ignition of silane

8/11/2019 Ignition of Silane

1/98

SEMATECHTechnology Transfer 96013067A-ENG

Ignition Characteristics of Releases

of 100% Silane


2/98

1996 SEMATECH, Inc.

SEMATECHand the SEMATECH logoare registered service marks of SEMATECH, Inc.

Hewlett Packardis a registered trademarks of Hewlett-Packard Company.

UNIXis a registered trademark of Novell, Inc.


3/98

Ignition Characteristics of Releases of 100% SilaneTechnology Transfer #96013067A-ENG

SEMATECHMarch 7, 1996

Abstract: This project studied the effect of ventilation on the ignition characteristics of releases of puresilane through commercially available restricted flow orifices (RFOs). Two scenarios of accidental

silane releases in a ventilated cabinet were simulated: the first involved a leak from a regulated

pressure line to a process tool; the second simulated a leak from a pigtail at a nearly full cylinder.

For the high pressure release, demonstration tests were performed in a gas cabinet. Additional

experiments were aimed at defining the effect of the line exit condition on ignition. Test results

indicated that the prompt ignition characteristics depend on the initial pressure of the silane and on

the conditions of the exit portion of the line rather than on ventilation velocity, as previously

believed.

Keywords: Safety, Bulk Gases, Emissions Control

Authors: Francesco Tamanini, Jeffrey L. Chaffee

Approvals: Michael Visokey, ESH Project ManagerPhyllis Pei, Program Manager

Ray Kerby, Director of ESH

Laurie Modrey, Technical Information Transfer Team Leader


4/98


5/98

iii

Technology Transfer # 96013067A-ENG SEMATECH

Table of Contents

1 EXECUTIVE SUMMARY........................................................................................................1

2 INTRODUCTION......................................................................................................................2

3 TECHNICAL BACKGROUND................................................................................................23.1 Literature Review..............................................................................................................3

3.2 Silane-Related Work by FMRC ........................................................................................4

3.2.1 Reactivity and Ignition Characteristics of 10/90 Silane/Nitrogen Blends..............4

3.2.2 Ignition/Combustion Characteristics of 100% Silane Releases..............................5

4 PROJECT OBJECTIVES..........................................................................................................6

5 TECHNICAL APPROACH.......................................................................................................7

5.1 Proposed Test Matrix ........................................................................................................7

5.1.1 Process Delivery Pressure Releases with Cross Flow............................................7

5.1.2 RFO Testing Within a Ventilated Enclosure..........................................................7

5.1.3 Demonstration Test in Gas Cabinet........................................................................75.2 Experimental Set-Up and Facilities...................................................................................7

5.3 Instrumentation..................................................................................................................9

5.4 Data Acquisition..............................................................................................................10

6 TEST RESULTS......................................................................................................................10

6.1 Shakedown Tests.............................................................................................................10

6.1.1 Scope of Testing ...................................................................................................10

6.1.2 Results with Preliminary Test Set-Up ..................................................................10

6.1.3 Results with Set-Up for Process Pressure Releases..............................................11

6.2 Process Delivery Pressure Releases with Cross Flow.....................................................12

6.2.1 Experimental Procedure and Phenomenology......................................................126.2.2 Detailed Test Results ............................................................................................15

6.3 RFO Testing Within a Ventilated Enclosure...................................................................17

6.3.1 Experimental Procedure and Phenomenology......................................................17

6.3.2 Detailed Test Results ............................................................................................19

6.3.3 Additional Tests with Restart of Silane Release...................................................22

6.4 Demonstration Tests in Gas Cabinet ...............................................................................23

6.4.1 Experimental Set-Up and Procedure.....................................................................23

6.4.2 Detailed Test Results ............................................................................................23

7 DISCUSSION..........................................................................................................................24

7.1 Summary of Overall Behavior of Silane Releases...........................................................24

7.2 Initial Release of Line Inventory.....................................................................................257.3 RFO-Controlled Jet Flow ................................................................................................26

7.4 Average Silane Concentration in Exhaust Flow..............................................................28

8 RECOMMENDATIONS.........................................................................................................30

9 CONCLUSIONS......................................................................................................................31

10 REFERENCES ........................................................................................................................32

APPENDIX A: DATA PLOTS.....................................................................................................33


6/98

iv

SEMATECH Technology Transfer # 96013067A-ENG

List of Figures

Figure 1 Overall View of Ventilated Enclosure..........................................................................8

Figure 2 Detailed Schematic of Piping Used for Process Pressure and RFO Release

Tests.............................................................................................................................. 9

Figure 3 Simplified Schematic of Supply Piping Used for Process Pressure Release

Tests............................................................................................................................ 11

Figure 4 Typical Test Results for a Simulated Silane Release from a Process Pressure

LineCase of No Prompt Ignition.............................................................................13


LineCase of Prompt Ignition...................................................................................14

Figure 6 Simplified Schematic of Supply Piping Used for RFO Release Tests .......................18

Figure 7 Typical Test Results for a Simulated Silane Release from a High Pressure

Line (RFO Case). No Prompt Ignition........................................................................20Figure 8 Calculated Silane Concentration in the Case of Perfect Mixing of a Release

with a Ventilation Flow ..............................................................................................29


7/98

v


List of Tables

Table 1 Silane Releases at Process Delivery Pressure with Cross Flow .................................16

Table 2 Silane Releases through RFOs into a Ventilated Enclosure.......................................21

Table 3 Prompt Ignition Summary of Silane Releases in Cross Flow from a Line atProcess Delivery Pressure...........................................................................................25

Table 4 Silane Flow Rates Through Restricted Flow Orifices Based on the Use of the

ISA Equation...............................................................................................................29


8/98

vi


Acknowledgements

Necessary guidance was provided to the project by the members of the SEMATECH PTAB and by

the SEMATECH Project Manager, Michael Visokey. Equipment and gases were contributed to the

project by Praxair, through the good auspices of Mr. Marcelo Viera, who was also very helpful in

making available a proprietary program to calculate silane flows through RFOs. The authors of thereport are extremely grateful for these contributions.


9/98

1


1 EXECUTIVE SUMMARY

This project studied the effect of ventilation on the ignition characteristics of releases of pure

silane through commercially available restricted flow orifices (RFOs). Two scenarios of

accidental silane releases in a ventilated cabinet were simulated in the tests: the first involved aleak from a regulated pressure line (@ 30-200 psig [3.1-14.8 bar]) to a process tool; the second

simulated a leak from a pigtail at a nearly full cylinder (about 1400 psig [97.6 bar]). For the high

pressure release, demonstration tests were performed in a gas cabinet. In all cases, silane was

released through a 1/4-in. (6.4 mm) outside diameter tube, with an RFO (0.010 or 0.020 in. [0.25

or 0.51 mm]) some distance upstream of the exit point.

The tests at process delivery pressure have shown that the prompt ignition characteristics of the

release are a function of the initial silane pressure, but are essentially independent of the

ventilation rate and the size of the RFO in the line. More specifically, prompt ignition was

observed at an initial line pressure of 30 psig (3.1 bar), whereas jets from a line initially at

200 psig (14.8 bar) generally failed to ignite for ventilation rates of 200, 100, and 50 lfpm (1.02,0.51 and 0.25 m/s). Limited testing at 50 psig (4.4 bar) line pressure provided the only evidence

that a reduction in the ventilation rate from 200 to 100 lfpm (1.02 to 0.51 m/s) may have reduced

the propensity of the release to ignite promptly.

The high pressure releases, at initial line pressures from 790 to about 1450 psig (55.5 to 101 bar),

never resulted in the prompt ignition of the jet, regardless of the ventilation rate (200 or 100 lfpm

[1.02 or 0.51 m/s]) or the size of the RFO (0.010 or 0.020 in. [0.25 or 0.51 mm]). This was

confirmed by the demonstration tests in the gas cabinet, which also failed to produce prompt

ignition.

