 parts per million(ppm). The U.S. National Ambient Air Quality Standard (NAAQS), however, wasnever met throughout the United States. Serious nonattainment areas for ozone(O3) such as in the Northeast and in California still exist and new ones were addedrecently. In 1997, the U.S. EPA further reduced the ground-level ozone standarddue to late health data and increased health concerns to 0.08 ppm ozone at an 8-haverage and introduced a new PM 2.5 standard of 65 µm/m3 at a 24-h average.That increased the nonattainment areas in the United States substantially (Fig. 5).The various state implementation plans for combustion emissions in the UnitedStates, therefore, call for substantial further NOx reductions of the U.S. inventoryof approximately 25 million tons of NOx per year (U.S. EPA).


FIG. 5 Nonattainment areas in the United States for ozone (dark areas) based on the EPA’s newstandards for ground-level ozone and PM 2.5.

In addition to major cities in California and the Northeast, cities such as At-lanta, GA, Chicago, IL, Dallas, TX and Houston, TX became serious or extremenonattainment areas. East coast states, after years of complaining about Mid-westcombustion pollution transfers to the East, finally received an ozone transfer regu-lation for about 20 states, requiring midwest states to reduce NOx substantially.According to NESCAUM data presented at the ICAC Forum 1998, the NOx emis-sion in the midwest was caused by utilities (�33%), by independent power produc-ers (14%), and by mobile equipment (on-road 34% and off/nonroad 14%). Otherestimates (i.e., California) are more like 30% each for stationary, on-road, andnonroad combustion equipment after 20 years of stationary emission source regula-tions in that State. Figure 6 shows the different emission source categories by areain Texas in 1996.

Technology

As stated earlier, SCR technology has been advanced considerably in Europe inrecent years; however, the basic SCR process is still the same. The SCR processequipment layout is pictured in Fig. 7, using an engine as an example for combus-tion equipment.


FIG. 6 NOx emission by source category in tons per year at selected areas of Texas in 1996. Example:Houston, clockwise: 2% area, 20% nonroad, mobile, 25% on-road, and 53% stationary emis-sion sources. (Source: TNRCC, 2000.)

FIG. 7 The SCR process, using a diesel engine SCR application as an example. (Courtesy of Siemens.)


The exhaust gas generated by a diesel engine passes from the engine throughthe exhaust pipe or duct, through the SCR reactor and other exhaust gas systemcomponents such as the muffler into the atmosphere. The exhaust gas duct, leadingto the reactor, houses the gas flow straightener, turning vanes, static mixers, andsome sensor probes. The SCR reactor houses the SCR catalyst and other sensorand instrumentation probes. Depending on the exhaust gas temperature, mass flow,raw or uncontrolled, and permitted or controlled NOx concentration and other vari-ables, a certain amount of a reducing agent is injected and homogenously mixedupstream of the catalyst bed.

There are several reducing agents in use today. Anhydrous ammonia (NH3),a toxic, hazardous, and flammable gas, is used primarily in stationary industrialand utility applications. A safer ammonia reducing agent, aqueous ammonia, con-taining 25–29% ammonia gas in demineralized water, was introduced to SCR ap-plications in the 1980s. The mobile SCR system application developments of the1990s, however, required a much safer reducing agent, aqueous urea, from whichammonia is generated through hydrolysis, yielding two NH3 and one CO2 in theexhaust duct upstream of the SCR catalyst. When subsequently passing throughthe SCR catalyst bed, the reducing agent NH3 reacts with the NOx to yield molecu-lar nitrogen (N2) and water vapor (H2O) with very limited ammonia slip, typicallyin the 2–30-ppm range as secondary emission. Some typical chemical reactionsof the SCR process are shown by Fig. 8a. Figure 8b shows the reversible sidereactions and their equilibrium at threshold temperatures (EESI/Steuler, a leadingSCR system supplier) at a specific chemical exhaust gas composition.

Secondary Reactions and Emissions

The effects of high SOx levels in the exhaust gas is given in Fig. 8b.A European university study on the effects of high SOx levels in the exhaust

gas duct concluded that temperature and SO2 � SO3 control is key to avoid ammo-nia salt emissions. Figure 8b shows a few of the reversible side reactions andtheir equilibrium threshold temperatures (EESI/Steuler), using an exhaust gas ofa specific chemical composition: 72% N2, 15% CO2, 10% H2O, 3% excess O2, 200ppm NO, 10 ppm NO2, 1000 ppm SO2, 50 ppm SO3, and 200 ppm HCL.

Ammonia slip has been a concern of health experts and air quality regulatorswhen considering SCR exhaust emission reductions. United States air quality regu-lators for gas turbine SCR applications have imposed extremely costly low levelsof NH3 slip emission limits of 2–5 ppm. However, pollution control experts pointout that this ‘‘NH3-slip hype’’ is not based on facts: European and U.S. NH3 Emis-sion Inventory Studies in the Netherlands, Germany, and California show that,depending on the life stock concentration, 60–75% of the total local NH3 emissionis caused by animal droppings. Also, depending on weather conditions and thetype and intensity of fertilizer usage, the area NH3 emission may be as high as 5–30%. The industrial applications such as air conditioning, nitric acid production,and SCR-based NH3 slip emission do not account for more than 0.5–2% of thetotal NH3 emission inventory (Bavarian-EPA, Cal. ARB, and other studies). In aU.S. study, a well-supported estimate of a SCR-based NH3 slip rate of 2–5 ppmfor 200,000 MW utility applications would just cause an NH3 emission increase


Reducing Agent Urea: Ammonia generation through hydrolysis in the exhaust or flue gas, upsteamof the SCR catalyst.

1. (NH2)2 CO � H2O → 2 NH3 � CO2

NOx Reduction with Ammonia

2. 4NO � 4NH3 � O2 → 4N2 � 6H2O3. 6NO � 4NH3 → 5N2 � 6H2O4. 2NO2 � 4NH3 � O2 → 3N2 � 6H2O5. 6NO2 � 4NH3 → 7N2 � 12H2O

(a)

Effects of high SOx levels in the exhaust gas

Temperature (°C) Chemical Reactions

1. 22 2NH3 � H2O � SO2 i (NH4)2SO3

2. 219 2NH3 � H2O � SO3 i (NH4)2SO4

3. 209 NH3 � H2O � SO3 i NH4HSO4

4. 113 NH3 � HCl i NH4Cl5. 88 NH3 � H2O � (1/2)O2 � 2NO2 i 2NH4NO3

(b)

(c)

FIG. 8 (a) Typical chemical reactions of the SCR process. (b) Equilibrium threshold temperaturesat which ammonia salts is formed/dissolved. (c) NH3/NOx ratio and the effects of overinjectionand underinjection of ammonia. (Courtesy of L. Pruce.)

of less than 10,000 tons per year equal to less than 1% of the total U.S. NH3

emission inventory.

System Controls

The component selection will vary depending on the type of combustion equip-ment, emission reduction, control/monitoring system requirements, and so forth.


Basically, two types of reducing agent injection control are being used. The firstis the open-loop, feed-forward control with a computer-based ‘‘map,’’ a correlationfunction of NOx emission generated at certain combustion equipment loads, re-quiring a calculated amount of reducing agent to achieve the permitted NOx emis-sion rate. Such a system is also called a Predictive Emission Monitoring System(PEMS). Maximum reduction rates at minimum unreacted ammonia (slip) can beachieved by adding to the rapidly responding open-loop system a slower closed-loop system. The feedback systems are, in most cases, certified systems, incorpo-rating continuous emission monitoring systems (CEMS) or sensors.

SCR System Design Considerations

The following basic design considerations are common to all SCR systems:

• Proper exhaust gas flow and temperature distribution at the front face of thecatalyst bed

• Proper selection of the catalyst material formulation, the catalyst bed, and hous-ing design

• Proper selection of materials, system components, and control system design

The SCR system component selection and system configuration may vary con-siderably depending on the application. Utility applications for large boilers or gasturbines are therefore custom engineered. Also, Independent Power Producer (IPP)and industrial applications such as cogeneration and standby generation sets arestill primarily custom engineered. However standardized, volume-produced smallgas turbine and engine SCR applications will be more cost-effective and thus morecommon in future. The new mobile SCR applications for diesel and lean-burngas engines will be such standardized and volume-produced SCR systems. Thisindustrial approach, developed and adopted for the first time in the HDD truckprogram by Siemens and the European trucking industry, is already benefiting thecustom-engineered SCR systems as well.

SCR System Component Summary

The Exhaust or Flue Gas Piping or Duct

This is usually heat insulated, made of heat-resistant nonscaling steel such as mo-lybdenum alloys, and connects the combustion equipment with the SCR reactor.Such inlet ducts with the SCR reactors downstream are pictured in Fig. 9.

The inlet duct or piping houses the gas flow straightener, flow turning vanes,sensor probes, and injection spray heads or nozzle as well as the static mixers.Only in very special cases, bypass valves are incorporated in ducts to cope withcertain operating conditions such as cold start-ups, alternate fuels, and so forth,which could harm the SCR catalyst of earlier developments. An even flow rate


(a)

(b)

FIG. 9 (a) SCR inlet duct and catalyst location, gas turbine application. (b) SCR inlet duct and catalystlocation, stationary engine application.

and temperature distribution at the front face of the catalyst bed are most crucialand prerequisites of any well-performing SCR system design. As a case in point,in a gas turbine project in southern California, even major improvements in theammonia/exhaust gas mixing system and a major increase in the catalyst volumecould not solve the original problem. The uneven gas flow rate and temperaturedistribution at the front face of the catalyst still deviated by 20–40% due to im-proper turning vanes and lack of static mixers that the contractor had refused toinstall. To obtain a Permit to Operate, the permitted NOx reduction rate was finallynegotiated and increased by the local regulator without hearing expert witnesses.Figure 10 pictures such a flow modeling case and how an even gas flow can beachieved through turning vanes and a static mixer in front of the SCR catalystbed.

The Reactor Housing and Stack Duct

These are normally heat insulated, made of heat resistant nonscaling molybdenumalloy steel rather than stainless steel, which is sensitive to stress corrosion inwelded areas. They house the SCR catalyst and any oxidation catalyst or sounddeadening material as may be required by the specific application. In addition, theyhouse instrumentation probes for monitoring operating conditions and emission


FIG. 10 Flow modeling: before and after computer and scale modeling, improving the uneven (32%)to a more acceptable (9%) flow rate deviation at the front face of the catalyst bed.

reductions. Manholes, loading and unloading equipment, and excess doors for han-dling catalyst modules are required in larger, stationary SCR applications as well.Most custom-engineered SCR systems have space available for adding a row orlayer of catalyst at a later date to reduce the emission even further or to meetcertain catalyst replacement strategies. (See Fig. 11.)

The SCR Catalyst

This is the ‘‘heart’’ of SCR systems and much research and development workwent into catalyst material formulations, structures, and production processes overthe last 20 years. Initially, catalyst ‘‘fouling’’ was a major problem, with the fol-lowing concerns:

• Masking, washable flue gas deposits reduce the reactive catalyst surface tempo-rarily

• Poisoning, an irreversible degradation of the catalyst surface• Plugging, dust clogging of the catalyst, causing an increase in back pressure

and/or reducing catalyst reactivity/ performance.

The operating temperature window was limited to 570–850°F (300–450°C)to avoid ammonia sulfate formations at low reactivity and ammonia combustionproblems at high temperature. These problems gave the SCR technology in theearly developments prior to the 1990s a bad name. Since then major R&D and


FIG. 11 Reactor housing with SCR catalyst in shelf system. (Courtesy of EESI/Steuler.)

application engineering work was carried out in Europe. Also, a pollution con-trol industry, trucking industry and air quality regulator partnership conductedlab/bench and field tests with several technical universities and certified test labsin Europe. Various catalysts from different European and U.S. manufacturers wereevaluated in late 1980s to early 1990s Some test results from the technical univer-sity RWTH-Aachen, Germany of 1989–1991 are shown in Table 1 and Fig. 12.Upon the completion of these tests, the best performing ‘‘Catalyst A,’’ the dieselcatalyst from Siemens was selected by the European trucking industry for the fieldtests in the early 1990s.

Today, several different types of SCR catalyst are commercially available fromseveral suppliers. The diesel SCR catalyst did overcome past SCR problems relatedto high-sulfur fuel, a phosphorus and zinc compound containing lubricating oils,arsenic resistance, and high-dust loads and allows operating temperatures as lowas 300°F (150°C) and as high as 1020°F (550°C).

Like most oxidation catalyst guaranties today, the SCR catalyst service lifewas guaranteed for 1 year only during early developments. Today, the standardprocess or performance guarantee for SCR systems worldwide is 20,000 operatinghours or 3 years, whichever occurs first, in stationary and several hundred thousandkilometers in mobile, on-road applications, maintaining the emission reduction,ammonia slip, and so forth.

Tank and Piping System for the Reducing Agent Anhydrous Ammonia,Aqueous Ammonia, or Aqueous Urea

Anhydrous ammonia, a hazardous, toxic, and flammable gas, requires professionalhandling. The operator of supply piping and anhydrous ammonia storage tanks is


TA

BL

E1

Eur

opea

nH

DD

Tru

ckT

ests

,C

atal

yst

Sele

ctio

n

Cpi

SV(m

3 /hm

3 )A

V(m

3/h

m2 )

A/V

(m2/m

3 )

A.

Hom

ogen

eous

,ex

trud

edV

2O5/T

iO2/W

O3

4551

006.

085

0B

.C

oate

dsu

bstr

ate

(CS-

base

d)V

2O5/T

iO2

100

8200

5.8

1400

C.

CS

base

dV

2O5/T

iO2

100

8200

6.8

1250

D.

CS

base

dV

2O5/T

iO2/W

O3

100

8200

6.8

1250

E.

Hom

ogen

eous

extr

uded

zeol

iteFe

/mol

ecul

arsi

eve

2140

007.

057

0F.

CS

base

dC

u/ze

olite

100

8200

6.8

1250

Not

e:C

pi�

Cha

nnel

spe

rin

ch,

SV�

spac

eve

loci

ty,

AV

�ar

eave

loci

ty,

A/V

�ar

ea/v

olum

e.


(a) (b)

FIG. 12 NOx (a) and VOC/HC (b) reduction rates at different temperatures during 30–50 operatinghours with NH3 slip of �30 ppm.

required to strictly adhere to safety instruction of the manufacturers as well as tolocal, state, and federal regulations.

Aqueous ammonia is safer but requires the same handling and storage precau-tions as for anhydrous ammonia. In Europe, 25% ammonia concentration in demin-eralized water is standard. In the United States, sometimes 27–29% ammoniaconcentrations are common, requiring pressurized tank certifications and routinemaintenance in some states (i.e., New Jersey). Ammonia with a concentration of20%, however, usually does not require such certified tanks.

Engineers with little or no corrosion engineering experience sometimes specifycarbon steel rather than SS316 for piping and lower-grade stainless steel for storagetanks in case aqueous ammonia is used, not realizing that aqueous ammonia ishighly corrosive in the vapor phase. This tank corrosion has been causing extensivedowntime due to rust particles, clogging valves, spray nozzles, and so forth. Again,this gave the SCR technology a bad name in the United States. Examples of SCRsystems with anhydrous and aqueous ammonia are shown in Fig. 13.

Aqueous urea, as salt dissolved in demineralized water, is nonhazardous, non-toxic, and nonflammable and is primarily used as fertilizer and animal feed. Itrequired extensive R&D and application engineering development to make ureaSCR systems work properly and reliably. The urea SCR system engineering devel-opment work in Europe was a prerequisite for applying the SCR technology tomobile, on-road, and off/nonroad applications, now benefiting the new and retrofitstationary combustion applications as well. Data on aqueous urea is presented inFig. 14.

The Ammonia and Urea Supply, Metering, and Injection Systems

These systems are due to the corrosive nature of the reducing agents made ofSS316 steel. To achieve proper mixing of the reducing agent into the flue gas, thereducing agent is heated and diluted with hot air or recycled exhaust gas prior to


(a) Anhydrous Ammonia Tank

(b) Aqueous Ammonia Tank

FIG. 13 (a) Gas turbine application with PEMS and CEMS controls. (b) Gas/diesel (dual fuel) 4.8-MW engine cogeneration plant. (Courtesy of EESI/Steuler.)

being injected into the exhaust gas stream in front of the catalyst bed. Alternately,aqueous ammonia and aqueous urea may be co-injected and atomized with com-pressed air directly into the exhaust pipe. Because the dissolved urea salt maycrystallize again and deposit around the spray nozzles, temperature control andother design considerations have to be met. Also, aqueous urea requires a certainreaction time to convert to ammonia in the exhaust duct, determining the injectionnozzle location upstream of the static mixer.


Aqueous Urea: (NH2)2 CO � H2O, industrial gradeSafety Aspects: Virtually nonodorous, nonhazardous, and nonflammableConsumption: Molecular weight ratio NO2: Urea � 1:0.87

NO: Urea � 1:1.00Concentration: Urea in demineralized water (%) 30 – 40%Crystallization: �11.4°C or �11.5 F at 32.5% concentrationSalting-out temperature: � 0.6°C or �33.0 F at 40% concentrationSpecification:

Typical industrial grade Value DimensionExample, Urea 40 � 0.5 %pH 8.0–9.0Density at 15°C 1105–1125 g/cm3

Viscosity (at 25°C) 1.38 mPa sSpecific heat (at 25°C) 3.26 kJ/kg KElectrical conductivity (1.2–1.3) � 10�4 m S/cmTemperature range without 0–35°C (32–95 F)

temperature controlBiuret �0.5 %Fe �0.1 mg/kgPO4 �0.5 mg/kgMg �0.1 mg/kgCa �0.1 mg/kg

FIG. 14 Aqueous urea data. (Courtesy of Siemens.)

There are basically two types of supply, metering, and injection system designscurrently being used. The first is a constant high-pressure common rail pipingsystem with needle valves or unit electronic hydraulic injectors. The alternate sys-tem incorporates corrosion-resistant, variable speed or variable stroke meteringpumps of different designs.

The reducing agent is metered and injected continuously in accordance withthe signals received from the microprocessor or Process Logic Controller (PLC)-based SCR electronic operating control system. Such system may feature eithera feed-forward PEMS-based injection control, operating on historical computermemory data, or the feed-forward control plus a slower, but fine-tuning feedbackcontrol. The feedback system incorporates either a Continuous Emission Monitor-ing System (CEMS) or sensors. Such custom-designed SCR systems used to oc-cupy an entire row of control cabinets. The mobile SCR system developments re-duced such controls to less than laptop computer size. (See Fig. 15.)

The Integrated, On-Line SCR Operating Controland Emission Monitoring System

This features a Predictive Emission Monitoring System (PEMS), which operateson historical emission data measured at different loads of the combustion equip-ment during (for example, engine bench tests or gas turbine, engine, or boiler SCRsystem) start-ups.

The correlation function of engine-load values versus NOx emission valuesgenerated is combined with an algorithm in the software of the SCR control system,calculating the amount of reducing agent required to reduce the NOx to the specified


(a)

(b)

FIG. 15 (a) SCR control system cabinets with PEMP and CEMS. (Courtesy of EESI/Steuler.)(b) Laptop-size SCR operating and injection control system. (Courtesy of Siemens.)

permitted emission rate. An example of this ‘‘mapping’’ process is shown in Fig.16a–16c.

Figure 16a shows the highest temperature of 900–950 F at 1000 rpm and me-dium torque, a typical truck operating condition. Figure 16b shows that the highestNOx emission is generated at medium speed and torque as well.


(a) (b)

(c) (d)

FIG. 16 (a,b) PEMS mapping: correlation functions of uncontrolled NOx and operating temperaturesof a 12-L HD diesel truck engine rated 400 HP. (c, d) PEMS mapping: correlation functionsof NOx reduction rates and break-specific fuel consumption. (Courtesy of Siemens.)

Figures 16c and 16d show that the highest NOx reduction takes place at mediumspeed and medium torque (Fig. 16c), where the lowest fuel consumption isachieved (Fig. 16d) as well. Engine test standards should therefore emphasize thoseoperating conditions.

The control system monitors all SCR system functions such as tank level con-trol with high, reordering, and low limit, operating parameters such as exhaust gastemperature and pressure drop of the SCR catalyst as well as all required mainte-nance and trouble-shooting management functions. Thereby, the system communi-cates with, for example, the electronic engine management system and the remotecentral control panel or On-Board Diagnostic (OBD) system by CAN bus or via4–20-mA or 0–5 V analog signals.

In many stationary applications, the local air quality regulator requires the useof certified CEMS in accordance with U.S. EPA regulations such as 40 CFR Part60, Appendix B for IC engines. Thereby, the accurate but expensive gas-analyzer-based CEMS are often more capital-intensive than the SCR emission reductionequipment, achieving NOx reductions of up to 90%. Again, this gave the SCRtechnology a bad name.


Catalyst Selection Process

The type of SCR application determines the material and component selection.Not all SCR catalysts and system components perform well in all SCR applications.The SCR catalyst selection begins with a careful analysis of the performance andguarantee requirements. Then, the SCR catalyst meeting the performance charac-teristics best is selected. In some cases, this is not sufficient and the SCR catalystis optimized through changes in catalyst material formulations; for example, reduc-ing the reactivity to avoid exhaust duct corrosion through SO2 � SO3 conversion(SO3 � H2O � H2SO4), or changing the catalyst structure such as the cell densityof honeycomb catalyst from i.e. 200 to 300 cells/in.2, to reduce the weight andthe size of the SCR catalyst for a mobile on-road application where it matters most.In several cases, oxidation catalyst manufacturers in the United States also includedoxidation catalysts upstream of the (SCR) catalyst, causing high corrosion throughup to 80% SO2 � SO3 conversion and reducing the NOx emission reduction byconverting more than 50% of NO � NO2, as shown in Fig. 17a–c.

SCR Catalyst and System Performance Parameters

Depending on the application, different operating conditions and performance re-quirements have to be met. The following is a partial list of parameters that mayhave to be considered while designing a SCR system:

• Fuels and fuel analysis: No. 1, 2, 3, and 6 fuel oil, natural gas, digester orlandfill gas, wood chips, chemical waste, liquid or gas, as well as the chemicalanalysis thereof

• Fuel operation: different fuels as percentage of total operating time• Flue gas analysis: NO/NO2 ratio, uncontrolled NOx and emission such as CO,

VOC/HC, PM and NH3-slip as well as the emission reduction requirements• NOx—in: raw/uncontrolled emission at different loads of the combustion equip-

ment• NOx—out: reduced, permitted NOx rate [kg or lb. per hour, ppm vd (dry by

volume)] at stack or tail pipe• Exhaust gas mass flow rate (kg or lb. per hour, N m3 per hour or scfm) and gas

density• Flow rate (m/s) and uneven gas flow and temperature deviations (%) at front

face of catalyst bed, requiring computer and/or scale-flow-modeling and staticmixer applications

• Maximum allowable pressure drop of the catalyst bed• Pressure drop, total from combustion equipment to stack or tail pipe exit• Sound attenuation requirements (dB A)• Free-flow, area-, and space-velocity data• Catalyst space availability: maximum cross section and length• Temperature range, max. and min. at front face of catalyst at different loads• H2O concentration in the flue gas• O2 concentration in flue gas


(a)

(b)

(c)

FIG. 17 (a) Metal-substrate based noble metal oxidation catalysts. (Courtesy of Miratech/Hug.) (b)Catalytic reactivity of noble metal catalyst for SO2 � SO3. (Courtesy of Dudoco 1995.) (c)NOx reduction rate at increased pre-oxidized NO � NO2 at �500 F (250°C). (VDI-Report1995.)


• O2 concentration of air quality regulator’s standard for emission rate calcula-tions at 5%, 7%, or 15% O2

• Acceptable process guaranties and equipment warranties• Calculated catalyst bed dimensions and number of catalyst modules• Oxidation catalyst requirements• Particulate filter requirement• Other

There are several SCR and Oxidation Catalyst and Application EngineeringCalculations required to layout a SCR system. For some basic calculations equa-tions are provided in Fig. 18.

SCR Catalyst Material

There are basically three types of material used for SCR catalysts, noble, basemetal, and zeolite. The material formulations and manufacturing processes are usu-ally proprietary developments of the manufacturers. The reaction of NOx with am-monia takes place at the catalyst macropore surface, which may amount to 60 m2/per gram of material of noble or base metal catalyst materials. In the case of theceramic-zeolite-type catalysts, the exothermic reaction takes place inside the vastmicropore structure of over 200 m2 per gram of zeolite.

Noble catalyst metals are platinum, rhodium, and palladium. They can be usedfor both NOx reductions and for the oxidation of VOC/HC, CO, PM. Due to thehigh cost, primarily oxides of base metal are being used for SCR catalysts. Noblemetal oxidation catalysts may be used upstream and/or downstream of the SCRcatalyst: for upstream, to enhance the SCR NOx reduction by partially oxidizingNO to NO2, which in some cases, however, is counterproductive if SO2 is convertedto SO3 as well; for downstream, to reduce possible ammonia slip spikes and CO/HC not oxidized by the SCR catalyst. In HD diesel engine SCR applications usingthe Siemens diesel SCR catalyst, oxidation catalysts are generally not recommendedbecause additional PM would be generated when burning sulfur fuel. Also, today’sefficient HD diesel engines emit only minimal amounts of CO and VOC/HC.

Base-metal-based SCR catalysts contain oxides of base metals such as titanium(TiO2), vanadium (V2O5), tungsten (WO3), and additive and ceramic binders. V2O5

is highly reactive and used in small amounts of up to �2% only. Catalysts witha high V2O5 content are used in the production of sulfuric acid (H2SO4) as well,which would also form in exhaust gas ducts if SO2 oxidizes to SO3 catalytically(SO3 � H2O � H2SO4). Base-metal-type SCR catalysts have nondiscrete mac-ropores and channels, adsorbing ammonia, which is desorbed in a subsequent oper-ation. This allows the adsorption of unreacted ammonia spikes (ammonia slip)rather than passing through the stack as secondary emission. Ammonia slip ratesas low as 3–10 ppm have been achieved in continuous operations. The advanceddiesel SCR catalyst development allows NOx emission reductions at temperaturesas low as 300 F (150°C). Although originally developed for NOx reduction only,the advanced SCR catalysts is able to simultaneously reduce VOC/HC by up to95%, PM by up to 50% and NOx by up to 95% at no extra cost. In Fig. 21, theplate-type and the extruded-type base metal catalyst packaging are pictured.


(a) The K values, the reactivity value of catalysts, may vary greatly, depending on type, material,and structure of catalyst

Keff. �SVAV

� ln(1 � n), (1)

where SV is the space velocity, the volume of the exhaust gas flow (N m3) per hour at normal conditionsdivided by the volume of the catalyst (m3) resulting in (1/h), AV is the area volume, the catalyst surfacearea per catalyst volume (m2/pm3), ln is the natural logarithm, and n is the emission reduction rate (i.e.95% � 0.95).

(b) Catalyst Volume (Vcat.)

Vcat. �VEGF (N m3/h)

SV (1/h)(m3) (2)

SV � Keff. �AV

ln(1 � n)see Eq. (1)

where VEGF is the total exhaust gas flow at standard or normal condition (N m3/h) and SV is the spacevelocity. [For each proprietary catalyst formulation and structure variation, the manufacturer has devel-oped proprietary space velocity table values for reactivity and NOx reduction rate (1/h).]

(c) Exhaust Gas Flow, (VEGF)

VEGF � VEGF min. � NGcons. (N m3/h) (3)

VEGF min. � VAIR min. � NGcons. (N m3/h) (4)

VAIR min. �2(Cx � Hx)

21(N m3/h) (5)

l �21

(21 � O2act.)(%) (6)

where VEGF min. is the volume of air consumption times O2 content times NG consumption (N m3/h),VAIR min. is the volume of air with 21% O2 required to oxidize total HC (N m3/h), l is the percentageO2 in the air used during combustion, Cx Hy is the various hydrocarbons (HC) of the NG analysismaking up the total HC, O2 act. is the actual oxygen (O2) concentration of the exhaust gas, and NGcons.

is the natural gas consumption (N m3/h).

(d) Reducing Agent Consumption, Example Aqueous NH3 Consumption

NH3 cons. � VEGF(N m3) � NOx reduction (ppm)

� 3.3/1,000,000 (kg/h) using aqueous (7)

ammonia with a 25% ammonia concentration,

NOx—Reduction � (NOx in � NOx out)(ppm), (8)

ppm NO � ppm NO � 30 (molecular weight)/22.4 (mg/N m3)

ppm NO2 � ppm NO2 � 46 (molecular weight)/22.4 (mg/N m3), (9)

ppm NH3 � ppm NH3 � 17 (molecular weight)/22.4 (mg/N m3),

ppm SO2 � ppm SO2 � 64 (molecular weight)/22.4 (mg/N m3),

O2act vs. stand � (ppmvd compound)21 � O2 stand.

21 � O2 actual.

(ppm vd) (10)

FIG. 18 (a) Reactivity of SCR catalysts; (b) catalyst volume calculation; (c) Exhaust gas flow calcula-tions, natural gas combustion; (d) reducing agent consumption.


FIG. 19 Pellet-type catalyst. (Courtesy of SCAQMD.)

Zeolite, also called molecular sieve, is a ceramic material. Some zeolite struc-tures occur naturally; others like the ZSM5 of Mobile Oil is produce synthetically.Oil refineries heavily depend on them for their gasoline cracker, synthetic lubrica-tion oil, and gasoline from natural gas processes. The extruded honeycomb-typezeolite-based SCR catalyst has a very large micropore structure of over 2000 ft2

or 200 m2 per gram of material. NOx and NH3 are attached to the micropore surfaceupon passing through the discrete pore openings of �6–10 A in size. This enor-mous sponge effect compensates for major spikes of NH3 and NOx during rapidload changes. The exothermic reaction of ammonia and NOx takes place inside themicropore structure through electrostatic forces. The reaction products, N2 and H2Ovapor, are disposed of, back into the exhaust gas. This reaction is relatively slow,requiring a higher volume of catalyst than, for example, base metal catalysts. How-ever, the zeolite catalyst has superior resistance to many compounds such as heavymetals, which are unable to enter the micropore structure through the discrete open-ings, and thereby extending the service life of the SCR catalyst considerably. Oneexample of the over 1000 known different zeolite crystals is shown in Fig. 20.

Some other more novel combinations of zeolite with noble or base metal mate-rials are presently being researched for PM and other emission reductions.

SCR Catalyst Structure

There are three types of SCR catalyst structure. The pellet-type catalyst, the ex-truded monolithic, honeycomb-type catalyst using either oxides of base metals orzeolite, and the coated-substrate-type catalyst, incorporating either corrugated foil


FIG. 20 A zeolite crystal.

or plate-type stainless-steel sheet metal or extruded Corderite ceramic substrates.There is also a novel catalyst development incorporating fiber-based substrates.(See Figs. 20–22).

Pellet-type catalysts are filled in containers through which exhaust gas ispassed. The pulsing exhaust gas flow, however, cause the pellets to vibrate, abrade/erode, and dust. The dust settles, clogging the catalyst bed and prevents an evengas flow. Due to the erosion of the pellets, the catalyst bed shrinks and unreactedexhaust gas will bypass together with the injected ammonia over the top of thecatalyst bed into the atmosphere. Thus, pellet catalysts do not work most of thetime and were replaced as soon as the honeycomb-type catalysts became available.However, there are still several such reactors operating in southern Californiatoday.

FIG. 21 Left: Coated stainless-steel mesh/expanded metal-substrate-based plate-type catalyst forhigh-dust applications; right: extruded, honeycomb, monolithic, base metal diesel catalyst.(Courtesy of Siemens.)


FIG. 22 Macropore structure of extruded base metal catalyst.

The extruded, monolithic, honeycomb-type SCR catalyst has low back pressureand is widely used for gas turbines and boilers, engines, and other applications.The higher the number of extruded channels per square inch (cps), the higher thereactivity/active surface area and the smaller the catalyst for a specific application.Advanced base-metal-type SCR catalysts are available with 14–300 cps with chan-nel wall thicknesses of 0.3–1.08 mm containing the macropore structure. Thisadvanced catalyst development allows NOx emission reductions at temperaturesas low as 300 F (150°C). The lack of a ‘‘sponge effect’’ may also be the reasonfor the lower emission reduction rates achieved by the coated-substrate-based cata-lyst at temperatures below 480 F (250°C); see Fig. 23.

The coated-type SCR and oxidation catalyst has usually three layers: the corro-sion-resistant substrate (such as the extruded Corderite monolith, corrugated stain-less-steel foil or mesh plates, the aluminized washcoat to which the third layer,the catalytically active material, is bond. The corrugated foil substrate is primarilyused for noble metal catalysts, whereas the Corderite monolith is used for nobleand base metal. The plate-type catalyst has been developed for flue gases, con-taining high-dust loads, such as the hard coal utility boiler, industrial and municipalsolid-waste incineration, and other industrial applications. Long-term operatingexperiences in Europe showed that the erosion of the reactive catalyst material atthe face of the SCR catalyst bed will terminate upon the exposure of the stainless-steel substrate, extending the service life. Due to the smaller macropore structureand surface area and thus absorption capability, the coated-type catalyst is lessreactive.

Conclusion

It would be beyond the scope of this introduction to the SCR technology to gointo further details of the process and the application engineering (i.e., reviewbasically 10–20-year-old designs for coal and gas utility boilers). The future of


(a)

(b)

FIG. 23 (a) Extruded monolithic honeycomb catalyst macropore structure versus the reduced coated-metal substrate-based macropore structure. (b) Standardized test: relative reactivity of ex-truded (�1) versus coated-catalyst structures depending on thickness of coating (�0.6 and0.2) at space velocity of 60,000/h. (Courtesy of Siemens.)

the SCR technology lies in distributed power generation applications such as gener-ation sets, cogeneration sets, and mobile on-road and non-road applications. Thereare already close to 1000 IC engine and turbine applications in service worldwidetoday (Intermacom AG). This number could multiply when HDD trucks and othermobile SCR applications come to market in 2001 through 2010. In the followingsection, a few examples of SCR projects are summarized. However, because somepast design, application engineering, and operation deficiencies gave the SCR tech-nology a bad name in the United States, SCR systems engineers will have to paymore attention to design and application engineering details in future (Table 2).


TABLE 2 Why Certain SCR Systems Have Not Performed in the United States

Advanced (SCR)Failure Cause Technology Solutions

Catalyst1. Clogging and bypass of Pellet type catalyst Honeycomb or plate-type

NOx and NH3 slip catalyst2. Reactivity loss Masking, poisoning, or de- Special lean burn/diesel cat-

lamination of catalyst alyst, homogeneous ma-coating with metal or terial, allowing up to 3%Corderite-based sub- sulfur fuel and an op-strates erating temperature win-

dow of 300–1020 F3. Emission spikes at rapid Little to no adsorption/ Diesel catalyst with micro-

load changes desorption capability of pore structure/‘‘spongecatalyst (nonmonolithic effect’’ with adsorption/catalysts) at �500 F desorption features(250°C)

System Design1. Clogging of injection Heavy corrosion/particle All stainless-steel storage,

system valves or noz- volume due to carbon delivery and injectionzles steel aqueous ammonia system

tank and piping material2. Clogging of catalyst Carbon steel reactor hous- Heat-resistant steel such as

ing, scaling/ particles low Molly steeldue to temperaturecycling

3. Low emission reduction Uneven gas flow at front Gas flow modeling, scaleface of catalyst bed or in- model tests, and lowsufficient mixing of ex- back-pressure static mix-haust gas/NH3 ers such as Parmix TM/

TM SiemensControls and Other1. High emission spikes at Relying only on a down- Feed-forward PEMS-based

load changes stream CEMS with long control with optionalfeedback/response time feed back CEMS or

sensor-based control2. Not cost effective Including a fully certified Electric–chemical sensor-

CEMS, which is often based accurate spotmore costly than the check analyzer with peri-SCR emission reduction odic emission testing bysystem for NOx, VOC, third partyand PM itself

3. Politics: Operator’s Operators avoiding fines, Independent test lab certifi-good references but bad shutdowns, and lawsuits cation, confirming equip-performance of air pollu- of poorly maintained sys- ment manufacturer’stion control equipment tem or new/unproven long-term performancein actual operation technology, ‘‘pro- claims during a 3-year

moted’’ by the regulator performance guarantee


Summary of SCR Application Case Studies

Diesel/Coal-Slurry Fueled Diesel Engine Cogeneration Plant,University of Alaska, 1999

System Overview

The University of Alaska, Fairbanks (UAF) has a 9000-kW electric Fairbanks–Morse reciprocating engine, which currently is fueled on No. 2 diesel fuel uponstart-up in Summer 2000. UAF, in partnership with Fairbanks–Morse and the De-partment of Energy (DOE), has been constructing a demonstration project to testthe feasibility of pulverized coal mixed in water, a coal slurry as an alternativefuel. Fairbanks–Morse had already developed and tested the engine modificationsin their research facility in Wisconsin so that the long-term field tests could com-mence in 2000. The engine will provide electrical power for the campus. Powernot used by the campus is sold to the local electric utility. An exhaust waste heatrecovery system generates steam, which is used for heating and cooling campus-wide.