Additional results were obtained during shakedown testing or in variations of the pre-assigned

release schedule following some of the production tests. Many of these additional experimentswere aimed at defining the effect of the line exit condition on ignition. In the experimental set-

up, a 5 in. (127 mm) section of line was placed downstream of the shut-off valve used to start the

release. The ignition behavior of the silane jets was not affected by nitrogen (versus air) in this

exit stub. However, partially reacted silane in the same stub (as would be the case after flow shut-

off at the end of a release with no ignition) consistently resulted in prompt ignition when the flow

was restarted.

The tests provided no evidence to support the widely accepted concept that, for a given release

scenario, there is a minimum ventilation velocity that assures a silane leak will ignite promptly.

In fact, the results appear to show that the prompt ignition characteristics depend on the initial

pressure of the silane and on the conditions of the exit portion of the line. This decoupling

between ventilation velocity and ignition of the jet is not surprising, since the exit velocity of theinitial burst of silane is orders of magnitude greater than that of the ventilation flow. As an

alternative to ventilation velocity (in lfpm), safe conditions for silane storage in gas cabinets

should be determined by requiring a minimum ventilation rate (in scfm), chosen to limit the

maximum concentration of silane to a percentage of the lower explosive limit (LEL).


10/98

2


2 INTRODUCTION

The presence of hydrides in the semiconductor industry has prompted the development of

hardware and the establishment of operating practices to address the safety issues associated with

the use of these materials. Silane (SiH4), which is the most common of this family of compounds,

is used in several manufacturing processes of electronic products as well as in the glass industry

and at research facilities. Silane is pyrophoric in that under certain conditions it ignitesspontaneously upon contact with air. When it does not ignite promptly, the unburned silane can

form a metastable mixture with air, whose delayed ignition can lead to a very rapid reaction.

Safety concerns have resulted in code requirements for restricted flow orifices (RFOs) to limit

the maximum rate of silane discharge in an accidental release and to set certain levels of

ventilation near likely leakage points, presumably to promote prompt ignition of the release,

prevent accumulations of unburned silane, and minimize the likelihood of a severe explosion.

The degree of attention paid to the potential hazards of silane and the consequent implementation

of safe handling practices are the primary reasons behind the relatively low property losses and

personnel injuries from silane accidents [1]. On the other hand, there is incomplete technical

justification for some of the accepted practices and for the 1994 Uniform Fire Code (UFC)requirements on RFO sizing (0.010 in. [0.25 mm]) and minimum ventilation rates (200 lfpm

[1.02 m/s]) at unwelded pipe fittings and connections. The first requirement reduces the use of

silane inventories because of the need to maintain a sufficiently high source pressure to satisfy

the flow demand of the process. The second increases operating costs by prescribing possibly

oversized local exhaust ventilation systems.

The increasing use of bulk silane storage is an additional incentive to design silane protection

measures on a more scientific basis. This is not only because of more immediate practical

considerations (e.g., limitations on the maximum size for RFOs may uneconomically restrict

systems designed to support higher volume production), but because technological changes often

alter the risk profile in unanticipated ways, particularly when the physical behavior of a reactive

system is poorly understood. This study presenting new experimental data on the characteristics

of high pressure silane releases represents an important contribution towards addressing these

issues.

3 TECHNICAL BACKGROUND

Although the ignition and combustion characteristics of silane releases in air have been the

subject of several published studies, a quantitative understanding of the effects of different

parameters on the consequences of an accidental release has remained elusive. Thus the

conditions under which a silane leak could lead to prompt ignition and a jet flame, instead of

accumulating unreacted material, potentially followed by an explosion, are uncertain.Furthermore, the type of reaction that would occur in the explosiona deflagration, a detonation,

or bulk autoignition and rapid volume reactionis not clearly understood. The development of

reliable explosion protection measures for silane-handling systems requires that the conditions

leading to the various modes of combustion be identified.


11/98

3


3.1 Literature Review

The following overview of the published literature summarizes the general knowledge of silane

behavior and provides some perspective on the efforts undertaken in this project to resolve

outstanding issues. The review paper by Britton [2] published in 1990 provides an excellent point

of reference for the state of knowledge on some of the more fundamental aspects of silane

behavior. This information should be generally considered current, except for recent resultsobtained at Factory Mutual Research Corporation (FMRC), which are discussed below.

Silane is a pyrophoric gas that will spontaneously ignite upon contact with air. For spontaneous

ignition, however, certain silane concentrations, turbulence, and temperature of the mixture must

be satisfied. By using appropriate experimental procedures, stable quiescent silane/air mixtures

can be formed and their combustion behavior can be studied under piloted ignition. From these

experiments, the LEL of silane in air has been found to be about 1.37%, with lower values if

some of the air is replaced with a diluent (nitrogen, helium, argon, carbon dioxide). The flames in

mixtures with less than 1.5% silane produce temperatures around 700 K (764F) and arecharacterized by a low burning velocity and low pressure rise under unvented conditions. While

these conditions have not been systematically tested before FMRCs work, it appeared that stablemixtures could be formed in air at silane concentrations well above the LEL. The maximum

concentration value for stable mixtures would presumably be a function of parameters such as

temperature, the presence of components other than air (diluents), and flow strain (turbulence).

If pure silane is released through an opening, the jet reportedly can ignite if the exit velocity is

below a critical value. This value depends on the temperature of the jet (and of the surrounding

environment); on the diameter of the orifice; and, presumably, on the details of the release

geometry. Typical values of critical velocities have been reported [2] for ambient temperatures

near 0C (32F) in the range 10-20 m/s (1970-3940 lfpm), with values up to 50 m/s (9850 lfpm)

for a release from a 1/2 in. (12.7 mm) tube at a temperature of 6C (43F). These numbers appear

to contradict test results reported by Hazards Research Corp. (HRC) from work done for IBM inthe late 1970s [3]. In this case, 100% silane releases promptly ignited for discharges in quiescent

air through orifices of 0.38, 1, and 4 mm (0.015, 0.039 and 0.158 in.) diameter from 5 and 500

psig (1.34 and 35.5 bar) sources. The exit velocity of the 5 psig discharge should have been

around 210 m/s (41,340 lfpm), which is much higher than the critical velocities reported by

Britton [2]. This apparent discrepancy remains unresolved, but may be associated with the

procedures used in the tests for flow start-up.

At least conceptually, the existence of a critical jet velocity for autoignition can be justified on

physical arguments. Near the jet origin (the point of release), while the silane concentration is

most likely higher than the value at which stable mixtures are possible, the flow strain from shear

(which is proportional to the jet velocity to diameter ratio) is also high, because velocities are

high and the jet diameter is small. The flow strains quenching (i.e., turbulent mixing) of the

chemical reaction is probably the main reason why the system does not autoignite.At increasing

distances from the jet source, velocity and silane concentration both decrease and the jet diameter

increases (therefore, the flow strain decreases). If no autoignition occurs, this means that in the

jet there is no combination of sufficiently high silane concentration (and temperature) and low

flow strain to satisfy the autoignition condition. If the flow rate through the orifice is decreased,

the concentration distribution and the jet geometry (jet diameter as a function of distance) remain

essentially the same. The velocity, however, and the flow strain both decrease. Autoignition then


12/98

4


occurs when the drop in overall strain in the flow allows for a sufficiently low strain to develop

at a location in the jet where the silane concentration is above the critical value for a stable

mixture. This qualitative description explains why signs of reaction are typically noted during the

gradual start-up or shutdown of a silane release.