An SCR system was proposed to reduce NOx emission. Because of the rela-tively high sulfur content of the fuel and the required low SO2 to SO3 conversion,a zeolite catalyst was selected. When operating on coal slurry, the engine produceslarge amounts of particulate matter (PM). In order to deal with the high-PM load,the extruded monolithic honeycomb SCR catalyst was designed with a larger thannormal pitch or channels per square inch (CPSI). This allows the PM to passthrough more easily. The system was designed with a vertical exhaust flow, fromtop to bottom, allowing the PM to pass through and being collected in an ashhopper below the catalyst housing. A soot-blowing system has been installed abovethe catalyst. The soot blower automatically blows down the PM from the catalystat regular intervals when the engine is running on coal slurry. When running ondiesel, the exhaust gas PM or soot concentration is low enough to operate withoutthe soot blower.

The SCR system is designed to reduce NOx emissions independently of thefuel used. Because there is a large difference in NOx emission, exhaust temperature,and exhaust flow rate between the two fuels, the system had to be designed witha wide operating range. When running on diesel fuel, exhaust temperature andflow, and NOx emissions are higher than when running on coal slurry. A systemof duplex metering pumps was furnished. The reducing agent flow is adjusting byvarying the speed of the metering pumps. When operating on coal slurry, only onepump is required. When operating on diesel, both pumps will run.

The system is designed to use aqueous ammonia (20% NH3 in demineralizedwater by weight) as the reducing agent. Ammonia was selected over urea becauseit is easier to source in the region, and because urea would require a great deal oftemperature control during winter operation. Injection rate control utilizes a feed-forward system, which sets pump speed based on engine load. A control feedbackis used which measures NOx emissions upstream and downstream of the catalyst;this data are then used to ‘‘trim’’ or fine-tune the reducing agent injection. Thecontinuous emission monitoring system (CEMS) analyzer (furnished by UAF) al-ternately reads emissions upstream and downstream of the catalyst and thereby


provides the feedback necessary to ensure that the SCR system is operating at peakemission reduction with minimal ammonia slip.

The system is designed to reduce NOx emissions by 90%, with ammonia slipof 10 ppm or less. Currently, since the system start-up in Summer 2000, the systemis operating within these parameters, on diesel fuel.

During commissioning, a 90% NOx reduction and almost zero ammonia slipwas measured. The construction on the coal–water slurry production facilities willstart in 2001. The engine is scheduled to run primarily on coal slurry by late 2001or early 2002.

In Fig. 24, the process and instrumentation diagram (P&ID) and modified sum-mary excerpts of the operating and maintenance manual are presented. Figure 25shows the Table of Contents of the summary excerpt of the Operating and Mainte-nance Manual.

Design Parameters

The SCR system is designed based on the parameters shown in Fig. 26.

SCR Catalyst and Housing

CER-NOx (SCR) Abatement Catalyst. The CER–NOx* (SCR) NOx abate-ment catalyst is a honeycomb-type, molecular sieve, all-zeolite catalyst. Zeolitesare crystalline microporous ceramic solids with pore openings of 3–10 A. Themicropores provide over 1500 ft2 of surface area per gram of zeolite material.

The SCR catalyst is located in the exhaust gas stream downstream of the Fair-banks–Morse engine, before the heat recovery boiler. Each catalyst module hasdimensions of �152 mm � 152 mm � 1000 mm long, without wrapping. Eachmodule has openings with a 6-mm pitch/cps (channels per square inch) to allowparticulate matters (PM) to pass through without plugging. Seventy-two modulesare wrapped in a stainless-steel cartridge, which protects them from mechanicaldamage. The catalyst cartridge weight is 1510 kg (3329 lbs.) and its dimensionsare 1920 mm � 960 mm � 1365 mm height.

Chemical Process. Ammonia reacts with NO and NO2 within the zeolite cata-lyst micropore structure to form nitrogen gas (N2) and water vapor (H2O). Nitrogenoxides and the injected aqueous ammonia are removed from the exhaust gasthrough adsorption into the catalyst micro pores of the zeolite, based on the concen-tration gradient. Electrostatic forces generated inside the micropores decrease theactivation energy for the reduction process, thus allowing reactions to occur in atemperature range of 300° C/570° F to 480° C/900° F.

The reaction releases energy, which forcibly expels the reaction products N2

and H2O from the micropores.

* CER-NOx is a trademark of EEST/Steuler.


FIG

.2

4Pr

oces

san

din

stru

men

tatio

ndi

agra

m(P

&ID

)of

the

Ala

ska

coge

nera

tion

plan

t.(C

ourt

esy

ofE

ESI

/Ste

uler

.)


CER-NOx (SCR) NOx ABATEMENT SYSTEMOverview and Component Description

Table of Contents

1. Design Parameters

2. (SCR) Catalyst & Housinga) CER-NOx (SCR) Abatement Catalystb) Chemical Processc) (SCR) Reactor Housing

3. Aqueous Ammonia Supply & Injection Systema) General Descriptionb) Ammonia Metering Panelc) Compressed Air Subsystemd) Electric Sub-panele) Ammonia Injection Lance Assemblyf) Static Mixerg) Ammonia Pump Stationh) Ammonia Storage Tanki) Ammonia Tank-Truck Unloading Station

4. Operating System Controla) Conditionb) Analysisc) Action

5. Operating Routinesa) Start Upb) Normal Operationc) Shut Down

6. Maintenance Routines

7. Daily Log, Visual Inspection

FIG. 25 Table of Contents of the CER–NOx SCR system.

SCR Reactor Housing. The housing is fabricated of A36 carbon steel. It in-cludes the following:

• A 90° inlet transition from the exhaust duct to a turning vane assembly to distrib-ute the exhaust gas evenly across the catalyst bed

• A shelf system with one row for catalyst• One soot-blowing system to blow ash off the catalyst bed during coal–water

fuel operation. The system includes control valves, isolation valves, and drivemotors, controlled from the Co-gen plant control system

• Bolt-on access doors provided in order to load and unload catalyst; catalyst isloaded into the housing after the housing has been installed

Aqueous Ammonia Supply and Injection System

General Description. The reducing agent supply and injection system consistsof a storage tank, an ammonia pump station, a metering panel, an injection spraynozzle, and interconnecting piping. A static mixer is welded into the inlet of thereactor housing to properly mix the reducing agent and to evenly distribute the


• Emission source: Fairbanks–Morse dual-fuel engine, 9000 kW(e)/ �12,500 BHP

• Fuel type: either diesel fuel or coal–water slurry fuel

Combustion Process Information:Diesel Fuel Coal Slurry Fuel

Exhaust flow rate, lb/hr 155,800 105,400Exhaust temp., max, °F 837 800Exhaust temp., design, °F 665 765

CO2, weight % 8.6 11.3O2, weight % 13.3 11.9H2O, weight % 3.5 6.6N2, weight % 74.4 70.2NOx, volume 0.18 vol.%, wet 600 ppmvSO2 0.016 vol. %, wet 79 ppmv

Ash/particulate, lb/hr 2.2 350

—SCR System PerformanceNOx Reduction, % 90% —Ammonia slip 10 ppmvd (15% O2)SO2 to SO3 conversion 0.1% maxPressure drop, max., for clean system 1.75″ water column, measured from inlet hood to

end of catalyst bed

—Reducing Agent Aqueous ammonia, nominal 25% in water

Technical grade in demineralized water only!—Reducing Agent consumption 53.1 gal/h 21 gal/h—Atomizing Air Requirements

Pressure Minimum 70 psi, maximum 120 psiConsumption 32 SCFM

FIG. 26 Design parameters for the SCR system. (Courtesy of EESI/Steuler.)

exhaust gas across the front face of the catalyst. The metering panel is controlledby the Co-gen plant’s Distributed Control System (DCS).

The reducing agent is supplied via stainless-steel pipe to the metering panel.An ammonia pump station provides pressurized (�40 psi) ammonia to the ringline. The metering panel contains metering pumps, which controls how much re-ducing agent is injected into the gas stream. Refer also to Fig. 24.

Ammonia Metering Panel. The metering panel is divided into two subsys-tems: one for reducing agent control, the other for control of atomizing compressedair. The reducing agent subsystem consists of the following components:

• One fine filter (7 µm) on the pump outlet.• Two metering pumps (P-103, P-104). During coal–water fuel operation, only

P-103 is operated. When the engine runs on distillate fuel, P-103 is run at fullspeed and P-104 is operated at varying speed to provide the balance of thereducing agent volume required.

• Two variable-speed drives (SIC-103, SIC-104), one for each metering pumpmotor. Each drive is controlled by the DCS via a 4–20-mA signal (for speedsetting) and dry contacts (for starting/stopping the drive). A 240-VAC single-


phase input to each drive is converted to a 240-VAC three-phase, variable-frequency output to the metering pump motor. The drives are located in anelectrical subpanel mounted on the side of the metering panel.

• One normally closed (energized to open) solenoid shutoff valve (FV-101). Thisvalve gives a positive shutoff of the reducing agent feed when the system isturned off. The valve is controlled by the DCS. The valve is opened when ametering pump is started and closed when the metering pump is stopped.

• One pressure gauge (PI-105) on the pump discharge to monitor injection pres-sure.

• Two relief valves (PSV-103, 104) to protect the metering pumps.• Two pulsation dampeners to absorb pulsations from the diaphragm pumps.• One magnetic-inductive flow meter (FE/FIT-101) which provides feedback to

the DCS to close the feed rate control loop.

Compressed Air Subsystem. This subsystem consists of the following com-ponents:

• One pressure regulator (PCV-201), to maintain air pressure at 43 psig.• One filter/dryer, to remove particulate matter and moisture from the compressed

air.• One normally closed (energized to open) solenoid valve (FV-201). This valve

closes off the airflow when the system is not running. It is controlled automati-cally by the DCS.

• One pressure switch (PSL-201). If the air supply pressure drops below 50 psig,the pressure switch contacts open, signaling the DCS.

Electrical Subpanel, Mounted on the Metering Panel. This subpanel con-sists of the following components:

• One 20-A disconnect switch, to provide a local disconnect of the 240-V powersupply to the metering panel

• One 24-VDC power supply to provide power for the solenoid valves and flowmeter

• Three circuit breakers to distribute AC power to the two variable speed drives(SIC-103 and SIC-104) and 24-VDC power supply

• One circuit breaker to switch on the outlet of the 24-VDC power supply• Two loop isolators for the variable-speed drives (SIC-103 and SIC-104) speed

setting input• Four 2PDT pilot relays with 24-VDC coils, to take control inputs from the DCS

for SIC start/stop control.

Ammonia Injection Lance Assembly. The injection lance is inserted intothe exhaust duct upstream of the first static mixer. The lance’s spray head shallpoint in direction of flow. The lance consists of an air/liquid atomizing nozzle anda carrier pipe, which transfers the atomized reducing agent into the exhaust duct.


Static Mixer. One static mixer is provided, which is welded into the duct up-stream of the reactor-housing inlet. The mixer ensures proper ammonia/exhaustgas mixing.

Ammonia Pump Station. The ammonia pump station supplies aqueous am-monia to the ring line feeding the metering pumps. It consists of the following:

• One supply pump, P-102, diaphragm type with fixed speed and capacity of 90gal/h.

• One motor, M-102, 1/3 hp, 115-VAC single phase, TEFC to drive P-102.• One pulsation dampener, to remove pulsations from the pump outlet.• One inlet strainer, 190 µm, to remove larger particle from the aqueous ammonia.• One pressure relief valve, PSV-101, to protect the pump.• One pressure/vacuum gauge (PI-103) on the inlet and one pressure gauge (PI-

104) on the pump outlet.• One back-pressure control valve (PCV-102) to set ring line pressure; the return

side of the ring line is connected to the PCV.• One control panel, with HOA switch, main disconnect, motor contactor, in

NEMA 4 enclosure. This panel also houses the ammonia tank level indicator/switch (LIS-101) which indicates the tank level in gallons.

Ammonia Storage Tank. The aqueous ammonia storage tank holds 8000 USgals, is made of 304 stainless steel, and is resistant to the highly corrosive ammoniavapor above the liquid. The tank is designed for atmospheric pressure. It is suppliedwith ball valves, a safety valve, and a level transmitter, shown on Fig. 24.

Ammonia Tank-Truck Unloading Station. The unloading station is used totransfer aqueous ammonia from the delivery tank truck to the bulk storage tank.It consists of the following:

• Centrifugal transfer pump, P-101, with capacity of 100 gal/min.• A 3 HP, 3450 rpm, 460VAC motor, M-101.• Flow switch, FE/FS-101, interlocked to the motor, to shut down the pump if

flow is lost.• One pressure/vacuum gauge (PI-101) on the inlet and a pressure gauge (PI-

102) on the pump outlet.• Cam-and-groove fittings for the tanker to connect.• Control panel, NEMA 4 enclosure, with manual start/stop control switch and

interlock with the flow switch above and the ammonia tank level switch (LIS-101). Upon a ‘‘high tank level’’ alarm, the unloading pump will be shut down.

Operating System Control

The SCR system is controlled by the central DCS of the Co-gen plant. The generaloperating sequence should follow the program in Table 3. See the detailed programdescription for pin assignments and input/output requirements.


TABLE 3 Operating System Control

Condition Analysis Actions

On start of engine Begin purge of injection Send ‘‘start’’ signal to P-lance 102

Start pump P-102Advice from Engine con- DCS decision on which me-

trol if running on distil- tering pump to runlate or coal–water fuel

Exhaust gas temperature Exhaust temperature for Open ammonia valve FV-(570 F) (from TIT-301) SCR Process OK, ammo- 101

nia injection permitted Start metering pump P-103(if CWF) and P-104 (ifdistillate fuel)

Exhaust gas temperature Exhaust temperature for Stop metering pumps(896 F) (from TIT-301) SCR Close ammonia valve FV-

Catalyst too high 101Ammonia storage tank low The ammonia tank empty Stop metering pumps;

level close ammonia valveFV-101

Atomizing air pressure low Failure of compressed air Stop metering pumps;system close ammonia valve

FV-101Ammonia flow below min Failure of metering pump; Stop metering pumps

value (from FIT-101) allow 1 min time delayand metering pump start from start of pump be-signal is given, and me- fore writing alarmtering pump speed signalabove min value

On shutdown of engine Stop metering pumps;close ammonia valveFV-101; close air valveFV-201; stop ammoniapump P-102

Source: EESI/Steuler.

Operating Routines

Start-up. Upon a complete equipment shutdown, the following restart checkshave to be performed: No maintenance is being performed at the NOx abatementsystem, a visual inspection shows no disconnected or broken pipes, wires, and soforth, and the reducing agent supply is available.

• Open all hand valves in the reducing agent supply and injection lines• Atomizing compressed air supply (compressor) available• Open all hand valves of the atomizing compressed air supply and aqueous am-

monia injection lines• Verify that the DCS available• Verify power is available to metering panel• Start the system via the DCS upon determine permissive conditions to start


injection based on temperature, lack of alarm signals, and other operating pa-rameters

Normal Operation. Under normal operating conditions, the DCS controls thespeed of the reducing agent pumps in accordance with the NOx signal receivedfrom the continuous emissions monitor-based feedback. It is generally not requiredto make adjustments or continuously monitor the system.

Shutdown. If the system will be shut down for maintenance, and so forth, thefollowing steps have to be taken:

• Shut down aqueous ammonia injection and allow system to purge lines andlances with compressed air

• Shut off power supply to the metering panel• Close air and ammonia hand valves at inlet and outlet of metering panel

Maintenance Routines

In general, routine maintenance on the CER-NOx SCR system is limited to occa-sional visual inspections, pump oil changes, and periodic replacement or cleaningof filters. Usually, the system can be quickly checked out in conjunction with othermaintenance activities. Table 4 outlines how often certain maintenance proceduresshould be performed. Note that these intervals assume continuous operation. Underintermittent operation, these intervals could be extended. The operator shall deter-mine if the intervals can be lengthened based on operating experience.

Daily Log

Table 5 is the inspection checklist.

Gas Turbine Combined Cycle Power Plant Rated 2 � 170 MW,Natural Gas/Liquid Fuel Fired at Bridgeport Harbor, Connecticut

Overview

Cogeneration and distributed power generation will be the preferred and most fuel-efficient way of the future to generate heat and electricity as fuel cost increasesand the utility industry becomes fully deregulated. Rather than ‘‘destroying’’ theso-called ‘‘waste heat’’ in cooling systems, the local heat requirement will be thedriving output/product and the by-product, the excess electricity can easily bewired away. Fuel efficiency has been a driving force of air-pollution control regu-lator outside the United States in the power generation community, especiallyin Japan and Europe. Fuel efficiencies of over 85% versus 50–60% max. in


TABLE 4 Maintenance Routine

Operation Daily Weekly Monthly Biannual Annual

Visual inspection (see: Visual Inspec- Xtion Daily Log)

Check reducing agent supply tank level XCheck/replace metering pump (P-103/ X

104) and ammonia transfer pump(P-102) oil. Replace every 1000 hof operation

Drain differential pressure gauge X(PDIT-301) condensate lines forSCR reactor

Record differential pressure indication Xfor SCR reactor

Clean/replace reducing agent in meter- Xing panel

Clean atomizing air filter in metering Xpanel

Clean reducing agent strainer at ammo- Xnia pump station

Check/replace metering pump and Xtransfer pump diaphragms

Check metering panel and pump sta- Xtion pulsation dampener charge

Check/rebuild metering panel solenoid Xvalves (FV-101/201)

Check metering panel pressure relief Xvalve setting (PSV-103/104)

Check air pressure switch (PSL-201) Xsetting

Check injection lance nozzle for Xcleanliness

Calibrate all 4–20-mA loops to DCS X

Source: EESI/Steuler.

combined-cycle, gas/steam turbine projects have been achieved. A Swedish utilitycompany operates a 1200-MW cogeneration power plant, providing heating to thetown nearby. The Saarbruecken utility company in Germany, reducing CO2 emis-sion by over 15% in the prior 10 years, primarily through cogeneration, receivedthe U.N. Environmental Award at the summit of world leaders in Rio, Brazil in1990.

In the United States, utility companies have rarely opted for the highly fuel-efficient cogeneration plant alternative yet. The new or upgraded power plants,however, usually incorporate gas turbines and downstream steam turbines, utilizingthe steam generated from the ‘‘waste heat’’ of the gas turbine for electricity genera-tion.


TABLE 5 Inspection Checklist

InitialInspection check list when done

1 Inspect reducing agent storage tank, and stainer for good physical con-dition.

Verify no leaks exist.2 Inspect supply line between reducing agent storage tank and metering

panel.Verify no leaks exist and that isolation valves are fully open.

3 Inspect atomizing air source. Verify that no leaks exist in air supplylines between atomizing air source and metering panel, and that iso-lation valves are open, and correct pressure is available (75–120 psi

4 Check metering panel for proper position of its isolation valves (fullyopen), proper draining of condensate drains, that no physical dam-age exists (internal or external of enclosure), and that no air or re-ducing agent is leaking inside or outside of enclosure. If system isrunning, that it is making normal operation noises, and when fin-ished; that the door is closed and latched.

5 Inspect air and reducing agent lines between metering panel and injec-tion lance for good physical condition. Verify no leaks exist andthat isolation valves are fully open.

6 Check that injection lance is in good physical condition (no bent orcracked hook-up ports).

7 Inspect insulation on exhaust duct and SCR housing for condition andproper installation. Verify that no obvious exhaust leaks exist (indi-cated by soot tracks at insulation joints).

8 Check thermocouple probe at inlet of SCR housing that no exhaustleaks exist and that termination cover is in place.

9 Inspect the SCR differential pressure gauge and its associated blockand bleed valves. Verify that they are not damaged and are properlypositioned (high- and low-pressure input valves open, bypass valveclosed, condensate lines closed).

10 Verify that the DCS is operating normally and that no alarms are dis-played.

Project Summary

The (2) Siemens V 84.3A gas turbine project is rated 170 MW electric per gasturbine. The process and instrumentation diagram (P&ID) of the SINOx exhaustgas cleaning system is shown in Fig. 27. The SCR system design is based on theexhaust gas data listed in Figs. 28a and Fig. 28b. The system ensures State ofConnecticut EPA compliance with the emission reduction of 91% to 4.5 ppmvdat 15% O2, dry basis for nitrogen oxides (NOx, as NO2).

The SINOx (SCR) system features the use of either aqueous urea or aqueousammonia as the reducing agent to meet future, more restricted safety regulations.The SINOx system was delivered as two preassembled units.


FIG. 27 P&ID of a (2) gas turbine power generation plant.

A brief technical description of the systems function is provided in Fig. 29 anda system maintenance schedule is given in Fig. 30.

Mobile, Portable and Other Applications

Background Summary

California ARB (Air Resources Board) Strategy for Additional EmissionReductions by 2007/2010. During the University of California–Irvine Tech-nology Meeting on October 6, 1999, the Cal.-ARB presented its objectives to sub-stantially reduce NOx, HC, and PM emissions. To achieve California SIP (U.S.EPA’s State Implementation Plan) goals by the year 2007/2010 emission of on-road and off/nonroad vehicles and equipment could be reduced by market incentiveand monetary incentive programs. Retrofit emission reduction applications for die-sel locomotives, diesel-powered coastal vessels and construction/mining equip-ment, portable generation sets, and various agriculture (i.e., irrigation/pump drives)and garden equipment should be prime targets.

Summary of Recent Market Data on Retrofit Emission Reduction TargetMarkets. Over 50% of the total U.S. NOx emission inventory of �25 milliontons per year and close to 60% of PM emissions is generated by portable, on-road,and off-road vehicles and equipment.


(a)

(b)

FIG. 28 (a) Project data, operating conditions of the (2) gas turbines; (b) project data, system utilities.


FIG. 29 Brief description of the SINOx system process.

• A total of 2.2 million tons/year of NOx from the United States representing8.6% were assumed from Class 8 HD diesel trucks. However, due to excess‘‘off-test-cycle’’ emissions of 15.758 million tons during 1988 through 1998,caused by 1.328 million Class 8 trucks, an additional 1.3 million tons of NOx

per year had to be added, for which engine manufacturers were fined in a con-


FIG. 30 System maintenance schedule.

sent decree. This increased HDD truck NOx emission resulted in over 3.5 mil-lion tons of NOx emission per year, equal to 12% of the total U.S. NOx inven-tory.

• The Consent Decree of the U.S. Justice Department and the U.S. EPA, the Stateof New York, and Cal.-ARB with the engine manufacturers did not incorporateany short-term remedies but only engine-rebuilt solutions for this large fleetof highly fuel-efficient HDD trucks. One reason, according to industry, is theunrealistic U.S. FTP (Federal Transient Protocol) test cycle with heavy empha-sis on low torque/low rpm and high rpm, whereas most truck operations take


FIG. 31 Product summary, SINOx SCR System. (From Intermacom AG, Feb. 20, 2000, Draft forthe WebPages of Cal.ARB.)


place in medium torque/rpm. In comparison, Japanese and European tests cyclesemphasize medium torque/rpm, almost neglecting the low torque/low rpm oper-ation. One of these European tests shall now complement the U.S. EPA FTPtests. New trucks however, are required to meet NOx � HC emission levels of2.4 gr./BHP h by 2002/2004.

• Off-road NOx and PM emission from diesel-powered locomotives, ships, con-struction, mining, and agriculture equipment as well as portable equipment suchas generation sets have not been a U.S. EPA priority even though they accountfor approximately one-quarter of the total U.S. NOx inventory. By 2010, U.S.diesel locomotives are required to meet 5.6 gr./BHP h. California has an earliertarget date and fleet averaging provisions. Also, marine applications lag behind,even though less than 2 gr./BHP h NOx emission rates have been demonstratedin close to 100 marine diesel engine applications (with engines rated 300 toover 10,000 BHP) in Europe. Portable generation sets are not required toachieve better than 5.9 gr./BHP h NOx emission rates in California, whereassome of the same engines in trucks have to meet 2 gr./BHP hr NOx in 2002/2004. According to Cal.-ARB there are 72,064 portable and stationary dieselengines without emission controls in California, rated 110–600 BHP.

Reports on Technology Evaluations. In recent publications of UC-Davis’ITS (University of California, Institute for Transportation Studies) and Diesel FuelNews, various emission control strategies for HD diesel engines were discussed.Two most promising retrofit technologies were identified achieving over 70–80%NOx and VOC/HC and substantial PM emission reduction: The UREA–SCR andthe NOx Absorber Technology. All other retrofit/post treatment technologies areeither years away from any commercialization or achieve only 20–40% NOx reduc-tion rates.

• The NOx Absorber Technology has been tested by Cummins, using the Euro-3(13 Mode) Steady State Diesel Engine Tests and 5 ppm sulfur fuel, achieving80% NOx reduction at a 8.5% fuel penalty. In a Marathon–Ashland Petroleumcommentary to U.S. EPA, it is claimed that only a 20,000-mile service life with5–15 ppm sulfur fuel could be expected. Currently, no 15 ppm sulfur fuel iscommercially available on a large-scale basis in the United States. The U.S.EPA does expect that such 15 ppm sulfur fuel will be readily available in theUnited States prior to 2007. In addition, lube oils with phosphor and sulfur com-pounds have to be reformulated and tested to avoid loss in engine service life.

• The UREA SCR Technology for the simultaneous reduction of NOx (70–85%),VOC/HC/AirToxics (80–95%), and PM (up to 50%) has already been used inHDD truck field tests in Europe and the United States in the 1990s. The technol-ogy will be commercially available by 2001, has been field tested for 4 years,and demonstrate no fuel penalty. The service life is expected to be over 300,000miles and the initial target price is estimated to be US$ 2000–3000 per HDDtruck. The European truck manufacturers and Siemens pioneered this technol-ogy. By the end of 2001, Siemens will go into the SCR system production forDaimler–Chrysler and MAN’s new ultralow-emission HD diesel trucks. Thevolume-produced SINOx Systems can then be used for mobile, transportable,


(a)

(b)

FIG. 32 (a) Stationary and mobile applications for the diesel SCR catalyst. (b) Simultaneous reductionof NOx, VOC/HC, and PM at stationary and mobile applications. (Courtesy of Siemens.)

and stationary engines rated 100–600 BHP. However, to use the prefabricated,off-the-shelf SINOx products as retrofits kits, engine model/application adapta-tions through local factory trained and licensed dealers are required.

The SINOx SCR systems are used in coal and gas-fired utility boilers and gasturbines, stationary and portable generation sets, cogeneration, and various mobile,on-road and off/nonroad applications. Figure 32 shows the range of SINOx Appli-cations.

Jet Fuel Pipeline Pump Station for the New York Area Airports, NJ

This SCR application was discussed in two U.S. engineering journals in 1998.Three 3000-BHP gas engines with low NOx emission exceeded the plants VOC/HC/AirToxics, CO, and NOx emission limits, set by the State of New Jersey. Figure33a shows one of the three engine enclosures with the SCR system and exhauststack located in front. In Fig. 33b, the engine data are listed. The system designaspects are listed in Fig. 33c. The process guaranteed emission limits are shownby Fig. 33d.


(a)

(b)

(c)

(d)

FIG. 33 (a) Engine enclosure with SCR system and stack in front; (b) Data on the three gas enginesof the jet fuel pump station; (c) SCR system design data; (d) Pump station permitting andprocess guarantee data. (Courtesy of Siemens.)


Coastal and Ocean-Going Vessels and Other Marine Applications

Two of the vessels equipped with SINOx diesel SCR Systems achieving very lowemission rates of equal or less than 1 gr/pBHP h are pictured in Fig. 34.

Heavy-Duty Diesel Truck Field Tests in Germany and the United States,1995–1999

SINOx (SCR) System Overview. The SINOx after-treatment system is a fullydeveloped and tested diesel exhaust emission reduction system. Figure 35 picturesthe first HD diesel truck on the left with a round muffler, replaced by a roundSINOx diesel SCR System in the early 1990s. The right photo shows the Daimler–Chrysler HD diesel truck barrel-type SCR reactor/muffler, going into productionat the end of 2001.

The UREA–SCR System developed by Siemens (Siemens Westinghouse) andthe European trucking industry uses aqueous urea as a reducing agent for the SCRprocess to reduce NOx. The process also reduces VOC/HC, and PM at the sametime. The manufacturer has performed laboratory and fleet tests of the SINOxsystem. Approximately 20 European Mercedes Benz (Daimler–Chrysler), IVECO,and MAN heavy-duty Class 8 trucks were tested in the field. In late 1998, someof these trucks had accumulated over 300,000 miles in common carrier operations.The urea infrastructure, the risks of tampering with the UREA (SCR) system, andother initial operating reliability and other concerns were resolved to the Europeantrucking industry’s and air quality regulators’ satisfaction. The diesel fuel usedduring the tests had a sulfur content of max. 500 ppm (0.05%). Figures 36a and36b show program objectives and the aqueous urea pump station, respectively.

FIG. 34 SCR marine applications with extremely low NOx emission. (Courtesy of Siemens.)


FIG. 35 Pilot-type and preproduction model SCR system for Class 8 HD diesel trucks. (Courtesy ofSiemens.)

(a)

(b)

FIG. 36 (a) SCR truck program objectives; (b) SCR aqueous urea pump station. (Courtesy of Sie-mens.)


The SCR system feasibility tests included bench tests using European Station-ary Cycle (ESC) (�OICA) and European Transient Cycle (ETC), and the FederalTransient Protocol (U.S. FTP) Test Cycle in the United States. NOx, VOC/HC,and CO of the exhaust gas were measured upstream and downstream of the catalyst,using sample gas conditioning systems, heated sample gas lines, and various gasanalyzers. The following analyzers were used: chemiluminescent detector (CLD)for NO and NO2, flame ionization detector (FID) for VOC/HC, a nondisperse infra-red analyzer (NDIR) for CO and CO2, and a magnetopneumatic analyzer for O2.Fourier transformation infrared spectroscopy (FTIR) allowed simultaneous, real-time monitoring for multiple gas components. The FTIR was used downstream of

(a)

(b)

FIG. 37 (a) European steady-state cycle (ESC) test results; (b) European transient cycle (ETC) testresults. (Courtesy of Siemens.)


the catalyst for the exhaust gas and for detecting potential traces of secondaryemissions, such as ammonia slip (NH3), laughing gas (N2O), cyanic acid (HCN),formaldehyde (CH2O), and so forth. Also, a microwave process analyzer was usedto measure ammonia slip during steady-state tests and for periodic verification/calibration tests of the FTIR analyzer. For analytical hydrocarbon tests, bag sam-ples were taken and analyzed with gas chromatographs.

Test results of an European HD diesel engine with a SINOx System are summa-rized and compared with emission standards, based on the European steady-statecycle ESC (OICA) and the European transient test cycle (ETC) as shown in Figs.37a and 37b, respectively. The NOx emission reduction was 80% in steady-state

(a)

(b)

FIG. 38 (a) Urea SCR steady-state/13-mode test results, DDC-S 60, 12 L/400 HP HDD truck; (b)Urea SCR U.S. EPA’s FTP test results, DDC-S 60 engine. (Courtesy of Siemens.)


and 80% in transient tests. Hydrocarbons are simultaneously reduced by over80%.

Ammonia slip is limited to single-digit ppm, and no other undesirable second-ary emissions were detected with the sophisticated analyzer setup described above.This was confirmed in U.S. bench tests in Detroit in 1998, using the most popularClass 8 HD diesel truck engine in the United States, the Series 60 Detroit dieselengine, as shown in Figs. 38a and 38b. The emissions measured downstream ofthe SINOx diesel catalyst was approximately 50% lower than the next lower tierof the European emission limits for NOx, VOC/HC, and PM [18]; that is, theemission limits achieved in Europe and later in the United States already met theyear 2005 European and the 2002/2004 U.S. EPA emission limits in 1999.

SINOx System Description

The exhaust gas generated by the diesel engine is fed into the SINOx catalyst,which is integrated into the exhaust gas pipe system (Fig. 39). A certain amountof the reducing agent, the aqueous urea (30–40%) is injected upstream of the SCRcatalyst. The reducing agent storage tank has a volume of �5% of the fuel tankvolume.

The control system data bus provides the data link between the HDD engine’scontrol unit and the SINOx control unit. A flow control unit in the metering panelmeters the amount of reducing agent. The SINOx system is activated by the signals‘‘engine on’’ (provided by the Engine Management System) and exhaust gas ‘‘tem-perature above set point’’ (temperature sensor in exhaust pipe). The amount ofreducing agent is metered and in accordance with the computer-memory-basedpredicted NOx emission and injected into the gas pipe upstream of the SINOx

FIG. 39 SCR system layout. (Courtesy of Siemens.)


(a)

(b)

FIG. 40 Utility boiler application.


catalyst. The SINOx control system uses a correlation function/map PEMS of theNOx mass flow, relative to the load, which was generated during engine benchtests. The reducing agent injection rate can be increased or decreased, dependingon the required NOx yield and/or the measured NOx concentration downstream ofthe catalyst via a feedback control.

The SCR Process

Aqueous urea is converted into ammonia through hydrolysis downstream of theinjection nozzle. In the presence of the highly reactive, homogeneous, extrudedSINOx (SCR) diesel catalysts, the nitrogen oxides (NOx) react with the ammonia(NH3) to yield nitrogen (N2) and water vapor (H2O). The operating temperaturerange is approximately 300–1020F. An oxidation catalyst is not required if anSCR diesel catalyst is used, because the fuel-efficient heavy-duty diesel enginespractically produce no CO and very little VOC/HC and PM, which are furtherreduced by 80–90% and 20–30%, respectively. Oxidation catalysts are even unde-sirable if sulfur-containing fuel is used.

The SINOx system is controlled automatically by a programmable micropro-cessor-based process logic control system. Active monitoring or supervision of thesystem by the operator is not required.

Summary on a Urea SCR Application in a Coal-Burning UtilityOperation, American Electric Power Gavin Plant at Cheshire, OhioRated 2 X 1300 MW

American Electric power (AEP), an international energy company based in Colum-bus, OH is capable of producing 38,000 MW electricity in 11 U.S. states and hasa customer base of more than 4.8 million. AEP and other power generators in theMidwest and Southeast are required by court order to reduce their NOx emissionby May 2003. AEP’s Gavin Plant has two 1300-MW generating units and is thelargest utility station in Ohio This AEP Urea SCR project is designed to reduceNOx by approximately 70% to 0.15 lbs./pBTU of the coal-burning plant’s utilityboilers at a cost of US$ 175 million. This project uses urea rather than anhydrousor aqueous ammonia as the reducing agent to overcome local safety concerns aboutthe transport, handling, and storage of ammonia. (See Fig. 40.)

Acknowledgments

The author would like to thank Siemens powergeneration’s KPW Group, Redwitz,Germany and Alpharetta, GA (USA) for supporting this project. The author isespecially thankful to Dr. Juergen Zuerbig, head of Siemens’ SCR Business world-


wide and Dr. Raimund Mueller,* Head of the U.S. SCR Business and their stafffor the support received and for the permission to incorporate Siemens material,used in prior papers and public presentations for this project. Without sharing theirinside information about the European advancements in the SCR technology, inparticular in mobile SINOx diesel catalyst applications in recent years, this projectwould not have been possible.

Also, the author would like to thank EESI/Steuler for the support received andthe permission to use material from public and internal documentation. The authoris especially thankful to Nick Detor, Engineering and Project Manager of EESInc., Cerritos, CA and Hans J. Wagner, Senior Manager of Steuler’s equipmentDivision, Hoehr–Grenzhausen, Germany who provided valuable information aboutsome CER-NOx zeolite SCR catalyst applications.

Furthermore, the author would also like to thank Dr. Ray Anthony, professorand head of the Chemical Engineering Department, Texas A&M University andhis staff for their valuable guidance and editing effort. Finally, the author wouldlike to thank the many professionals in the air pollution control community in theUnited States and Europe, with whom he has been associated over the years. Theinformation shared privately, at conferences, and while serving with the author oncommittees in California, such as the Scientific Review Committee on Best Avail-able Control Technology (BACT) at SCAQMD, on IC engines at Cal. ARB, andat joined projects at the University of California, was most helpful.

Bibliography

H. Lueders, R. Backes, G. Huethwohl, D.A. Ketcher, R.W. Horreck, R.G Hurley, and R.H.Hammerle, ‘‘An Urea Lean NOx Catalyst System for Light Duty Diesel Vehicles,’’SAE 95-2493 (October 1995).

R. Mueller, H. Roemich, and M. Joisten, ‘‘An Experience Report on Reducing Emissionsof Criteria Pollutants of Stationary and Mobile Sources,’’ ICAC Forum ’98, CuttingNOx Emissions. (1998).

G. Fraenkle, C. Havenith, and F. Chmela, ‘‘Test Cycle Development EURO3 for HD DieselEngines,’’ CSAT, 1996.

‘‘UREA–SCR Test Achieves Big Cuts in NOx with Fuel Savings over Competing Technolo-gies,’’ Diesel Fuel News (April 1999).

J. Koeser, ‘‘SCR-De-NOx Katalysatoren,’’ Vulkan Verlag, Essen, Germany, 1992.N. Fritz, R. Mueller, J. Zuerbig, and W. Mathes, ‘‘On-Road Demonstration of NOx Emission

Control for Diesel Trucks with SINOx Urea SCR System,’’ SAE Report 1999-01-0111(1999).

S. Eidloth, H. Roemich, M. Joisten, and A. Silini, ‘‘Exhaust Gas Aftertreatment SystemsOnboard Seagoing Vessels,’’ Marine Conference and Exhibition, Brussels, 1999.

G. Lepperhoff, G. Huetwohl, Q. Li, and F. Pischinger, ‘‘Untersuchung der NOx-Reduzierungim Abgas von Dieselmotoren durch SCR-Katalysatoren’’ [NOx Emission Reduction of

* Dr. Raimund Mueller became Manager of Mobile SINOx (SCR) applications worldwide in early2000.