The failure to quantitatively define the ignition/combustion characteristics of silane are the main

reason for the seemingly erratic behavior of releases of this material. Additional HRC work [4],also sponsored by IBM, has contributed to the perception that randomness may be a factor in the

phenomenology of silane leaks in practical industrial situations. In these HRC tests, releases from

a 500 psig (35.5 bar) source through a 1 mm (0.039 in.) orifice inside a gas cabinet with gas

cylinders in place led to explosions that destroyed the cabinet. In one case, the leak ignited a few

seconds after the beginning of the release. Based on the flow rates of ventilation air through the

cabinet (500 cfm [0.24 m3/s]) and the flow of silane, the silane concentration should have been

about 2%, assuming perfect mixing. In the second nominally identical test, the explosion took

place about 5 seconds after the end of a 10-second release, at a time when the ventilation flow

should have provided two air changes in the gas cabinet volume. Tests where the same release

issued from a point near the inlet of the exhaust duct from the cabinet were essentially

uneventful. These results indicate that the mixing inside the cabinet is strongly influenced by the

geometry and that the potential for pocketing exists even at the high ventilation rates normally

present in these types of equipment.

When the release is affected by cross flow and/or confinement, matters become more complex

and the details of the mixing process cannot be determined on the basis of a well documented

textbook situation, as in the case of a forced free jet. However, the same conceptual picture used

to describe the simpler case presented by a free jet can also apply to more complicated mixing

situations. In essence, conditions for autoignition will still require the simultaneous presence of

sufficiently high silane concentrations and sufficiently low turbulence (flow strain). It is doubtful

that the autoignition behavior of the mixture can be satisfactorily characterized through studies

that address the complex flow configuration (release in cross flow into a confinement), without athorough understanding of the simpler flow systems. Much previous work and the more recent

activity at FMRC have been based on this premise.

3.2 Silane-Related Work by FMRC

3.2.1 Reactivity and Ignition Characteristics of 10/90 Silane/Nitrogen Blends

Recent work [5] at FMRC has further contributed to the understanding of silane hazards by

focusing on 10/90 silane/nitrogen mixtures. This work has generated new quantitative

information on the reactivity and the venting requirements of silane. In this FMRC program,

stable mixtures with silane concentrations as high as 3.7% were reliably generated to perform

unvented and vented explosion tests in a 1.35-m3(48-ft3) vessel. As would be expected, the data

show that the reactivity of the mixture (indicated by the burning velocity and by the peak rate of

pressure rise) is a strong function of the silane concentration. In addition, results indicate that

previously published burning velocity data, obtained using a burner technique, are about half the

values that should be used for vent sizing calculations.

At the highest concentration of the reported data (3.7%), the normalized peak rate of pressure

rise, K, was measured to exceed 750 bar m/s. This is almost double the value that characterizes

the behavior of propane under vented conditions. In the silane data, however, there is no evidence


13/98

5


of the acoustically-induced flame instabilities that were found to be important in venting

hydrocarbon/air mixtures. For this nitrogen-diluted silane source, bulk autoignition did not

appear to occur for any of the conditions of the experiments. In rough approximation, a mixture

produced by mixing of a 10/90 silane/nitrogen blend with air has the same venting requirements

as propane/air when the silane concentration reaches about 3%. This type of quantitative

information is necessary to design adequate protection, but has not typically resulted from othersilane studies.

The same project produced unpublished data on the ignition behavior of releases of the 10/90

silane/nitrogen blend. The tests were performed at an ambient temperature of 25C (77F) and arelative humidity of 48%, using orifices ranging from 1.8 to 7.7 mm (0.070 to 0.305 in.) in

diameter. The release was always started at a sufficiently high flow to prevent spontaneous

ignition. The flow was then gradually reduced until the jet ignited. This ignition was found to

correspond to a constant exit velocity of about 6 m/s (1180 lfpm), essentially independent of the

diameter of the orifice. Blow-out conditions were also determined by gradually increasing the

exit velocity; an average value of exit velocity of 34 m/s (6690 lfpm) was found for orifices

ranging from 1.1 to 3.3 mm (0.043 to 0.130 in.) in diameter. The data indicated a somewhathigher blow-out velocity for larger orifices.

The releases reported in the HRC study[3] are inconsistent with these FMRC data. In the HRC

tests with a 10/90 silane/nitrogen blend, immediate ignition was observed for a release into free

air through a 0.015 in. (0.38 mm) diameter orifice from a source at 5 psig (1.34 bar). At this

pressure, the exit velocity is about 210 m/s (41,340 lfpm), which is much higher than the critical

velocity for prompt ignition observed in the FMRC tests. The same mixture was also reported by

HRC to immediately ignite when discharged into a 28 ft3(0.79 m

3) box through orifices of 0.015,

0.039, and 0.158 in. diameter (0.38, 1 and 4 mm). In these tests, the cavity just upstream of the

orifice was initially flushed with nitrogen and then charged with the test mixture, where it was

presumably held at atmospheric pressure before the discharge run. While this procedure may

explain the discrepancy with FMRC results, the discrepancy itself points to the difficulties inobtaining reliable results on the behavior of silane.

3.2.2 Ignition/Combustion Characteristics of 100% Silane Releases

As part of an effort addressing cleanroom protection problems, FMRC investigated issues related

to the behavior of pure silane releases. The work completed to date has provided data on the

reactivity of silane/air mixtures from piloted ignition tests at all concentrations for which stable

conditions (no spontaneous ignition) can be established. More specifically, stable mixtures of

silane have been made successfully in the FMRC 5.1-liter sphere, for concentrations up to just

over 4% (by volume).

In these tests, the reactivity of the mixtures was characterized from measurements of fundamentalburning velocity and maximum rate of pressure rise. The results show that the burning velocity is

an increasing function of the silane concentration, starting at about zero near the LEL (about

1.4%) and reaching almost 5 m/s (985 lfpm) at concentrations near 4%. Essentially the same

trend was observed in tests with mixtures in air of a simulated 10/90 (silane/nitrogen) blend,

which covered silane concentrations up to about 3.5%. For these lean mixtures (the silane

concentrations at stoichiometric conditions are 9.5 and 5.1%, respectively for pure silane and for

the 10/90 blend), these data indicate that the reactivity of the mixture is determined only by the


14/98

6


amount of the limiting component (the fuel) and is unaffected by the fact that a portion of the

excess oxygen is replaced by nitrogen in mixtures produced with the 10/90 blend.

When the silane concentration is greater than about 4.5%, the mixture is metastable and will

undergo bulk autoignition after a certain delay. Ignition delay data were obtained for these

conditions at silane concentrations up to 38%, with shorter ignition delays at higher

concentrations. The maximum pressure developed by the reaction was fairly constant over therange of conditions tested and equal to about 150 psia (10.3 bar). This is consistent with the

predictions from a chemical equilibrium program, which indicate that the constant-volume

explosion pressure should be relatively constant over the range of the tests. Because of

limitations in the instrumentation and data acquisition set-up, the experiments did not produce

adequate information on the characteristics of these self-ignited explosions. It was, therefore, not

possible to conclusively determine the nature of the process, i.e., fast deflagration versus

detonation or, as it is more likely, volume reaction from bulk ignition.

While limited, these data present a consistent picture of the behavior of quiescent silane/air

mixtures. Mixtures below about 1.4% are non-reactive and cannot be ignited. This confirms

generally available knowledge on the lower concentration limits for the flammability of silane inair. Between 1.4 and about 4.5%, mixtures can react if an ignition source is provided. In this

case, the reactivity increases with the silane concentration and the laminar burning velocity can

reach values (up to 5 m/s [985 lfpm]) in excess of those of worst-case hydrogen/air mixtures. If

the silane concentration is above about 4.5%, then the mixture is metastable and will self-ignite

after a certain delay. Data for this situation indicate that, in the case of the 5.1-liter vessel, this

ignition delay can range from 2 minutes for the lean mixtures to about 15 seconds at the upper

end of the concentration range tested. These results will be used to guide the interpretation of the

data obtained in the course of this project.

4 PROJECT OBJECTIVESThe purpose of the project was to generate reliable data to define the conditions of source

pressure and external ventilation needed for releases of pure silane from high- and regulated-

pressure piping to promptly ignite.

The specific project objectives were as follows:

1. Perform releases of 100% silane through a 10-ft length of 1/4 in. outside diameter (OD)

tubing, with cross ventilation of 200 and 100 lfpm, to evaluate process delivery pre-purged

pressures of 30, 50, 200, and 200 to 0 psig. The silane source was a 790 psig cylinder fitted

with a commercially available RFO of 0.010 in. diameter equipped with a sintered metal

filter. Additional tests at 50 lfpm cross ventilation and/or with a maximum flow duration of 5

seconds were optional.2. Perform releases of 100% silane through a 14 in. length of 1/4 in. OD tubing, with cross

ventilation of 200 and 100 lfpm, through commercially available RFOs (0.010, 0.014, and

0.020 in. diameter), equipped with a sintered metal filter and with supply pressures of 790,

1200, and 1650 psig. Additional tests at 50 and 0 lfpm were optional. Also perform one final

test in a commercially available ventilated gas cabinet for one set of conditions among those

tested.


15/98

7


5 TECHNICAL APPROACH

5.1 Proposed Test Matrix

Three sets of tests were identified to achieve the stated objectives.

5.1.1 Process Delivery Pressure Releases with Cross Flow

This set of tests was intended to address the accidental discharge from a regulated pressure line

to a process tool. The planned simulation involved releasing silane into a ventilated enclosure

from a pre-pressurized line 10 ft (3.0 m) long, 1/4 in. (6.4 mm) outside diameter, supplied from a

source at 790 psig (55.5 bar) through a 0.010 in. (0.25 mm) diameter RFO. The matrix called for

20 tests for the following parameter values:

Line Pressure: 30, 50, and 200 psig (3.1, 4.4, and 14.8 bar)

Ventilation Flow: 50, 100, and 200 lfpm (0.25, 0.51, and 1.02 m/s)

In addition, initial testing was also planned to determine the effect of the presence of air or

nitrogen in the exit section of the line.

5.1.2 RFO Testing Within a Ventilated Enclosure

This set was intended to address the accidental high pressure discharge from a pigtail next to a

cylinder. The planned simulation involved releasing silane into a ventilated enclosure from a pre-

pressurized line 14 in. (356 mm) long, 1/4 in. (6.4 mm) outside diameter, connected to a source

through RFOs of two sizes. The matrix called for 18 tests for the following parameter values:

RFO Size: 0.010 and 0.020 in. (0.25 and 0.51 mm)

Source Pressure: 790, 1200, and 1500 psig (55.5, 83.8, and 104.4 bar)

Ventilation Flow: 100 and 200 lfpm (0.51 and 1.02 m/s)It was different from the initial project objectives in that the intermediate size RFO (0.014 in.)

was excluded because it was believed that the two remaining sizes would have provided data

bracketing the condition of the one excluded.

5.1.3 Demonstration Test in Gas Cabinet

This final demonstration test was intended to confirm that, for one set of representative

conditions, the behavior of the silane release in a realistic geometry would be the same as the

behavior in the ventilated enclosure.

5.2 Experimental Set-Up and Facilities

The experimental apparatus used for the tests is shown in Figure 1. The 3 ft (0.91 m) high

chamber has a cross section of 4.5 ft2(0.42 m2), 1.5 by 3 ft in plan view (0.46 by 0.91 m). The

side walls of the injection chamber are made of one 1/8 in. (3.2 mm) steel plate (on the silane

injection side), two 26-gauge (0.45 mm) galvanized steel plates, and one 1/8 in. (3.2 mm)

polycarbonate plate. The first plate (1/8 in. [3.2 mm] steel) is bolted to the supporting frame,

whereas the remaining three are held in place by strip magnets. The chamber overpressure

required to "blow out" these plates is not known with accuracy but is estimated to be very low,

most likely around a few inches of water. A perforated metal plate with an open area of about


16/98

8


30% is located at the entry cross section of the chamber to minimize the effect of external flows

(wind).

The ventilation flow was provided by a suction blower capable of a maximum flow rate of

1350 scfm (0.64 m3/s). The silane was injected through a 1/4 in. (6.4 mm) outside diameter tube

protruding for 2 in. (51 mm) into the enclosure. The injection point was located in the middle of

the short dimension (1.5 ft [0.46 m] side) and at 22.5 in. (0.57 m) above the open bottom of theenclosure to minimize entrance effects and possible silane leakage at the highest release rates.

A detailed view of the piping used to supply the silane is shown in Figure 2. All testing was

carried out using written procedures, both for safety and consistency in the results.

The lines were purged of nitrogen before the system was activated and after every test day. The

lines were charged with silane before each test.

Figure 1 Overall View of Ventilated Enclosure


17/98

9


Figure 2 Detailed Schematic of Piping Used for Process Pressure and RFO Release

Tests

5.3 Instrumentation

The experimental facility was instrumented to document the conditions of the releases. This

included measurements of

1. Pressure and temperature of the silane at the simulated source (channel tagsPSIL1 and TSIL1).

2. Pressure and temperature of the silane near the release point (channel tags PSIL2

and TSIL2).

3. Gas temperatures at two locations on the geometric axis of the release: at 6 and

18 in. (152 and 457 mm) from the exit of the 1/4 in. line (channel tags TENCL1

and TENCL2).

4. Gas temperature in the exhaust duct above the ventilated enclosure (channel tag

TDUCT).

5. Enclosure pressure (channel tag PLOCKER, gas cabinet test only).

All temperatures were measured with type-K thermocouples (Chromel-Alumel), which have auseful range of -2001250C (-3282282F). In addition, environmental conditions wererecorded through measurements of ambient pressure, temperature, humidity, and wind speed.

The ventilation flow was set by measuring the flow in the exhaust duct using a Pitot tube. The air

velocity near the release point was checked by using a hand-held anemometer.


18/98

10


5.4 Data Acquisition

A UNIX-based Hewlett-Packard system was used for data acquisition. This system can handle 24

analog and 16 digital channels of data, at acquisition rates up to 100,000 readings/sec. Software

is available for online monitoring of instrumentation and post-test data display, plotting, and

analysis. In most tests, this equipment was operated at a data acquisition rate of 100 scans/sec,

for subsequent averaging to bring the effective rate to 10 scans/sec. This was high enough toensure adequate time to capture significant occurrences. Events (times of valve operation) were

recorded as part of the data. All tests were video taped using a Hi-8 camera.

6 TEST RESULTS

6.1 Shakedown Tests

6.1.1 Scope of Testing

Extensive shakedown testing was carried out as part of the test procedures development and to

resolve some issues before production testing. The first set of tests was run using a version of the

apparatus that was somewhat different from the equipment for the production experiments. These

tests also involved releases into the enclosure with most of the side walls removed and no forced

ventilation. They can, therefore, be assumed to have taken place in a nominally quiescent

environment. Despite these departures from the conditions specified for the project objectives,

the data in all cases were still useful.

6.1.2 Results with Preliminary Test Set-Up

Sixteen tests were performed with a preliminary version of the apparatus, which functionally

resembled the set-up used for the RFO tests (see Figure 1 and Figure 6). The main difference was

that the actual RFO (0.020 in. [0.51 mm] diameter, fitted with filter) in all tests was located rightafter the pressure regulator attached to the silane supply and that the shut-off valve was about

32.3 ft (9.8 m) downstream of the RFO. Also, in the first nine tests, a remotely operated nitrogen

purge was installed in the exit stub at a cross located just downstream of the shut-off valve. In all

tests, the supply pressure ranged from 790 to 850psig (55.559.6 bar) and the regulated pressure

was set at 200400 psig (14.828.6 bar)

The main objective of these shakedown runs was to study the effect of the nitrogen purge on the

ignition behavior of the silane release. The results were somewhat inconclusive, possibly because

of inadequate experimental control over the test conditions. For example, in some cases, no

ignition was observed at initial pressures of 200 psig (14.8 bar), whereas "almost" prompt

ignition was recorded in releases starting at 400 psig. In these cases, ignition was not truly

prompt, since it occurred after a delay of 1 to 2.5 seconds after the release began. In the tests

where no ignition took place after opening the shut-off valve, the silane flow was slowed down

by lowering the regulated pressure upstream of the RFO until ignition was observed at the exit

point. The corresponding silane exit velocity was estimated to range from 3.5 to 5 m/s.