Diesel Engine Exhaust Gas, a Comparative Analysis of European and U.S. (SCR) Cata-lysts], Report, Technical University RWTH–Aachen, Germany (1996).

Vols. 6 & 7.‘‘Encyclopedia of Chemical Processing and Design,’’ Marcel Dekker, Inc.,New York, 1978.

Intermacom A.G., various in-house studies, 1990–1999.F.O. Witzel, ‘‘Massnahmen fuer einen wirtschaftlichen und sauberen Betrieb der Schiffs-

Diesel-Motoren’’ [Methods to Operate Ship Diesel Engines Efficiently with Low Emis-sion], VDI Rep. 15 (198), (1998).

BACT Guidelines, South Coast Air Quality Management District (SCAQMD) (1998).M. Joisten and R. Mueller, ‘‘Experience with the Reduction of NOx Emissions from Diesel

Cogeneration Plants and Industrial Incineration Facilities,’’ Power Gen Europe, 1995.M.K. Khair, ‘‘Demonstration of Advanced Emission Control Technologies Enabling Diesel-

Powered Heavy-Duty-Engines to Achieve Very Low Emission Levels,’’ MECA 1998.J. Kolar and H. Gleis, ‘‘NOx-Minderung in Rauchgasen,’’ VDI, (1987).R. Mueller and M. Grove: ‘‘Advanced SCR System Technology for the Simultaneous Re-

duction of NOx, VOC/Airtoxics and PM Emission,’’ California Air Pollution ControlOfficers Association (CAPCOA) Engineer Symposium, 1999.

J. Peckham, ‘‘80% NOx Reduction with 8.5% Fuel Penalty, (NOx) Adsorber Test,’’ DieselFuel News (October 1999).

‘‘Exhaust Emission Reduction in Existing Ships,’’ Diesel & Gas Turbine Worldwide (April1995).

F. Fleischer, ‘‘NOx Reduction Technology Challenges Marine Diesel Builders,’’ Diesel &Gas Turbine Worldwide.

R.H. Thring and R. Hull, ‘‘NOx Control Technology Data Base for Gas-fueled Prime Mov-ers, Southwest Research Institute, 1987.

L. Bruce, ‘‘Reducing NOx Emission,’’ Power (1981).J. Davis and G. Duponteil, ‘‘Using SCR for NOx Control, Power (1986).J.H. Wasser and R.B. Perry. ‘‘Diesel Engine NOx Control with SCR,’’ EPA Stationary

Source Symposium, 1984.C. Havenith and R.P. Verbeek, ‘‘Transient Performance of a UREA DeNOx Catalyst for

Low Emission Heavy-Duty Diesel Engines,’’ SAE Paper 970185, 1997.Diesel Net, various reports 1998–1999. http:/ /www.dieselnet.com.Diesel Fuel News, various reports (1998/1999).N. Kato, N. Kokune, B Lemire and T. Walde, ‘‘Long Term Stable NOx Sensor with Inte-

grated In-Connector Control Electronics,’’ SAE Paper 1999-01-0202 (1999).D. Simbeck, ‘‘The Future of Distributed Power Generation,’’ Power Generation in the 21st

Century, Energy Frontiers & U.S. Department of Energy, (1997).W.R. Miller, J.T. Klein, R. Mueller, W. Doelling, and J. Zuerbig, ‘‘The Development of

UREA-SCR Technology for U.S. Heavy-Duty Trucks,’’ SAE Paper 2000–01–0190M. Kirchner, ‘‘Die Bedeutung von Ammoniak and Ammonium fuer Mensch und Umwelt,’’

GSF-Forschungszentrum, 1999.‘‘Use of Ammonia in NOx Pollution Control,’’ ICAC, 1999.M. Grove and W. Sturm, ‘‘NOx Reduction with the CER-NOx (SCR) Process,’’ ASME

Proceedings, South West Research Institute, October 1988.M. Grove and W. Sturm, ‘‘NOx Abatement System for Glass Melting Furnaces,’’ Ohio State

University, November 1988.

MANFRED GROVE

150 Nanophase Materials in Chemical Process

Nanophase Materials in Chemical Process

Introduction

The selection of construction materials for chemical process equipment is a cri-tical factor in the efficient operation of a chemical process plant. Engineeringmaterials must satisfy a variety of criteria in order to operate successfully in aprocess environment. The mechanical properties of a material must be satisfactorywith regard to tensile strength, elastic modulus, and resistance to fracture, wear,fatigue, and creep across a broad range of temperatures and/or pressures. Thechemical property of corrosion resistance is also a significant factor to consi-der in process material selection. Additionally, the effects of process thermalconditions and temperature cycling on material properties are relevant to thedurability of the material in a working plant. Beyond these generic materialselection considerations, process-specific material properties are regularly re-quired.

The demand for new materials with an ever-expanding range of properties hasled to the exploration of nanophase materials. Nanophase materials are in a uniqueportion of the ordered length scale of materials between molecular and bulk do-mains. As the domain size of a given material increases across the nanometer sizeregime of between 1- and 100-nm molecular properties become increasing collec-tive and bulklike in nature. Thus, many bulk properties of a material consideredto be inherent constants are, in fact, size-dependent variables across the nanophasesize regime. The practical effect of material properties being adjustable acrossthe nanophase size regime is that material domain size becomes another modeof modifying process material properties, along with the conventional modes ofcomposition and treatment conditions.

This article highlights nanophase material developments relevant to chemicalprocess construction and operation. Materials science has experienced substantialprogress in the synthesis and characterization of nanophase materials [1]. Nano-phase materials have properties which vary from those of the bulk including reac-tivity [2], magnetism [3], melting temperature [4], and mechanical strength [5].Although the high surface area of nanophase materials has long been exploited incatalysis, promising results suggest broad application of nanophase materials inall aspects of future chemical processes.

Nanophase concepts and technologies can be applied in many different config-urations, including active nanoclusters, nanocrystalline materials, nanocomposites,nanotubules, and even ‘‘nanoparticles of vacuum,’’ in which desired materialsproperties are obtained by introducing nanometer-scaled voids or pores in materials[6,7]. These structural configurations combine with the fundamental aspects ofnanophase materials to provide tools for the tailoring of desirable mechanical andchemical properties.

Nanophase Materials in Chemical Process 151

Nanophase Properties: Origin and Overview

Nanophase properties are largely a function of an incomplete electronic band struc-ture and surface sites comprising a non-negligible percentage of the total atoms.These characteristics of nanophase systems will be manifest and helpful in under-standing the material properties discussed in this article.

Molecular Origin

The central effect of an incomplete electronic band structure in a system is thatthe smaller the dimensions, the higher the energy of the lowest excited state. Froma philosophical standpoint, nanocrystals can be examined as either a summationof a large, but finite, number of interacting atoms or as a small piece of materialexcised from a bulk crystal. The former approach relies on atomic orbital hybrid-ization to build a band structure. Because the evolution from atomic orbitals tothe band structure associated with the bulk follows directly from atomic hybridiza-tion, a linear combination of atomic orbital–molecular orbitals (LCAO–MO)method is invoked to explain nanophase behavior.

Changes in electronic structure as a function of size are most easily illustratedby the use of a one-dimensional system [8]. An infinite chain of N ‘‘atoms,’’ eachpossessing one π orbital, containing one electron and separated by a distance, a,in the absence of any interatomic interactions, has N degenerate orbitals, φ i, eachwith an orbital energy, e, given by

e � � φ i|H |φ i � (1)

where H is the electronic Hamiltonian. When the individual units interact, theresonance energy is given by

β � � φ i|H |φ j � (2)

The result of a unit of the chain interacting with adjacent units is to lift the degener-acy of orbitals, as shown in Fig. 1.

The energy levels of these one-electron orbitals are derived using the Huckelapproximation of neglecting orbital overlap:

E( j, N)� e � 2β cos� jπ(N � 1)�, j � 1, 2, . . . , N

(3)

� e � 2β cos(kj a)

where kj � j π/(N � 1) a. As N approaches infinity, k, the wave vector, and E(k)become continuous variables from 0 to π/a and e � 2β to e � 2β, respectively.


FIG. 1 The result of adjacent orbital interactions serves to lift the degeneracy of the orbitals resultingin the energy levels of N degenerate orbitals splitting into a band.

It is now apparent that π/a is the limit of the Brillouin zone and each ‘‘atom’’ inthe chain is comparable to a unit cell in the semiconductor. For a finite chain, theeigenvalues are discrete and energetically separated by increasing amounts as thechain shortens, as shown in Fig. 2. The extreme eigenvalues asymptotically ap-proach e � 2β as the chain length increases.

This model system shows how adjacent eigenvalue spacings increase as thesystem shrinks, as in the case of a metal, such as Pt [9]. The addition of a two-unit repetition pair and Jahn–Teller distortion serves to open a gap between thehighest occupied and lowest unoccupied molecular orbitals (HOMO–LUMO) inthe middle of the dispersion curve at k � π/a′, where a′ is the repetition pairspacing, as shown in Fig. 3.

The dependence of the electronic properties on particle size shows dramaticchanges between the dimensions of atoms and the bulk material [9]. As the preced-ing model suggests, the electronic band structure incrementally grows toward the

FIG. 2 The band structure for a one-dimensional chain of N atoms, each with a single electron andan interatomic separation a. Energy in units of overlap energy. (—): an infinite energy chain;� N � 9 eigenvalues; � N � 5 eigenvalues.


FIG. 3 The effect on a uniform chain of the introduction of a two-atom repeat unit and a Jahn–Tellerdistortion to produce a HOMO–LUMO gap at k � π/a′.

bulk as the system size increases. Likewise, one might infer that bulk thermody-namic characteristics arise from an additive stabilization of atomic properties.

Surface to Volume Ratio

The surface of a nanocrystal contains a non-negligible fraction of the total atoms.For example, a 1-nm-radius nanocrystal of arbitrary composition contains about200 atoms, nearly 60% of which occupy surface sites. The effect of a surface canbe incorporated into thermodynamic models to account for the modified behaviorassociated with fine particulate metals and molecular clusters. The surface freeenergy produces a size dependence in the chemical potential, µ, of the particle.For a constant density liquid droplet, the chemical potential is given by

µ � µ∞ �2γρR

(4)

where γ is the surface tension, ρ is the density, and R is the droplet radius. Thesame equation describes crystalline solids with shapes satisfying the Curie–Wulffequation:

γ i

hi

� construction. (5)

where γ i is the surface energy of face i and hi is the distance from that face to theparticle center of mass.


Chemical bond energy and coordination differences also exist between surfaceand bulk lattice sites within a material [10,10a]. With the large percentage of con-stituent atoms or molecules of a nanophase material present on surface sites orwithin a few atomic layers of the surface, nanophase materials often are understoodas a manifestation of surface properties which are masked in studies of bulk materi-als. By way of example, Cu nanoparticles show a significant change in interatomicspacing, owing to surface tension effects and strained surface geometries associ-ated with nanophase systems, with the spacing increasing from 2.22 to 2.50 to2.56 between Cu2, 0.5-nm nanoparticles, and the bulk, respectively [11].

Properties

Melting

Melting is a fundamental, but not simple, process of a crystalline solid at a distincttemperature, Tm undergoing a phase transformation to a liquid. The complexity inthis seemingly mundane process lies in the atomistic nature of the transformation.Although thermodynamics rigorously defines Tm as the temperature at which theGibbs free energy of the solid and liquid phases is equivalent, there is no specifica-tion as to the mechanism of melting. Thermodynamics is also mute as to the kineticissue of the rate at which the transformation occurs. As such, thermodynamic theo-ries of melting assume a homogenous process occurring simultaneously throughoutthe system. Because all real solids contain extrinsic (thermodynamically metasta-ble) defects such as surfaces, dislocations, and grain boundaries, which often areboth thermodynamically and kinetically more reactive to transformation, it standsto reason that they play a significant role in melting. The loss of long-range orderis a characteristic of melting, and in the high concentration limit, extrinsic defectsrepresent a similar disruption of the long-range order.

Metallic and semiconductor nanocrystals have been shown to possess reducedmelting temperatures when compared with the bulk. The depression in melting andannealing temperature is evident throughout the nanocrystal size regime, but mostprofound in the size range from 1 to 6 nm. Melting studies on a range of nanocrystalshave established that the melting temperature is size dependent in the nanometer-sizeregime and is approximately proportional to the inverse particle radius, regardless ofthe material identity, as shown in Fig. 4. The size-dependent melting temperature ofmetallic nanocrystals has included studies of Au [12], Pb and In [13], Al [14], andSn [15]. Even though semiconductors have directional and somewhat covalent bondsand the crystal facets of binary semiconductors do not all exhibit bulk stoichiometry,the size-dependent melting temperature has been shown in the direct band-gap semi-conductors CdS [16] and GaAs [17]. Nanocrystals of the indirect semiconductor Sialso exhibit size-dependent melting [18]. The reduction in melting temperature as afunction of nanocrystal size can be enormous. For example, 2-nm Au nanocrystalsmelt at about 300°C, as compared to 1065°C for bulk gold [12].

Softening of Nanocrystalline Metals

The properties of bulk materials fail to hold true when dealing with nanophasematerials. For example, nanocrystalline solids often are sought for their increased


FIG. 4 The melting temperature of materials as a function of particle size, plotted as a fraction ofnanocrystal melting temperature Tm to the bulk melting temperature Tb.

hardness and/or ductility [19]. The Hall–Petch effect is observed in polycrystallinemetals with random grain orientations, where both hardness and yield stress typi-cally increase as the grain size decreases. Yet contrary to the Hall–Petch effect,computer simulations of nanocrystalline copper show that as the grain size de-creases down to values less than approximately 7 nm, hardness and yield decreaseas the grain size decreases [19].

Nanophase Materials

Intermetallic Compounds

Nanophase materials technology can be used to reinforce an intermetallic compoundwith a ceramic to obtain desirable characteristics of each type of material. For exam-ple, the reinforcement of titanium and aluminum intermetallic compounds (Ti3Al,TiAl) with TiC improves the toughness and retention of strength across a broadrange of temperatures while continuing to provide a lightweight material with goodoxidation resistance [20]. Mechanical alloying is a viable method for homogeneouslydistributing nanophase particles of TiC in an amorphous Ti–Al matrix [20], therebycreating one such in situ ceramic-reinforced intermetallic compound.

Ceramic reinforcement SiC–Si3N4 composite powders have been synthesizedthrough a sintering process, resulting in nanophase SiC particles distributed inter-granularly and intragranularly throughout the Si3N4 matrix [21]. This resulted ina dramatic improvement in high-temperature strength and creep resistance [21].


Chemiresistors

High-performance chemiresistors have been developed through the use of a colloi-dal nanophase structure known as a metal–insulator–metal ensemble (MIME) [22].The resistance of a chemiresistor changes when the device is placed in the presenceof specific chemicals, as both adsorbed and absorbed vapors affect the transfer ofelectrons across the matrix [22]. A thin transducer film having gold nanoclustersterminated with alkanethiol ligands is deposited onto an interdigital microelectrode[22]. By selecting the absolute and relative sizes of the gold nanocluster and theligand thickness, chemiresistors can be developed to target-specific chemical spe-cies while ‘‘ignoring’’ others. Thus, one can selectively detect the presence ofvarious chemicals through changes in electrical resistance.

FIG. 5 Vapor response isotherms of the Au: C8 (1 : 1) MIME sensor to toluene, tetrachloroethylene(TCE), 1-propanol, and water, based on 15°C vapor pressures. The inset displays the tolueneresponse down to a 2.7 ppmv vapor concentration. (From Ref. 22.)


FIG. 6 A simple two-dimensional sketch of the gold cluster film morphology illustrating the goldcore, alkanethiol ligand shell, and region of lower ligand density. (From Ref. 22.)

One such construct has been shown to be extremely sensitive to toluene andtetrachloroethylene (TCE) vapors, but insensitive to water vapor [22]. Such achemiresistor serves as the basis for a new class of miniature chemical sensingdevices that remain unaffected by typical environmental conditions, such as humid-ity [22].

Polymer Films

Although the coating of lenses is an old practice, high-performance broad-bandantireflection coatings have not been commonplace due to materials limitations[23]. Antireflective (AR) coatings function through destructive interference of re-flected light (see Fig. 7A) [23].

The mathematics of this destructive interference, combined with a review ofthe available dielectrics, show that single-layer antireflective coatings cannot pos-sess the desired refractive indices [23]. However, nanoporous films, with pore sizesmuch less than the incident light wavelength can be tailored to specific refractiveindices. The refractive index is dependent on the pore volume ratio of the film[23]. Nanoporous films have been prepared by demixing a binary polymer blendduring spin-coating, and then using a selective solvent to remove one of the sourcepolymers [23].

Nanocluster technology is being applied to embed metal particles in polymer


FIG. 7 (A) Reflection of light from both interfaces of an AR layer. For a given wavelength andincidence angle, light transmission is maximized when the two reflected beams interfere de-structively. (B) Preparation of a binary polymer film. Initially, both polymers (black and gray)and the solvent form one phase. During spin-coating, phase separation sets in, and after evapo-ration of the solvent, a lateral phase morphology is obtained. (C) The film is exposed to asolvent that is selective for one of the polymers, producing a porous film. (From Ref. 23.)

films in order to provide new properties of strength, magnetism, or infrared absorp-tion capabilities [24]. Iron, iron oxide, silver, and copper have been placed success-fully into permanent polymer lattice sites [24]. In such constructs, a small quantityof embedded metal allows the film to remain transparent while still affecting itsperformance characteristics [24]. Applications for such materials include magnetic‘‘watermarks’’ and heat-absorbent window coatings [24].

FIG. 8 Atomic force microscope (AFM) images of two porous PMMA films �110 nm thick. Afterspin-casting of a PS–PMMA–THF mixture onto silicon oxide surfaces, the PS phase wasremoved by washing the sample in cyclohexane. (A) Films prepared from higher Mw PS andPMMA show average structure sizes of �1 µm. (B) If low Mw PS and PMMA are used, thelateral structure size is reduced to �100 nm. Although the film in (A) appears opaque, thenanoporous film in (B) is transparent, with a low effective refractive index. (From Ref. 23.)


TABLE 1 Summary of the Magnetic Data Obtained on FourDifferent Samples

Sample Tb (K) Hc (G) µ(eff) (emu/g)

Co 70 1800 0.115Co:Pt 140 2700 0.0598Co:Pt3 130 2000 0.0357Au at Co:Pt 30 1000 0.0166

Magnetics

As the information revolution continues, the need for reliable, more compact stor-age media continues to grow. As researchers look to alloys to improve upon themagnetic properties of traditional materials, nanophase metallic alloys are viewedas attractive candidates because of its ability to control properties as a function ofdomain size [25]. The magnetic properties of nanophase cobalt, cobalt–platinum,and gold-coated cobalt–platinum appear to be promising high-performance materi-als showing dramatic effects in the blocking temperatures (Tb) and coercivities(Hc) (see Table 1) [25].

High-density media storage requires the nanoparticles to approach monodisper-sity and a defect-free state. Nanometer-scaled ferromagnetic particle arrays havereversible rotation and magnetic switching behaviors associated with magneticcore diameters and shape anisotropy that affords a variety of new device designoptions [26]. At temperatures above 30 K, magnetization reversal is dependent onthermal activation within a volume that increases with particle diameter [26]. Sucha model is useful in predicting blocking temperatures [26].

Catalysts

Properties

The ability to conduct molecular studies of surface behaviors of nanometer-sizedparticles and/or nanoporous materials has led to an understanding of surface effectsin catalysis [27]. Improved scanning tunneling microscopy (STM), sum frequencygeneration (SFG), vibrational spectroscopy, and atomic force microscopy (AFM)have provided information about surface reactions and catalysis at relatively highpressures, as compared to previous instrumentation [27]. The ability to producenanometer-scaled ordered arrays and then to conduct detailed studies of their sur-face behaviors at, or near atmospheric pressures approaching actual catalysis con-ditions has led to a greater understanding of the pressure dependence of metalnanocluster catalysis [27]. SFG studies at high pressures reveal that CO moleculesoccupy binding states on Pt(111) that are not present at low pressures, whereasadsorbates that are readily detectable in low-pressure surface studies often are sta-tionary spectators or totally absent during high-pressure catalytic reactions. Fur-ther, turnover rates are negligible compared with the overall reaction rate [27].These findings illustrate the need for pressure-specific studies of the catalytic be-


havior of metal nanoclusters, because the reaction pressures greatly influence theoutcomes in a way that cannot be extrapolated from low-pressure studies [27]. Inadditional, the level of surface defects, or other irregularities, can have similarlyprofound effects on catalysis behaviors [27].

At the same time, investigations continue into the ‘‘living metal polymer’’model, in which nanocluster growth is considered analogous to polymer growth[28] and autocatalytic surface growth once M(0) nanoclusters have been nucleated[29]. Of interest are the developments in size-control and size-prediction methodol-ogies, which have been tested successfully in monometallic nanoclusters such asiridium(0) [29]. Such models are predicted to be applicable to bimetallic, trimetal-lic, and multimetallic transition metal nanocluster systems [29].

An example of this process of slow, continuous nucleation, followed by auto-catalytic surface growth, is shown in Scheme 1 and Fig. 9. It is the ratio of therate of growth to the rate of nucleation [R � k2(nanocluster active sites)/k] [29]that can be used to predict, and consequently control, nanocluster size [29]. Theresearch into these methodologies also has led to an endorsement of the ‘‘magic

(a) Nucleation

A →k1

B

(1) (COD)Ir(P2W15Nb3O62)8� � 2 acetone (COD)Ir(acetone)2� � P2W15Nb3O62

9�

(2) (COD)Ir(acetone)2� � 2.5H2 → Ir(0) � H� � S � 2 acetone

(3) nIr(0) → Ir(0)n

(b) Autocatalytic Surface Growth

A � B →k2

2B

(4) (COD)Ir(P2W15Nb3O62)8� � 2.5H2 � Ir(0)n → Ir(0)n�1 � H� �P2W15Nb3O62

9� � S

Net Reaction:

300 [(COD)IR(P2W15Nb3O62)8�] � 750H2 →

Ir(0)300 � 300 P2W15Nb3O629� � 150 H� 300S

SCHEME 1 Minimum mechanism for the formation of Ir(0) nanoclusters, consisting of (a) slow,continuous nucleation (steps 1–3), rate constant k1 for the pseudoelementary step,A → B, followed by (b) fast autocatalytic surface growth (step 4), rate constant k2 forthe pseudoelementary step A � B → 2B. Nucleation and growth are separated in timebecause k1 �� k2 [B], which, in turn, is a key to the observed formation of a near-monodisperse (� �15%) particle size distribution. (From Ref. 9.)


FIG. 9 Idealized, roughly-to-scale representation of a P2W15Nb3O629� polyoxoanion and Bu4N� stabi-

lized Ir(0)�300 nanocluster, [Ir(0)�300(P4W30Nb6O12316�)�33](Bu4N)�300Na�228. The Ir(0) atoms are

known (by electron diffraction) to be cubic close packed as shown. For the sake of clarity,only 17 polyoxoanions are shown, in their monomeric form, and the �300 Bu4N� and �228Na� cations have been deliberately omitted. (From Ref. 29.)

number’’ theory of nanoclusters, which states that because closed-shell structuresare stable, they will be more common in the size distribution of completed na-nocluster formations [29].

Further exploration of the catalysis properties of transition metal colloids indi-cates that their behavior is extremely reaction-specific and dependent on manyfactors. For example, in the synthesis of silicone polymers, the catalytic perfor-mance of bimetallic colloids of Au(core)/Pt(ligand) and Pd(core)/Pt(ligand) wascompared to the performance of Pt nanoclusters formed during the induction periodof typical industrial reactions [30]. Previously, Au–Pt had been shown to providemarked improvement over Pt alone in the semihydrogenation of 2-hexyne into cis-2-hexene [31]. However, in the synthesis of bis(trimethylsiloxyl)octamethylsilane(BTMOS), the Au–Pt showed no improvement over Pt alone, whereas Pd–Ptshowed marked improvement [30]. These experiments were designed in order tostudy the behavior of a bimetallic colloid with a more electronegative core thanthe ligand (Au–Pt) and compare it to that of one with a more electropositive corethan ligand (Pd–Pt) [30]. For this particular commercial synthesis, the more elec-tropositive core seemed to yield superior results. However, in light of the Au–Ptsemihydrogenation mentioned, experimental verification is required before suchfindings can be extended to other catalytic systems [30].

Nanocluster agglomeration is prevented through surface stabilization using li-gands, polymers, or surfactants [32]. Using a combination of STM and transmis-


FIG. 10 A schematic diagram showing how STM and TEM can be used in concert to determinesurfactant stabilizer thickness. Thicknesses observed in this manner are consistent with stan-dard MM2 force-field calculation theoretical values. (From Ref. 32.)

sion electron microscopy (TEM) to observe and examine the size relationshipsbetween surfactant stabilizers and metal cores, it has been shown that shell thick-ness is independent of core size and is directly dependent on the size of the surfac-tant ion, as shown in Fig. 10 [32].

Synthesis

Any discussion of nanophase materials in chemical processes must address thetopic of catalysis. The high surface area to volume ratios associated with nanophasematerials makes them exceptional catalysts as compared to bulk materials. Al-though nanophase metal catalysts supported on substrata have been industry stan-dards for high-throughput reactions, in recent years particle size control has strivedto produce stoichiometrically defined particles. Such a well-defined nanophase cat-alyst offers the prospect of engineering the ratio of edge atomic surface sites ona particle surface to alter reaction product efficiencies and mixtures. To this end,transition metal nanocluster systems have been synthesized which behave as isola-ble and compositionally well-defined soluble heterogeneous catalysts [33]. Further,new methodologies and a better understanding of nanoclusters in general are call-ing into question previously identified catalysts in previously explored reactions[33].

One such example is a hydrogenation of benzene, in which the accepted ion-pair catalyst, [(C8H17)3NCH3]�[RhCl4]�, has been discredited, with the catalystshown more likely to be Cl�- and [(C8H17)3NCH3]�-stabilized Rh(0) nanoclusters[33]. This indicates that in the cases of other putative homogenous catalysts wherea facile heterogeneous M(0) catalyst is well established, catalysis by even traceamounts of possibly highly active nanocluster catalysts cannot be ruled out[33].

The development of industrial catalysts,many of which are composed of metalparticles on oxide substrata, is a topic of great commercial interest [34]. Fabricationand engineered control in order to deposit tailored metal clusters on oxide surfacewith an ordered structure remain a goal of catalyst research [34]. Lithographic


nanostructure fabrication technologies of the semiconductor industry can be ap-plied to create such engineered clusters and to provide viable model systems forindustrial-supported catalysts [34]. This model of catalytic structure and behaviorhas been tested against ethylene hydrogenation on a platinum nanocluster [34].

In a study of the thermal stability of supported silver catalysts, these modelsystems and fabrication techniques were used to demonstrate that the presence ofoxygen is the key factor influencing the thermal stability of the silver nanoclusters[34]. No migration of silver clusters was observed at �700°C in the absence ofoxygen [34]. Yet, with increasing annealing temperature in oxygen, silver clustersurface oxidation occurs at �200°C [34].

Methods of nanostructure lithographic fabrication also can be applied to othersurface science studies. When combined with AFM, these developments are usedto determine surface mechanical properties, such as the elastic modulus, on nano-meter-scaled samples [34]. Methodologies also have been developed to measurethe yield of an ion-sputtering process [34].

In addition to the synthesis and use of ligand-stabilized metal clusters, signifi-cant benefits are realized by polymer stabilization of nanocluster catalysts [35].The platinum nanoclusters’ catalytic performance on polymer-based supports hasbeen compared to performance on oxide supports [35]. Polymeric supports showa marked increase in the stability of the catalysts, especially at room temperatures[35].

Although stabilizers control particle size and prevent agglomeration, clustersurface passivation is often the result, with the net effect of reducing catalyticperformance [36]. One proposed solution to this dilemma is the use of dendrimersto act simultaneously as monodisperse synthesis templates and stabilizers [36].Researchers have partitioned transition metal ions, such as Pt, into the interior ofpolyamidoamine (PAMAM) and have achieved an encouraging combination ofparticle size control, particle stability, and electrocatalytic behavior and perfor-mance [36]. (See Fig. 11.)

Nanophase catalysts also function in decomposition reactions of harmful green-house gases,such as CO2[37]. Experiments with different ferrites (XFe2O4) haveshown that with appropriate selection of X and particle size, CO2 is broken downinto carbon and oxygen with virtually no CO by-product [37]. The decompositionof CO2 also produces quantities of methane. Comparisons of Zn, Ni, and Co ferritesand of varying particle sizes yielded dramatic results, with ZnFe2O4 showing themost promise [37].

In addition to purely chemical catalysis, nanocluster materials also function aselectrocatalysts. In the synthesis of sodium chlorate, the Dimensionally StableAnode (DSA) has realized extensive energy savings on one side of the oxidationreaction, but high activation overpotentials on the cathodic side still contribute tolarge energy losses [38]. The experimental production of a solid Ti2RuFe nanocrys-talline cathode has shown promising results. In fact, once the nanocrystalline mate-rial was pressed into a usable cathode, the activation overpotential was reduceddramatically when compared to a standard iron electrode [38]. However, thismethod of electrode fabrication failed to produce anticipated current density in-creases expected from the very small crystal size [38].

Template methods, combined with chemical vapor deposition (CVD), also canbe used in the preparation of carbon nanotubules, with diameters ranging from 20


FIG. 11 Schematic illustration of fourth-generation (G4) PAMAM dendrimers having EOH andNH2 terminal groups, synthesis of Pt nanoparticles within the hydroxyl-terminated den-drimer template, and attachment of the composite to an electrode surface. Between 12 and60 Pt2� ions can be loaded into a single dendrimer and, upon reduction with BH4

�, an en-trapped cluster containing the same number of atoms results. The dendrimer-encapsulatedPt nanoparticles are electrocatalytically active. (From Ref. 36.)

to 200 nm [39]. Such tubules can be filled with metal catalyst nanoclusters of thetypes previously discussed to display interesting and potentially useful electrocata-lytic properties [39]. A basic method used to synthesize metal-nanocluster-filledcarbon nanotubes includes carbon deposition by CVD onto an alumina templatemembrane, followed by immersion in a metal ion solution. Air-drying and reduc-tion by hydrogen gas completes the formation of the metal nanoclusters within thenanotubules. Alumina is removed by HF immersion [39]. The mechanical andelectrochemical properties displayed by metal-filled nanotubules hold promise forthe fuel cell industry [39].

Glass Silicates

A physicochemical analysis of nanophase crystalline silicalite-1 shows that it hasmany properties in common with micrometer-sized crystalline silicalite-1 [40].These common properties include a refined structure and concentrations of tetra-propylammonium species incorporated during the synthesis [40]. Yet, nanophasesilicates have nondegenerate spectroscopic features such as the characteristicframework infrared (IR) vibration (550 cm�1) of the micrometer-sized crystal split-ting into a doublet (at 555 cm�1 and 570 cm�1) in the nanophase material [40]. Inaddition, the nanophase material exhibits a high concentration of defect sites, a


FIG. 12 Log-log plot of the bulk (K), shear (G), and Young’s (E) moduli for bulk and nanophasea-SiO2, as a function of the density relative to the bulk density. The solid lines are the bestleast-squares fits for each of the moduli. (From Ref. 41.)

strain in the crystallites along the a crystallographic direction, as well as a two-stage dinitrogen physisorption in the low-pressure region [40].

The pore morphologies and mechanical behaviors of nanophase amorphousSiO2, investigated by molecular-dynamics (MD) simulations, show that the bulkamorphous densities of the various nanophase a-SiO2 glasses are characterized bydifferent pore sizes and distributions, yet the morphology of the pores, defined interms of the fractal dimension of pores and the roughness exponents of pore–silicainterfaces, is similar across various densities [41]. This consistent short-range order(SRO) of nanophase silica glass of various densities differs little from the SROof bulk glass, with both structures consisting of corner-sharing Si(O1/2)4 tetrahedra[41]. However, analysis of the first sharp diffraction peak (FSDP) shows a signifi-cant difference in the intermediate-range order (IRO), dependent on the density,which is controlled synthetically in the nanophase silica glass [41].

In terms of mechanical properties, the elastic moduli of both the bulk and thenanophase a-SiO2 clearly are density dependent [41], as shown in Fig. 12. Themoduli (M) scale as M � (ρ-bar)3.5 � 0.2, where ρ-bar is the ratio of the sampledensity to the density of bulk silica glass [41]. An understanding of this power-law dependence can lead to very specific tailoring of physical and mechanicalproperties for various uses.

Glass Ceramics

Nanophase glass–ceramic technology also is a rapidly growing field with a widevariety of commercial applications. Nanophase microstructures composed of crys-tals �100 nm in size are achieved through efficient nucleation and slow crystalgrowth, providing an impressive uniformity of microstructure and the controlledmechanical properties not available in glass–ceramics of larger crystalline micro-structures [42].

Transparent nanophase glass–ceramics achieve near-zero coefficients of ther-mal expansion, along with high thermal stability, high thermal shock resistance,and, of course, transparency [42]. The requirements to achieve transparency havepreviously been recited [42]. Glass–ceramics with these properties typically utilizelithium-stuffed β-quartz crystals to satisfy scientific and commercial applications,


such as telescope mirror blanks, stove cook tops, cookware, woodstove windows,fire doors, and other technical devices [42].

Near-zero coefficients of thermal expansion result from adjusting the variouscomposition components and percentages in stuffed lattices [42]. ‘‘Stuffed’’ β-quartz crystals are so named because Al3� replaces Si4� in the β-quartz frameworkof interlinked helixes of SiO2 tetrahedra, with the charge balance being maintainedby ions that stuff the interstitial tetrahedral cavities [42].

Other interesting subclasses of transparent glass–ceramics, many of which alsohave a ‘‘stuffed’’ microstructure, consist of materials with coefficients of thermalexpansion near that of silicon [42], materials such as transparent mullite or spinelglass–ceramics that can serve as superior-performance host media for luminescenttransition metal ions such as CR3� [42], and oxyfluoride glass-ceramics used ashosts for optically active rare earth (RE) cations, because of their low phononenergies and broad transparency in the IR region of the spectrum, used for theamplification of light in telecommunications systems [42].

Conclusions

As we have seen, nanophase materials show great promise for many areas of chem-ical process. Their unique properties, brought about by a high surface area to vol-ume ratio and incomplete electronic band structures, give them extraordinary flex-ibility with regard to the desired tailoring of mechanical and chemical properties.Composition and configuration options offer exciting opportunities to take nano-phase materials out of the laboratory and into innumerable industrial applications.

References

1. Handbook of Nanophase Materials, A. N. Goldstein (ed.), Marcel Dekker, Inc., NewYork, 1997.

2. R. L. Whetten, D. M. Cox, D. J. Trevor, and A. Kaldor, Phys. Rev. Lett., 54, 1494–1497 (1985).

3. S. N. Khanna, Effect on Properties of Reduced Size and Dimensions, in Handbookof Nanophase Materials (A. N. Goldstein, ed.), Marcel Dekker, Inc., New York, 1997,pp. 1–13.

4. M. Wautelet, J. Phys. D: Appl. Phys., 24, 343–346 (1991).5. K. A. Gschneidner, Solid State Phys., 16, 275–426 (1964).6. N. Herron and D. L. Thorn, Adv. Mater., 10, 1173–1184 (1998).7. G. Schmid, M. Baumle, M. Geerkens, I. Heim, C. Osemann, and T. Sawitowski,

Chem. Soc. Rev., 28, 179–185 (1998).8. Y. Wang and N. Herron, J. Phys. Chem., 95, 525–532 (1991); J. K. Burdett, Prog.

Solid State Chem., 15, 173–255 (1984).9. S. N. Khanna, J. P. Bucher, J. Buttet, and F. Cyrot-Lackmann, Surface Sci., 127, 165–

174 (1983).


10. T. Dannhauser, M. O’Neil, K. Johansson, D. Whitten, and G. McLendon, J. Phys.Chem., 90, 6074–6076 (1986).

10a. G. A. Somorjai, Introduction to Surface Chemistry and Catalysis, John Wiley & Sons,New York, 1994, Chap. 2.

11. P. A. Montano, G. K. Shenoy, E. E. Alp, W. Schulze, and J. Urban, Phys. Rev. Lett.,56, 2076–2079 (1986).

12. Ph. Buffat and J-P. Borel, Phys. Rev. A, 13, 2287–2298 (1976).13. C. J. Coombes, J. Phys., 2, 441–449 (1972).14. J. Eckert, J. C. Holzer, C. C. Ahn, Z. Fu, and W. L. Johnson, 2, 407–413 (1993).15. C. R. M. Wronski, Br. J. Appl. Phys., 18, 1731–1737 (1967).16. A. N. Goldstein, C. M. Echer, and A. P. Alivisatos, Science, 256, 1425–1427

(1992).17. A. N. Goldstein, ‘‘Thermodynamic Properties of Semiconductor Nanocrystals,’’ Ph.D.

thesis, University of California at Berkeley, 1993, pp. 140–165.18. A. N. Goldstein, Appl. Phys., A, 62, 33–37 (1996).19. J. Schiotz, F. D. Di Tolla, and K. W. Jacobsen, Nature, 391, 561–563 (1998).20. N. Q. Wu, G-X. Wang, W. Li, J. M. Wu, and Z. Z. Li, Mater. Lett., 32, 259–262

(1997).21. D. F. Carroll, A. W. Weimer, S. D. Dunmead, G. A. Eisman, J. H. Hwang, G. A.