The main finding from this phase of testing was the discovery of a discrepancy in the flow rate

through the filter-fitted RFOs used in the experiments. The installed RFOs provided only a

fraction (in some cases less than one-third) of the nominal flow rate estimated for that size

orifice. The flow through the RFO was estimated using the data on repressurization of the line


19/98

11


downstream of the RFO after closing the shut-off valve at the end of a test. Since this flow

discrepancy could complicate the performance of the experiments, all subsequent tests used

RFOs without filters.

6.1.3 Results with Set-Up for Process Pressure Releases

Thirteen more shakedown tests were carried out using the set-up required for the process pressurereleases (see Figure 3 below). These tests, intended to resolve the question of the effect of the

nitrogen purge in the exit stub, used a simulated source of 790 psig (55.5 bar) and a 0.010 in.

(0.25 mm) diameter RFO (without filter). Ambient conditions were 1415C temperature and

5060% relative humidity for the first eight tests, 2022C and about 40% for the last five. As inthe case of the first set of shakedown experiments, the silane was released into a partial enclosure

(only the exit nozzle panel and one side panel installed) without ventilation.

In the first five tests in which the exit stub was purged of nitrogen, the jet promptly ignited in the

two runs at initial pressures of 200 and 250 psig (14.8 and 18.2 bar). No ignition took place in the

three tests with an initial pressure of 300 psig (21.7 bar). The following three tests with air in the

exit stub were carried out at initial pressures of 200, 250, and 300 psig (14.8, 18.2 and 21.7 bar),respectively. These releases did not ignite. The last five tests were done at initial pressures of 200

and 150 psig (14.8 and 11.3 bar), both with and without nitrogen purge, and 100 psig (7.9 bar)

with air in the exit stub.

Figure 3 Simplified Schematic of Supply Piping Used for Process Pressure Release

Tests

No prompt ignitions were observed. In these tests, the silane ignited when the pressure upstream

of the RFO was turned down, corresponding to exit velocities of 3-3.5 m/s.

This set of tests was performed with good control of the conditions of the experiment. The results

tend to support the conclusion that the presence of air in the exit stub does not increase thechances of silane to promptly ignite. Since the enclosure was open to the surroundings on two

sides, the release was not completely isolated from external disturbances. It is not believed,

however, that this had any effect on the prompt ignition of the jets. From the higher initial

pressure runs there is some indication of increased susceptibility of the jet to ignite promptly

when nitrogen is in the exit stub. This effect, however, should be confirmed through further

testing. Experimental scatter may be responsible for some of these data, since one of the

production tests (see Test #11 in Table 1), which were all run with air in the exit stub, resulted in

prompt ignition at a 200 psig (14.8 bar) initial pressure. For this project, this potential uncertainty


20/98

12


did not warrant additional study. Furthermore, these results support the conclusion that air in the

exit line (which is more realistic in terms of accident simulation) does not artificially enhance the

tendency of silane releases to ignite promptly. Because of this, all production tests were

performed with air in the exit section of the line.

6.2 Process Delivery Pressure Releases with Cross Flow

6.2.1 Experimental Procedure and Phenomenology

A simplified schematic of the silane supply piping used for this set of tests is shown in Figure 3.

The silane source is fitted with a shut-off valve (V1) installed at the cylinder. This is followed by

a first stage of pressure regulation (PR1) and an instrumented fitting to monitor pressure (p1) and

temperature (T1). An RFO is located at the end of a 29-ft (8.84-m) long 1/4 in. (6.4 mm) diameter

line. A second step of pressure regulation (PR2), right after the RFO, controls the pressure in the

following 10-ft (3.05-m) length of 1/4 in. (6.4 mm) line. At the end of this line is a second

pressure (p2) and temperature (T2) monitoring station and a shut-off valve (V2). Finally, a short (5in. [127 mm] long) stub of 1/4 in. (6.4 mm) tubing is used for injecting the silane in the

ventilated enclosure. The 6 in. (152 mm) dimension indicated in Figure 3 is the distance from the

exit to the midpoint of the shut-off valve. The line volumes upstream and downstream of the

RFO were estimated to be 152 and 52 ml (9.3 and 3.2 in3), respectively.

In all tests of this group, the silane pressure upstream of the RFO was set equal to 790 psig

(55.5 bar), while the 10-ft (3-m) line upstream of shut-off valve V2was filled with silane at a

pressure of 30-200 psig (3.1-14.8 bar) as regulated by the setting of PR2. The exit stub was

cleaned of silica deposits and flushed with air before each release.

A release is initiated by the sudden opening of valve V2. The resulting burst of silane into the

ventilated enclosure either leads to immediate ignition or to the formation of a non-reacting jet.In either case, the outflow of silane rapidly (


21/98

13



LineCase of No Prompt Ignition


22/98

14



LineCase of Prompt Ignition


23/98

15


At the opening of shut-off valve V2, the section of line downstream of the RFO empties very

rapidly, as indicated by the sudden drop in pressure p2, and a burst of silane is discharged in the

ventilated enclosure. At the same time, the source pressure (p1) starts to decrease slightly,

dropping eventually to about 650 psig (45.8 bar), because of a difference between static and

dynamic set point of the pressure regulator (PR1). After the initial inventory is discharged, the

silane flow is controlled by the RFO. Since, in this case, the jet did not ignite, the valve (V1) onthe silane supply is closed at about 10 seconds into the release. The flow is then sustained by the

inventory of silane in the 29-ft (8.84-m) long section of line upstream of the RFO. Eventually,

the exit velocity of the jet drops to a sufficiently low value for the release to ignite. This occurs at

about 82 seconds into the test, at a time when the estimated exit velocity is 2.6 m/s. Ignition is

marked by a rise in the temperature (Tencl,1) measured by the thermocouple on the jet axis at 6 in.

(152 mm) from the end of the exit stub. At about 94 seconds into the test, the shut-off valve (V2)

is closed, causing a repressurization of the line downstream of the RFO (see pressure p2and

temperature T2).

The prompt-ignition case shown in Figure 5 is from Test #17, which had 30 psig (3.1 bar) initial

pressure, 0.010 in. (0.25 mm) RFO and a ventilation flow of 50 lfpm (0.25 m/s). In this test, the

silane ignition follows the opening of the shut-off valve (V2), as indicated by the rapid rise of the

temperature in the enclosure near the open end of the exit stub (Tencl,1). Burning continues for

about 10 seconds, at which time the shut-off valve (V2) is closed, causing the line downstream of

the RFO to be recharged. This is indicated by the return of pressure p2to the pre-test value of 33

psig (3.3 bar) and by the increase in T2, because of compression of the silane in the line. The data

in the bottom graph show that very high gas temperatures develop in the enclosure (up to 1000C

[1832F] measured by Tencl,1) and that temperature rises significantly in the gas flowing through

the exhaust duct (temperatures of 170190C [338-374F] measured by Tduct).

6.2.2 Detailed Test Results

Twenty-three silane release tests to simulate failure of a 1/4" line to a process tool wereperformed on three consecutive days. As required by the test plan, these experiments involved

releases from a pressurized line (30-200 psig [3.1-14.8 bar]) into a ventilated enclosure. All

production runs were performed with air in the exit stub.

Tests were carried out for two RFO sizes (0.010 and 0.020 in. [0.25 and 0.51 mm] diameter, no

filter), three ventilation rates (200, 100, and 50 lfpm, corresponding to 900, 450, and 225 scfm

[0.42, 0.21, and 0.106 m3/s]), and three initial line pressures (30, 50, and 200 psig [3.1, 4.4 and

14.8 bar]). The results of the tests are reported in Table 1, while full data plots are provided in the

Appendix. These data support the following observations for the main features of the ignition

behavior of the silane releases.