Cochran, D. W. Susnitzky, D. R. Beaman, and C. L. Conner, AIChE J., 43 (suppl.),2624–2635 (1997).

22. H. Wohltjen, and A. W. Snow, Anal. Chem., 70, 2856–2859 (1998).23. S. Walheim, E. Schaffer, J. Mlynek, and U. Steiner, Science, 283, 520–522 (1999).24. Chem. Eng. Prog., 16 (Dec. 1997).25. E. E. Carpenter, C. T. Seip, and C. J. O’Connor, J. Appl. Phys., 85, 5184–5186 (1999).26. S. Wirth, S. von Molnar, M. Field, and D. D. Awschalom, J. Appl. Phys., 85, 5249–

5254 (1999).27. X. Su, J. Jensen, M. X. Yang, M. B. Salmeron, Y. R. Shen, and G. A. Somorjai,

Faraday Discuss, 105, 263–274 (1996).28. M. L. Steigerwald and L. Brus, Acc. Chem. Res., 23, 183–184 (1990).29. M. A. Watzky and R. G. Finke, Chem. Mater., 9, 3083–3095 (1997).30. G. Schmid, H. West, H. Mehles, and A. Lehnert, Inorg. Chem., 36, 891–895 (1997).31. G. Schmid, H. West, J-O. Malm, J-O. Bovin, and C. Grenthe, Chem. Eur. J., 2, 1099–

1108 (1996).32. M. T. Reetz, W. Helbig, S. A. Quaiser, U. Stimming, N. Breuer, and R. Vogel, Science,

267, 367–369 (1995).33. K. S. Weddle, J. D. Aiken III, and R. G. Finke, J. Am. Chem. Soc., 120, 5653–5666

(1998).34. M. X. Yang, D. H. Gracias, P. W. Jacobs, and G. A. Somorjai, Langmuir, 14, 1458–

1464 (1998).35. W. Yu, H. Liu, and X. An, J. Mol. Catal. A, 129, L9–L13 (1998).36. M. Zhao and R. M. Crooks, Adv. Mater., 11, 217–220 (1999).37. S. Komarneni, M. Tsuji, Y. Wada, and Y. Tamaura, J. Mater. Chem., 7, 2339–2340

(1997).38. M. Blouin, D. Guay, J. Huot and R. J. Schulz, J. Mater. Res., 12, 1492–1500 (1997).39. G. Che, B. B. Lakshmi, C. R. Martin, and E. R. Fisher, Langmuir, 15, 750–758 (1999).40. R. Ravishankar, C. Kirschhock, B. J. Schoeman, P. Vanoppen, P. J. Grobet, S. Storck,

W. F. Maier, J. A. Martens, F. C. De Schryver, and P. A. Jacobs, J. Phys. Chem. B,102, 2633–2639 (1998).

41. T. Campbell, R. K. Kalia, A. Nakano, F. Shimojo, K. Tsuruta, P. Vashishta, and S.Ogata, Phys. Rev. Lett., 82, 4018–4021 (1999).

42. G. H. Beall and L. R. Pinckney, J. Am. Ceram Soc., 82, 5–15 (1999).


Bibliography

Bloechl, P. E., C. Joachim, and A. I. Fisher (eds.), Computations for the Nano-scale, KluwerAcademic, Dordrecht, 1993.

Fendler, J. H. (ed.), Nanoparticles and Nanostructured Films: Preparation, Characteriza-tion and Applications, Wiley–VCH, Weinheim, 1998.

Goldstein, A. N. (ed.), Handbook of Nanophase Materials, Marcel Dekker, Inc., New York,1997.

Komoreni, S., J. C. Parker, and G. J. Thomas (eds.), Nanophase and Nanocomposite Materi-als: Symposium held Dec., 1–3, 1993, Boston, MA, USA, Material Research Society,Pittsburgh, 1993.

AVERY N. GOLDSTEINDAVID M. FISHBACH

Process Safety and Risk ManagementRegulations: Impact on Process Industry

Introduction

Chemical accidents have been caused by a number of reasons, including humanerror, design flaws, lack of process and engineering knowledge, equipment failure,and natural disasters. The danger posed to the employees of a chemical plant aswell as the public is illustrated by the accidents that have occurred in onshore aswell as offshore chemical process industries. Figure 1 provides accident statisticsfor 1989 from the Accidental Release Information Program (ARIP) of the U.S.Environmental Protection Agency [1]. ARIP statistics cover catastrophic and un-planned releases of chemicals into the atmosphere. However, these statistics under-line the fact that a large number of accidents and catastrophic releases occur be-cause of design flaws, wrong equipment specifications, and lack of or disregardof operating and maintenance procedures. The boardroom perspective on the causeof these accidents and what to do about them varies. The total number of processplant accidents cannot be estimated accurately because of underreporting. How-ever, it is clear that the number of accidents is large and many people, both workersand the public, are affected adversely by these accidents. For example, in 1991,the National Response Center received over 16,300 calls reporting the release orpotential release of hazardous chemicals [3]. Another study [4] analyzed the EPA’sEmergency Response Notification System database of chemical accident notifica-tions. The study found that from 1988 through 1992, an average of 19 accidents

Process Safety/Risk Management Regulations 169

FIG. 1 U.S. Environmental Protection Agency Statistics Accidental Release Information Program—1989. (Reprinted with permission from Ref. 2.)

occurred each day (i.e., more than 34,500 accidents involving toxic chemicals oc-curred over the 5-year period). The promulgation of the Toxic Release InventoryReporting requirements [5] as part of the Clean Air Act Amendments of 1990 ledto the submittal of toxic release information which clearly delineated the numberand extent of toxic chemical releases and their potential impact on the public andthe environment. In addition to the industry and government agencies, the univer-sity has a critical role in changing this situation.

In addition to statistics and the sheer number of facilities involved, a numberof highly publicized chemical plant incidents in the 1970s and 1980s focused atten-tion on management systems and technologies. For example, the 1974 Flixboroughaccident occurred because a temporary pipe was used to replace a reactor whichhad been removed for repair [6]. The temporary piping was not properly designedand supported merely on scaffolding. A management of change system could verywell have prevented the incident. The causes behind the 1984 Bhopal accident,which involved the release of methyl isocyanate and caused thousands of fatalities,have been investigated quite extensively with varying conclusions. However, theneed for inherently safer design considerations is quite unanimous. Bhopal andmany other process plant incidents including the 1984 Mexico City disaster [7]also emphasize the need for application of structured management systems forhazard recognition and identification. According to the official report [8] followingthe 1988 Piper Alpha disaster in a North Sea offshore rig, a pump relief valve wasremoved for overhaul and the open end blanked. Another shift not knowing therelief valve was missing, started up the pump. However, this primary reason doesnot obviate the fact that a number of other factors contributed to the extensivedamage. Among other things, the Phillips 1989 explosion [9] in the high-densitypolyethylene plant demonstrates the need to adhere to operating procedures andimplementing appropriate management systems for contract workers. Many pro-cess plant accidents in the 1970s and 1980s also exposed the need for managementsystems to ensure process and equipment integrity.

Change in population demographics, increasing awareness of process plant

170 Process Safety/Risk Management Regulations

hazards, and, above all, the continuing threat of a chemical catastrophe continueto provide the impetus for governments to develop legislation for eliminating orminimizing the potential of such accidents. International efforts include the SevesoDirective covering members of the European Community. Many other nations alsohave similar laws, such as the Sedesol guidelines in Mexico for performing processrisk audits and the post-Bhopal accident prevention law in India. The World Bankhas developed guidelines for identifying and controlling hazards, and the Interna-tional Labor Organization has developed a code of practice for preventing majoraccidents. In 1990, the U.S. Congress enacted the Clean Air Act Amendments(CAAA) into law. The CAAA directed the Occupational Safety and Health Admin-istration (OSHA) and the Environmental Protection Agency (EPA) to develop reg-ulations to reduce the frequency and severity of chemical plant accidents. In keep-ing with the congressional mandate given in Section 304 of the CAAA, OSHApromulgated the Process Safety Management (PSM) regulation on February 24,1992. The PSM regulation is intended to protect workplace employees. Similarly,as mandated by Section 301(r) of the CAAA, EPA promulgated its risk manage-ment program regulation in 1996, to protect the public and the environment. Inthe United States, federal agencies are not the only government regulators activein the chemical accident prevention arena. Several states have empowered theirhealth, safety, and environmental agencies to create regulations requiring compa-nies to establish and practice specific programs to improve safety.

The Impact of the Industrial Revolution

The eighteenth century was the beginning of technological development, whichaffected society and commerce in ways that are felt even today. This technologicaldevelopment, known as the industrial revolution, was one of the main revolutionsof this era.

The industrial revolution changed manufacturing by changing the way peopleworked. For one thing, it brought work out of the home and centralized it in thebeginning to small and simple plant operations and increasingly to large and com-plex operations. The industrial revolution grew more powerful each year as newinventions and manufacturing processes added to the efficiency of machines andincreased productivity. The industrial revolution has had far more impact on theworld than the political revolutions of the era, because the Industrial Revolutioneffects on society are longer lasting. For example, today we have automobiles,television, and computers, all made possible by this revolution. Without the indus-trial revolution, we would not have the technologies that we have today and neitherwould we have the standard of living that we enjoy.

Before the introduction of machines and the factory setting, goods were manu-factured by hand, in single homes or cottages, where the owner worked side byside with his employees. This changed with the introduction of machines and massproduction. The major advances in technology, particularly in the use of steam,in the later half of the eighteenth century has its roots in devices that were inventedearlier in the era.


Each advancement in the industrial revolution has brought with it a certainamount of risk and hazardous activity. Before the industrial revolution, the chemi-cal hazards we strive to manage safely today did not exist because none of thesechemicals were used. As global competition increased, so did the need for usingincreasingly complex processes and new and exotic chemicals at different op-erating conditions. Demographic changes and the unplanned growth of bedroomcommunities around industrial belts have also created a higher risk.

Notwithstanding the benefits of the industrial revolution, there has also beenconsiderable dialogue about the risks posed by these industrial facilities. The sub-ject of risk-benefit analysis has been discussed to some extent but is not quiteformalized. Based somewhat on post-accident emotions and partly on a realizationthat industry can and should operate safely, there has been a groundswell of opin-ions from various stakeholders for structured programs to improve safety.

Government Regulations

The history of safety regulations in the United States can be traced back to 1899,when the United States government issued the River & Harbor Act. This act prohib-ited the creation of any obstruction not authorized by Congress, to the navigablecapacity of any waters of the United States except on plans authorized by theSecretary of the Army. The act was promulgated expressly to protect the nation’swaterways from excessive dumping. Subsequent to the River and Harbor Act,Congress has passed numerous laws, which impose environmental or safety regula-tions on businesses.

In 1936, the federal government enacted the Walsh–Healy Act to establishfederal safety and health standards for activities relating to federal contracts. TheWalsh–Healy Act led to early research into the identification and control of occu-pational diseases. The ideas behind this act are the basis of many of today’s occupa-tional health and safety regulations.

During the period between 1936 and 1970, a number of other regulations werepromulgated. For example, the Federal Water Pollution Control Act, the AtomicEnergy Act, the Metal and Non-Metallic Mine Safety Act, and the Federal CoalMine Health and Safety Act. Although some progress was made, these regulationswere never sufficiently supported to carry out a satisfactory program. This pro-duced relatively inconsistent and ineffective results.

In 1970, Congress promulgated the Occupational Safety and Health Act(OSHAct). As a result of this landmark legislation, OSHA and the National Institutefor Occupational Safety and Health (NIOSH) were established within the Depart-ment of Labor and the Department of Health and Human Services, respectively.OSHA’s mission is to ‘‘Assure so far as possible every working man and womanin the nation safe and healthful working conditions.’’ The OSHAct allows OSHAto set and enforce standards that require employers to maintain safe and healthfulworkplaces. NIOSH, on the other hand, does not have any regulatory or enforcementauthority, but it is charged with the responsibility of training professionals and withthe research and recommendation of new regulations to the Secretary of Labor.


Environmental issues affecting the public health and the environment also re-ceived widespread attention. As a result, the Environmental Protection Agency(EPA) was established in 1970 to protect the nation’s public health and environ-ment. The EPA is responsible ‘‘to find ways to clean up and prevent pollution,ensure compliance and enforcement of environmental laws, assist states in environ-mental protection efforts, and scientific research and education to advance the na-tion’s understanding of environmental issues.’’ In 1970, the EPA promulgated theClean Air Act, followed by amendments to the act in 1977 and 1990.

The Toxic Substances Control Act (TSCA), passed in 1976, gave the EPA theability to track and study the 75,000 industrial chemicals produced or imported tothe United States. The TSCA is a federally enforced law and is not delegated tothe states. Under this act, the EPA has the authority to ban the manufacture ordistribution in commerce, limit the use, require labeling, or place other restrictionson chemicals that pose unreasonable risk. Asbestos, chlorofluorocarbons, and poly-chlorinated biphenyls are some of the chemicals regulated by the EPA underTSCA.

In 1977, the International Safe Container Act established uniform structuralrequirements for international cargo containers designed to be transported inter-changeably by sea and land carriers. In 1983, the Surface Transportation AssistanceAct established protection from reprisal by employers for truckers and certain otheremployees in the trucking industry involved in activity related to interstate com-mercial motor vehicle safety and health.

In December 1984, the release of 40 metric tons of methyl isocyanate from apesticide manufacturing plant in Bhopal, India caused the deaths of over 2000people and injuries to another 100,000 [10]. As a direct consequence of this inci-dent, the U.S. Congress in 1986 promulgated the Emergency Planning and Com-munity Right-to-Know Act (EPCRA). The EPCRA requires manufacturers, users,and storage facilities to keep records about quantity, use, and release of hazardousmaterials and make these records available for public record. The EPCRA providedpathways for better understanding of chemical hazards and called for communityemergency response procedures at the local and state levels. Subsequently, theEPCRA led to the formation of Local Emergency Planning Committees (LEPCs)and State Emergency Response Commissions (SERCs). The LEPCs are voluntaryorganizations at the local level and are responsible for developing local emergencyresponse plans in coordination and collaboration with local industry. The SERCs,on the other hand, are state organizations responsible for coordinating the localemergency response plans and administering state programs. EPCRA’s reportingrequirements and emergency planning and notification provisions established acoordinated effort among EPA, state governors, SERCs and LEPCs, owners/opera-tors of regulated facilities, and local fire departments. LEPCs receive chemicalinventory information, analyze the hazards, and develop local emergency responseplans. The LEPCs are responsible for disseminating this information to the publicand serving as a focus for community awareness and action. The SERCs are ap-pointed by the governors and consist of state emergency, environmental, and healthagencies, public interest associations, and others with emergency management ex-perience. The LEPC’s makeup is specified by law, typically consisting of the fol-lowing:


• Representatives of elected state and local officials• Law enforcement officials, civil defense workers, and firefighters• First aid, health, hospital, environmental, and transportation workers• Representatives of community groups and the news media• Owners and operators of industrial plants and other users of chemicals, such

as hospitals, farms, and small businesses

The EPCRA extended right-to-know beyond the workplace and into the commu-nity. This information has stimulated communication between industries and com-munities and encouraged industries to store smaller inventories of hazardous sub-stances, discharge less, and substitute less hazardous chemicals. One majordrawback of this initiative is the unfunded and voluntary nature of the LEPCs. Asa result, LEPCs in many counties are marginally active or do not exist at all.

In 1989, OSHA published recommended Safety and Health Program Guide-lines. These voluntary guidelines identify four general elements that are critical tothe development of a successful safety and health management program. These aremanagement commitment and employee involvement, worksite analysis, hazardprevention and control, and safety and health training.

In 1990, EPA analyzed chemical incidents in the early to mid-1980s and com-pared them to the Bhopal incident. The analysis concluded that 17 incidents re-leased sufficient volumes of chemicals that could have been more severe than Bho-pal if the weather conditions and plant location were different. Thus, the CleanAir Act Amendments of 1990 contained specific mandates requiring OSHA andEPA to establish regulations to protect workplace employees and the public andthe environment, respectively. OSHA fulfilled its mandate in 1992 by promulgatingthe process safety management regulation. The EPA, on the other hand, promul-gated the Risk Management Program regulation in 1996. The Clean Air ActAmendments of 1990 also established the Chemical Safety and Hazard Investiga-tion Board.

State and Local Government Roles

During the late 1800s and early to mid-1900s, the majority of worker-safety lawswere enacted by the state and local governments and thus varied widely in theirextent and enforcement.

By statute, individual states have the option of seeking the delegation of mostfederal safety regulations. The state may request delegation from the federal gov-ernment and submit a state implementation program. The state implementationprogram must, in content and enforcement, be, at the minimum, as stringent asthe federal regulation.

Twenty-three states and two U.S. territories operate their own OSHA-approvedoccupational safety and health programs. These ‘‘State plan States’’ are integralpartners in OSHA’s mission of assuring the safety and health of the nation’s work-ers. They are not required to operate programs identical to those of the federal


OSHA, but they have the flexibility to operate programs that reflect their own state-specific issues and concerns, provided their programs are ‘‘at least as effective’’ asthe federal OSHA program.

States must set job safety and health standards that are ‘‘at least as effectiveas’’ comparable federal standards. Most states adopt standards identical to federalones. States have the option to promulgate standards covering hazards not ad-dressed by federal standards.

A state must conduct inspections to enforce its standards, cover public (stateand local government) employees, and operate occupational safety and health train-ing and education programs. In addition, most states provide free on-site consulta-tion to help employers identify and correct workplace hazards. Such consultationmay be provided either under the plan or through a special agreement under Section21(d) of the act.

To gain OSHA approval for a ‘‘developmental plan,’’ the first step in the stateplan process, a state must assure OSHA that within 3 years, it will have in placeall the structural elements necessary for an effective occupational safety and healthprogram. These elements include appropriate legislation, regulations and proce-dures for standards setting, enforcement, appeal of citations and penalties, and asufficient number of qualified enforcement personnel.

Once a state has completed and documented all its developmental steps, itis eligible for certification. Certification renders no judgment as to actual stateperformance, but merely attests to the structural completeness of the plan.

At any time after initial plan approval, when it appears that the state is capableof independently enforcing standards, OSHA may enter into an ‘‘operational statusagreement’’ with the state. This commits OSHA to suspend the exercise of discre-tionary federal enforcement in all or certain activities covered by the state plan.

The ultimate accreditation of a state’s plan is called ‘‘final approval.’’ WhenOSHA grants final approval to a state under Section 18(e) of the act, it relinquishesits authority to cover occupational safety and health matters covered by the state.After at least 1 year following certification, the state becomes eligible for finalapproval if OSHA determines that it is providing, in actual operation, worker pro-tection ‘‘at least as effective’’ as the protection provided by the federal program.The state also must meet 100% of the established compliance staffing levels(benchmarks) and participate in OSHA’s computerized inspection data system be-fore OSHA can grant final approval.

Rulemaking Process

Regulations by all U.S. federal agencies are developed in a similar manner, withsome minor variations. As an example, the rulemaking process followed by OSHAis discussed here.

The Occupational Safety and Health Administration can begin standards-set-ting procedures on its own initiative or in response to petitions from other parties,including the Secretary of Health and Human Services (HHS), NIOSH, state and


local governments, any nationally recognized standards-producing organization,employer or labor representatives, or any other interested person.

If OSHA determines that a specific standard is needed, any of several advisorycommittees may be called upon to develop specific recommendations. There aretwo standing committees, and ad hoc committees may be appointed to examinespecial areas of concern to OSHA. All advisory committees, standing or ad hoc,must have members representing management, labor, and state agencies, as wellas one or more designees of the Secretary of HHS. The two standing advisorycommittees are as follows:

• National Advisory Committee on Occupational Safety and Health (NACOSH),which advises, consults with, and makes recommendations to the Secretary ofHHS and to the Secretary of Labor on matters regarding administration of theact

• Advisory Committee on Construction Safety and Health, which advises the Sec-retary of Labor on formulation of construction safety and health standards andother regulations

Recommendations for standards also may come from NIOSH, established by theact as an agency of the Department of HHS.

The National Institute for Occupational Safety and Health conducts researchon various safety and health problems, provides technical assistance to OSHA,and recommends standards for OSHA’s adoption. While conducting its research,NIOSH may make workplace investigations, gather testimony from employers andemployees, and require that employers measure and report employee exposure topotentially hazardous materials. NIOSH also may require employers to providemedical examinations and tests to determine the incidence of occupational illnessamong employees. When such examinations and tests are required by NIOSH forresearch purposes, they may be paid for by NIOSH rather than the employer.

Once OSHA has developed plans to propose, amend, or revoke a standard,it publishes these intentions in the Federal Register as a ‘‘Notice of ProposedRulemaking’’ or often as an earlier ‘‘Advance Notice of Proposed Rulemaking.’’

An ‘‘Advance Notice’’ is used, when necessary, to solicit information that canbe used in drafting a proposal. The Notice of Proposed Rulemaking will includethe terms of the new rule and provide a specific time (at least 30 days from thedate of publication, usually 60 days or more) for the public to respond.

Interested parties who submit written arguments and pertinent evidence mayrequest a public hearing on the proposal when none has been announced in thenotice. When such a hearing is requested, OSHA will schedule one, and will pub-lish, in advance, the time and place for it in the Federal Register. After the closeof the comment period and public hearing, if one is held, OSHA must publish inthe Federal Register the full, final text of any standard amended or adopted andthe date it becomes effective, along with an explanation of the standard and thereasons for implementing it. OSHA may also publish a determination that no stan-dard or amendment needs to be issued.

The Occupational Safety and Health Administration continually reviews itsstandards to keep pace with developing and changing industrial technology. There-fore, employers and employees should be aware that, just as they may petition


OSHA for the development of standards, they also may petition OSHA for modifi-cation or revocation of standards.

Technology and Research Base

The National Institute for Occupational Safety and Health was established by theOccupational Safety and Health Act of 1970. NIOSH is part of the Centers forDisease Control and Prevention (CDC) and is the only federal institute responsiblefor conducting research and making recommendations for the prevention of work-related illnesses and injuries. The Institute’s responsibilities include the following:

• Investigating potentially hazardous working conditions as requested by employ-ers or employees

• Evaluating hazards in the workplace, ranging from chemicals to machinery• Creating and disseminating methods for preventing disease, injury, and dis-

ability• Conducting research and providing scientifically valid recommendations for

protecting workers• Providing education and training to individuals preparing for or actively work-

ing in the field of occupational safety and health

Although NIOSH and OSHA were created by the same act of Congress, they aretwo distinct agencies with separate responsibilities. OSHA is in the Departmentof Labor and is responsible for creating and enforcing workplace safety and healthregulations. NIOSH is in the Department of Health and Human Services and is aresearch agency.

The National Institute for Occupational Safety and Health identifies the causesof work-related diseases and injuries and the potential hazards of new work tech-nologies and practices. With this information, NIOSH determines new and effec-tive ways to protect workers from chemicals, machinery, and hazardous workingconditions. Creating new ways to prevent workplace hazards is the job of NIOSH.

In 1980, the U.S. Congress created the Agency for Toxic Substances and Dis-ease Registry (ATSDR) to implement the health-related sections of laws that pro-tect the public from hazardous wastes and environmental spills of hazardous sub-stances. ATSDR is charged with assessing the presence and nature of healthhazards at specific sites, to help prevent or reduce further exposure and the illnessesthat result from such exposures, and to expand the knowledge base about healtheffects from exposure to hazardous substances.

In 1984, amendments to the Resource Conservation and Recovery Act of 1976(RCRA), which provides for the management of legitimate hazardous waste stor-age or destruction facilities, authorized ATSDR to conduct public health assess-ments at these sites, when requested by the EPA, states, or individuals. ATSDRwas also authorized to assist the EPA in determining which substances should beregulated and the levels at which substances may pose a threat to human health.

With the passage of the Superfund Amendments and Reauthorization Act of


1986 (SARA), ATSDR received additional responsibilities in environmental publichealth. This act broadened ATSDR’s responsibilities in the areas of public healthassessments, establishment and maintenance of toxicological databases, informa-tion dissemination, and medical education.

The Process Safety Management Program

The 14 elements of the OSHA Process Safety Management (PSM) regulation (29CFR 1910.119) were published in the Federal Register on February 24, 1992 [11].The objective of the regulation is to prevent or minimize the consequences ofcatastrophic releases of toxic, reactive, flammable, or explosive chemicals. Theregulation requires a comprehensive management program: a holistic approach thatintegrates technologies, procedures, and management practices.

The process safety management regulation applies to processes which involvecertain specified chemicals at or above threshold quantities, processes which in-volve flammable liquids or gases on-site in one location, in quantities of 10,000lbs. or more (subject to few exceptions), and processes which involve the manufac-ture of explosives and pyrotechnics. Hydrocarbon fuels, which may be excludedif used solely as a fuel, are included if the fuel is part of a process covered by thisregulation. In addition, the regulation does not apply to retail facilities, oil or gaswell drilling or servicing operations, or normally unoccupied remote facilities.

The management system required by OSHA’s process safety management reg-ulation envisions a holistic program with checks and balances aimed at creatingmultiple barriers of protection. This principle is shown in Fig. 2. The performance-based approach does not prescribe specific methods and approaches, thus givingfacilities the flexibility for customizing the methods to best meet their needs andorganizational structures. The process hazard analysis (PHA) is the heart of the

FIG. 2 Holistic implementation of the process safety management program.


program and impacts or interfaces with all of the other elements. However, it mustalso be pointed out that all elements of the program must be implemented in theirentirety to get the maximum benefit and accomplish the ultimate objective (i.e.,reduce the frequency and severity of chemical plant accidents). Some of the otherconcepts that are apparent from Fig. 2 are as follows:

1. The process safety information and incident history are important inputs tothe PHA and must be compiled before the PHA.

2. Employee participation is important not only for the whole program but alsoprovides critical information during the PHA.

3. Results of the PHA should be used in modifying and or developing operatingprocedures, mechanical integrity program, emergency response program, andother impacted elements.

4. Irrespective of the PHA done earlier, each process change should be evaluatedby the management of change program, and, if necessary, an appropriate haz-ard analysis should be done.

5. Pre-start-up safety review is an essential procedure for new or modified pro-cesses.

Each element of the process safety program is discussed in more detail here andthe interface with other elements of the process safety management is discussed.

Employee Participation

This element of the regulation requires developing a written plan of action regard-ing employee participation, consulting with employees and their representativeson the conduct and development of other elements of process safety managementrequired under the regulation, and providing to employees and their representativesaccess to process hazard analyses and to all other information required to be devel-oped under this regulation.

Process Safety Information

This element of the PSM regulation requires employers to develop and maintainimportant information about the different processes involved. This information isintended to provide a foundation for identifying and understanding potential haz-ards involved in the process.

The process safety information covers three different areas (i.e., chemicals,technology, and equipment). A complete listing of the process safety informationthat must be compiled in these three areas is shown in Table 1. This informationis intended to provide a foundation for identifying and understanding potentialhazards involved in the process.

The information in Table 1 is essential for developing and implementing aneffective process safety management program. The fundamental concept is thatcomplete, accurate, and up-to-date process knowledge is essential for safe andprofitable operations. The information contained in the first column of Table 1should be available from the Material Safety Data Sheets (MSDS) for the hazard-


TABLE 1 Process Safety Information

Chemicals Technology Equipment

Toxicity Block flow diagram or pro- Design codes employedcess flow diagram

Permissible exposure limit Process chemistry Materials of constructionPhysical data Maximum intended inven- Piping and instrumentation

tory DiagramsReactivity data Safe limits for process pa- Electrical classification

rametersThermal and chemical sta- Consequence of deviations Ventilation system design

bility data Material and energy bal-Effects of mixing ances

Safety systemsRelief system design and de-

sign basis

ous chemicals, which are used as primary or intermediate feedstocks or are pro-duced as products at the plant.

The information contained in the second column of Table 1 pertains to thetechnology of the process itself. The block flow diagram can be replaced by aprocess flow diagram. The process chemistry information must contain the basicchemical reactions involved and a brief description of the chemistry involved. Themaximum intended inventory refers to the maximum amount of any regulatedchemical that may be expected to be present in the whole facility at any time. Thesafe limits for process parameters refer to the upper and lower bounds for theprocess parameters outside of which the process would be hazardous. For example,in the case of a distillation process, the upper and lower limits of the processparameters outside which the operation of the process could cause significant dam-age to the tower or other attached equipment would have to be stated. In thisexample, the process parameters for which upper and lower bounds are to be speci-fied are temperature, pressure, composition, and flow rate. The consequence ofdeviation from these stated bounds must also be compiled.

Safe upper and lower limits for process parameters and equipment are alsonecessary for calibrating instrumentation. It is important to understand the distinc-tion between process parameter limits and equipment limits. Safe upper and lowerlimits for process parameters can be defined as follows:

• For nonreactive processes, process parameter safe limits are defined based onequipment design ratings, relief device set points, and upstream and downstreamconditions of the process and/or equipment.

• For reactive processes, process parameter safe limits are defined based on theprocess chemistry information or any restricted physical conditions for the reac-tion as well as the criteria used for nonreactive processes.

In contrast, equipment limits are specified by the manufacturer based on materialsof construction and design basis. Processes must be operated and maintained withinboth the process parameter and equipment limits.


The final type of information that must be compiled pertains to the equipmentused in the process. The intent is to assure that all equipment used in the processmeets appropriate standards and codes such as those published by the AmericanSociety of Mechanical Engineers (ASME), the American Petroleum Institute (API),the American Institute of Chemical Engineers (AIChE), the American National Stan-dards Institute (ANSI), the American Society of Testing and Materials (ASTM), andthe National Fire Protection Association (NFPA). Accepted industry practices canbe used to decide which standards apply. In cases where standards do not exist,generally acceptable engineering practices can be used. The materials of constructionfor each equipment item and the design codes employed can be compiled as separatelists or may be listed in the Piping and Instrumentation Diagram (P&ID). The P&IDs must represent the facility exactly as it exists with flanges, valves, and all otherconnections shown. The different electrical classifications must also be compiled fordifferent parts of the facility. A simple plot plan showing the different areas ofelectrical classification would be considered to be in compliance with the regulation.Information must also be compiled on any ventilation system. This informationwould indicate the areas in the facility that are ventilated and the nature of ventilation.A listing of all safety systems must also be compiled that are available to the workers.This listing should include any and all equipment that is available for protection ofthe workers from any hazard or emergency. Information should also be availableon the location of these safety systems and the procedures to use these systems.

As is apparent from the foregoing discussion, compilation of the process safetyinformation database represents a major challenge. This is complicated even morebecause the process safety information must also be kept up-to-date and accurateand made accessible to employees. Many plants have therefore implemented orare in the process of implementing electronic data management systems to manage,access, and use these data. The completion and accuracy of process safety informa-tion is crucial to the implementation of other PSM elements, including PHAs andmechanical integrity.

The PSM regulation requires that process equipment should comply with gen-erally accepted engineering practices. It is therefore not only important to compileall equipment information but also to ensure that it complies with consensus stan-dards. OSHA is recognizing that there are consensus standards for design andfabrication, installation, maintenance procedures, and inspection and testing.Therefore, equipment constructed in accordance with codes, standards, or practicesno longer in use should be evaluated to ensure that they are compatible with ex-isting standards. For example, a multilayered vessel built years ago should be eval-uated to ensure that it complies with today’s engineering standards.

Process Hazards Analysis

This element of the PSM regulation requires facilities to perform a PHA. The PHAmust address the hazards of the process, previous hazardous incidents, engineeringand administrative controls, the consequences of the failure of engineering andadministrative controls, human factors, and an evaluation of effects of failure ofcontrols on employees. This element requires that the PHA be performed by oneor more of the following methods or any other equivalent method:


• What If• Checklist• What If/checklist• Hazard and operability (HAZOP) studies• Failure modes and effects analysis (FMEA)• Fault-tree analysis

The regulation suggests a performance-oriented requirement with respect tothe PHA so that the facility has the flexibility to choose the type of analysis thatwill best address a particular process. PHAs may not be performed unless completeProcess Safety Information is available for the process.

Process hazard analyses and the results of PHAs can and will impact the devel-opment, implementation, and practice of other elements of the PSM regulation. APHA would help facilities identify hazards and ways to address them. For example,a 1989 explosion and fire at a facility in Baton Rouge, Louisiana led to a partialloss of pressure, power, and fire water because the power, steam, and water lineswere colocated with the lines carrying flammable gases [12]. The losses compli-cated and prolonged the process of responding to the release, thereby increasingthe damage caused by the release. Similar problems occurred at a facility in Norco,Louisiana, where an explosion led to the loss of all utilities. A thorough and prop-erly done PHA should identify these types of potential hazards and allow facilitiesto determine how to mitigate the problems. PHAs also identify situations wheremajor accidents due to control failure (e.g., pressure gauges, overfill alarms) couldbe prevented by redundant or backup control or by frequent maintenance and in-spection practices.

Many other elements of the PSM program should flow from, or at least berevised based on, the results of the PHA. Existing standard operating procedures,training and maintenance programs, and pre-start-up safety reviews may need tobe revised to reflect changes in either practices or equipment that derive fromthe PHA. The PHA may help define critical equipment that require preventivemaintenance, inspection, and testing programs. It may also help a facility focusits emergency response programs on the most likely and most serious release sce-narios.

The PSM regulation also requires that the PHAs be updated and revalidated,based on their completion date. First, the PSM regulation requires that the processhazard analysis shall be updated and revalidated. Thus, the whole PHA should notonly be examined to verify that the PHA is consistent with the current process,but it also should be analyzed to examine if the original analysis is valid. Second,the PSM regulation requires that the updating and revalidating be done by a teamsimilar in qualifications to the team that conducted the original PHA. The intentof a PHA revalidation is to ensure that the PHA team evaluates the previous PHA,examines the extent of any changes that might have occurred since the PHA wasimplemented (or last reviewed), and decides what work is needed to make thePHA current.

Philosophically, in an ideal process safety management system, the followingprocess safety concepts should work as checks and balances to accomplish thePSM program objectives (i.e., minimize the frequency and consequences of cata-strophic releases):


• Initial process hazards analysis: Identify process hazards and develop mitigationtechniques (technology, equipment, and procedures).

• Management of change: Assess the safety and health impacts of process changesand ensure that process changes are analyzed (e.g., HAZOPed).

• Revalidation of process hazards analysis: Update and revalidate the PHAs toassure that the process hazard analysis is consistent with the current process.

Operating Procedures

The operating procedures must be in writing and provide clear instructions forsafely operating processes, must include steps for each operating phase, operatinglimits, safety and health considerations, and safety systems. Procedures must bereadily accessible to employees, must be reviewed as often as necessary to assurethey are up-to-date, and must cover special circumstances such as lockout/tagoutand confined space entry. The employer must certify annually that the operatingprocedures are current and accurate. The synergism and commonality of operatingprocedures to maintenance procedures is in safe work practices. These safe workpractices include lockout/tagout, confined space entry, opening of process equip-ment or piping (i.e., hot tapping), and control of entrance into the battery limits.Even though the operations department is involved with all of these practices,maintenance also plays a very vital role in ensuring that these procedures are fol-lowed during all maintenance tasks. Many incidents have resulted from inadequatesafe work practices or a failure to follow procedures when they exist.

Training

The regulation requires that facilities certify that employees responsible for op-erating the facility have successfully completed (including means to verify under-standing) the required training. The training must cover specific safety and healthhazards, emergency operations, and safe work practices. Initial training must occurbefore assignment. Refresher training must be provided at least every 3 years.

Even though this element of the PSM regulation pertains to operations staffonly, it is important to remember that operations and maintenance training shouldbe coordinated.

Contractors

The PSM regulation identifies responsibilities of the employer regarding contrac-tors involved in maintenance, repair, turnaround, major renovation or specialtywork, on or near covered processes. The host employer is required to considersafety records in selecting contractors, inform contractors of potential process haz-ards, explain the facility’s emergency action plan, develop safe work practices forcontractors in process areas, periodically evaluate contractor safety performance,and maintain an injury/illness log for contractors working in process areas. Inaddition, the contract employer is required to train their employees in safe work


practices and document that training, assure that employees know about potentialprocess hazards and the host employer’s emergency action plan, assure that em-ployees follow safety rules of facility, and advise host employer of hazards contractwork itself poses or hazards identified by contract employees.

In the contractor paragraph, OSHA has used a belt and suspender approach.Both the host employer and contract employer have specific responsibilities thatthey must fulfill. The need for flexibility, quick turnarounds, and specialized ser-vices is the main reason why process plants are contracting out increasingly sig-nificant portion of their daily work, particularly maintenance work to contractors.Contractors, in general, receive less training and often perform more hazardoustasks in process plants as compared to direct-hire workers [13].

Pre-Start-up Safety Review

This element of the PSM regulation requires a pre-start-up safety review of allnew and modified facilities to confirm integrity of equipment, to assure that appro-priate safety, operating, maintenance, and emergency procedures are in place, andto verify that a process hazard analysis has been performed. Modified facilities forthis purpose are defined as those for which the modification required a change inthe process safety information.

Usually, changes occur during maintenance and, therefore, maintenance per-sonnel should be well versed in pre-start-up safety review procedures. Maintenanceshould ensure that all necessary procedures have been completed prior to start-up.

Mechanical Integrity

This element of the PSM regulation mandates written procedures, training for pro-cess maintenance employees, and inspection and testing for process equipment,including pressure vessels and storage tanks, piping systems, relief and vent sys-tems and devices, emergency shutdown systems, pumps, and controls such as mon-itoring devices, sensors, alarms, and interlocks. PSM calls for correction of equip-ment deficiencies and assurance that new equipment and maintenance materialsand spare parts are suitable for the process and properly installed.

Hot Work Permit

This element of the PSM regulation mandates a permit system for hot work opera-tions conducted on or near a covered process. The purpose of this element of theregulation is to assure that the employer is aware of the hot work being performedand that appropriate safety precautions have been taken prior to beginning the work.Because welding shops authorized by the employer are locations specifically desig-nated and suited for hot work operations, the regulation does not require a permitfor hot work in these locations. Additionally, hot work permits are not required incases where the employer or an individual to whom the employer has assigned theauthority to grant hot work permits is present while the hot work is performed.


Management of Change

This element of the regulation specifies a written program to manage changes inchemicals, technology, equipment, and procedures which addresses the technicalbasis for the change, impact of the change on safety and health, modification tooperating procedures, time period necessary for the change, and authorization re-quirements for the change. The regulation requires employers to notify and trainaffected employees and update process safety information and operating proce-dures as necessary.

Incident Investigation

This element of the regulation requires employers to investigate as soon as possible(but no later than 48 hrs) incidents which did result or could have resulted incatastrophic releases of covered chemicals. The regulation calls for an investigationteam, including at least one person knowledgeable in the process (a contractoremployee, if appropriate), to develop a written report of the incident. Employersmust address and document their response to report findings and recommendationsand review findings with affected employees and contractor employees. Reportsmust be retained for 5 years.

Emergency Planning and Response

This element requires employers to develop and implement an emergency actionplan according to the requirements of 29 CFR 1910.38(a) and 29 CFR 1910.120(a),(p), and (q).

Compliance Audits

This element of the regulation requires employers to certify that they have evalu-ated compliance with process safety requirements every 3 years and specifies reten-tion of the audit report findings and the employer’s response. The employer mustretain the two most recent audits.

Trade Secrets

Similar to the trade secret provisions of the hazard communication regulation, thePSM regulation also requires information to be available to employees from theprocess hazard analyses and other documents required by the regulation. The regu-lation permits employers to enter into confidentiality agreements to prevent disclo-sure of trade secrets.

As is apparent from the foregoing discussion, the process safety managementregulation requires a systems approach for managing safety. Segments of the haz-ardous chemicals industry have for sometime practiced some or all of the requiredprograms. The promulgation of the regulation formalized the requirements and


established minimum criteria. This is both good and bad. The regulation now re-quires everyone to establish the management systems and apply the technologiesneeded to comply with the regulation. However, because of the same reason, thereis a tendency to look for ‘‘paper compliance’’ as compared to making real improve-ments in safety programs and technologies.

The Risk Management Program

In 1996, EPA promulgated the regulation for Risk Management Programs for Chem-ical Accident Release Prevention (40 CFR 68). This federal regulation was mandatedby Section 112(r) of the Clean Air Act Amendments of 1990. The regulation requiresregulated facilities to develop and implement appropriate risk management programsto minimize the frequency and severity of chemical plant accidents. In keeping withregulatory trends, the EPA required a performance-based approach toward compli-ance with the risk management program regulation. The eligibility criteria and re-quirements for the three different program levels are given in Table 2.

The EPA regulation also requires regulated facilities to develop a Risk Manage-ment Plan (RMP). The RMP includes a description of the hazard assessment, pre-vention program, and the emergency response program. Facilities submit the RMPto the EPA and, subsequently, it is made available to governmental agencies, thestate emergency response commission, and the local emergency planning commit-tees and is communicated to the public.

The RMP regulation defines the worst-case release as the release of the largestquantity of a regulated substance from a vessel or process-line failure, includingadministrative controls and passive mitigation that limit the total quantity involvedor release rate. For gases, the worst-case release scenario assumes the quantity isreleased in 10 min. For liquids, the scenario assumes an instantaneous spill andthat the release rate to the air is the volatilization rate from a pool 1 cm deep unlesspassive mitigation systems contain the substance in a smaller area. For flammables,the scenario assumes an instantaneous release and a vapor cloud explosion using a10% yield factor. For alternative scenarios (note: the EPA used the term alternativescenario as compared to the term more-likely scenario used earlier in the proposedregulation), facilities may take credit for both passive and active mitigation systems.

Appendix A of the final regulation lists endpoints for toxic substances to beused in worst-case and alternative scenario assessment. The toxic endpoints arebased on ERPG-2 (Emergency Response Planning Guidelines—Level 2) or levelof concern data compiled by EPA. The flammable endpoints represent vapor cloudexplosion distances based on overpressure of 1 psi or radiant heat distances basedon exposure to 5 kW/m2 for 40 s.

Impact on Process Industry

In general, the impact of any regulation on any part of the regulated communitycan be related to several factors which include, but are not limited to the following:


TABLE 2 Eligibility Criteria and Compliance Requirements for Different ProgramLevels; EPA’s Risk Management Program Regulation

Program 1 Program 2 Program 3

Program Eligibility CriteriaNo off-site accident history Process not eligible for pro- Process is subject to OSHA

gram 1 or 3 PSM (29 CFR1910.119)

No public receptors inworst-case circle

Emergency response coordi- Process is SIC code 2611,nated with local re- 2812, 2819, 2821,sponder 2865, 2869, 2873,

2879, or 2911Program Requirement

Hazard Assessment Hazard Assessment Hazard AssessmentWorst-case analysis Worst-case analysis Worst-case analysis5-Year accident history Alternative releases Alternative releasesCertify no additional 5-Year accident history 5-Year accident history

steps neededManagement Program Management Program

Document management Document managementsystem system

Prevention Program Prevention ProgramSafety information Process safety informa-

tionHazard reviewProcess hazard analysisOperating proceduresOperating proceduresTrainingTrainingMaintenanceMechanical integrityIncident investigationIncident investigationCompliance auditCompliance auditManagement of changePre-start-up safety reviewContractorsEmployee participationHot work permits

Emergency Response Pro- Emergency Response Pro-gram gram

Develop plan and pro- Develop plan and pro-gram gram

• The degree to which a particular entity or entities is already attempting to reacha particular level of achievement

• The additional cost associated with regulatory overhead not directly associatedwith the goals of the program (e.g., the cost of record keeping)

• The degree to which already successful programs are compromised by the regu-lation


• The technical feasibility of compliance• The impact on compliance with other regulations• The emphasis on compliance rather than impact• The ability to maintain a profitable operation• The ability to absorb the associated overhead cost when offshore competition

does not have those costs• The degree to which enforcement and other emphasis by the regulating authority

promotes real improvement as opposed to ‘‘paperwork’’ compliance

It must also be noted that painting all industry or even one segment of industrywith one brush is very inappropriate. Many companies and/or groups of companiesare and have been for many years investing considerable time and resources intoprocess safety and risk management prior to any regulatory requirement to do so.Also, many companies and industry groups have initiated voluntary programswhich go further than regulations require. In some cases, these efforts have beenin partnership with government and nongovernment groups and in other cases theyhave been independent of government participation.

Compliance Programs

At first glance, it would seem that ‘‘compliance’’ must imply compliance withregulations. However, there are many examples of responsible industry doingsomething because it was ‘‘the right thing to do.’’ Following the Texas City disas-ter in 1949, the industries on the Houston Ship Channel voluntarily formed ChannelIndustries Mutual Aid (CIMA) in order to help each other and the public duringemergencies. There were no regulatory drivers. Responsible chemical companieshave been responding to transportation emergencies involving their products forat least 50 years. Safety programs for worker protection were in place years beforeOSHA. For example, dating back to the 1960s and 1970s, Dow Chemical Companyhas had an excellent reactive chemicals program. A positive result of regulatoryaction has been that companies who have not voluntarily invested in these kindsof programs are forced to make the investment. Unfortunately, when compliancewith regulation becomes the issue, administrative overhead may increase signifi-cantly. Furthermore, a one-size-fits-all approach to regulation and compliance mayforce a company to abandon a program that has proven successful over time simplyto meet a regulation. It is very difficult to explain to management and workersthat a highly successful, popular program has to be changed or eliminated becauseof a rigidly written regulation.

Many companies had employed some form of lockout/tagout procedure foryears that worked. The one size fits all regulatory mentality caused those programsto be changed at great cost and effort. It can be argued that the cost was not justifiedby the result. The changed requirements also caused confusion in the case of someworkers, which might actually have increased the risk. The good news is that thosesame responsible companies extend their policies to their facilities wherever theymay be. For example, members of the American Chemistry Council apply theResponsible Care Code internationally.


Finally, when the ultimate goal becomes compliance rather than results, it ispossible to be in technical compliance and not in functional compliance. For exam-ple, several persons both in and out of government during work on the NationalResponse Team’s integrated contingency plan effort commented that in many casesfacility and vessel owners hired contractors to produce response plans under theOil Pollution Act of 1990 that were judged by government contractors to meet allplanning requirements but were not useful for response. In other words, compli-ance, not planning and response, was the goal.

Enforcement

Enforcement can be a two-edged sword. As mentioned earlier, compliance issuescan sometimes outweigh the issues of functionality or statutory and regulatoryintent. Very often, with both government and industry, the issue becomes ‘‘didyou meet the requirements’’ not ‘‘have you increased or reduced the risk.’’ At thesame time, it must also be said that in some cases, the threat of enforcement isthe only leverage regulators have. Sometimes, however, the threat of enforcementmay actually serve to increase overall risk. There is currently a great deal of con-cern that efforts to mix enforcement and response at the incident command level inoil spills may discourage the cooperative spirit necessary for an effective response.Overall, the availability of enforcement has probably had a positive impact onprocess safety and risk management. The threat of enforcement has probably hada chilling effect on cooperative efforts that could possibly make an even largerimpact.

Operational Issues

Operationally, regulations have had a net positive impact in that they have forcedmany companies not already committed to environmental and safety excellenceto operate better. In many cases, this improvement has actually been transferredto the bottom line. Improved operating discipline actually will increase quality,productivity, morale, and product yield. There are occasional situations, however,that require procedures that may cause significant problems. For example, therehas never been a catastrophic explosion of an underground fuel tank. That is whymany fire codes prohibit aboveground tanks at retail facilities. The undergroundstorage tank regulations have made it very expensive to maintain and install under-ground tanks. In this case, there is an obvious conflict between two competinginterests. Some air regulations require inspection of the seals on floating roof tanks.It is very difficult to clear those tanks well enough to eliminate emissions during theinspection process. In this case, one regulation may result in violation of another oran actual increase in emissions. Elimination of chlorofluorocarbons has caused aconcurrent elimination of Halon-type fire-extinguishing systems. Another exampleis the requirement to test marine facility foam systems by actually flowing foam.These tests have a high likelihood of discharging some foam to the water—a viola-tion of the Clean Water Act.


Management Systems for Compliance

Management systems for regulatory compliance have evolved over a period oftime. In the early stages, there was a tendency to manage process safety as a sepa-rate function. With experience, industry has learned that is not the most efficientor effective approach. In order to gain the most advantage, safety must be inte-grated into the plant’s everyday activities through each phase of the plant’s lifecycle. The following provides a discussion of some example components of themanagement systems necessary for the effective implementation of process safetyand risk management programs.

Integrated Systems Versus Compartmentalized Systems

Management systems for compliance must be integrated and cannot be compartmen-talized. First, although statutes and the accompanying regulations are often con-ceived and written as separate entities, they often have impacts on other areas thatmay or may not be intentional. One example can be found in the list of chemicalsregulated under the Clean Air Act Amendments that included ethylene glycol. Underthe reportable quantity (RQ) regulations of the Superfund Amendments and Reautho-rization Act, ethylene glycol defaulted to a 1-lb RQ. This had the effect of creatinga large number of unnecessary and wasteful emergency release reports from peoplewho understood the regulations and in one sense made criminals out of almost every-one who had a radiator leak. Only an integrated system could respond to that quirk.The real thrust of integrated management systems must be to include all safety andenvironmental concerns in the entire process from conception to design to construc-tion and commissioning to start-up and operation to shutdown and demolition. Onlythis approach can insure compliance, but more importantly safety and environmentalexcellence. Individuals in the organization must understand their respective roles.Therefore, the management system for compliance should be integrated verticallywithin the organization and horizontally across regulatory regimes.

Process Design

It is critical that companies have a system in place that assures compliance fromthe time of conception to eventual start-up of any process. There are multipleregulations that impact every facet of a new process. The process design mustlook at air emissions and permitting, wastewater and stormwater handling, wastehandling, generation and reduction, Toxic Substance Control Act implications,Risk Management Plan implications, and other potential areas of concern. Workersafety, including repetitive motion injuries, should be an issue from the start. Softissues such as public perception must be considered. All of these issues also applyto modifications to existing facilities. They may also apply to changes in operatingprocedures and other ‘‘soft’’ changes. For all of these reasons and for OSHA com-pliance, a well-conceived management of change (MOC) system must be inte-grated into the culture.


Very subtle, often seemingly innocuous changes can have significant regulatoryimplications. An operator of a water-treatment facility may have several 1-tonchlorine cylinders on site but never have more than one connected to the process.That operator may decide that the greatest risk occurs when changing the cylinderout. The obvious solution would be to manifold several cylinders together. Thatchange would then make the operator subject to the provisions of Section 112r ofthe Clean Air Act.

Equipment Maintenance

For all of the reasons listed earlier, equipment maintenance must be part of themanagement system. It is possible that regulatory issues may actually discouragea good maintenance program. Many maintenance activities, especially in olderplants, require opening processes to the atmosphere in order to make the equipmentsafe to work. Air regulations may prohibit, or at least discourage, this activity.There is no way of quantifying how often needed maintenance is delayed or can-celed because of environmental concerns. Because of this, a relatively small emis-sion or environmental upset resulting from maintenance was avoided at the costof an eventual catastrophic event.

Prevention and Reduction

Responsible industry has always had prevention and reduction programs. As statedearlier, regulations have largely served to ‘‘level the playing field’’ between responsi-ble and irresponsible industry. As early as the 1960s, one chemical company execu-tive stated on the ‘‘Today’’ show that his company was in the business of makingsalable product out of raw materials, and if a molecule left the plant as waste to theenvironment, it represented lost profit. Further, in recent years, all companies havebegun to recognize that they must be good neighbors by perception as well as byreality. Initiatives such as Responsible Care illustrate this recognition.

Response Mechanism

Again, responsible companies have had response systems in place for many years.Several things have occurred to alter and, in most cases, improve those systems.The OSHA PSM regulation and the Hazardous Waste Operations regulation(OSHA 1910.120) have elevated the training requirements for responders. Al-though experienced responders already had most of the required training and/orexperience, the regulations created minimums that have been very effective. Plan-ning requirements under a multitude of federal and state regulations have, howeverresulted in considerable duplication. PSM, RMP, Community Right to Know,OPA90, RCRA, and others all mandate response planning. The vessel responseplan requirements proposed by the U.S. Coast Guard will add still further require-ments. This duplicative regulatory system resulted in numerous plans designedsolely to meet compliance requirements and often had little to do with response.The good news is that the effort of the National Response Team to publish the


Integrated Contingency Planning Guidance has relieved much of that problem. Inthe case of OPA90 in particular, oil spill response is much better than it was duringthe Exxon Valdez spill. The most lasting improvement has been in the responsepartnerships formed between industry and government at all levels.

Risk Communication

For many years, much of industry did a fairly poor job of telling people aboutpotential risks. This has been significantly improved by Community Right toKnow, 112r of the Clean Air Act, and voluntary programs such as ResponsibleCare. Much remains to be done in this area. Industry and government tend torespond to public perception of risk without doing a good job of education abouttrue risk. Very often, uninformed public outcry has forced questionable, if nottotally incorrect, reaction on the part of industry and government. All too often,groups with hidden agendas have aggravated this process. Fears about dioxins thathave not been scientifically confirmed have resulted in actions like Times Beach;this is in spite of the record of Seveso in which there are still no documented long-term effects of the exposure. The challenge is creating an interactive dialogue withthe public to a point where risk perception is based not only on emotion but alsoon some level of scientific reality. Zero risk is a myth. However, zero negativeimpact is a vision for which industry and government must strive.

Small Business Issues

In the United States, a significant portion of the economy consists of small and mid-sized companies. We must remember that an accident from such a small facility hasthe likelihood of severe consequences and can damage the whole industry ‘‘licenseto operate’’ just like an accident in a plant operated by a large multinational com-pany. However, the large facility probably has resources, training, and equipmenteither to prevent the accident in the first place or respond to the consequences if itdoes occur. On the other hand, the small facility probably lacks awareness, training,information, and other things. A recent accident in the United States resulted inmultiple fatalities and total destruction of a small plant. Preliminary investigationshighlight the lack of reactive chemicals knowledge among the plant personnel.The challenge therefore is to develop collaborative efforts with government agen-cies (both state and federal), professional and trade organizations, and industry forsafety programs aimed at improving safety in the small and mid-sized companies.

Future Developments

In an increasingly global economy in which the developed nations will competeeven more strongly with the less developed nations, several things must happen.


First, multinational companies must insist that the best practices of worker, public,and environmental protection be followed throughout the world. Product steward-ship is one effective method for ensuring the implementation of the same standardsat small and medium-sized businesses in the United States as well as in facilitiesoverseas. Governments should also promote those best practices. Industries andgovernments not doing so should face sanctions in the international marketplace.It is imperative that those companies trying to do what is right are not penalizedby unfair competition. Partnerships between key players in this arena should beencouraged and rewarded. Examples include Integrated Contingency Planning andthe OSHA Voluntary Protection Program. The public should be represented inthese programs. All of the stakeholders involved in these issues must work toidentify and eliminate barriers to improvement. These barriers and issues include,but are not limited to the following:

• Tort regimes that discourage sharing of lessons learned and near-miss informa-tion

• Regulations and requirements that are duplicative or not based on science• Public policy that results in ‘‘knee-jerk’’ statutory and regulatory response to

single events• Companies and industry groups that are willing to conduct meaningful dialogue

with stakeholders• Most importantly, a willingness to accept and work toward a vision in which

no facility has a negative impact on its workers, the public, or the environment.

Future developments in the United States with regard to process safety and riskmanagement programs may quite likely be based on risk–benefit analyses. Thereis also number of efforts underway to develop stakeholder dialogue and arrive atconsensus opinions regarding national safety goals and targeted improvementsin safety performance. It is quite clear that the need to operate safely is recog-nized as a competitive advantage and a positive contributor to the bottom line.The regulatory regime and requirements will also keep changing as more infor-mation becomes available. Thus, industrial programs and practices will have tokeep pace with the changing clime and consensus standards and targeted safetygoals.

Summary and Conclusions

Safety regulations in the United States have mirrored the industrial revolution. Theindustrial revolution brought prosperity along with the use of hazardous processes.As our understanding of the hazards associated with these processes developed,procedures and practices were put in place to limit or eliminate the damage. Gov-ernment programs and industry initiatives spurred improvements in the scienceand technology needed for the recognition of hazards and associated risks. Manage-ment systems have been developed to implement safety programs and industrypractices.


Government regulations continue to be a significant driver for safety programs.As such, one of the main objectives of the management systems is to ensure com-pliance. However, it is also quite clear that profitability is directly related to safetyand loss prevention. Thus, the management systems for safety are intricately tiedinto the operational management. It is also quite apparent that government regula-tions alone cannot accomplish our ultimate safety goals. Safety must be tied intoprofitability and business objectives. Also, the use of risk assessments and risk–benefit analysis will play a significant role in the future safety programs.

References

1. U.S. Environmental Protection Agency, Accidental Release Information Program,EPA, Washington, DC, 1989.

2. M. Mannan, A. Akgerman, R. G. Anthony, R. Darby, P. T. Eubank, and K. R. Hall,‘‘New Challenges in Chemical Engineering: Integrating Process Safety into ChemicalEngineering Education and Research,’’ Chem. Eng. Ed., 33(3), 198–209 (1999).

3. U.S. Environmental Protection Agency, A Review of Federal Authorities for Hazard-ous Materials Accident Safety: Report to Congress Section 112(r)(10) Clean Air ActAs Amended, EPA, Washington, DC, 1993.

4. National Environmental Law Center and United States Public Interest ResearchGroup, Accidents Do Happen: Toxic Chemical Accident Patterns in the United States,United States Public Interest Research Group, Washington, DC, 1994.

5. Code of Federal Register, 40 CFR 372, U.S. Environmental Protection Agency, ToxicChemical Release Reporting, Community Right-to-Know, FDA, Washington, DC,June 26, 1991.

6. T. A. Kletz, Lessons from Disaster, Gulf Publishing Company, Houston, TX, 1993.7. Skandia International, Bleve—The Tragedy of San Juanico, Skandia International,

Stockholm, 1985.8. W. D. Cullen, The Public Inquiry into the Piper Alpha Disaster, Her Majesty’s Statio-

nery Office, London, 1990.9. U.S. Department of Labor, The Phillips 66 Company Houston Chemical Complex

Explosion and Fire, U.S. Department of Labor, Washington, DC, 1990.10. P. Shrivastava, Bhopal—Anatomy of a Crisis, Ballinger, Cambridge, MA, 1987.11. ‘‘Process Safety Management of Highly Hazardous Chemicals; Explosives and Blast-

ing Agents; Final Rule, 29 CFR Part 1910, Department of Labor, Occupational Safetyand Health Administration, Washington, DC, February 24, 1992,’’ Fed. Reg., 57(36),6356–6417 (1992).

12. Risk Management Programs for Chemical Accidental Release Prevention; ProposedRule; 40 CFR Part 68, Environmental Protection Agency, Washington, DC, October20, 1993, Fed. Reg., 58(201), 54,190–54,219 (1993).

13. John Gray Institute, Managing Workplace Safety and Health: The Case of ContractLabor in the U.S. Petrochemical Industry, John Gray Institute, Lamar University Sys-tem, 1991.

M. SAM MANNANJIM MAKRIS

H. JAMES OVERMAN

194 Recent Advances in Biomaterials

Recent Advances in Biomaterials

A vast number of materials (collectively termed biomaterials) are used in clinicalapplications, from drug delivery (capsules, transdermal patches) to hip replace-ments to materials in hemodialysis machines to heart valves. These materials havehad a tremendous impact on patients’ lives, allowing many injured or ill peopleto lead a near-normal life in many instances. As one would imagine, these biomate-rials represent a significant health care cost. The annual sales of medical devices,diagnostics, and pharmaceutical products using biomaterials in some capacity ex-ceed $100 billion per year in the United States alone. With new materials anddevices being developed to treat disease and injury, the market for biomaterialsis likely to increase significantly in the next decade.

Specifically, a biomaterial can be classified as a material such as a polymer,metal, or ceramic that is in intimate contact with a biological environment suchas blood, tissue, or individual cells. The biocompatibility of a biomaterial is definedas the ability of a material to function with a specific host response. For example,a blood contacting material should not cause thrombosis or complement activation,and a hip implant should be load-bearing and non destructive to the surroundingtissue. The study of biomaterials is a multidisciplinary endeavor, integratingknowledge from fields such as engineering, chemistry, biology, and medicine. Be-cause of the high impact of these materials on patient care, this area has been anintensely researched for the last 20 years.

Although current biomaterials have had a large impact in medicine, they consistmainly of ‘‘off-the-shelf ’’ materials not originally designed for clinical use. As aresult, complications have and continue to occur. For example, leakage, hardening,and rupture of silicone gel breast prostheses have lead to billions of dollars inlitigation related to either real or perceived patient health problems attributed tothe material. The early use of cellulose in hemodialysis tubing resulted in anaphy-lactic shock and death in several patients. As a result of these incidents and numer-ous others, many materials manufacturers no longer permit the clinical use of theirmaterials; thus, device manufacturers have feared a shortage of materials. To helpalleviate any future shortage of biomaterials, President Clinton signed the Biomate-rials Access Assurance Act of 1998 to limit product liability in the biomaterialsarea.

To eliminate many of the complications and expand the use of biomaterials inmedicine, research has focused on developing materials specifically designed forthe clinic. These materials are being designed using an advanced knowledge of cellbiology and tissue–material interactions. Many of these new materials representchemical and physical modifications of existing materials or are entirely novel intheir chemical structure. Here, we will focus on new material development in threehigh-impact areas of biomaterials research: cell–biomaterial interactions, tissueengineering, and drug delivery. This discussion is by no means exhaustive, asmany significant advances have also been made in other research areas that usebiomaterials.

Recent Advances in Biomaterials 195

Cell–Biomaterial Interactions

Because the lack of biocompatibility of most conventional biomaterials can beattributed to adverse cell–material interactions, a large amount of research hasbeen focused on understanding and controlling what happens at the cell–materialand tissue–material interfaces. Most mammalian cells are anchorage dependent;that is, they must adhere to a substrate to grow and differentiate. Adhesion to asurface is mediated by protein adsorption onto the surface, followed by cell–receptor interactions with the adsorbed proteins. As the adherent cells grow andproliferate, they produce their own extracellular matrix (ECM) on which to adhere.Surfaces that minimize protein adsorption, such as those modified with poly (ethyl-ene glycol), also minimize cell adhesion and spreading [1]. Cells on these surfacesare sparse in number, very loosely attached, and appear rounded.

In addition to purely thermodynamic considerations such as protein adsorption,cell adhesion is strongly influenced by receptor–ligand interactions between inte-grins on the cell surface and ECM proteins adsorbed on the substrate. These pro-teins, examples of which include laminin, fibronectin, and collagen, contain peptidesequences, such as RGD (R � arginine, G � glycine, D � aspartic acid), specificfor integrin binding. Materials modified with RGD peptide sequences alone arecapable of integrin binding and thus permit cell attachment and growth [2]. Zhanget al. have used the RAD (A � alanine) sequence to promote cell adhesion tooligopeptides terminated with cysteine residues [3]. Via sulfhydryl interactionswith a gold-coated substrate, these oligopeptides can form self-assembled mono-layers (SAMs) on the surface of the substrate and thus create a good adhesionsubstrate for a variety of cell types.

In addition to integrin–cell adhesion, cells such as hepatocytes possess recep-tors for asialoglycoprotein receptors for β-d-galactose residues present in the ECM[4,5]. Griffith and colleagues took advantage of these unique interactions to createan adhesion surface for hepatocytes. Starting with poly (ethylene glycol) or PEGstar polymers (PEG chains radiating from a divinylbenzene core), they addedd-galactose moieties to the hydroxy terminus of PEG and cross-linked the gelsusing electron beam irradiation. Hepatocytes readily adhered to these surfaces eventhough the ligand density was an order of magnitude lower than in similar galac-tose-modified polyacrylamide gels. The reason for this strong adhesion was thatthe flexibility of the PEG chains allowed necessary multiple ligand–receptor inter-actions to occur, even though the ligand density was low.

In the past few years, Stupp and colleagues [6–8] have developed a novel classof block copolymers that self-assemble into highly ordered supramolecular structureson surfaces. These materials were diblock polymers termed ‘‘rod–coil’’ and are com-posed of a rigid rod segment of biphenyl esters and a flexible segment (the coil)composed of a block copolymer of polystyrene and polyisoprene (Figs. 1 and 2).When cast on surfaces, these polymers self-assembled to form identical nanometer-scale clusters. The self-organization of the polymers on a surface was based on thecrystallization of the biphenyl ester rod segments, as can be seen in the base of thestructure depicted in Fig. 2. Molecular modeling studies suggested that these aggre-gates were essentially supramolecular ‘‘mushrooms’’ with the ‘‘stem’’ consisting of


FIG. 1 Structure of self-assembling rod–coil polymer. (Adapted from Ref. 6.)

the crystallized rod segments and the ‘‘head’’ comprised of the coil segments. Whenannealed at high temperature, the isoprene segments cross-linked to form films stablein a variety of solvents. Because of the highly ordered and stable nature of theseclusters, they hold great potential in the design of new biomaterials.

Microfabrication techniques such as photolithography and microcontact print-ing have also been used to create surfaces with unique topographies and spatiallydefined chemistries. Such surfaces have had a profound influence on cell–materialinteractions. Researchers at Cornell, for example, have developed micromachinedprobes that show minimal adverse tissue response when implanted in the cerebralcortices of rats [2]. Others have shown that photolithography can define regionsof cell adhesion, controlling both cell position and migration [9].

Microcontact printing or µ-Cp, a soft lithography technique, has received ex-tensive attention as a method of both chemical patterning [10] and protein pat-terning surfaces [11] to control cell adhesion and migration. As described earlier,Whitesides and Lauffenberger used oligopeptide SAMs to create surfaces for celladhesion [3]. These peptides can be patterned by µ-Cp, resulting in the spatialcontrol of cell adhesion and migration and creating unique geometries suitable for

FIG. 2 Molecular graphics of a self-assembled rod–coil polymer structure containing nine layers ormonomer units of isoprene. (From Ref. 6.)


the study of cell–cell interactions on these surfaces. Earlier studies demonstratedthat there are geometric limits to the extent cells can be confined without causingapoptosis (programmed cell death) [12–14]. Besides µ-Cp, Takayama et al. havealso used the low-mixing laminar-flow conditions in micromachined channels topattern cocultures of eukaryotic cells on a single surface [15].

Going beyond SAMs, a four-step soft lithographic process based on microcon-tact printing (µ-Cp) of organic monolayers, hyperbranched polymer grafting, andsubsequent polymer functionalization has resulted in polymer patterns that directthe grown of mammalian cells such as IC-21 murine peritoneal macrophages [16],human umbilical vein endothelial cells, and murine hepatocytes. The functionalunits on these surfaces were three-dimensional cell ‘‘corrals’’ that have walls 50nm in height and lateral dimensions on the order of 60 µm. The corrals havehydrophobic, methyl-terminated n-alkanethiol bottoms, which promote cell adhe-sion, and walls consisting of hydrophilic poly(acrylic acid)/poly(ethylene glycol)(PAA/PEG) layered nanocomposites that inhibit cell growth. Cells seeded on pat-terned surfaces adhere and grow within the corrals, but they do not span the PAA/PEG corral walls (Fig. 3). Cell viability studies indicate that cells remain viable

FIG. 3 Murine macrophages cultured on a micropatterned surface of poly(acrylic acid)/poly(ethyleneglycol). (From Ref. 16.)


on the patterned surfaces for up to 21 days, and microscopy studies demonstratethat cell growth and spreading does not occur outside of the corral boundaries.

Other significant insights have been made in understanding tissue–materialinteractions, particularly in influencing the foreign-body response. When a foreignbody such as a medical implant is introduced into a host, the natural tendency ofthe surrounding tissue is to degrade or extrude the implant. If the host cannoteliminate the foreign body, a chronic inflammatory reaction results and the objectis encapsulated in fibrous tissue. This capsule poses a difficult problem in thedevelopment of cell-based therapeutic devices and engineered tissues [17]. It pre-sents a mass-transfer barrier and therefore limits the concentration of nutrients andoxygen reaching the transplanted cells. As a result, these cells frequently die fromhypoxia or lack of nutrients, particularly cells that are highly metabolically active(e.g., pancreatic islets and hepatocytes) [18]. The ability of the capsule to limitmass transfer has been demonstrated in several studies. In one early study, fibrouscapsules were generated by implanting materials in vivo followed by the harvestingof the capsule for permeability studies, with results of these studies showing lowerpermeability for thicker capsules [19]. In vivo studies of transplanted pancreaticislets showed that islet viability decreased dramatically with increasing distancebetween blood vessels and islets [17,20]. This result is not surprising because ina healthy pancreas, islets are very closely associated with capillaries [21]. Increaseddistance from blood vessels also resulted in lower insulin production from trans-planted islets [18].

Devices containing transplanted cells must therefore develop capillary net-works with transcapillary mass-transfer rates high enough to ensure survival ofthe cells. Prevascularization of polymer scaffolds for tissue engineering was at-tempted as a method to develop a vascular network in and around the implant andpromote the survival of cells transplanted into the scaffold at a later date [22]. Anumber of studies have also shown that the microarchitecture of an implantedmaterial has a great influence on its local tissue response [23]. In suture materials,the topography of the surface affected macrophage involvement, with round su-tures resulting in less macrophage involvement than sutures with surface irregulari-ties on the scale of 10–15 µm [24]. In another study, surface features greater than10 µm appeared to attract foreign-body giant cells to the surface of the implant.The presence of these inflammatory cells around the implanted material resultedin thickening of the fibrous capsule [23]. In addition to implant topography, theanatomical site of implantation influences fibrosis. Acrylic fibers implanted intothe subcutaneous tissue of rats showed a higher degree of fibrosis than those im-planted into the abdominal fat pad. When islets were introduced, oxygen presentinside these hollow fibers was also consumed by islets in minutes [20].

Studies by Padera and Colton [25] and Brauker et al. [26] have shown thatcertain microporous materials will allow blood vessels to grow and be maintainedat the tissue–material interface and in some cases within the pores of the material(Fig. 4). However, this is not true for all porous polymer membranes, even thosewith similar porosities and chemistries. What appears to be driving the host re-sponse is not necessarily the chemistry of the material, but the microstructure ofindividual features within the material onto which host cells can attach. Materialsthat are microporous but contain large planar features prompted an avascular hostresponse while the same material lacking these planar features and having a more


FIG. 4 Top: Micrograph of a tissue section containing a PVDF membrane with a 0.22-µm pore size.Arrows indicate foreign-body giant cells at the tissue material interface. Note the lack of bloodvessels near the tissue–material interface. Bottom: Micrograph of a tissue section containinga 5-µm PVDF membrane. Arrows indicate blood vessels either at the tissue–material interfaceor within the material.


fibrous structure prompted neovascularization at the tissue–material interface. Ifthe mechanism of this response can be elucidated, it may then become possibleto fabricate materials with a microstructure designed to promote specific levels ofneovascularization.

Tissue Engineering

Tissue engineering is an interdisciplinary field combining fundamental principlesfrom biology and engineering to understand the structure and function of tissue.Tissue failure or loss due to disease or injury accounts for over $400 billion intotal annual health care costs in the United States alone. For some ailments suchas burns, kidney failure, and liver failure, tissue transplantation is the conventionaltreatment. However, there exists a serious shortage in available donor tissue. Forexample, nearly 40,000 people die each year waiting for a suitable liver for trans-plantation. To address this need, tissue engineering has developed as a disciplinewith the goal of restoring, maintaining, and improving the function of living tissue.

Tissue engineering has taken on three basic forms. The first is the developmentof biohybrid artificial organs such as implantable islet-containing bioreactors ormicroencapsulated islets for the treatment of diabetes. The second is the controlledrelease of growth factors and cytokines to promote tissue repair or regenerationusing existing mechanisms within the body. The third approach, which is addressedhere, uses cells from the desired tissue cultured in vitro on a three-dimensionalbiodegradable scaffold. Once these cells have reached a critical density on thescaffold, they are then transplanted into the subject at the desired location. Theywill then hopefully continue to grow and mature into the desired tissue, the scaffoldhaving degraded with time. This approach has met with some success, particularlyfor the generation of skin and cartilage. However, there are several significantobstacles that must be overcome before it can be used for the repair or replacementof many types of tissue.

Many of these tissue substitutes take the form of cells from the tissue of interestcultured on biomaterials, frequently microporous biodegradable polymer scaffolds.For example, if one wished to form a tissue-engineered liver, hepatocytes (livercells) are cultured in vitro on a polymer scaffold to a high cell density. The tissue–material construct is then implanted into the abdomen of the patient. As the tissuecontinues to grow, the polymer scaffold degrades and is eventually replaced byhealthy and functioning liver tissue. Many readers may remember the televisionnews image of a human ear growing on the back of a rat, an example of tissueengineering of cartilage. The ear began as cartilage cells (chondrocytes) culturedon a polymer scaffold in the shape of a human ear.

For some types of tissue such as cartilage and skin, tissue-engineered productshave moved outside the laboratory and into the marketplace. As an example, tissue-engineered skin developed from biodegradable scaffolds are available from Ad-vanced Tissue Sciences. Protein Polymer Technologies is developing silk/elastingels that can be freeze-dried to form sponges for wound healing and organ-fillerapplications. Other tissue-engineered products are currently in development, in-


cluding encapsulated cells and cell-based bioreactors for the treatment of Type Idiabetes and liver failure, vascular grafts, and bone replacement.

For other types of tissue such as liver and nervous tissue, many challengesare present, including the integration of blood vessels and nerves, adverse tissueresponses to the biodegradable polymer, and engineering tissue containing multiplecell types. To promote blood vessel growth in engineered tissue, researchers suchas David Mooney of the University of Michigan are developing polymer scaffoldsprepared in supercritical carbon dioxide that release blood-vessel-promotinggrowth factors such as vascular endothelial growth factor (VEGF) [27]. VEGFbioactivity was retained and controlled-release sustained in excess of 30 days.VEGF, also known as vascular permeability factor (VPF), is a protein composedof two identical subunits which has a molecular weight of 45 kDa. This proteinis a secreted cytokine that is specific for vascular endothelial cells and has beenshown to have angiogenic activity in vivo [28–31]. Numerous studies have alsoshown that VEGF is secreted by macrophages and that VEGF expression is en-hanced in macrophages under hypoxic conditions [31]. Other researchers such asKeith Gooch of the University of Pennsylvania are forming capillaries in vitro thatconnect to existing blood vessels when implanted.