24/98

16


Table 1 Silane Releases at Process Delivery Pressure with Cross Flow

Test # RFO

Diameter

[in.]

Ventil.

Rate

[lfpm]

Initial

Press.

p2[psig]

Rel.

Hum.

[%]

Amb.

Temp.,

Tencl,1[C]

Silane

Temp.,

T2[C]

Prompt

Ignition

[Y/N]

Veloc.@

Ign.

[m/s]

1 0.010 200 30 30 25 28 Yes -

2 " " 202 32 25 22 No 2.3

3 " " 203 32 26 20 " 0.9

4 " " 30 35 24 24 " 2.6

5 " " 52 34 25 23 Yes -

6 " " 54 36 25 24 " -

7 " " 29 35 24 24 " -

8 " 100 31 35 23 22 " -

9 " " 32 35 24 22 " -

10 " " 202 34 23 18 No n/a

11 " " 204 77 23 18 Yes -

12 " " 202 77 23 18 No 2.0

13 " " 202 77 23 18 " 0.8

14 " " 53 77 23 22 " 0.5

15 " " 52 77 23 22 " 0.6

16 " 50 31 53 30 30 Yes -

17 " " 33 53 30 30 " -

18 " " 203 53 30 24 No 4.5

19 " " 203 41 30 24 " n/a

20 0.020 100 30 49 23 23 Yes -

21 " " 203 48 25 19 No 5.2

22 " " 202 48 24 20 " n/a

23 " " 29 48 24 24 Yes -

NOTE: Silane source at nominal pressure of 790 psig. Reported silane temperatures (T2) are the lowest valueright after opening the shut-off valve.

6.2.2.1 Prompt Ignition

The prompt ignition characteristics of the releases appear to be a function of initial pressure and

show no clear dependence on the ventilation rate. The limited data obtained for the 0.020 in.

(0.51 mm) RFO at 100 lfpm (0.51 m/s) show no effect of the RFO diameter. This is not

surprising, since the exit velocity of the initial jet is orders of magnitude greater than the velocity


25/98

17


of the air flow into which it enters, and the RFO size affects only the rate of outflow established

after the initial discharge.

6.2.2.1.1 Initial Pressure of 200 psig

With one exception (Test #11), silane charges at initial pressures of 200 psig (14.8 bar) did not

ignite promptly. The anomalous test involved a release that ignited in a cross flow of 100 lfpm(0.51 m/s). The other three tests (#10, 12, and 13) for the same conditions consistently showed no

prompt ignition. Furthermore, no prompt ignition was observed for the tests at cross flows of 200

and 50 lfpm (1.02 and 0.25 m/s).


With one exception (Test #4), charges at initial pressures of 30 psig (3.1 bar) did ignite promptly.

The anomalous test involved a release that did not ignite in a cross flow of 200 lfpm (1.02 m/s).

The other two tests (#1 and 7) for the same conditions consistently showed prompt ignition.

Prompt ignitions were consistently observed for the tests at cross flows of 200 and 50 lfpm (1.02

and 0.25 m/s).


Tests at an initial charge pressure of 50 psig (4.4 bar) ignited promptly (Tests #5 and 6) in the

200 lfpm (1.02 m/s) cross flow, whereas the same initial pressure in the 100 lfpm (0.51 m/s)

cross flow (Tests #14 and 15) did not. This is the only evidence that a reduction in cross flow

velocity likewise reduces the chances of silane to ignite promptly. It is interesting to note,

however, that the 30 psig (3.1 bar) releases still ignited promptly at the much lower ventilation

rate of 50 lfpm (0.25 m/s). If considered of sufficient importance, the result at the 50 psig (4.4

bar) condition should be confirmed through additional testing.

6.2.2.2 Ignition During Silane Line Discharge

During discharge of the high pressure section of the 1/4 in. (6.4 mm) line, silane ignited at exit

velocities ranging from 0.5 to 5 m/s (1001000 lfpm). There does not appear to be any obvious

dependence of the exit velocity on the parameters varied in the tests, particularly the ventilation

rate. These ignitions occurred relatively late in the discharge process, as indicated by the

calculated initial exit velocities at a supply pressure of 790 psig (55.5 bar), which were 31.3 and

125.2 m/s (6160 and 24,650 lfpm) for RFO sizes of 0.010 and 0.020 in. (0.25 and 0.51 mm),

respectively.

6.3 RFO Testing Within a Ventilated Enclosure

6.3.1 Experimental Procedure and Phenomenology

For these tests, the released silane was injected in the ventilated chamber shown in Figure 1. The

supply piping used was similar to that in the experiments simulating process pressure releases,

except for changes to simulate the different conditions of the postulated accident scenario. A

schematic view of the modified piping is shown in Figure 6.


26/98

18


Figure 6 Simplified Schematic of Supply Piping Used for RFO Release Tests

As in the earlier set-up, the silane source is fitted with a shut-off valve (V1) installed at the

cylinder. This is followed by a pressure regulator (PR1) and an instrumented fitting to monitor

pressure (p1) and temperature (T1). The restricted flow orifice (RFO) is located at the end of a 31ft. (9.4 m) long 1/4 in. (6.4 mm) diameter line, which is followed by an 8 in. (203 mm) section of

1/3 in. (6.4 mm) tubing terminated by a shut-off valve (V2). A second pressure (p2) and

temperature (T2) monitoring station is provided just upstream of this valve. The silane is again

injected into the ventilated enclosure through a short (5 in. [127 mm] long) stub of 1/4 in. (6.4

mm) tubing.

As with the low pressure release tests, the 6 in. (152 mm) dimension indicated in Figure 6 is the

distance from the exit to the midpoint of the shut-off valve. The volumes upstream and

downstream of the RFO were estimated to be 162 and 7.8 ml (9.9 and 0.48 in3), respectively. The

hold-up volume of 7.8 ml downstream of the RFO is equivalent to that of a section of 1/4 in.

tubing 18.7 in. (475 mm) long, which is slightly longer than the 14 in. (356 mm) distance fromthe RFO to the end of the exit stub. The difference, because of the volume of the instrumented

cross located upstream of the shut-off valve (V2), is conservative in that it makes the inventory of

compressed silane greater than it would be in a straight section of tubing.

In these tests, the silane pressure upstream of the RFO was set equal to 790 psig (55.5 bar), or to

the highest pressure (1350-1450 psig [94.1-101 bar]) available in the supply. The exit stub was

cleaned of silica deposits and flushed with air before each release.

As with the first set of tests, the sudden opening of the shut-off valve (V2) released silane into the

ventilated enclosure. At these high initial pressures, the resulting burst of silane is expected to

form a non-reacting jet. Following this initial short burst, the outflow of silane rapidly

(


27/98

19


temperatures in the enclosure at 6 in. (152 mm) from the end of the exit stub (Tencl,1) and in the

center of the exhaust duct (Tduct).

In this test, opening the shut-off valve (V2) releases the rapid discharge of the silane inventory

downstream of the RFO. The resulting expansion causes the temperature (T2) to drop to close to

the boiling temperature of silane (-112C) and to remain there during the release. The lowtemperature of the silane jet is reflected in the gas cabinet by the data recorded for Tencl,1and is

also apparent, even though to a lesser extent, in the drop in temperature of the gases leaving the

enclosure through the exhaust duct (see Tduct). When the shut-off valve (V2) is closed,

temperatures and pressures in the lines return to near their pre-test values. The conditions of the

source (see p1and T1) remain quite constant during the entire release.