Significant advances have also been made in the engineering of other tissues.For nervous tissue, Schmidt and colleagues found that electrically conducting poly-mers such as polypyrrole aid the reconnecting of severed nerves [32]. For bone,Anseth and colleagues have developed photo-cross-linkable polyanhydrides thathave excellent mechanical properties and can be fabricated into complex shapessuch as screws [33]. Oberpenning and colleagues recently tested a tissue-engineered bladder (Fig. 5) that, when implanted in dogs, was able to retain urinenormally for up to 11 months [34]. This bladder, based on smooth-muscle cells andurethelial cells cultured on a biodegradable poly (lactide-co-glycolide) scaffold,represents the first construction of a hollow organ by tissue engineering. Finally,a number of systems have been developed for the in vivo polymerization of amatrix for tissue engineering, including poly(ethylene glycol) gels that can be inter-facially polymerized to prevent intimal thickening following ballon angioplasty[35] and polymer solutions containing chondrocytes that can be polymerized trans-dermally [36,37].

Drug Delivery

Drug delivery research in the controlled release of proteins has become increasinglyimportant as many new drugs are either protein or peptide based. Two prime exam-ples are human growth hormone and erythropoeitin for the treatment of anemia;both of these proteins are billion dollar drugs. Most of these biomolecules are unsta-ble in vivo and as a result must be administered by multiple injections. To stabilizethese drugs in vivo and to control the rate at which they are delivered and thelocation of delivery, new polymers, both degradable and nondegradable, are beingdeveloped. These materials are designed to control the mass transfer of drug to thesurrounding tissue and to be biocompatible. Many of these materials can maintain a


FIG. 5 Radiographic cystograms of a tissue-engineered bladder implanted in a dog. (From Ref. 34.)

constant release of drug over many days and months. Examples of such degradablematerials include polyesters such as poly(glycolic acid) used in resorbable sutures,polyanhydrides, and poly(ethylene glycol)–polyester copolymers. These materialsare chemically designed such that their degradation products are nontoxic.

One example is Gliadel, the first new FDA-approved treatment of gliabastoma(a deadly form of brain cancer) in 20 years [38]. This product is composed of abiodegradable polymer containing the anticancer drug carmustine. After the braintumor is surgically removed, the space formerly occupied by the tumor is linedwith polymer wafers containing the drug. As the polymer degrades over time, theanticancer drug is released directly to the brain in concentrations that cannot be


achieved by administering the drug via the bloodstream. As a result, the reoccur-rence of the disease is diminished and chemotherapy side effects are reduced be-cause the drug is delivered only where it is needed.

Drug delivery products for vaccine delivery, anemia, and cancer treatment areexpected to be available shortly. Other research in drug delivery is currently fo-cused on developing targeted polymer delivery systems (such as biodegradablemicroparticles to the lungs) and developing polymer delivery vehicles for genetherapy.

Hydrogels, both degradable and nondegradable, have been of particular inter-est. These materials are highly swollen with water and can possess equilibriumwater content in excess of 90% by weight. With the exception of swelling-con-trolled or osmotic delivery systems, polymeric materials control drug release bylimiting Fickian diffusion of the drug through the solvent-swollen network. Thisprocess is highly dependent on polymer chain mobility. Mikos and colleagues re-cently provided a comprehensive review of both nondegradable and degradablehydrogel systems for drug delivery [39].

Some specific examples are instructive. For example, Khare and Peppas havedescribed the effects of copolymerization, ionic strength, pH, and buffer composi-tion on the release of biomolecules [40]. Particularly in the case of anionic gels,these factors influence the equilibrium water content of the polymer network andhence the mesh size in the gel. Brannon-Peppas and Peppas describe research onthe effects of comonomer composition and hydrophobicity on swelling [41]. Net-work hydrophilicity controls swelling via water absorption, and because equilib-rium water content is directly proportional to biomolecule mass-transfer rates outof the gel, extended release profiles can be achieved by increasing the hydropho-bicity of the gel. In polyelectrolytic gels, the ionic character is pH dependent andcan exhibit substantial changes in swelling caused by shifts in pH [41]. Unlesshighly localized changes in pH occur, these pH-induced changes in swelling maynot occur in vivo.

DNA–polycation complexes have received extensive attention in recent yearsas an alternative to virus-mediated transfection of therapeutic gene therapy vectorsinto mammalian cells. Polycations investigated to date include polylysine [42],cationic liposomes [43], polyethylene imine [44], and poly(amidoamine) (PA-MAM) dendrimers [45]. PAMAM dendrimers are unidispersed and have a highpositive-charge density at their surface. Electrostatic complexes of DNA and den-drimers were shown to transfect a variety of cell lines, including fibroblasts, CHO,Rat2, lymphoma, and hepatoblastoma [45]. In the Rat2 cell line, transfection washighly dependent on the generation of the dendrimer, which dictates the surfacecharge and size of the molecule. Transfection efficiency as compared to simpleexogenous DNA was found to increase exponentially with the generation of thedendrimer, reaching a plateau after the ninth generation of 20,000 to 40,000 timesmore efficient. Although chloroquine and DEAE–dextran enhanced transfection,they were not required and transfection, as determined from the expression of aluciferase gene in the vector, occurred at levels easily measured.

Recently, a novel system was described that may have significant applicationsin drug delivery. Discher and colleagues [46] described the formation of poly-mersomes, vesicles made of amphiphilic diblock copolymers of poly(ethylene ox-ide) and poly(ethylethylene) (Fig. 6). These structures were found to be structurally


FIG. 6 Polymersomes. (A) Schematic representation of diblock copolymer assembly; (B) micrographof diblock copolymer vesicles, rodlike structures (black arrow) and micelles (gray arrow).(From Ref. 46.)

tougher than liposomes and significantly less permeable to water. With furthermodification, these materials may be developed into new vehicles for nucleic-acid- and protein-based drugs.

References

1. N. Desai and J. Hubbell, ‘‘Biological Responses to Polyethylene Oxide Modified Poly-ethylene Terephthalate Surfaces,’’ J. Biomed. Mater. Res., 25, 829–843 (1991).

2. J. Turner, W. Shain, D. Szarowski, M. Andersen, S. Martins, M. Isaacson, and H.Craighead, ‘‘Cerebral Astrocyte Response to Micromachined Silicon Implants,’’ Exp.Neurol., 156(1), 33–49 (1999).

3. S. Zhang, L. Yan, M. Altman, M. Lassle, H. Nugent, F. Frankel, D. Lauffenburger,G. Whitesides, and A. Rich, ‘‘Biological Surface Engineering: A Simple System forCell Pattern Formation,’’ Biomaterials, 20, 1213–1220 (1999).

4. L. Griffith and S. Lopina, ‘‘Microdistribution of Substratum-Bound Ligands Affects


Cell Function: Hepatocyte Spreading on PEO-Tethered Galactose,’’ Biomaterials 19,979–986 (1998).

5. S. Lopina, G. Wu, E. Merrill, and L. Griffith-Cima, ‘‘Hepatocyte Culture on Carbohy-drate-Modified Star Polyethylene Oxide Hydrogel,’’ Biomaterials, 17, 559–569 (1996).

6. S. Stupp, S. Son, H. Lin, and L. Li, ‘‘Synthesis of Two-Dimensional Polymers,’’Science 259, 59–63 (1993).

7. L. Radzilowski and S. Stupp, ‘‘Nanophase Separation in Monodisperse RodcoilDiblock Polymers,’’ Macromolecules, 27, 7747–7753 (1994).

8. S. I. Stupp, V. LeBonheur, K. Walker, L. S. Li, K. E. Huggins, M. Keser, and A.Amstutz, ‘‘Supramolecular Materials: Self-Organized Nanostructures,’’ Science, 276,384–389 (1997).

9. T. Matsuda and T. Sugawara, ‘‘Control of Cell Adhesion, Migration, and Orientation onPhotochemicallyMicroprocessedSurfaces,’’ J.Biomed. Mater.Res., 32, 165–173(1996).

10. Y.XiaandG.Whitesides, ‘‘SoftLithography,’’Angew.Chem. Int.Ed., 37, 550–575(1998).11. C. James, R. Davis, M. Meyer, A. Turner, S. Turner, G. Wither, L. Kam, G. Banker,

H. Craighead, M. Isaacson, J. Turner, and W. Shain, ‘‘Aligned Microcontract Printingof Micrometer-Scale Poly-l-lysine Structures for Controlled Growth of Cultured Neu-rons on Planar Microelectrode Arrays,’’ IEEE Trans. Biomed. Eng., BE-47(1), 17–21 (2000).

12. C. Chen, M. Mrksich, S. Huang, G. Whitesides, and D. Ingber, ‘‘Geometric Controlof Cell Life and Death,’’ Science, 276, 1425–1428 (1997).

13. C. Chen, M. Mrksich, S. Huang, G. Whitesides, and D. Ingber, ‘‘Micropatterned Sur-faces for Control of Cell Shape, Position, and Function,’’ Biotechnol. Prog., 14, 356–363 (1998).

14. R. Singhvi, A. Kumar, G. Lopez, G. Stephanopoulos, D. Wang, G. Whitesides, andD. Ingber, ‘‘Engineering Cell Shape and Function,’’ Science, 264, 696–698 (1994).

15. S. Takayama, J. McDonald, E. Ostuni, M. Liang, P. Kenis, R. Ismagilov, and G.Whitesides, ‘‘Patterning Cells and Their Environments Using Multiple Laminar FluidFlows in Capillary Networks,’’ Proc. Natl. Acad. Sci. USA, 96, 5545–5548 (1999).

16. P. Ghosh, M. Amirpour, W. Lackowski, M. Pishko, and R. Crooks, ‘‘A Simple Litho-graphic Approach for Preparing Patterned, Micron-Scale Corrals for Controlling CellGrowth,’’ Angew. Chem., 38(11), 1592–1595 (1999).

17. C. K. Colton, ‘‘Implantable Biohybrid Artificial Organs,’’ Cell Transplant, 4(4), 415–436 (1995).

18. K. Dionne, C. Colton, and M. Yarmush, ‘‘Effect of Hypoxia on Insulin Secretion byIsolated Rat and Canine Islets of Langerhans,’’ Diabetes, 42, 12–21 (1993).

19. R. Wood, E. LeCluyse, and J. Fix, ‘‘Assessment of a Model for Measuring DrugDiffusion Through Implant-Generated Fibrous Capsule Membranes,’’ Biomaterials,16, 957–959 (1995).

20. J. Bodziony, ‘‘Bioartificial Endocrine Pancreas: Foreign-Body Reaction and Effec-tiveness of Diffusional Transport of Insulin and Oxygen After Long-Term Implanta-tion of Hollow Fibers into Rats,’’ Res. Exp. Med., 192, 305–316 (1992).

21. S. Bonner-Weir, ‘‘Morphological Evidence for Pancreatic Polarity of Beta-Cell withinIslets of Langerhans,’’ Diabetes, 37(5), 616–621 (1988).

22. H. Wald, G. Sarakinos, M. Lyman, A. Mikos, J. Vacanti, and R. Langer, ‘‘Cell Seed-ing in Porous Transplantation Devices,’’ Biomaterials, 14(4), 270–278 (1993).

23. C. Campbell and A. V. Recum, ‘‘Microtopography and Soft Tissue Response,’’J. Invest. Surgery, 2, 51–74 (1989).

24. T. Salthouse and B. Matlaga, Biomaterials in Reconstructive Surgery (L. Rubin, ed.),Mosby, St. Louis, MO, 1983, pp. 40–45.

25. R. Padera and C. Colton, ‘‘Time Course of Membrane Microarchitecture-Driven Neo-vascularization,’’ Biomaterials, 17(3), 277–284 (1996).


26. J. Brauker, V. Carr-Brendel, L. Martinson, J. Crudele, W. Johnston, and R. Johnson,‘‘Neovascularization of Synthetic Membranes Directed by Membrane Microarchitec-ture,’’ J. Biomed. Mater. Res., 29, 1517–1524 (1995).

27. P. Eiselt, B. Kim, B. Chacko, B. Isenberg, M. Peters, K. Greene, W. Roland, A. Loebsack,K. Burg, C. Culberson, C. Halberstadt, W. Holder, and D. Mooney, ‘‘Development of Tech-nologies Aiding Large-Tissue Engineering,’’ Biotechnol. Prog., 14(1), 134–140 (1998).

28. S. Hopkins, J. Bulgrin, R. Sims, B. Bowman, D. Donovan, and S. Schmidt, ‘‘ControlledDeliveryof VascularEndothelial GrowthFactor PromotesNeovascularization and Main-tains Limb Function in a Rabbit Model,’’ J. Vasc. Surgery, 27(5), 886–894 (1998).

29. K. Norrby, ‘‘Vascular Endothelial Growth Factor and de Novo Mammalian Angiogen-esis,’’ Microvasc. Res., 51, 153–163 (1996).

30. M. Peters, B. Isenberg, J. Rowley, and D. Mooney, ‘‘Release from Alginate Enhancesthe Biological Activity of Vascular Endothelial Growth Factor,’’ J. Biomater. Sci.Polym. Ed., 9(12), 1267–1278 (1998).

31. D. Shweiki, A. Itin, D. Soffer, and E. Keshet, ‘‘Vascular Endothelial Growth FactorInduced by Hypoxia May Mediate Hypoxia-Initiated Angiogenesis,’’ Nature, 359,843–845 (1992).

32. C. Schmidt, V. Shastri, J. Vacanti, and R. Langer, ‘‘Stimulation of Neurite Outgrowth Usingan Electrically Conducting Polymer,’’ Proc. Natl. Acad. Sci. USA, 17, 8948–8953 (1997).

33. K. Anseth, V. Shastri, and R. Langer, ‘‘Photopolymerizable Degradable Polyanhy-drides with Osteocompatibility,’’ Nature Biotechnol., 17, 156–159 (1999).

34. F. Oberpenning, J. Meng, J. Yoo, and A. Atala, ‘‘De Novo Reconstitution of a Func-tional Mammalian Urinary Bladder by Tissue Engineering,’’ Nature Biotechnol., 17,149–155 (1999).

35. J. West and J. Hubbell, ‘‘Separation of the Arterial Wall from Blood Contact UsingHydrogel Barriers Reduces Intimal Thickening After Ballon Injury in the Rat: TheRoles of Medial and Luminal Factors in Arterial Healing,’’ Proc. Natl. Acad. Sci.USA, 93(23), 13,188–13,193 (1996).

36. J. Elisseeff, K. Anseth, D. Sims, W. McIntosh, M. Randolph, and R. Langer, ‘‘Trans-dermal Photopolymerization for Minimally Invasive Implantation,’’ Proc. Natl. Acad.Sci. USA, 96, 3104–3107 (1999).

37. J. Elisseeff, K. Anseth, D. Sims, W. McIntosh, M. Randolph, M. Yaremchuk, and R.Langer, ‘‘Transdermal Photopolymerization of Poly(ethylene oxide)-Based InjectableHydrogels for Tissue-Engineered Cartilage,’’ Plast. Reconstr. Surgery, 104(4), 1014–1022 (1999).

38. M. Wu, J. Tamada, H. Brem, and R. Langer, ‘‘In Vivo Versus In Vitro Degradationof Controlled Release Polymers for Intracranial Surgical Therapy,’’ J. Biomed. Mater.Res., 28(3), 387–395 (1994).

39. M. T. Ende and A. G. Mikos, ‘‘Diffusion-Controlled Delivery of Proteins from Hy-drogels and Other Hydrophilic Systems,’’ in Protein Delivery: Physical Systems(S. A. Hendren, ed.), Plenum Press, New York, 1997, pp. 139–165.

40. A. Khare and N. Peppas, ‘‘Swelling/Deswelling of Anionic Copolymer Gels,’’ Bioma-terials, 16(7), 559–567 (1995).

41. L. Brannon-Peppas and N. Peppas, ‘‘Dynamic and Equilibrium Swelling Behaviorof pH Sensitive Hydrogesl Containing 2-Hydroxyethylmethacrylate,’’ Biomaterials,11(11), 635–644 (1990).

42. J. Kim, I. Kim, A. Maruyama, T. Akaike, and S. Kim, ‘‘A New Non-viral DNA DeliveryVector: The Terplex System,’’ J. Controlled Release, 53(1–3), 175–192 (1998).

43. O. Hottiger, T. Dam, B. Nickoloff, T. Johnson, and G. Nabel, ‘‘Liposome-Mediated GeneTransfer into Human Basal Cell Carcinoma,’’ Gene Ther., 6 (12), 1929–1935 (1999).

44. W. Godbey, K. Wu, and A. Mikos, ‘‘Poly(ethylenimine) and Its Role in Gene Deliv-ery,’’ J. Controlled Release, 60(2–3), 149–160 (1999).

Extractive Distillation 207

45. J. Kukowska-Latallo, A. Bielinska, J. Johnson, R. Spindler, D. Tomalia, and J. Baker,‘‘Efficient Transfer of Genetic Material into Mammalian Cells Using Starburst Poly-amidoamine Dendrimers,’’ 93, 4897–4902 (1996).

46. B. Discher, Y. Won, D. Ege, J. Lee, F. Bates, D. Discher, and D. Hammer, ‘‘Poly-mersomes: Tough Vesicles Made from Diblock Copolymers,’’ Science, 284, 1143–1146 (1999).

MICHAEL V. PISHKO

Recent Development of Extractive Distillation:A Distillation Alternative

Introduction

Distillation is the most commonly used method for recovering and purifying petro-chemicals and chemicals in the industry. The difference in boiling points betweenthe key components to be separated is the means for separation. The ease of separa-tion is conveniently measured by the relative volatility (α) between the key compo-nents, which is defined as:

α �Y1/X1

Y2/X2

where X1 and X2 are the mole fraction of components 1 and 2, respectively, in theliquid phase and Y1 and Y2 are those in the vapor phase. In fact, α is one of the majoreconomic factors for distillation. Colburn and Schoenborn [1] gave the followinggeneralized correlation for the approximate number of theoretical plates requiredfor a separation of products each of 99% � purity:

Number of theoretical plates �4

α � 1

In general, distillation becomes uneconomical when 0.95 � α � 1.05, because alarge number of plates requires very high capital investment and a large refluxratio requires very high operating cost. Under this situation, a solvent-enhanceddistillation, such as extractive distillation (ED), becomes economically attractiveand practical.

The basis for ED is to increase α by introducing a high-boiling, polar solvent tothe distillation column. The characteristics, design, and operation of an extractivedistillation column (EDC) have been thoroughly discussed in the literature [2–5].

208 Extractive Distillation

Depending on applications, the solvent-to-feed weight ratio (S/F) can vary from3 to as high as 20, so the EDC is normally operated using substantially higher amountof liquid (nonvolatile solvent) than the conventional distillation. Also, due the to differ-ence in solubility of the key components in the polar solvent, a certain portion of theEDC may have two liquid phases. Significant progress has been made in the industryrecently in terms of understanding and handling these special situations in the EDC.

Hydrodynamic Behavior of a Packed EDC

The presence of the nonvolatile solvent in the EDC not only substantially increasesliquid flow (L) but also reduces the vapor flow (V) by preferentially absorbing themore polar components in the vapor stream. Therefore, extractive distillation isnormally operated under significantly higher liquid to vapor ratios (L/V) as com-pared to conventional distillation.

Depending on applications, the solvent-to-feed weight ratio (S/F) varies from 3/1 to as high as 20/1. Hydrodynamic behavior of a distillation operation with a high-L/V condition has not been significantly reported in the literature. Nevertheless, suchinformation on packed columns used for ED operation was reported by Brown andLee [6]. Two types of packing were tested; a random dumped packing [0.63-cm pro-truded metal packing (Pro-Pak)] and a structured packing (Koch-Sulzer BX).

Shown in Fig. 1, the tests on Pro-Pak random packings were conducted in a 0.15-

FIG. 1 Schematic diagram of ED pilot plant for hydrodynamic study of packings.


TA

BL

E1

ED

Run

sw

ithD

iffe

rent

Tow

erPa

ckin

gs

Ext

ract

(%)

Raf

finat

e(%

)So

lven

t-K

ettle

Run

Pack

ing

to-f

eed

tem

p.R

eflux

Tol

uene

Tol

uene

Hep

tane

Hep

tane

no.

code

ratio

(°C

)ra

tiore

cove

rypu

rity

reco

very

puri

ty

1P

6.5

174

4.4

99.6

99.5

95.0

95.4

1S

6.5

174

3.2

99.1

99.8

98.3

93.3

2P

6.5

182

2.1

88.8

100.

010

0.0

59.8

2S

6.5

182

3.7

99.3

100.

010

0.0

95.3

3P

10.0

182

3.6

99.2

99.9

99.3

94.4

3S

10.0

182

3.4

99.1

100.

010

0.0

94.0

4P

10.0

193

1.1

72.8

100.

010

0.0

46.7

4S

10.0

193

1.6

86.7

100.

010

0.0

59.3

5P

6.5

172

4.4

99.7

99.7

96.8

96.5

5S

6.5

172

3.2

99.1

100.

010

0.0

93.4

Not

es:

Pack

ing

code

s:P

�pr

otru

ded

pack

ing

(Pro

Pak)

;S

�st

ruct

ured

pack

ing

(Koc

h–Su

lzer

BX

).A

llra

tios

and

perc

ents

wer

em

easu

red

byw

eigh

t.


m-diameter EDC with 5.7-m total packed height loaded in a duo-column system: Thetop and bottom halves of the column were connected in series by insulted vaporand liquid lines. During the operation, pipe distributors were installed at the top andat 40% above the bottom of each half-column. The feed (a mixture of 88/12 wt%toluene/heptane) was fed to the location at 55% above the bottom of the entire EDC,and the extractive solvent (N-methyl pyrrolidone) was fed at 88% above the bottomof the EDC. During a normal run, hydrocarbon was introduced into the EDC at arate of 0.015–0.022 m3/h at a temperature slightly below the hydrocarbon bubblepoint and a pressure of 35 kPa. Lean solvent was fed to the top of the upper halfof ED tower at a subcooled temperature of 121°C. The solvent-to-feed ratio (S/F)varied from 6.5 to 10 and the kettle temperature varied between 171°C and 193°C.

Koch–Sulzer BX structured packing was also tested in the same EDC under thesame condition as the Pro-Pak testings. Distributors were placed only at the top ofeach half-column. A total of 31 elements with 5.3-m total packing height were fittedin the duo-column EDC. A summary of the selected runs is presented in Table 1.

To investigate the pressure drop in the EDC under high-L/V condition, a plot ofthe difference between the measured pressure drop and the predicted pressure drop isgiven in Fig. 2. It was found that the actual pressure drops measured in this study aresignificantly higher than the ones predicted by the packing vendor’s correlations [7,8].The difference between the predicted and actual pressure drop for protruded packingare at least twice that of the structured packing under the same conditions.

It was also found that the height equivalent to a theoretical plate (HETP) onboth packings was underpredicted by the vendor’s correlations [7,8]. As shown inFig. 3, the difference in HETP for the structured packing and protruded packing arerespectively 5 and 10 in. The actual HETP was roughly estimated by the Fenske–Underwood correlations, utilizing the composition of feed and products, and anaverage relative volatility.

Then, computer simulations were run to determine the number of theoreticalstages, using a rigorous computer algorithm capable of simulating distillation pro-cesses with multicomponents and multiphases. Experimental activity coefficients

FIG. 2 Comparison of predicted and actual pressure drop.


FIG. 3 Comparison of predicted and actual HETP.

for N-methyl pyrrolidone/n-heptane/toluene were used as input to a Renon activitycoefficient model, which was used to determine vapor–liquid equilibrium con-stants. In the computer simulation, an initial trial solution to the computer algorithmwas made by assuming constant molal overflow and a guessed (assumed) column-temperature profile. A final solution was then made by convergence on the inputspecification values, vapor–liquid equilibria, and heat and material balances. ANewton–Raphson convergence technique was used.

The gas capacity factor (Fs � the superficial gas velocity times the square rootof the gas density) is an important parameter in determining the loading on thecolumn. Figure 4 shows that the pressure drop for both packings increase withincreasing Fs, but protruded packing consistently shows a pressure drop three times

FIG. 4 Pressure drop versus gas capacity factor.


FIG. 5 HETP versus gas capacity factor.

higher than structured packing. The gas capacity factor is also plotted againstHETP in Fig. 5. It appears that Fs has no effect on HETP for both packings overthe range tested. The HETP for protruded packing is slightly higher than that ofstructured packing (approximately 1–5 in higher). It is estimated from vendor in-formation that the column was at 12% of flood, using Koch–Sulzer BX structuredpacking [7], and at 32–37% of flood, using the protruded packing [8].

Liquid loading in an EDC is much higher than the conventional distillationcolumn, so it could be the limiting factor in the design of an EDC. The pressuredrop is plotted against the liquid loading (kg/h/m2) in Fig. 6 for both packings.

FIG. 6 Pressure drop versus liquid loading.


FIG. 7 HETP versus liquid loading.

Liquid loading shows a positive linear relationship with pressure drop for the pro-truded packing and shows a minimum pressure drop at liquid loading around 8070–8310 kg/h/m2, but otherwise shows little correlation.

Liquid loading is also plotted against HETP for both packings in Fig. 7. Al-though literature sources predict a decrease in HETP with increased liquid loading[9,10], the structured packing had a constant HETP and the protruded packing haslittle correlation over the liquid loading range investigated. The flow parameter,(L/V)(ρv/ρ l)0.5, where ρv and ρ l are the density of vapor and liquid, respectively,is an important parameter when considering the flood point or maximum loadingon the column. Assuming the densities (ρv and ρ l) to be relatively constant in therange investigated, L/V is plotted against pressure drop and HETP. Figure 8 showsthe pressure drop against the maximum L/V predicted in the computer simulation(normally at solvent feed stage).

The column packed with the structured packing shows a constant pressure dropover a L/V range of 15–27, whereas the column packed with protruded packingshows a slight decrease in pressure drop with increasing L/V, although the datapoints were scattered. A very good linear correlation was found when L/V deter-mined by the maximum liquid and vapor loadings in the column is plotted againstthe pressure drop, as shown in Fig. 9. Both packings show a pressure drop of about0.1 in. of water per foot of packing for every 0.2 increase in L/V. This is consistentwith established principles, as the stage with the maximum L/V plays the mostimportant role in determining pressure drop in a column and will be used for sizingthe column.

The HETP is also plotted against L/V for the protruded packing in Figs. 10and 11 for the solvent feed stage and for the stage with maximum vapor and li-quid loadings, respectively. Both figures show an increase in HETP with increas-ing L/V. However, there was no change in HETP with L/V for the structuredpacking.

It is concluded that both random and structured packings show a loss of effi-


FIG. 8 Pressure drop versus maximum predicted L/V (on solvent feed tray).

ciency under ED operation (or under very high L/V) as compared to conventionaldistillation. Both pressure drop per foot of packing and HETP are underpredictedby vendor correlations for both types of packings, although the predicted resultsfor the structured packing is closer to the actual results than those of the protrudedpacking. Also, compared to random packing, the structured packing appears to beless affected by the changes in gas capacity, liquid loading, and L/V. Further litera-ture data are required to confirm the superiority of the structured packing for extrac-tive distillation services.

FIG. 9 Pressure drop versus L/V determined by maximum liquid and vapor loadings.


FIG. 10 HETP versus L/V (on solvent feed tray) for protruded packing.

Handling Two Liquid Phases in EDC

One of the considerations in ED technology is the handling of possible formationof two liquid phases in a certain portion of the EDC where the less soluble compo-nents are concentrated. The occurrence of a second liquid phase is caused by thefact that some of the less polar components have significantly lower solubility inthe polar solvent than the more polar components.

One way to solve the problem of two liquid phases in the ED tower is to select

FIG. 11 HETP versus L/V determined by maximum liquid and vapor loadings for protruded packing.


a polar solvent, which has enough solvency to dissolve all the components in themixture under process condition. In general, however, solvents with a high selectiv-ity for compounds to be separated will have a reduced solvency (capacity), andvice versa. Therefore, in order to eliminate two liquid phases, one may have tocompromise the solvent selectivity, sometimes, to a great extent.

A better way is to cope with two liquid phases in the EDC, without sacrificingthe solvent selectivity, for following reasons:

1. Although two liquid phases normally reduce the solvent selectivity in a three-phase equilibrium (vapor–liquid–liquid) condition in the EDC, it can be com-pensated by intrinsic selectivity of a highly selective solvent. For example,the performance of sulfolane (SULF) was compared with those of N-formylmorpholine (NFM), N-methyl pyrrolidone (NMP), 2-pyrrolidone (2PD), anddimethyl sulfoxide (DMSO). The rough comparison was made through theirabilities to enhance the relative volatility of n-heptane over benzene (an aro-matic and nonaromatic separation) in a one-stage equilibrium cell. Table 2shows that although two liquid phases were observed using sulfolane as thesolvent, it still gave a better performance than other solvents when a singleliquid phase existed in the mixture.

2. Two liquid phases present no ill effects on the efficiency of small tray orpacked towers with diameter from 0.08 to 0.46 m. However, in a larger tower,the heavy liquid phase tends to accumulate on the tray if the liquid phases arenot well mixed. This problem can be eliminated by tray designs promotinggas agitation, causing the two liquid phases to behave as a homogeneous liquidthat followed general correlations for pressure drop, liquid holdup, brothheight, downcomer liquid level, and fractional entrainment. For larger packed

TABLE 2 Comparison of the ED Solvents for n-Heptane and Benzene Separation

Phase composition† (wt%)Liquid

Solvent S/F X1 Y1 X2 Y2 α1/2 phases

SULF 1.0 16.24 28.94 83.76 71.06 2.10 2SULF 3.0 16.32 43.76 83.68 56.24 3.99 2NFM 1.0 16.23 26.39 83.77 73.61 1.85 1NFM 3.0 16.23 36.84 83.77 63.16 3.01 1NMP 1.0 18.60 27.74 81.40 72.26 1.68 1NMP 3.0 18.60 35.71 81.40 64.29 2.43 1DMSO 1.0 15.21 26.97 84.79 73.03 2.06 1DMSO 3.0 15.21 37.54 84.79 62.46 3.35 12PD 1.0 14.74 24.32 85.26 75.68 1.86 12PD 3.0 14.74 34.76 85.26 65.24 3.08 1

Notes: SULF: sulfolane; NFM: N-formyl morpholine; NMP: N-methyl pyrrolidone; DMSO: dimethylsulfoxide; 2PD: 2-pyrrolidone.† X1 and X2 are the liquid compositions of n-heptane and benzene, respectively, and Y1 and Y2 are thevapor compositions of n-heptane and benzene, respectively (all in weight% on solvent-free basis).α1/2 � (Y1/X1)/(Y2/X2), the relative volatility of n-heptane over benzene.


columns, the liquid–liquid redistributor should be specially designed to allowseparate distribution of the two liquid phases [11].

Computer simulations have been developed which are capable of accuratelypredicting the development of two liquid phases in the EDC and the summary wasreported [12]. In one approach, the simulation algorithm starts from linearizedpressure, temperature and concentration profiles, and feed conditions given by theprogram operator. New estimates of composition are solved using the materialbalance and equilibrium relationship for each tray. Then, the equilibrium constantsare reestimated and a new temperature gradient is established to calculate a tray-by-tray energy balance. Accumulated errors are calculated for the energy, material,and equilibrium balances. Appropriate column operation restraints are factored inat this point. A correction factor is found for the temperature, rate profiles, andliquid composition profile by inverting the accumulated error matrix. These correc-tion factors are used to form new estimates of composition to start the process againuntil the correction factors are small enough to call the components converged.

Multicomponent vapor–liquid and liquid–liquid equilibria solutions are re-quired for the algorithm. Two activity coefficient models, NRTL and UNIQUAC,are readily extendable to multicomponent systems and capable of such solutions.Experimental activity coefficients, γ, at infinite dilution are used for calculatingbinary parameters for the NRTL equation. These parameters are then tested usingexperimental liquid–liquid ternary data, experimental vapor–liquid equilibriumdata, and data from pilot plant or commercial plant. The NRTL equation is usedin the algorithm to calculate activity coefficients and is given by the followingequations:

ln γ1 � x 22�τ21� G21

x1 � x2G21�

2

�τ12G12

(x2 � x1G12)2�ln γ2 � x 2

1�τ12� G12

(x2 � x1G12)�2

�τ21G21

(x1 � x2G21)2�where

ln G12 � �β12τ12, ln G21 � �β21τ12

τ12 �η12 � S12T

RT, τ21 �

η21 � S21T

RT

where Gij, η ij, Sij, τ ij, and β ij are empirical constants, γ i is activity coefficient, Ris the gas constant, T is the absolute temperature, and xi is the liquid-phase molefraction of component i. A Newton–Raphson-based flash algorithm checks for twoliquid phases by checking Gibbs free energies for possible second liquid-phasecomponents. If two liquid phases are indeed present, regular solution theory pro-vides a method of combining the liquid-phase activity coefficients.


Some Unique Applications of ExtractiveDistillation

Extractive distillation technology has been practiced and continuously improved com-mercially since World War II when it gained commercial recognition in the recovery ofhigh-purity butadiene, isoprene, and C4 olefins from the C4 and C5 petroleum streams.However, ED technology for recovering high-purity heavier petrochemicals from thepetroleum streams has gained commercial importance only recently.

Many ED solvents have been studied during the past 50 years to determinetheir selectivity for purifying heavier hydrocarbons. As shown in Table 3, a numberof solvents selective for aromatic recovery from petroleum streams have been listedin the literature [13]. However, none of the solvents listed in Table 3 have gainedcommercial importance for BTX aromatic recovery from the petroleum streams.

The modern state-of-the-art ED technologies for BTX aromatic recovery arebased on several solvent systems: SULF, NFM, and NMP. In most cases, proprie-tary cosolvents are added to the base solvents to enhance the solvent performance.

The modern ED processes can compete very favorably with, for example, liq-uid–liquid extraction based on sulfolane, which has dominated the BTX aromaticrecovery field for many years. However, in the following cases, ED technologymay be the preferred or the only choice.

Feeds with High Aromatic Content

The recovery of BTX aromatics from pyrolysis gasoline, which contains 80% � ar-omatics is one example. The high aromatic content tends to prevent the interfaceformation between the raffinate and the extract phases in the liquid–liquid extrac-tor, making the process inoperable.

Today, two of the leading ED processes for BTX aromatics recovery are of-fered by GTC Technology Corporation (GT-BTXSM process) and Krupp Koppers(Morphylane process) [14,15]. The following is an example using the GT-BTXprocess to recover high-purity BTX aromatics from the full-range (C6–C8) pyroly-

TABLE 3 ED Solvents for Aromatics Recovery from Petroleum Streams

Furfural Acetonyl acetone Nitrobenzene Nitrotoluene Phenol

Aniline Dichloroethy- Phenyl cellosolve Cresollether

Phenol-cresol Nonanoic acid Sodium-o-xylenesulfonate �H2O

Phthalic anhydride 2-Ethylhexanol Propylene glycolHexyleneglycol 2-Ethyl hexy- o-Chloroamine o-Phenetidine

lamineo-Chlorophenol o-Nitrophenol o-Phenyl phenolMethyl salicylate Dimethyl ani- Dipheyloxide

line–aniline


sis gasoline, which contains around 90% aromatics. This type of feed is unsuitablefor liquid–liquid extraction (LLE), because the aromatic content is so high that itprevents the formation of the interface between the extract and raffinate phases,which is necessary for LLE operation. A part of the raffinate stream from the LLEunit is often recycled to reduce the aromatic content in the feed stream (to ensurethe interface formation) and, thus, reduces the efficiency of the process.

To illustrate the application of ED technology to high aromatic containing feed,two different feedstocks were tested in a pilot plant consisting of a 60-tray extrac-tive distillation column and a packed solvent stripper for solvent recovery. Thesimplified feed compositions are given in Table 4.

The pyrolysis gasoline feed is introduced to the middle portion of the EDCnear its bubble point. Lean solvent is fed near the top of the EDC at about 10°Cbelow the column temperature to generate internal reflux to improve the columnperformance. The solvent preferentially extracts the more polar components in themixture, allowing the nonaromatic components to rise as vapor to the top of thecolumn as the raffinate product. The bottoms of the column consist of the solventand the aromatic components; these are fed to a solvent stripper (containing 9.5m of random packing) to separate the solvent from the extract products. The leansolvent is then recycled to the top of the EDC. A schematic diagram is presentedin Fig. 12. The analyses for the product streams are summarized in Table 5.

The solvent-to-feed ratio (weight) for both feeds was 3.0. The recovery forbenzene, toluene, and mixed xylenes, were respectively 96.5–97.0%, 99.0%, and99.9% by weight. In order to determine the purity of BTX aromatics producedwithout fractionating the extract product, a more detailed analysis was carried outto determine the nonaromatics in the feedstock. Table 6 shows the major compo-nents and their boiling points in Feed No. 1.

From Table 5, the nonaromatics in the extract (product) stream was 0.54 wt%,and the split of bottom (extract) to feed in the ED column was 0.87. The nonaromaticsin the extract is equivalent to 0.47 wt% (0.54 wt% 0.87) of the heaviest nonaromat-ics in the feed. According to Table 6, these nonaromatics were iso-nonanes (0.16 wt%)and the heavy portion of iso-octanes (0.31 wt%). According to material balances,commercial grade of benzene (with 99.9 wt% purity), toluene (with 99.0 wt% purity),and mixed xylenes (with 98.5 wt% purity) can be recovered from the extract productby distillation. The impurities in benzene and toluene products were the 0.31 wt%heavy iso-octanes in the feed, whereas the impurities in mixed xylenes were the 0.16wt% iso-nonanes in the feed. These experimental results demonstrate the effectivenessof a ED process (GT-BTX process) in recovering chemical grade BTX from a full-range (C6–C8) pyrolysis gasoline containing high aromatics.

TABLE 4 Composition of Pyrolysis Gasoline for BTXAromatics Recovery

Component (wt%) Feed No. 1 Feed No. 2

Nonaromatics 11.50 7.89Benzene 49.28 48.83Toluene 27.79 29.38C8-aromatics 11.35 13.90C9

�-aromatics 0.08 0.00


FIG. 12 Schematic diagram of ED process.

Heavy Aromatics Recovery

Liquid–liquid extraction is more effective to extract lighter aromatics, such asbenzene and toleune, but less effective to extract heavier aromatics, such xylenesand C9

�-aromatics. The latter compounds have relatively less solubility in the ex-tractive solvents than the former compounds. Extractive distillation, on the otherhand, will provide higher recovery for the heavier aromatics, which tend to staythe solvent at the bottom of the ED column, due to their higher boiling points.However, the solvent boiling point has to be high enough to allow a clean separa-tion between the extract product and the solvent by stripping or distillation in thesolvent recovery column. The performance of ED technology for recovering C8

to C9� aromatics has not been significantly reported in the literature.

To determine the performance of GT-BTX process, two heavy aromatic feedswith composition shown in Table 7 were investigated in a pilot plant consisting

TABLE 5 Composition of Product Streams from the ED Unit for High Aromatic Feeds

Feed No. 1 Feed No. 2Component

(wt%) Raffinate Extract Raffinate Extract

Nonaromatics 84.84 0.54 80.93 0.44Benzene 13.30 54.66 15.69 52.22Toluene 1.87 31.66 3.23 32.06

C8-Aromatics 0.11 13.04 0.15 15.31C9

�-Aromatics 0.00 0.10 0.00 0.00


TABLE 6 Components and Their Boiling Points of Feed No. 1

Boiling pointComponent Wt% (°C)

Cyclopentane 0.22 49.32-Methylpentane 0.55 60.33-Methylpentane 0.36 63.3n-Hexane 1.21 68.7Methyl cyclopentane 3.57 71.8Benzene 49.65 80.1Cyclohexane 1.29 80.72-Methylhexane 0.30 90.12,3-Dimethyl pentane 0.19 89.83-Methylhexane 0.21 91.81-cis-3-Dimethyl cyclopentane 0.20 90.81-tr-3-Dimethyl cyclopentane 0.12 91.71-tr-2-Dimethyl cyclopentane 0.24 91.9n-Heptane 0.32 98.42,2-Dimethylhexane 0.07 106.8Methyl cyclohexane 0.47 100.92,4-Dimethylhexane 0.52 109.4Toluene 27.50 110.6n-Octane 0.07 125.7Iso-octane 1.36 107–119Ethylbenzene 6.38 136.2p-Xylene 0.42 138.3m-Xylene 3.32 139.1o-Xylene 1.18 144.4Iso-nonanes 0.16 122–151C9-aromatics 0.38 152–176

Total 100.26

of a 60-tray extractive distillation column and a packed solvent stripper for solventrecovery.

The solvent-to-feed ratio (weight) for both feeds was 3.0. Based on the feedanalyses, the kettle temperature and pressure of the ED column were adjusted toachieve the split of bottom-to-feed ratio of approximately 0.87 for Feed No. 1 and0.76 for Feed No. 2. The solvent stripper (solvent recovery column) was operated

TABLE 7 Composition of Heavy Aromatics Feeds for ED Study

Component (wt%) Feed No. 1 Feed No. 2

Nonaromatics 12.51 24.03Benzene 0.31 0.40Toluene 2.66 1.52C8-aromatics 63.82 59.91C9

�-aromatics 20.71 14.17


TABLE 8 Composition of Product Streams from the ED Unit for Heavy AromaticFeeds

Feed No. 1 Feed No. 2

Components Extract Extract(wt%) Raffinate Extract Recovery Raffinate Extract Recovery

Nonaromatics 85.10 0.11 0.75 87.75 1.09 3.34Benzene 1.95 0.00 0.00 1.51 0.00 0.00Toluene 3.02 2.59 83.29 2.33 1.31 63.38C8-aromatics 8.07 73.16 98.06 8.51 78.95 96.91C9

�-aromatics 2.15 23.81 98.34 0.05 19.03 98.76

with stripping gas under the proper level of vacuum to minimize the stripper tem-perature. The analyses of the production streams are presented in Table 8.

Shown in Table 8, under a solvent-to-feed ratio of 3.0, the overall aromaticrecovery and purity have achieved the commercial requirements. The overall aro-matic recovery for Feed No. 1 was 97.7 wt%, and for Feed No. 2, it was 95.7wt%, and the overall aromatic purity for Feed No. 1 was 99.89 wt%, and for FeedNo. 2, it was 98.91 wt%. It is also observed that under similar operating conditions,the process performance from Feed No. 1 is indeed better than that of Feed No.2, probably due to higher aromatic content in Feed No. 1 (87.5 versus 76 wt%).

Revamping Existing LLE Processes Using the ED Method for BTXRecovery

The ED method can be most effectively applied to revamping an existing LLEfacility for recovering BTX aromatics from reformate or pyrolysis gasoline. Thisapproach has been recently reported by Gentry and Kumar [16]. In a typical LLEprocess where the solvent has higher density than the feed, the feed is fed intothe lower section of the LLE column and flows countercurrently against the sol-vent. The solvent is introduced to the top of the column and flows downwardthrough the column to preferentially extract the more polar components in the feedinto the bottoms. The bottoms stream is then fed to a solvent recovery column torecover the purified more polar components in the overhead and the lean solventas the bottoms to be recycled to the LLE column.

BTX aromatics are recovered using LLE with sulfolane as the extractive solvent.A typical schematic diagram of this type of process is shown in Fig. 13. Reformatecontaining BTX aromatics and the extractive solvent are fed into the LLE columnaccording to the manner described in previous paragraph. The solvent extracts thearomatics and some nonaromatics into the bottoms, which is fed to the extractivestripping column. This column strips the nonaromatics from the solvent and aromatics,for recycle to the LLE column. The bottoms from the extractive stripping columncontain the aromatics and solvent are separated in a solvent recovery column. Leansolvent from the solvent recovery column is recycled back to the LLE column. Theraffinate stream from the top of LLE column contains some solvent, which is recoveredin a water-wash column. The water is returned to the system in a closed-loop recycle.


FIG. 13 Schematic diagram of LLE process using sulfolane solvent.

A fundamental aspect of this process is that the solvent exhibits a selectivity fa-voring lower-boiling components more than the high-boiling components. The solventselectivity favors the hydrocarbon species according to the following sequence: aro-matics � naphthenes/olefins � paraffins. Lighter nonaromatic impurities are morelikely to be coextracted due to their solvent affinity in LLE column, but they shouldbe the easiest to remove in the ED column due to their lower boiling point.

Adding just one ED column to the existing LLE process system can not onlysubstantially increase the process throughput but also improve the performance interms of product quality and purity. The following are some examples.

Case A: Purifying By-product Benzene

The xylene isomerization and toluene disproportionation units within the aromaticcomplex produce benzene as a by-product, but the quality of benzene is low andrequires reprocessing in the LLE unit to upgrade its purity. Figure 14 shows a newapproach using a hybrid of the LLE process with ED that bypasses part of thefeed around the original extraction section.


FIG. 14 A hybrid of LLE and ED process purifying by-product benzene.

In the hybrid scheme, the by-product (benzene-rich feedstock) is fed to thestand-alone ED column, where the maximum aromatics limit in the feed chargeis not a concern. The original feed to the extraction section is not affected by thisnew ED operation. The rich solvent from both LLE and ED operations are com-bined into the existing solvent-recovery column. The typical solvent-recovery col-umn can usually manage the higher capacity or can be redesigned to do so. Ifdesired, the raffinate from the two can be segregated for optimum disposition; theED raffinate rich in cyclohexane may be recycled to the reformer unit, whereasthe LLE raffinate rich in paraffins could be blended into the feedstock for a naphthacracker. The conventional LLE process can be retrofitted to use this hybrid process,without requiring extensive modifications, investment, or shutdown time. The pri-mary changes are modifying the solvent system to be compatible with both LLEand ED operations and to make the appropriate tie-ins to the ED column.

Case B: Optimize the BTX Product Distribution

Among the BTX aromatics, toluene has historically contributed the least upgradeover its alternative value in motor fuel (e.g., the petrochemical value of toluene


FIG. 15 A hybrid of LLE and ED process for selective purge of toluene.

is only marginally above its fuel blending value); thus, it is not profitable to extracttoluene. In a conventional LLE unit for BTX aromatics recovery, toluene is inevita-bly extracted along with benzene and xylenes.

Figure 15 shows a LLE and ED hybrid scheme where toluene is selectivelypurged from a BTX mixture to shift the product mix toward benzene and xylenesand to avoid some operating cost for toluene. As shown in Fig. 15, the reformatesplitter is retrofitted to include a side cut of the C7 and C8 components, which isfed to an ED column operated in parallel with the existing LLE unit. Operationof the ED column is made to intentionally purge the majority of the toluene alongwith the nonaromatics to the column overhead. Then, the toluene-lean ED columnbottoms are combined with the benzene-rich solvent mixture from the extractivestripper bottoms as feed to the solvent-recovery column. A number of benefits fromthis operation can be realized: (1) overall processing costs are reduced because thestream contains less toluene; (2) xylenes recovered should have higher purity be-cause the majority of the C8–C9 nonaromatics are purged with toluene in the EDcolumn; (3) substantial increase in overall process capacity by adding only oneED column to the existing LLE system.

Case C: Capacity Increases

Adding a new ED column to an existing LLE unit (as shown in Fig. 13) can doublethe capacity of the unit. As demonstrated in Fig. 16, the new ED column takes


FIG. 16 A hybrid of LLE and ED for substantial increase of throughput.

the fresh feed while the main liquid–liquid extractor takes the feed from the EDcolumn overhead raffinate stream. Solvent is divided between the ED column andthe raffinate extractor. The original extractive stripper column is converted into asolvent-recovery column operating in parallel with the existing solvent-recoverycolumn. Using a single new ED column to retrofit the existing LLE unit can createsignificant increase in throughput and improvement in product quality with mini-mized capital costs.

Styrene Recovery from Pyrolysis Gasoline

One of the most difficult separations, using the ED method, is the purification ofstyrene from the close-boiling C8 aromatic isomers. The method is based on theslightly higher polarity in styrene than the other C8 aromatics, due to the doublebond in the side chain of the styrene molecule. Recently, a proprietary ED processwas developed for the commercial application for recovering styrene directly frompyrolysis gasoline [17].

In developing this process, a number of extractive solvent candidates werescreened in the laboratory for selectivity, solubility, and other important properties,such as thermal stability, toxicity, corrosivity, boiling point, freezing point, andso forth. The selected solvent underwent an extensive pilot-plant test program todetermine its performance in the ED process, and, consequently, to optimize the


TABLE 9 Composition of Feed and Products of the ED PilotPlant for Styrene Recovery

Feed Overhead BottomComponent (wt%) (wt%) (wt%)

Dicyclopentadiene 0.26 0.40 �1 ppmBenzene 0.22 0.23 �1 ppmToluene 35.81 43.68 �1 ppmVinylnorbornenes 3.73 4.25 �1 ppmEthyl Benzene 12.03 13.47 �1 ppmp-Xylene 2.97 3.30 �1 ppmm-Xylene 7.45 8.80 �1 ppmCumene 395 ppm 431 ppm �1 ppmo-Xylene 4.52 5.79 �1 ppmStyrene 23.82 10.31 1.11Allylbenzene 390 ppm 269 ppm 23 ppmNonaromatics 9.08 9.72 �1 ppm2,5-Dimethylthiophene 172 ppm 195 ppm �1 ppmSolvent 0 0.37 98.88

key process variables to support the computer process simulation for scale-up andcommercial design.

The pilot-plant tests were conducted in a 7.6-cm-diameter ED column packedwith 7.3 m of knitted wire-mesh packing (Goodloe Style #773). The C8 cut ofpyrolysis gasoline with the composition shown in Table 9 was fed at a location4.3 m from the column bottom and the lean solvent was introduced at a location6.4 m from the column bottom. A schematic diagram of the pilot plant can berepresented by Fig. 12. The column was operated under reduced pressure to mini-mize the column operating temperature, and a proprietary inhibitor was added intrace amounts to the feed to prevent styrene polymerization in the column. Thekey variables, such as S/F, the reflux ratio, and the kettle temperature, were prop-erly adjusted to yield styrene with 99.9 wt% purity (on a solvent-free basis) in thecolumn bottoms. The composition of the column overhead and the bottom productsare also given in Table 9. In this operation, a cosolvent was used as the key toimprove the styrene purity from 95 to 99.9 wt%, to reduce the column temperatureby 35°C, and to decrease the solvent circulation by 20%. The bottom product (therich solvent) was routed to the solvent stripper column, where styrene was distilledoverhead, and the lean solvent exited from the bottom of the column for recycleto the ED column.

The separation strategy for recovering styrene from pyrolysis gasoline is toremove the majority of the light and heavy components in pyrolysis gasoline byconventional distillation and to use ED to remove the remaining close-boilingcomponents. Figure 17 shows a possible scheme for processing pyrolysis gaso-line, which includes styrene recovery by ED. The main benefits for recoveringstyrene from pyrolysis gasoline generated from a naphtha cracking complex areas follows:


FIG. 17 Schematic diagram for pyrolysis gasoline processing including styrene recovery.

1. The styrene component is upgraded from motor fuel value to petrochemicalvalue.

2. The xylenes can be upgraded from motor fuel value to the feedstock value toxylene isomer unit.

3. The overall hydrogen consumption to convert styrene to ethylbenzene is re-duced.

4. The catalyst fouling and operating costs in the selective hydrotreater are re-duced.

5. The potential debottleneck of the hydrotreating area can be avoided; it other-wise would be required for naphtha cracking expansion.

The economics of styrene recovery from pyrolysis gasoline depend to a greatextent on the quantity of styrene available in this stream. Units produce at 10,000to 15,000 metric tons per year of styrene in pyrolysis gasoline would be the candi-date for this technology. Table 10 shows that the economics for styrene recoverywith a pretax return on investment (ROI) of 41% can be achieved, based on a25,000-metric tons per year plant capacity.

Cyclohexane Recovery from Natural Gas Liquid or Naphtha

Cyclohexane exists naturally in naphtha and natural gas liquid (NGL) streams andis an important raw material for the nylon industry. As shown in Table 11, it isimpossible to recover high-purity cyclohexane from these streams by conventionaldistillation, because of the close-boiling C7 iso-paraffins in the streams. Because


TABLE 10 Economics for Styrene Recovery from PyrolysisGasoline

Typical U.S. Gulf Coast capital cost $ 20 millionStyrene value in pyrolysis gasoline $ 180 per metric tonStyrene product sales value $ 550 per metric tonProcessing cost $ 40 per metric tonGross margin $ 8.3 million per yearPretax return on investment 41%

Note: Basis: 25,000 metric tons per year styrene capacity from a world-scalenaphtha cracker. Values on styrene product, feedstock, processing cost, andcapital investment were calculated based on the 1997 published information.

cyclohexane and the close-boiling components in the feed mixture have only a verysmall difference in polarity and are relatively insoluble in the selective solvents, italso takes a very difficult ED process to do the separation. Because no effectivesingle ED solvent has been found, a mixed solvent was developed commerciallyto recover high-purity cyclohexane directly from an NGL fraction containing 85%cyclohexane [18].

The proprietary mixed solvent (the MIST solvent) was developed through ex-tensive test in the laboratory and the evaluation in a 150-mm-diameter ED pilotplant using a refinery NGL stream, which has an average composition shown inTable 11. In fact, the cosolvent in MIST solvent played a major role in the successof this process [18]. As demonstrated in Fig. 18, The overall cyclohexane recoverychanged from 100 to 56 wt% as the cosolvent concentration decreased from 30to 10 wt% under a constant kettle temperature of the ED column. Meanwhile, therecovery of 2,4-dimethylpentane in the raffinate stream (ED column overhead)increased from 87 to 96.3 wt% over the same composition range of the MISTsolvent. The higher 2,4-dimethylpentane in the raffinate stream, the higher cyclo-hexane purity in the extract (product) stream.

Table 12 summarizes the results for 80% and 90% cyclohexane recoveries atvarious cosolvent concentrations. At an 80% cyclohexane recovery, MIST solvent

TABLE 11 Average Composition and Boiling Point of theFeedstock for Pilot-Plant Testing

Boiling pointComponent Weight% (°C)

Cyclohexane 89.1 80.72,2-Dimethylpentane 1.3 79.12,4-Dimethylpentane 4.0 80.43,3-Dimethylpentane 0.1 86.02,3-Dimethylpentane 0.9 89.72-Methylhexane 1.6 89.93-Methylhexane 1.1 91.92,2,3-Trimethylbutane 0.8 80.8Dimethylcyclopentane 1.0 90.6n-Heptane 0.1 98.3


FIG. 18 Effect of cosolvent in MIST solvent on product purity and recovery.

containing 10% cosolvent appears to give the greatest cyclohexane purity. At 90%cyclohexane recovery, MIST solvent having 25% cosolvent had similar perfor-mance to that with 10% cosolvent. Finally, the MIST solvent with 25% cosolventshowed better performance than the one with 10% cosolvent. Based on the success-ful pilot-plant testing on the MIST solvent, a commercial plant for purifying 100metric tons per day cyclohexane was designed, constructed and started up in 1991.

TABLE 12 Effect of Cosolvent Concentration on Cyclohexane Recovery and Purity

2,4-Cyclohexane Cyclohexane Dimethylpentane

Cosolvent (wt%) Recovery (wt%) Purity (wt%) Recovery (wt%)

25 80 98.9 89.820 80 99.0 92.010 80 99.1 93.030 90 99.2 87.225 90 99.2 90.010 90 99.2 90.325 94 99.1 86.810 94 99.1 80.0

Polyolefins Produced by Single-Site Catalyst 231

References

1. A. P. Colburn and E. M. Schoenborn, Trans. AIChE, 43, 42 (1945).2. G. T. Atkins and C. M. Boyer, Chem. Eng. Prog., 45, 553 (1949).3. J. M. Chambers, Chem. Eng. Prog., 47, 555 (1951).4. K. H. Hackmuth, Chem. Eng. Prog., 48, 617 (1952).5. R. M. Butler and J. A. Bichard, U. S. Patent 3,114,783 (1963).6. R. E. Brown and F. M. Lee, ‘‘Effect of Packing on Distillation Columns with High

Liquid to Vapor Ratios,’’ AIChE Annual Meeting, Miami Beach, FL, 1992.7. R. Billet, ‘‘Packed Column Analysis and Design,’’ Department for Thermal Separa-

tion Processes, Ruhr-University Bochum, 1989, p. 11.8. Cannon Instrument Co., ‘‘Pro-Pak Protruded Metal Distillation Packing,’’ Bulletin

23, Cannon Instrument Co., State College, PA.9. R. Billet, ‘‘Packed Column Analysis and Design. Department for Thermal Separation

Processes,’’ Ruhr-University Bochum, 1989, p. 41.10. D. P. Kurtz, et al., Chem. Eng. Prog., 87,(2), 43 (1991).11. C. C. Herron, Jr., B. K. Kruelskie, and J. R. Fair, AIChE J., 34,(8), 1267 (1988).12. B. A. Todd and F. M. Lee, ‘‘Two Liquid Phases in Extractive Distillation for Aromatic

Recovery,’’ AIChE Summer National Meeting, Seattle, WA, 1992.13. M. Van Winkle, Distillation, McGraw-Hill, New York, 1967, p. 464.14. J. C. Gentry and F. M Lee, ‘‘New Extractive Distillation Process for Aromatics Recov-

ery,’’ AIChE Spring National Meeting, Houston, TX, 1995.15. G. Emmrich, ‘‘Morphlane: Operational Experience and Results Obtained with New

Plants,’’ NPRA Spring Meeting, San Francisco, 1995.16. J. C. Gentry and C. S. Kumar, Hydrocarbon Process, 69, (March 1998).17. F. M. Lee and J. C. Gentry, Hydrocarbon Eng., 3(6), 62 (1998).18. R. E. Brown and F. M. Lee, Hydrocarbon Process, 83 (May 1991).

FU-MING LEE

Structure, Properties, and Applicationsof Polyolefins Produced by Single-SiteCatalyst Technology

Introduction

Polyethylene is composed of only carbon and hydrogen (with some exceptions),which can be combined in number of various ways to make many different polyeth-ylenes. These can generally be grouped into six types:

232 Polyolefins Produced by Single-Site Catalyst

• LDPE, low-density polyethylene• EVA, ethylene vinyl acetate copolymers• HDPE, high-density polyethylene• LLDPE, linear low-density polyethylene• ULDPE, ultralow-density polyethylene (also known as Very low-density poly-

ethylene, VLDPE)• Single-site polyethylenes (e.g., substantially linear homogeneous polyethylene)

Polyethylene technology encompasses a range of crystallinity and meltingpoints. Figure 1 illustrates the approximate range of densities, crystallinities, andmelting points available within each major type of polyethylene. Specialty ethylenecopolymers such as acid copolymers and ionomers are not included. The totalpolyethylene consumption is expected to exceed 100 billion pounds by the year2000. Applications include films, extrusion coatings, injection-molded and blow-molded parts, fibers, adhesives, wire and cable coatings, large parts such as storagetanks made by rotational molding process, foams, and so forth.

The most recent advance in polyethylene technology is the development andcommercialization of polyethylenes produced by single-site-catalyst technology.Two major classes of single-site catalyst (SSC) technology developed for the poly-merization of ethylene and α-olefins are the metallocene catalyst (MTC) and theconstrained-geometry catalyst (CGC) systems. The use of these catalyst technolo-gies has allowed a very rapid development of olefin copolymers with a wide rangeof structures and related properties. This technology has initiated a major revo-lution for the polyolefin industry [1–4]. Several families of MTC- and SSC-technology-based polyolefin copolymers have been commercialized in the 1990s.These include polyolefin elastomers (e.g., ENGAGE from DuPont Dow Elasto-mers, LLC), polyolefin plastomers (e.g., AFFINITY from The Dow ChemicalCompany; EXACT from Exxon Chemical Company), EPDM (NORDEL IPfrom DuPont Dow Elastomers, LLC), enhanced polyethylene (ELITE from TheDow Chemical Company); gas-phase LLDPE (EXCEED from Exxon ChemicalCompany); slurry LLDPE (mPACT from Phillips Petroleum Company), and

FIG. 1 The approximate range of density, crystallinity, and melting point available within each majortype of polyethylene.


polypropylenes (ACHIEVE from Exxon Chemical Company). In addition tothese commercial activities, several other single-site-catalyst-related technologiesthat allow the copolymerization of α-olefins with polar comonomers are also underindustrial and academic development [5,6].

This article will focus on the solid-state structure, rheology, properties, andtypical applications of homogeneous polyolefin copolymers made by copolymer-ization of ethylene and α-olefins (e.g., 1-butene, 1-hexene, 1-octene) using SSCtechnology. Unlike conventional linear low-density polyethylenes (LLDPE) madeby copolymerization of ethylene and α-olefins with Ziegler–Natta (Z-N) catalysts,ethylene–α-olefin copolymers produced by SSC technologies have narrow compo-sition distributions (narrow molecular-weight and comonomer distributions).Hence, they are called homogeneous copolymers and they behave much more likeideal polymers. The polymerization kinetics and the resulting polymer and copoly-mer structures can be modeled. This significantly advances the fundamental under-standing of the structure–property relationships. The major advantage of single-site catalysts is its versatility in building well-defined molecular structures. Thiscapability allows the polymer and material scientists to design new polymers usinga molecular architecture approach and to develop new products with exceptionalspeed.

Preparation of Ethylene–α-Olefin Copolymersby SSC Technology

Two major families of high-efficiency SSC are commercially used for the pre-paration of polyethylene copolymers. These are a bis-cyclopentadienyl (Bis-Cp)single-site metallocene catalyst (also known as a Kaminsky catalyst) and a half-sandwich, constrained-geometry mono-cyclopentadienyl single-site catalyst(known as constrained-geometry catalyst, CGC, under the trademark of INSITEby The Dow Chemical Company). These two catalyst systems are illustrated inFig. 2. The Zr in the Zr-based MTC used by many polymer producers has an

FIG. 2 Structure of the Bis-Cp metallocene catalyst and the constrained geometry catalyst.


oxidation state of �4 and the Ti in the Ti-based CGC system has an oxidationstate of �2.

Single-site-catalyst technology polyolefin copolymers can be produced byhigh pressure, solution, gas-phase, and slurry polymerization processes. Typicalprocess conditions for making polyolefin copolymers in these processes are asfollows:

1. High-Pressure Process: Two types of commercial reactor, stirred autoclaveand a tube reactor, are used for the polymerization of ethylene and ethylene–α-olefin copolymers at high pressure. Ethylene and an α-olefin comonomerare usually compressed to at least 10,000 psi when fed into the reactor. Thepolymerization temperature usually exceeds 100°C.

2. Solution Process: Stirred reactors are usually used for the polymerizationof ethylene and ethylene–α-olefin copolymers in a solution phase. In mostcases, C6–C8 hydrocarbons are used as the solvent. The reactors are oper-ated under about 500 psi pressure and the polymerization is carried out at�60°C.

3. Gas-Phase Process: In the gas-phase process, ethylene and α-olefin como-nomers (1-butene and/or 1-hexene) are polymerized in the solid state in afluidized-bed reactor into a powder form. The polymer powder is then con-verted into pellet form by an extrusion process. The reactor pressure is usuallyset at �300 psi and the polymerization process is generally carried out at�90°C.

4. Slurry Process: In the slurry process, polymers are made in stirred or loopreactors with an organic liquid carrier (C 4–C6 hydrocarbons). The polymeriza-tion process temperature is generally carried out at �90°C at a reactor pressureof �300 psi. Products made in this process are in powder form and are con-verted into pellets using extrusion processes.

Molecular Structure of Polyethylene CopolymersMade by Single-Site-Catalyst Technology

Molecular-Weight Distribution

Ethylene–α-olefin copolymers prepared by single-site-catalyst technologies ex-hibit a narrow molecular-weight distribution (MWD). MWD, or polydispersity, isthe ratio Mw/Mn, where Mw is the weight-average molecular weight and Mn is thenumber-average molecular weight. Copolymers prepared by a conventionalZeigler–Natta catalyst have a MWD usually larger than 3, whereas the MWD ofhomogeneous polyolefins made by SSC technology is usually less than 2.5. Figure3 compares gel permeation chromatography (GPC) traces of a conventional, het-erogeneous ethylene–octene copolymer versus a homogeneous copolymer. Poly-mers with a narrow MWD in general have increased toughness and less ‘‘solventextractables.’’ On the other hand, the narrow MWD of a linear homogeneous co-


FIG. 3 GPC MWD comparison of two ethylene–octene copolymers having a 0.920-g/cm3 density.

polymer also results in poor melt processability (low melt strength, high extruderback-pressure, high-energy consumption during extrusion, etc.).

Crystallinity

The comonomer content in the copolymer has a profound effect on the propertiesof the polymer, including crystallinity, thermal, and mechanical properties. In thecase of ethylene–α-olefin copolymers, the crystallinity is measured and specifiedusing density. The density of the amorphous phase (ρa) and crystalline phase (ρc)of ethylene–α-olefin copolymers at room temperature is about 0.853 g/cm3 and1.000 g/cm3, respectively. Weight percent crystallinity can be obtained from themeasured density (ρ) as follows:

Wt% cryst. �ρc

ρ �ρ � ρa

ρc � ρa� (1)

The effect of comonomer content on the density of homogeneous ethylene–octeneand ethylene–butene copolymers, made using a CGC catalyst, is illustrated inFig. 4. The comonomer content was measured using nuclear magnetic resonance(NMR). Higher mole% butene is required to achieve the same density comparedto octene, especially at lower densities.

Comonomer Distribution

Because of the significant effect of the α-olefin comonomer on the properties ofthe polymer, it is critical to understand and to measure the comonomer distribution


FIG. 4 Densities of homogeneous ethylene–octene and ethylene–butene copolymers as a functionof mole% comonomer.

among the polymer molecules (intermolecular distribution) and along the backbone(intramolecular distribution). Intermolecular comonomer distributions can be mea-sured by the temperature rising elution fractionation (TREF) technique [7]. In thisprocedure, a heated solution of the polymer is placed in a column, packed withvery small stainless-steel shot, and slowly cooled to room temperature. The temper-ature is then increased while solvent flows through the column to a detector thatrecords the amount of polymer in the solvent at any given temperature. As illus-trated in Fig. 5, the intermolecular comonomer distribution for a homogeneousethylene–octene copolymer prepared by the CGC technology is much narrowerthan that of a conventional LLDPE, which is a mixture of polymer moleculeshaving different levels of an α-olefin comonomer incorporated in the polymerbackbone.

The narrow TREF curve for the homogeneous copolymer signifies that thenumber of comonomer units per unit chain length between the copolymer mole-cules is very similar. However, although the intermolecular comonomer distribu-tion for the homogeneous polymer is very narrow, this does not mean that theintramolecular comonomer distribution is uniform. The narrow intermolecularcomonomer distribution for the homogeneous polymer arises from the single-sitenature of the MTC and CGC. The uniformity of the intramolecular comonomerdistribution, however, is dictated by the reactivity ratio of the monomer and thecomonomer. Intramolecular comonomer distribution in homogeneous copolymersmade by SSC technology can also have an effect on the solid-state structure andproperties of the polymer (e.g., thermal properties, optics, etc.).

To address the issue of the effect of the intramolecular comonomer distributionon polymer properties of a homogeneous polymer, structural characteristics for


FIG. 5 Intermolecular comonomer distribution of ethylene–octene copolymers measured by TREF.(From Ref. 10.)

CGC ethylene–octene copolymers were modeled using the Monte Carlo simulation[8]. Figure 6 illustrates the intramolecular comonomer distribution for three CGCcopolymers, from 0.87 g/cm3 to 0.92 g/cm3 density, in terms of ethylene blocklength between the hexyl branches within a polymer molecule. As illustrated inFig. 6, the intramolecular comonomer distribution for the higher-density polymer(0.92 g/cm3), in terms of ethylene block-length distribution between the shortbranches that are formed from the α-olefin comonomers, is much broader thanthat of the lower-density copolymers. This is due to the fact that the lower-density

FIG. 6 Model prediction for intramolecular comonomer distribution of three CGC ethylene–octenecopolymers: ethylene block-length distribution.


copolymers have more comonomer units along the polymer backbone. This resultsin a shorter ethylene block length between the short branches, and the distributionis also narrower.

Long-Chain Branching

Homogeneous ethylene homopolymers and ethylene–α-olefin copolymers firstmade by CGC technology have one more unique molecular structural feature notfound in the original MTC technology polymers. This unique molecular structuralfeature is long-chain branching (LCB). Homopolymers and copolymers made bythe CGC technology at certain process conditions contain a controlled amountof long-chain branching. These polymers are referred to as substantially linearhomogeneous polyethylenes. To summarize this, the molecular structure of threeethylene–α-olefin copolymers (conventional LLDPE made by a Z-N catalyst, ho-mogeneous copolymers made by MTC, and substantially linear homogeneous co-polymers made by CGC) are schematically illustrated in Fig. 7. One possible mech-anism of LCB formation via CGC technology is as follows. Thermal terminationis one of the most effective ways to control molecular weight during the polymer-ization process. In the thermal termination process, polymeric molecules with avinyl chain end are formed due to a ‘‘β-hydride elimination’’ mechanism that istaking place in the thermal termination step (Fig. 8). Depending on the last mono-mer or comonomer insertion, the unsaturation could be vinyl, vinylidene, or transas shown in Fig. 8. The polymer with a vinyl chain end becomes one of the re-actants (in addition to ethylene and α-olefin) that reacts with the catalyst site andforms long-chain branching.

A polymer with LCBs made by the CGC technology has many unique rheologi-cal properties [9,10]. A few of the most significant features for the CGC-technologypolymer are its melt fracture resistance and the control of shear-thinning behavior,as illustrated in Figs. 9 and 10 respectively. In Fig. 10, I2 is the melt index (MI)of the polymer (flow rate in grams per 10 min at 2.16 kg weight, at 190°C) and

FIG. 7 Molecular structure comparison of three different ethylene–α-olefin copolymers. (EXACT isa trademark of Exxon Chemical Company. AFFINITY is a trademark of The Dow ChemicalCompany.)


FIG. 8 Reactor sources of unsaturation in polyethylene during polymerization of ethylene and α-olefin comonomers. (From Ref. 10.)

I10 is the flow rate in grams per 10 min at 10 kg weight, at 190°C. The I10/I2 ratiois a measure of shear-thinning behavior of the polymer melt.

Solid-State Structure and Morphologyof SSC-Technology Ethylene Homopolymersand Ethylene–α-Olefin Copolymers

The ethylene homopolymer and ethylene–α-olefin copolymers are semicrystallinepolymers. Three major factors that affect the crystal morphology and crystallinityof ethylene polymers in the solid state are (1) the amount and size of α-olefin inthe polymer, (2) the molecular weight of the polymer, and (3) the crystallization

FIG. 9 Extrudates of CGC polymer and LLDPE at 3.66 106 dyn/cm2 shear stress. (Irgafos is atrademark of Ciba-Geigy Corporation.)


FIG. 10 Rheology of CGC polymers measured at 190°C: effect of LCB on shear-thinning behaviorof three 1 MI polymers with various amount of LCB. LCB/10,000°C was estimated usinga kinetic model.

conditions (crystallization temperature at isothermal condition and/or cooling rateat nonisothermal crystallization conditions) [11]. A recent review article on poly-mer crystals was written by Phillips [12]. The most generally accepted crystalmorphology for ethylene polymers is lamellae formed from a folded-chain confor-mation [13–15].

For very low-density SSC-technology ethylene–α-olefin copolymers (whichcontain a high percentage of comonomer), the crystal morphology can be verydifferent from the conventional lamella model. For example, let us consider themorphological model of a 60/40 wt% ethylene–octene copolymer made by CGCtechnology. At this level of comonomer content, the polymer exhibits about 10%crystallinity, as measured by differential scanning calorimetry (DSC) and a densityof approximately 0.87 g/cm3. The intramolecular comonomer sequence distribu-tion of the hexyl branches from the octene comonomer, calculated from the reac-tive-ratio kinetic model (Fig. 6), illustrates that more than 90% of the octene unitsare less than 50 ethylene units apart. The regular ethylene block length betweenthe octene units is, therefore, much less than the minimum length to form onefolded unit to form the thinnest possible lamella (which is about 30 A thick). Fora chain with this molecular structure, it is expected that the polymer has to crystal-lize in a crystalline form different from the conventional folded-chain lamellamodel.

Figure 11 is a transmission electron micrograph of the crystal morphology ofa higher-density (0.920 g/cm3) ethylene–octene copolymer made by CGC technol-ogy. At this density and octene level, the copolymer clearly shows a lamellar mor-phology. Figure 12 is a transmission electron micrograph of a very low-densityethylene–octene copolymer (0.87 g/cm3). It clearly shows that the major crystallinestructure of this very low-density, low-crystallinity copolymer is ‘‘spotlike’’fringed micelle crystals. Polymers having higher crystallinities with the lamella-


FIG. 11 Transmission electron micrograph of a 0.920-g/cm3 ethylene–octene copolymer made byCGC technology.

FIG. 12 Transmission electron micrograph of a 0.87-g/cm3 ethylene–octene copolymer made by CGCtechnology.


FIG. 13 Classification of SSC ethylene–α-olefin copolymer based on crystal morphology. (FromRef. 16.)

type crystal morphology are expected to have higher moduli and undergo yieldingwhen deformed. Polymers having low crystallinities with the fringed micelle crys-tals, however, are expected to be more like elastomers and do not have well-definedyielding behavior. With these two different types of crystal structure, the SSC-technology homogeneous ethylene–α-olefin copolymers can, therefore, be classi-fied into four major domains [16], as illustrated in Fig. 13.

Thermal and Dynamic Mechanical Properties

Melting Behavior of SSC Polymers

Differential scanning calorimetry is one of the most popular methods for studyingthe thermal properties (melting and crystallization) of semicrystalline polymers.Structural information can be uncovered by a careful interpretation of DSC thermo-grams generated at different conditions.

Due to the narrow intermolecular comonomer distribution, SSC copolymersusually have a much narrower melting peak than their heterogeneous counterpartsproduced by multiple-site Z-N catalysts. Figure 14 shows the DSC melting curveof two ethylene–octene copolymers, both at similar density: one made by a conven-tional multiple-site Z-N catalyst and one by SSC technology. The heterogeneousLLDPE copolymer always has a peak melting point at around 120–130°C over abroad density range (�0.89–0.95 g/cm3). This is because the conventional hetero-geneous LLDPE copolymer is a reactor blend of molecules ranging from the un-


FIG. 14 DSC melting curve of SSC and LLDPE ethylene–octene copolymers at about 0.903 g/cm3

density. Samples were cooled and heated at 10°C/min.

crystallizable and poorly crystallizable type to the HDPE type containing only ofα-olefin [17]. These HDPE-type linear chains can, therefore, crystallize to a ratherlarge size crystal that has a melting point in that temperature range. The shape ofthe melting peaks of all the homogenous copolymers still represents a single peakor a single peak with a shoulder, very different from that of the heterogeneousLLDPEs, as illustrated in Fig. 15. The peak melting point of the homogeneouscopolymer, however, drops accordingly as the density/crystallinity of the polymeris lowered, as seen in Fig. 15. This characteristic has a very significant commercialvalue for heat-seal and oriented-shrink-film food-packaging applications.