Twenty-one tests simulating an accidental silane release from a pigtail were performed over three

days (Tests #24-32, Tests #33-38, and Tests #39-44). These experiments involved releases from

a pressurized line (790-1450 psig [55.5-101 bar]) into an enclosure with a cross flow of 200 or

100 lfpm (1.02 or 0.51 m/s). As with the tests for process pressure releases, all experiments wereperformed with air in the 5 in. (127 mm) long exit stub.

Testing was carried out for two RFO sizes (0.010 and 0.020 in. [0.25 and 0.51 mm] diameter, no

filter), two ventilation rates (200 and 100 lfpm, corresponding to 900 and 450 scfm [0.42 and

0.21 m3/s]), and three initial line pressures (790, 1200 and about 1400 psig [55.5, 83.8 and

97.6 bar]). The conditions of the tests are reported in Table 2, and data plots are in the Appendix.

In terms of results, the test series was uneventful. The following observations describe the

behavior of the releases:

1. No prompt ignitions took place in any of the tests.

2. In some of the tests, pops were heard at shut-off, with the sound somewhat louder during the

lower pressure releases (790 psig [55.5 bar]). In a couple of cases, sufficient pressure was

developed to dislodge the Lexan panel (without damaging it, however).


28/98

20


Figure 7 Typical Test Results for a Simulated Silane Release from a High Pressure

Line (RFO Case). No Prompt Ignition.


29/98

21


Table 2 Silane Releases through RFOs into a Ventilated Enclosure

Test # RFO

Dia.

[in.]

Ventil.

Rate

[lfpm]

Initial

Press. p2[psig]

Rel.

Hum.

[%]

Amb.

Temp.,

Tencl,1[C]

Silane

Temp.,

T2 [C]

Prompt

Ignition

[Y/N]

Ign. Pop

@ End

24 0.010 100 840 81 21 -19 No No

25 " " 1225 77 21 -66 " "

26 " " 1250 75 22 -66 " Yes*

27 " " 800 70 23 -7 " No*

28 0.020 " 800 66 23 -22 " Yes

29 " " 1250 67 24 -92 " No

30 " " 1280 60 25 -104 " "

31 " " 1260 60 25 -90 " Yes*

32 " " 790 60 24 -21 " Yes*

33 " " 1380 70 18 -105 " No

34 " " 1380 70 18 -105 " "

35 0.010 " 1450 57 21 -65 " "

36 " " 1450 53 22 -75 " "

37 " 200 800 49 22 -10 " Yes

38 " " 1440 49 22 -70 " No

39 " " 1310 38 18 -65 " "

40 " " 780 35 18 -17 " Yes

41 0.020 " 1360 30 20 -107 " No

42 " " 790 27 20 -27 " Yes

43 " " 790 23 20 -28 " No

44 " " 1320 23 20 -108 " Yes

NOTE: The asterisk (*) indicates tests with silane flow turn-down. The reported silane temperatures are the

lowest recorded value right after opening of the shut-off valve.

In summary, at these high initial line pressures, no prompt ignition appears possible, regardless

of the ventilation rate in the enclosure. This is consistent with the results from the tests for

process pressure releases and from the shakedown tests, which showed that prompt ignition

would generally not occur at initial line pressures of 200 psig (14.8 bar) or higher. The sudden

gas expansion upon opening the shut-off valve (V2) cools the silane to very low temperatures (as


30/98

22


low as -110C, the boiling temperature of silane), indicating possible condensation of thedischarge. Combined with the high velocity of the jet, this further reduces the chances of ignition.

6.3.3 Additional Tests with Restart of Silane Release

Six additional release tests were performed after some of the production tests to study the

behavior of the silane release under restart conditions. In these runs, which always followed testsin which silane did not ignite promptly, the 5 in. (127 mm) exit stub was left untouched (i.e., no

flushing with air). The restart of the release occurred about 4.5 to 15 minutes after the completion

of the previous test and was initiated with the lines up to the shut-off valve (V2), fully charged

with silane, but with the valve near the silane source (V1) closed.

These restart tests were carried out after Tests #36, 38, 39, 40, 41, and 44. As a result, most

involved releases at initial pressures of 1300 psig (90.7 bar) or higher, except the restart after

Test #40, which had an initial pressure of 790 psig (55.5 bar). These tests were initiated by

opening valve V2and by keeping it open for varying time intervals, during which the silane in the

lines was discharged (valve V1was left closed). No prompt ignitions were observed in any of the

runs when the flow was first initiated.In the first two tests (after Tests #36 and 38), the discharge was allowed to continue until the

lines emptied completely. No ignition of the silane was observed at any time during the

discharge. Because of the high initial pressure of these two releases (about 1450 psig [101 bar]),

the initial temperature of the silane jet was low (


31/98

23


6.4 Demonstration Tests in Gas Cabinet

6.4.1 Experimental Set-Up and Procedure

Five final demonstration tests were performed by simulating a high-pressure silane release into a

ventilated gas cabinet. The particular cabinet used was a commercially available model that could

accommodate three cylinders. The unit was 37 in. (0.94 m) wide and 74 in. (1.88 m) high. Thedepth of the cabinet was 15.5 in. (0.39 m) for the bottom 43 in. (1.09 m) and 17.5 in. (0.44 m) for

the remainder, yielding flow cross sections of 4.0 ft2(0.37 m2) and 4.5 ft2(0.42 m2), respectively.

The silane injection tube was installed in the middle of a side wall, at the approximate height (59

in. [1.50 m]) of the Compressed Gas Association (CGA) fitting on a compressed gas cylinder and

ended 2 in. (51 mm) inside the cabinet. A pressure transducer was installed on the same wall,

below the silane injection point, 44 in. (1.12 m) from the cabinet bottom. The procedures for

these demonstration tests were identical to those used in the RFO tests.


The five demonstration tests were carried out for ambient conditions of 2022C (6872F)temperature and 5762% relative humidity. All tests had a ventilation flow of 100 lfpm(0.51 m/s) and a 0.020 in. (0.51 mm) RFO installed 8 in. (203 mm) upstream of the shut-off

valve (see Figure 6). The initial silane pressure was about 800820 psig (56.257.6 bar) for the

first two tests and 12401300 psig (86.590.7 bar) for the last three. These demonstration runs

confirmed the results from the RFO tests (no prompt ignition when the silane release was first

started). The silane flow was restarted in some of the tests with the outcome detailed in the

following discussion (also refer to the data plots in the Appendix).

In the first test (#45) of this series, which was run at an initial pressure of 800 psig [56.2 bar]),

the silane flow was stopped after 10 seconds. No ignitions were observed either at start-up or

shut-down. The only indication of reaction was a very small temperature rise (about 0.7C

[1.3F])in the exhaust duct right after closing the shut-off valve (V2). At that time, the

temperature of the silane jet was -37.3C (-35.1F).

Except for a slightly higher initial pressure (820 psig [57.6 bar]), the second run (Test #46) was

essentially a repeat of the first. In this case, however, the excess flow valve (V1) was closed at the

end of the 10-second release while the shut-off valve (V2) was left open. During the ensuing,

gradually slowing release, the jet ignited briefly(about 12 sec) when the pressure upstream of

the RFO had dropped to near atmospheric, too low to be accurately measured by the high-range

pressure transducer. Another small flame appeared a few seconds later, when the flow was finally

interrupted by closing the shut-off valve.

Test #47 was the first of the releases performed for the highest source pressures available(1240 psig [86.5 bar]). A small ignition flash was observed at the end of the 10 second release,

which was terminated when the shut-off valve (V2) was closed. When the flow was interrupted,

the temperature of the silane was -101.6C [-151F].

Test #48 repeated the previous, until the flow was shut off after the initial 10-second release.

Both tests yielded a small ignition flash shortly after the flow was stopped. In this test, however,

the silane flow was restarted 10 seconds after the shut-off, at which time it ignited immediately,

followed by a rise of pressure in the cabinet sufficient to open the smaller of the two doors.


32/98

24


Subsequent examination of the gas locker and of the pressure data revealed that the latch of the

door that opened was incompletely engaged and that the pressure in the locker had reached only

0.27 psig (19 mbarg) about 25 msec after the re-opening the shut-off valve.