A plot of the melting points of LLDPE/VLDPE, LDPE, SSC ethylene–octene,and EVA resins as a function of density is shown in Fig. 16. All the samples werecooled and heated at 10°C/min using a Perkin-Elmer DSC-7. LDPE resins haveslightly lower melting peak temperature compared to SSC ethylene–octene copoly-mers made using CGC technology at a given density. The density of EVA resinsincreases as the vinyl acetate content increases, even though the degree of crys-tallinity decreases. This is due to the bulky nature of the vinyl acetate group in-creasing the amorphous phase density. Hence, the density of EVA resins cannotbe compared with the density of ethylene–α-olefin copolymer resins.

A plot of melting points of LDPE, LLDPE/VLDPE, and SSC ethylene–octeneand EVA resins as a function of resin crystallinity is shown in Fig. 17. Prior to1990, EVAs were the only cost-effective polyethylenes widely available for pack-aging applications requiring low-temperature sealing, low-temperature shrinkage,and toughness properties. This is due to their lower crystallinity (typically less thanabout 40%) and lower melting point (typically less than about 100°C). Beginning in


FIG. 15 DSC melting curve of SSC and LLDPE ethylene–octene copolymers at various densities.Samples were cooled and heated at 10°C/min.

FIG. 16 Peak melting temperatures of heterogeneous and homogeneous ethylene–octene copolymers,LDPE, and EVA resins versus polymer density.


FIG. 17 Peak melting temperatures of heterogeneous and homogeneous ethylene–octene copolymers,LDPE, and EVA resins versus polymer crystallinity by DSC.

early 1990, polyolefin plastomers (POPs) and polyolefin elastomers (POEs), madeusing a single-site catalyst, have become available over a wide range of crystallin-ity and melting points. When plotted against resin crystallinity, the melting pointsof EVA resins lie very close to the curve for the SSC ethylene–octene resins (Fig.17). Hence, SSC resins can potentially be used in various film applications whereEVAs are being used. Two applications of SSC resins, sealants, and oriented shrinkfilms where EVAs have been traditionally used will be discussed in detail later.

Many detailed studies of the thermal properties of SSC technology, homoge-neous ethylene–α-olefin copolymers, have been performed by many different insti-tutes [11,16,18]. The first publication showing the effect of increasing comonomercontent on the melting points of homogeneous copolymers was U.S. Patent3,645,992 by Elston of DuPont, Canada [19].

The DSC melting curve of a polyolefin elastomer (POE) at 0.87 g/cm3 densityand one melt index is shown in Fig. 18. The glass transition temperature of thePOE resin was about �54°C. The POE exhibited a very broad single melting peakwith a shoulder and a peak melting temperature of 59°C. The melting begins attemperatures as low as about �40°C. A substantial amount of crystallinity is mol-ten by room temperature. The broad melting range of the POE can be attributedprimarily to the intramolecular comonomer distribution. For example, the longerethylene runs may form a larger lamella crystal with a higher melting point andthe shorter one may form a small lamella crystal or even a fringe micelle crystalwith a lower melting temperature.


FIG. 18 DSC melting curve of ethylene–octene polyolefin elastomer at 0.87 g/cm3, made using CGCtechnology.

Dynamic Mechanical Properties

Dynamic mechanical properties of the homogeneous copolymers as a function ofdensity are illustrated in Figs. 19 and 20. The storage modulus decreases withincreasing temperature due to a decrease in crystallinity. The storage modulus de-creases sharply near the melting point. The γ-transition peak of all four copoly-mers and the homopolymer appear at about �120°C in the plot of tan δ versustemperature. In some literature the γ peak was designated as the true glass transitiontemperature (Tg) of polyethylene. The β-transition peak for the polyolefin elastomer(0.870 g/cm3) appears at about �40°C; for the polyolefin plastomer (0.898 g/cm3),it is about �25°C, and for the 0.920-g/cm3 copolymer, it is about �20°C. The0.954-g/cm3 polymer has a broad β peak, also at around �20°C. In most cases,the β-transition temperature can be related to the low-temperature performanceof ethylene-based polymers and copolymers. For example, below the β-transitiontemperature, the polymer usually becomes brittle. Therefore, most polyethyleneusers usually report the β-transition temperature as the practical glass transitiontemperature. Also, the glass transition temperature measured using DSC matcheswell with that of the β-transition temperature. For example, as shown in Fig. 18,the DSC Tg of the 0.87-g/cm3 elastomer is about �54°C. The β-transition tempera-ture of the 0.87-g/cm3 elastomer is about �40°C. The β-transition temperaturewould be higher than the DSC Tg due to the high frequency (10 rad/s) used in themeasurement. For the 0.87-g/cm3 elastomer, the α-transition is almost merged withthe β-transition. For higher-density SSC polyethylene, distinct α-transitions canalso observed. The mechanical properties and deformation behavior of these homo-geneous polymers will be discussed further in the next section.


FIG. 19 Storage modulus of homogeneous ethylene–octene copolymers and a homopolymer as afunction of temperature at 10 rad/s.

FIG. 20 Tan δ of homogeneous ethylene–octene copolymers and a homopolymer as a function oftemperature at 10 rad/s.


Deformation Behavior and Mechanical Propertiesof Homogeneous Ethylene–�-Olefin CopolymersMade with SSC Technology

Tensile Properties

The stress–strain curves of homogeneous ethylene–octene copolymers and a ho-mopolymer, made using CGC technology, are shown in Fig. 21. The polymerswere heated well above the melting point and cooled at 1°C/min. The yield regionof copolymers is also enlarged in Fig. 21. The density (crystallinity) of polymerprofoundly affected the response to deformation, as evident by the broad spectrumof tensile properties. At high densities, the deformation had characteristics commonto many semicrystalline thermoplastics with localized yielding and cold drawing[16]. For copolymers having medium densities (0.902, 0.91, and 0.918 g/cm3),significant strain hardening was observed. For copolymers having low densities(less than 0.885 g/cm3), the moduli were low and the deformation was essentiallyuniform (homogeneous) and elastomeric.

Differences in yield behavior of the four homogeneous polyethylenes at differ-ent densities are shown in the photographs in Fig. 22. These photographs weretaken at 150% engineering strain. These polymers exhibited deformation behaviorranging from necking at room temperature (density greater than about 0.910 g/cm3) to uniform deformation (density less than about 0.910 g/cm3). The effect ofincreasing comonomer content on the tensile deformation behavior and the correla-tion between the structural classification and the large-scale deformation behaviorhave been studied and discussed in detail in many publications [16,20–24] and,therefore, will not be further discussed in this article.

FIG. 21 Engineering stress–strain properties of homogeneous ethylene–octene copolymers and ahomopolymer, measured at a strain rate of 0.04 s�1. (From Ref. 16.)


FIG. 22 Photographs of profiles of deformation behavior of Type I to IV homogeneous ethylene–octene copolymers, taken at 150% engineering strain. (From Ref. 16.)

Modulus and yield strength are important properties for ethylene homopoly-mers and copolymers for many commercial applications, such as packaging film,injection molded containers, and so forth. The slope of the stress–strain curve atvery low strain (less than or equal to 2% strain) is measured as the modulus ofthe polymer. Secant modulus at 1% or 2% strain and Young’s modulus are theones most commonly used.

Moduli of the polyethylenes in the solid state are strongly controlled by poly-mer crystallinity, crystallite size distribution, and fabrication conditions, whichresults in orientation. Even with this complexity, there are some ‘‘rules of thumb’’which can be applied to the modulus. For example, it is generally true that theinitial modulus increases with density (crystallinity) of the polymer. There is alsoa direct relationship between initial modulus and crystallite size. It has been shown[24] that two polyethylenes of the same degree of crystallinity can have very differ-ent moduli (up to 100% difference), which is related to a difference in crystallitesize.

The moduli of a series of homogeneous ethylene–octene copolymers made byCGC technology, several homogeneous ethylene–butene copolymers made byMTC technology, and several conventional LLDPEs made by a Z-N catalyst over arange of polymer densities and comonomer content were measured for comparisonpurposes [25]. These samples were prepared by compression molding at similarheating and cooling conditions to impart the same heat history on each of thesamples. The data are summarized in Fig. 23. It seems that if the polymer samplesare prepared under similarly controlled conditions, the Young’s modulus of thesepolymers is simply a function of the polymer density. However, it should be notedthat it is very difficult to produce any ethylene copolymer having a density belowabout 0.885 g/cm3 using Z-N catalysts. Therefore, data for Z-N heterogeneous


FIG. 23 Young’s modulus versus density of ethylene copolymers determined on compression-moldedplaques.

LLDPE at below 0.885 g/cm3 are not available for comparison purposes. Basedon common knowledge of semicrystalline polymers, the very low-density hetero-geneous polymer made by Z-N catalysts, if one can make it, may have a highermodulus than the homogeneous copolymer at similar densities. This is becausethe heterogeneous polymer, no matter how low the density is, will always havesome polymer chains that have a very low level of comonomer incorporated. Thesepolymer chains will crystallize to form larger-sized crystals, which may result inhigher modulus than the homogeneous polymer at equivalent density.

Elastic Properties of Polyolefin Elastomers

Polyolefin elastomers (POEs), very low-density (crystallinity) homogeneous ethyl-ene–α-olefin copolymers, have a fringed micelle crystal morphology and exhibit avery different deformation behavior as compared to the higher-density copolymershaving lamella crystal structures. POEs are now commercially available from Du-Pont Dow Elastomers, LLC (ENGAGE). The small fringed micelle crystals dis-persed in the soft, amorphous matrix act as tie-points to anchor the amorphouschains during deformation and, therefore, result in an elastic recovery upon releaseof stress. The load/unload property under a tensile deformation mode for a repre-sentative octene-based POE (ENGAGE EG-8100, 0.870 g/cm3, 1 MI) is illustratedin Fig. 24.

The elastic recovery (permanent set) of various polyethylenes is illustrated inFig. 25. The slow-cooled (15°C/min) 30-mil-thick compression-molded sampleswere pulled to desired elongation at 10 in./min (gauge length � 2 in.). The samplewas held at the desired elongation for 30 s. The crosshead was brought back tothe initial grip separation at 10 in./min and held for 1 min. The sample was pulled


FIG. 24 Load/unload behavior of a representative octene-based polyolefin elastomer (ENGAGE EG-8100, 0.870 g/cm3, 1 MI). Strain rate, 2.25 min�1; specimens were compression molded andcooled at 15°C/min; test was run at room temperature.

again at 10 in./min until the load rises above zero. Percent permanent set is thepercent elongation at which the load rises above zero. The permanent set of homo-geneous polyethylenes increased with increasing density. At densities below 0.885g/cm3, the percent set is low (less than about 20%) at elongations less than 100%.However, the unrecovered strain became much higher beyond the 100% strain.

FIG. 25 Permanent set of various polyethylenes as a function of elongation.


This phenomenon can hypothetically be explained by a slip-link theory in whichthe fringed micelle crystals are treated as slippage links. Beyond 100% strain,these crystals started to slip, which resulted in permanent deformation. A detaileddiscussion of the slip-link model and the use of this model to predict the overallelastic properties of POE has recently been published [22]. New applications forthis class of elastic materials were developed commercially for wire and cableinsulation, shoe soles, elastic fiber and films, foams, and so forth.

Tie-Molecules in Ethylene–�-Olefin CopolymersMade by SSC Technology

Consideration of the crystalline domains has long dominated research to explainthe properties of semicrystalline polymers such as linear low-density polyethylene(LLDPE). Although crystalline domains control low-strain properties of semicrys-talline polymers such as modulus and yield stress, it has clearly been establishedby many researchers that large-strain properties such as stretchability, impact, tear,failure processes, and so forth are also controlled by the amorphous region, particu-larly by tie-molecules, the amorphous chains that bridge adjacent lamellae [26–33]. However, development of appropriate structure–property relationships insemicrystalline polymers has been perceived to be hindered by the inability toanalytically measure relative tie-chain concentration. In the past, relative tie-chainconcentration has been semiquantitatively characterized using techniques such asmeasurement of the brittle fracture strength [27], infrared dichroism after deforma-tion, and chlorination of films [26]. The relative tie-molecule concentration hasalso been estimated from chain dimensions and the semicrystalline morphology(topology) of the polymers [28–33].

Tie-molecules in the semicrystalline polymer are critical for enhancing me-chanical properties such as environmental stress crack resistance (ESCR), impact,tear, and tensile strength. A schematic diagram of the tie-molecule structure in anethylene–α-olefin semicrystalline copolymer with a lamellar morphology is illus-trated in Fig. 26. A main cause for rejection of a polyethylene chain from a crystalis the presence of imperfections along the chain backbone, which are usuallybranch points formed by the α-olefin comonomer. Without these branch points(e.g., HDPE), the major part of the polymer chain can possibly be incorporatedinto the same lamella crystal and, thus, few tie-molecules can be formed. This willresult in a polymer with very low mechanical strength.

Probabilities for forming tie-molecules in a homogeneous ethylene–α-olefin co-polymer made by SSC technology versus a heterogeneous copolymer were estimated[34]. Unlike the heterogeneous ethylene copolymer made by conventional Z-N cata-lyst technology, the SSC technology polymer has a homogeneous distribution of theα-olefin among the polymer chains. This allows all the polymer chains to crystallizesimilarly, which results in a narrow crystallite size distribution. The effect of density(crystallinity) on relative tie-chain formation probability and concentration is mod-eled [35] and the results are illustrated in Fig. 27. The probability of tie-chain forma-tion alone does not reflect the actual tie-chain concentration in semicrystalline poly-


FIG. 26 Schematic diagram of tie-molecule structure in an ethylene–α-olefin copolymer.

mers. The relative tie-chain concentration would also depend on the concentrationof ‘‘junction points,’’ which would approximately depend on the volume fractionof the crystallinity. The relative tie-chain concentration was obtained from the prod-uct of tie-chain probability and the volume fraction of the crystallinity.

Figure 27 illustrates the optimum density range and the molecular weight effecton tie-molecule formation. The Huang–Brown model used to calculate probabilityof tie-chain formation does not take into account the effect of the type of α-olefincomonomer. Intrinsic Elmendorf tear strengths on compression-molded 10-mil-thick films of various ethylene–α-olefin copolymers were measured [36] and thedata are illustrated in Fig. 28. Figure 28 shows that higher α-olefin copolymers(octene and hexene) have much better intrinsic tear strength than butene and pen-

FIG. 27 Relative tie-molecule concentration in a SSC-technology ethylene–octene copolymer at vari-ous molecular weights.


FIG. 28 Intrinsic tear strength of various SSC-technology ethylene–α-olefin copolymers.

tene copolymers, with octene being the highest of all. The optimum tear strengthfor all polymers studied is found to be in the density range between 0.89 and 0.92g/cm3. These experimental results are in good agreement with the tie-chain model.

Applications

In previous sections, three key distinctions have been noted between ethylene–α-olefin copolymers produced with single-site catalysts and ethylene–α-olefincopolymers produced with Ziegler–Natta catalysts: (1) an ability to incorporatehigher levels of the α-olefin to achieve low polymer density or crystallinity, (2) auniform comonomer distribution giving a lower melting point at a given density,and (3) a narrow MWD with a Mw/Mn of about 2. The ability to incorporate higherlevels of comonomer has made possible new product families (e.g., POP and POE).For example, POEs with densities less than 0.885 g/cm3 are available from SSC,whereas traditional Z-N ethylene–α-olefin copolymers typically are not availablebelow about 0.89 g/cm3 density. The narrow molecular-weight and comonomerdistributions contribute to several unique properties, including controllable meltingpoints, reduced extractables, reduced blocking, excellent optics, and excellent me-chanical properties.

The unique characteristics of the homogeneous ethylene–α-olefin copolymersdescribed make them useful for packaging applications. For example, the abilityto control melting and crystallization behavior, the excellent optics, and excellentmechanical strength make POPs an ideal candidate to compete with the traditionalhigh-performance sealants such as EVA copolymers containing 9–18% vinyl ace-tate and ionomers. Indeed, one important commercial application for plastomersis use as a high-performance sealant in multilayer coextruded or laminated films.Table 1 shows several examples of key properties of plastomers produced withSSC and the resulting packaging benefits.


TABLE 1 Benefits of Plastomers in Flexible Packaging

Polymer properties Packaging benefits

Toughness of a linear polyethylene Protection of goods, gauge reductionLow sealing temperatures Faster packaging line speedsLow extractables Better taste and odor characteristicsExcellent optics Package appealHigh oxygen transmission Breathable filmsElasticity High recovery and flexibilityHydrocarbon composition Polyolefin compatibility, recycle friendlyThermal stability Processing flexibilityMoisture insensitivity Bulk handling, good moisture barrier

Food-Packaging Film Applications

The primary distinguishing feature of food packaging is the need to prevent con-tamination of the food and to maintain or extend the shelf-life. The package actsas a barrier to the environment. Barrier properties will vary depending on the spe-cific type of food being packaged. The barrier of a flexible package can be compro-mised when (1) an hermetic seal of the package is not obtained during packaging,(2) the seal of the package is not maintained during the packaging, distribution,and sales of the product, or (3) the package itself fails due to inadequate puncture,tear, impact, or abrasion resistance. Seal performance, shrinkage, and abuse resis-tance are some of the critical performance requirements for many, if not all, ofthe food packaging applications to be described.

High-Performance Sealants

Polyethylenes are widely used as sealants in various packaging applications. Manyhigh-performance films are multilayer structures, where the individual layers per-form a specific function(s). High-performance sealants often must meet a numberof performance requirements, placing constraints on the polymer design. Beyondparameters like heat-seal-initiation temperature (HSIT), caulkability, seal throughcontamination, and hot-tack range, consideration is given to parameters like pro-cessability during extrusion, machinability (stiffness, coefficient of friction, perme-ability, appearance, cost, and/or FDA compliance. Types of polyethylenes usedas high-performance sealants include EVA, ULDPE, ethylene acrylic acid (EAA)copolymers, ionomers, and, recently, polyolefin plastomers.

One of the key attributes of a high-performance sealant is the low meltingpoint, which facilitates the key performance requirement of a low seal-initiationtemperature. When plotted against resin crystallinity (Fig. 17), the melting pointsof EVA resins lie very close to the curve for plastomers produced via SSC technol-ogy. Hence, plastomers can potentially be used in various films applications wherelow-melting-point polymers such as EVAs and ionomers are currently in use (e.g.,sealants, oriented shrink films, etc.). Other key attributes of a high-performancesealant are the hot-tack strength, hot-tack window, and heat-seal strength.


Hot-tack strength is a measure of the force required to separate a semimolten,just-made seal. This test is designed to predict the performance on vertical form,fill, and seal (VFFS) equipment, in which the strength of the seal after it has justbeen made may be the rate-limiting step in this automated process. High hot-tackstrength reduces the likelihood that the food product will break through the sealwhen being dropped into the forming package. The ultimate hot-tack strength isthe highest hot-tack strength achieved over a full range of sealing temperatures.Hot-tack strengths of ethylene–octene plastomers and EVA resins are comparedin Fig. 29. EVA polymers exhibit much lower hot-tack strength compared to plasto-mers. This is due to the very high degree of long-chain branching present in EVAcopolymers, which significantly lowers diffusion rates across the molten interface.In the case of EVA resins, an increase in melt index actually increases the hot-tack strength due to higher diffusion rates [37].

Heat-seal strengths of ethylene–octene plastomers and EVA resins are com-pared in Fig. 30. Interestingly, at the same melting point, EVA resins exhibit some-what higher heat-seal initiation temperatures compared to plastomer resins (at meltindex of approximately 1 or less). For example, although the melting point of0.896-g/cm3 plastomer is about 94°C and the melting point of ELVAX 3165 (18%VA EVA) is about 86°C, the 0.896-g/cm3 plastomer exhibited lower seal-initiationtemperature. This may due to the high degree of LCB significantly lowering diffu-sion rates in EVA resins. The ethylene–octene plastomer resins exhibit slightlyhigher seal plateau strengths compared to EVA resins. Polyolefins plastomers withtheir low, sharp melting points are thus well suited for high-performance sealantsbecause the ultimate seal and hot-tack strengths of the polymer are achieved atlow temperatures [38–41].

One of the applications of high-performance sealants is the box inner liners

FIG. 29 Hot-tack strength of ethylene–octene plastomers made using CGC technology and EVAresins as a function of seal bar temperature. Nylon–EAA–sealant (1/1/1.5 mil)-blown coex-trusion film was used.


FIG. 30 Heat-seal strength of ethylene–octene plastomers made using CGC technology and EVAresins as a function of seal bar temperature. Nylon–EAA–sealant (1/1/1.5 mil)-blown coex-trusion film was used.

for dry foods such as cake mix, cereal, and crackers. Whereas a high-oxygen barrieris critical to preserve the freshness of meat and cheese, a high-moisture barrier isessential for preserving the freshness of dry products. A typical plastic film struc-ture for cereal and cake-mix inner liners is a coextrusion of HDPE with a sealant.The sealants typically used for these applications are ionomers or EVAs containinghigh levels of vinyl acetate (VA) (i.e., 18% VA).

Again, because a strong, hermetic seal is critical for maintaining the barrier ofthe package, POPs are another option for replacing ionomers or EVA copolymers(18% VA). A comparison of the other critical sealant performance requirements fora polyolefin plastomer (POP 2), an ionomer (Zn ionomer), and an EVA copolymer(18% VA) are discussed below and summarized in Table 2. POP 2 and EVA copoly-mers (18% VA) have comparable seal-initiation temperatures (69°C and 72°C, re-spectively). They are followed by the Zn ionomer, with a seal-initiation temperatureof 84°C. Table 2 shows that POP 2 provides the highest ultimate hot-tack strengthof the three products, followed by the Zn ionomer and the EVA copolymer. Notethat the 18% VA copolymer provides very low ultimate hot-tack strength due tovery high levels of long-chain branching retarding diffusion across the interface andis generally not suitable for the packaging of heavy products, such as cake mixes.

Oriented Shrink Films

Oriented shrink films can be differentiated from conventional hot-blown shrinkfilms by their fabrication processes. The conventional blown shrink-film methodinvolves extrusion of the molten polymer through an annular die at a certain draw-down ratio and a certain blowup ratio and then cooling the bubble via ambient orchilled air. Orientation of the film takes place in a completely molten state and is


TABLE 2 High-Performance Sealants for Box Inner Liners

EVAResin POP (18% VA) Zn Ionomer 1

Melt index (dg/min) 1.6 0.8 5.5Density (g/cm3) 0.8965 Not available Not available2.0-Mil monolayer blown-film properties

Puncture (ft-lb./in.3) 270 222 104Tear Resistance (g) 300 118 190Haze (%) 1.3 1.5 2.9Odor intensity rating 1.2 Not available 2.7Water vapor transmission rate 6.2 Not available

(g mil/100 in.2 day atm) 2.01.5-Mil HDPE/0.5-mil sealant, blown

coextrusionHeat-seal initiation (°C) 69 72 84Ultimate hot-tack strength (N/in.) 7.5 3.4 4.4

primarily in the machine direction, due to the drawdown ratio, and secondarily inthe cross direction, as a result of the blowup ratio. In contrast, the oriented shrink-film process can be described generically by the following steps:

Extrusion–quenching–reheating–baxially stretching–cooling

The reheating step involves temperatures just above the glass transition tempera-ture in the case of amorphous polymers, or below the peak melting temperaturein the case of semicrystalline polymers.

There are two basic methods to produce biaxially oriented films. The firstmethod is the tenter frame process and the second method is often referred to asthe ‘‘double- bubble’’ or ‘‘trapped-bubble’’ process. More in-depth descriptionsof these processes can be found elsewhere [42,43]. Orientation by these processestakes place in a semimolten state (i.e., at a temperature below the melting pointof the semicrystalline polyolefin). As a result, a significantly higher degree of orien-tation is obtained compared to hot-blown shrink films. These orientation processesproduce high-tensile, high-modulus films with high shrink and shrink tension, aswell as excellent clarity [42,43]. Biaxially oriented films made via the ‘‘double-bubble’’ or ‘‘trapped-bubble’’ process are used to produce packages for very large,subprimal cuts of meat after slaughter. The packages are often referred to as barrierbags. The subprimal cuts of meat are inserted into a barrier bag that is generallysealed at one end, the air evacuated, and then the bag heat sealed. The package isthen heated, usually via hot water, to obtain shrinkage of the film around the meat[44]. Typical barrier bag structures range from three to five layers and minimallycontain an abuse/shrink layer, an oxygen barrier layer such as a poly(vinylidenechloride) (PVDC) or ethylene vinyl alcohol copolymer (EVOH), and a sealantlayer. In addition, adhesive layers may be used to tie the structure together.

The use of LLDPEs and ULDPEs alone or as blends with EVA copolymersin both the abuse and sealant layers have been taught in the patent literature [44,45].The broad comonomer distributions and melting ranges of LLDPEs and ULDPEscreate a sufficiently broad orientation window for the double-bubble process. Fur-


FIG. 31 Hot-water shrinkage at 90°C of films oriented on a T.M. Long stretcher under isothermalconditions at a 4.5 4.5 draw ratio at 5 in./s.

thermore, the excellent toughness of these linear polymers provides the high levelof abuse resistance needed to prevent punctures from the bones in subprimal meatsand to sustain impact abuse during the distribution process. ULDPE resins offerimprovements in shrinkage over higher-density LLDPE resins, due to their reducedcrystallinity. Further improvements in shrinkage can be obtained using POPs. POPscan offer low-temperature shrinkage improvements over ULDPEs at the same den-sity because of their lower melting points. Figure 31 illustrates the improvedshrinkage at 90°C for POPs at a density below about 0.908 g/cm3 [46]. In addition,it is assumed that the improved sealability and abuse-resistance properties for POPsin conventional blown films also translate to improvements in oriented shrink films.

As mentioned earlier, blends of EVA with ULDPE or LLDPE have been taughtfor use in barrier bag structures. A particular disadvantage of these blends is thepoor optics and reduced abuse resistance. As shown in Table 3, the POP/LLDPEblend provides significantly better optics and potentially better toughness in an ori-ented shrink film than the EVA/LLDPE blend. Both the POP and EVA have compa-

TABLE 3 Optics of Oriented Films

EVA (12% VA)/LLDPE POP*/LLDPE†

Blend ratio 30%/70% 30%/70%Haze (%) 7.6 0.5Shrinkage at 90°C (%) 16 16.5

Note: Films oriented on T.M. Long stretcher under isothermal conditions ata 4.5 4.5 draw rate at 5 in./s.* POP is an ethylene–octene copolymer having a melt index of 0.9 and den-sity of 0.898 g/cm3.† LLDPE is an ethylene–octene copolymer having a melt index of 1.0 dg/min and density of 0.920 g/cm3.


rable melting points and crystallinity. The poor optics of the EVA/LLDPE blendare believed to be due to the partial immiscibility of these two polymers [47].

Other important applications of homogeneous polyethylenes made using SSCtechnology include fresh-cut produce packaging [48–50], high-performance films[51–54], batch inclusion bags [55], cling layer in stretch films [56,57], extrusioncoating [58,59], and impact modification of polypropylenes [60–65].

Conclusion and Summary

Development and implementation of SSC technology have greatly expanded theproduct range of polyolefins. These homogeneous polymers have narrow molecu-lar-weight and composition distributions, resulting in enhanced mechanical, opti-cal, and heat-seal properties. Because of the simple and predictable molecularstructure of this class of polyolefin copolymers, building a model to predict proper-ties for these polymers became feasible. This allows the materials scientists todesign products by a molecular-architecture approach, using property and perfor-mance requirements as guidelines. With the advancement of the SSC technologyfor the production of polyolefin copolymers with well-defined molecular structures,the authors expect the polyolefin industry can take great advantage of this technol-ogy to design unique polymers and fulfill their customer needs in the future.

Homogeneous polyolefins provide excellent utility in numerous packaging appli-cations, ranging from retail meat packages to industrial stretch films. These resinsmay be fabricated into packaging films or containers via blown- and cast-film meth-ods, biaxial oriented film processes, extrusion coating, lamination, and injectionmolding. In packaging applications, the resins based on CGC technology have suc-cessfully replaced and/or upgraded EVA, ionomer, LDPE, LLDPE, and ULDPE.

Acknowledgments

The authors of this article would like to express their sincere appreciation to Profes-sor Eric Baer, Professor Anne Hiltner of Case Western Reserve University, andPradeep Jain, Kalyan Sehanobish of The Dow Chemical Company. They providedmany useful data in this article.

References

1. K. W. Swogger, Proceedings of the 1992 Int’l Business Forum on Special Polyolefins,SPO ’92, 1992, p. 115.

2. B. A. Story and G. W. Knight, Proceedings of Metcon ’93, 1993, p. 112.


3. C. S. Speed, B. C. Trudell, A. K. Mehta, and F. C. Stehling, Proceedings of PolyolefinsVII International Conference, 1991, p. 45.

4. S. Chum, C. I. Kao, and G. W. Knight, Plastics Eng., 51 (6), 21 (1995).5. L. K. Johnson, S. Meeking, and M. Brookhart, J. Am. Chem. Soc., 118, 267–268 (1996).6. L. K. Johnson, C. M. Killian, and M. Brookhart, J. Am. Chem. Soc., 117, 6414–6415 (1995).7. L. Wild, T. Ryle, D. Knobeloch, and L. J. Peat, J. Polym. Sci. Polym. Phys. Ed., 20,

441–455 (1982).8. S. Chum and J. Ruiz. Proceedings of Aspen World Modeling Conference, Boston, 1997.9. S. Y. Lai, J. R. Wilson, S. Chum, G. W. Knight, and J. C. Stevens, U.S. Patent

5,272,236 (1993).10. G. W. Knight and S. Y. Lai, Proceedings of the Society of Plastics Engineers RETEC

’93, 1993, pp. 226–241.11. M. Peeters, B. Goderis, C. Vonk, H. Reynaers, and V. Mathot, J. Polym. Sci., Part

B: Polym. Phys., 35, 2689–2713 (1997).12. P. J. Phillips, Rep. Prog. Phys., 53, 549 (1990).13. A. Keller, Phil. Mag., 2, 1171 (1975).14. P. H. Geil, Polymer Single Crystals, John Wiley, & Sons, New York, 1963.15. S. Krimm and T. C. Cheam, Faraday Discuss., 68, 244–250 (1979).16. S. Bensason, J. Minick, A. Moet, S. Chum, A. Hiltner, and E. Baer, J. Polym. Sci.,

Part B: Polym. Phys., 34, 1301–1315 (1996).17. V. F. B. Mathot and M. F. J. Pipers, J. Appl. Polym. Sci., 39, 979–994 (1990).18. J. Minick, A. Moet, A. Hiltner, E. Baer, and S. Chum, J. Appl. Polym. Sci., 58, 1371–

1384 (1995).19. C. Elston, U.S. Patent 3,645,992 (1972).20. S. Bensason, J. Minick, A. Moet, A. Hiltner, E. Bear, S. Chum, and K. Sehanobish,

Proceedings of the International SPE Annual Technical Conference ANTEC ’95,1995, pp. 2256–2262.

21. Y. Hwang, S. Chum, R. Guerra, and K. Sehanobish, Proceedings of the InternationalSPE Annual Technical Conference ANTEC ’94, 1994, pp. 3414–3418.

22. S. Bensason, E. Stepanov, S. Chum, A. Hiltner, and E. Baer, Macromolecular, 30(8),2436–2444 (1997).

23. R. Popli and L. Mandelkern, J. Polym. Sci., Part B: Polym. Phys., 25, 441–483 (1987).24. J. T. Graham, R. G. Alamo, and L. Mandelkern, J. Polym. Sci., Part B: Polym. Phys.,

35, 213–223 (1997).25. K. Sehanobish, R. Patel, B. Croft, S. Chum, and C. Kao, J. Appl. Polym. Sci., 51,

887–894 (1994).26. A. Lustiger and N. Ishikawa, J. Polym. Sci., Part B: Polym. Phys., 29, 1047–1055 (1991).27. N. Brown and I. M. Ward, J. Mater. Sci., 18, 1405–1420 (1983).28. Y. Huang and N. Brown, J. Polym. Sci., Part B: Polym. Phys., 29, 129–137 (1991).29. S. Hosoda and A. Uemura, Polym. J., 24, 939 (1992).30. N. Brown, X. Lu, Y. Huang, I. P. Harrison and N. Ishikawa, Plastics Rubber Compos-

ites Process. Appl., 17, 255–258 (1992).31. J. T. Yeh and J. Runt, J. Polym. Sci., Part B: Polym. Phys., 29, 371–388 (1991).32. Z. Zhou, X. Lu, and N. Brown, Polymer, 34, 2520–2523 (1993).33. L. L. Bohm, H. F. Enderle, and M. Fleibner, Adv. Mater., 4, 234 (1992).34. N. Kashiwa and A. Todo, Proceedings of Metcon ’93, 1993, p. 235.35. R. Patel, K. Sehanobish, P. Jain, S. Chum, and G. W. Knight, J. Appl. Polym. Sci.,

60, 749–758 (1996).36. T. Plumley, K. Sehanobish, R. M. Patel, S. Y. Lai, S. P. Chum, and G. W. Knight,

J. Plastic Film Sheet. 11, 269–278 (1995).37. J. W. Spink, TAPPI Polymers, Laminations, and Coatings Conference Proceedings,

1991, pp. 579–587.


38. N. F. Whiteman, J. A. deGroot, L. K. Mergenhagen, and K. B. Stewart, Proceedingsof the RETEC Conference, 1995.

39. S. A. DeKunder, Proceedings of the Future Pak ’97 Conference, 1997.40. M. F. Simpson and J. L. Presa, Proceedings of the Future Pak ’96 Conference, 1996.41. S. A. DeKunder and M. F. Simpson, Proceedings of the SME ’97, Barrier Technology

for the Food Packaging Industry, 1997.42. C. J. Benning, Plastic Films for Packaging, Technomic Publishing Company, Lancas-

ters, PA, 1983.43. W. R. R. Park and J. Conrad, Encycl. Polym. Sci. Technol. 2, 339–373 (1965).44. D. L. Newsome, U.S. Patent 4,457,960 (1984).45. S. Lustig, P. Forest, N. M. Mack, J. M. Schuetz, and S. J. Vicik, U.S. Patent 4,863,769 (1989).46. R. M. Patel, M. F. Langohr, K. L. Walton, and O. K. McKinney, WO 97/30111,

patent pending.47. R. M. Patel, P. Saavedra, C. Hinton, and J. A. deGroot, J. Plastics Film Sheet., 14,

344–355 (1998).48. G. L. Young, ‘‘Designing Packages for Fresh Cut Produce,’’ TAPPI Polymers, Lami-

nations, and Coatings Conference, 1995.49. J. J. Wooster, ‘‘Step-by-Step Procedures for Extending the Shelf-Life of Fresh Cut

Produce,’’ Proceedings of the MAPack ’95 Conference, 1995.50. J. J. Wooster, ‘‘New Resins for Fresh-Cut Produce,’’ TAPPI Polymers, Laminations,

and Coatings Conference, 1997.51. A. Sukhadia, J. Plastics Film Sheet., 14, 54–74 (1998).52. J. A. deGroot, ‘‘New Polyethylenes Via INSITE Technology,’’ Proceedings of the

SPO ’96 Conference, 1996.53. R. D. Maier, Kunstsoffe, 89, 120–132 (1999).54. L. L. Boehm and M. Fleißner, Kunststoffe, 88, 1864–1870 (1998).55. D. J. Falla and D. Walker, U.S. Patent 5,525,659 (1996).56. D. B. Ramsey and K. B. Stewart, U.S. Patent 5,840,430 and 5,789,029.57. J. Cook, ‘‘The Role of Metallocene Based Polyethylenes in Stretch Film,’’ Proceed-

ings of the SPO ’96 Conference, 1996.58. D. C. Kelley, S. L. Baker, and M. W. Potts, ‘‘DPT-1450: A Polyolefin Plastomer

for Extrusion Coating Applications,’’ TAPPI Polymers, Laminations, and CoatingsConference, 1994.

59. M. W. Potts and T. J. Pope, ‘‘Extrusion Coating vs. Film Lamination: How AFFINITYPolyolefin Plastomers Will Offer More,’’ Proceedings from Future Pak ’96 1996.

60. C. H. Silvis, D. J. Murray, T. R. Fiske, S. R. Betso, and R. R. Turley, US Patent5,688,866 (1997).

61. K. Sehanobish, S. Wu, L. A. Meiske, and P. S. Chum, U.S. Patent 5,861,463 (1999).62. T. C. Yu, Proceedings of the International SPE Annual Technical Conference ANTEC

’95, 1995, pp. 2374–2385.63. L. A. Meiske, S. Wu, K. Sehanobish, and J. Dibbern, Proceedings of the International

SPE Annual Technical Conference ANTEC ’96 1996, pp. 2001–2005.64. T. C. Yu, Proceedings of the International SPE Annual Technical Conference ANTEC

’96, 1996, pp. 1995–2000.65. N. R. Dharmarajan and T. C. Yu, Proceedings of the International SPE Annual Techni-

cal Conference ANTEC ’96, pp. 2006–2013, 1996.

RAJEN M. PATELSTEVE CHUM

encyclopedia of chemical processing and design: 69

Documents