The last test (#49) involved a slightly higher source pressure than the previous two (1300 psig

[90.7 bar]), but otherwise the initial behavior of the release was identical, since no ignition

occurred at flow start. About 1 second after the flow was shut off, a small explosion occurred inthe cabinet, which opened the access window on the larger door. The latch on the smaller door

had been re-adjusted after the previous test, so that both doors were properly closed during this

run. Post-test examination of the equipment and the data revealed that the plastic latch on the

window had failed and that the pressure in the cabinet had reached 0.11 psig (7.6 mbarg). The

silane flow was restarted about 14 seconds after the end of the first release; the jet ignited

immediately, and an overpressure of about 0.2 psig (14 mbarg) caused the (now unlatched)

access window to open again.

In summary, these demonstration tests confirmed and better quantified the behavior of the high-

pressure releases simulated in the RFO tests in the ventilated enclosure. Consistently, the silane

did not ignite at release but did at restart. This ignition eventually damaged the gas cabinet (abroken plastic lock), which was consistent with the limited pressure rise on the order of a few

tenths of a psi (a few tens of mbar) measured in the cabinet.

7 DISCUSSION

7.1 Summary of Overall Behavior of Silane Releases

The results from the silane releases in this study provide a consistent picture of the ignition

behavior of this material under the conditions reproduced in the tests; i.e., a sudden leak into a

ventilated enclosure from a charged 1/4 in. (6.4 mm) line equipped with an RFO (0.010 or 0.020

in. [0.25 or 0.51 mm]) at some distance upstream of the exit point.

The prompt ignition characteristics of the release have been found to be a function of the initial

silane pressure; they are essentially independent of the ventilation rate and of the size of the RFO

in the line. To better illustrate this point, the ignition data from the releases at process delivery

pressure already reported in Table 1 are repeated in Table 3 in summary form: no prompt

ignitions were observed in any of the high pressure tests for releases through RFOs (see Table 2).

The data support the conclusion that silane will promptly ignite if the initial line pressure is 30

psig (3.1 bar) or less, whereas it will not ignite if the pressure is 200 psig (14.8 bar) or greater.

These limits are applicable to the conditions of the tests (average ambient temperatures in the

2025C [6877F] range and relative humidities from 2080%); they are generally valid even

though a couple of exceptions were observed. More specifically, one 200-psig release promptlyignited and one 30-psig release did not. These anomalous tests do, however, display a behavior

that counters previously accepted notions about the effect of ventilation on prompt ignition: the

200-psig release (Test #11) ignited promptly, even though the ventilation was relatively low (100

lfpm). The 30-psig release (Test #4), on the other hand, failed to ignite, even though the

ventilation was high (200 lfpm).


33/98

25


Table 3 Prompt Ignition Summary of Silane Releases in Cross Flow from a Line at

Process Delivery Pressure

Initial Pressure

[psig]

Ventilation Rate

[lfpm]

Orifice Size

[inch]

Prompt Ignition

[Yes/No]Number of Tests

30 200 0.010 Yes (2), No (1) 3

100 " Yes 2

" 0.020 Yes 2

50 0.010 Yes 2

50 200 0.010 Yes 2

100 " No 2

200 200 0.010 No 2

100 " Yes (1), No (3) 4

" 0.020 No 2

50 0.010 No 2

A final observation is supported by the results of selected tests, during which the silane flow was

re-started after a previous release was interrupted. These experiments indicated that the ignition

behavior of the releases is affected by the presence of partially reacted silane in the exit stub; i.e.,

silane can be expected to ignite promptly regardless of the initial pressure of the system. This

happens if the flow is re-initiated seconds after the flow is shut off after a release that does not

ignite.

7.2 Initial Release of Line Inventory

The following discussion of jets entering a ventilated enclosure is offered as a means to clarify

the phenomenology of silane releases and as background information for the recommendations in

the next section. Three aspects of the jets are covered in the discussion: 1. the release of the

initial inventory in the pressurized line; 2. the characteristics of the jet established by the RFO-

controlled flow; and 3. the average silane concentration after the release mixes with the

ventilation flow.

The volume of the line located downstream of the RFO contains the pressurized silane charge

that enters the ventilated enclosure shortly after the shut-off valve is opened. The volume of this

inventory is equal to 52 ml (3.2 in.3) for the process delivery pressure releases and 7.8 ml (0.48

in.3) for the RFO and the cabinet tests.

In the process delivery tests at initial pressures of 200 psig (14.8 bar), the silane charge in the

52 ml hold-up volume expands to 838 ml (51 in.3) at standard conditions, which would produce

8.8 liters (539 in.3) of stoichiometric silane/air under near worst-case conditions. This amount of

silane, if ignited in a confined volume equal to that of the ventilated enclosure used in the tests

(13.5 ft3[0.38 m

3]), would produce a pressure rise of 0.21 barg (3.1 psig). In the RFO tests, the

line hold-up volume was smaller (7.8 ml [0.48 in.3]), but the pressure was higher (1400 psig

[97.6 bar]). In this case, the release would involve a standard volume of silane of 1720 ml (105


34/98

26


in.3), which would generate a stoichiometric silane/air mixture of 18.1 liters (1100 in.

3). If

ignited, this could cause a maximum pressure rise of 0.44 barg (6.3 psig).

In the demonstration tests in the gas cabinet, the volume of the ventilated enclosure was 25.9 ft3

(0.73 m3). For this configuration, the largest amount of silane (1605 ml [98 in.3]) was released in

the last test, when the initial pressure was 1300 psig (90.7 bar), corresponding to a stoichiometric

volume of 16.9 liters (1030 in.3). Ignition of this amount of silane could have produced apressure rise of 0.21 barg (3.1 psig). The actual test results have shown that the addition of a

volume of silane equal to about 0.22% of the enclosure volume produced ignition pressures that

did not exceed 0.3 psig (20 mbarg). The estimated pressures were probably not encountered in

the tests because they were based on the assumption of a completely sealed (unvented) enclosure.

Also, it is almost impossible that the entire silane charge could mix in stoichiometric proportions

with air.

7.3 RFO-Controlled Jet Flow

The opening of the shut-off valve initiates the flow of silane first as a burst from the release of

the initial inventory under pressure, then as a sustained flow controlled by the pressure of thesource and by the size of the RFO. In both cases, a jet is formed: a transient, developing jet

during the initial stages of the release, and a steady state jet once the flow is stabilized. The

following simplified analysis provides a set of estimates for the parameters that characterize the

mixing patterns established by the forced jet.

Under the assumption of isothermal and subsonic conditions, it can be shown that the rate of

volume entrainment by a turbulent forced jet is given by:

v = 0.25 D U ,0 [1]

where D is the initial diameter of the jet (0.18 in. [4.57 mm] using the 1/4 in. tube with 0.035 in.

[0.89 mm] thick walls), and U0is the initial velocity of the jet. For the flow [6] produced by a

pressure of 1350 psig (94.1 bar) upstream of a 0.020 in. (0.51 mm) RFO, the exit velocity isapproximately 325 m/s (63,980 lfpm) (i.e., assuming that the jet temperature is equal to ambient).

Equation [1] then says that the entrainment rate in the jet is:

v = 0.37m / m / s = 240scfm / ft .3 [2]

This estimate means that the jet pulls in a flow of 240 scfm (0.113 m3/s) for each ft (0.3 m) of its

development. So, for the case of a jet spanning the width of the ventilated enclosure of about 3 ft

(0.91 m), the full entrainment requirement is 720 scfm (0.34 m3/s). If the ventilation flow is

lower than this value, the jet cannot entrain the flow it needs; therefore, the material being

released at the jet source will not be diluted as it would in an infinite environment. In reality, the

effects of the confinement on flow dilution will be felt before the ventilation flow reaches the

value calculated above. Note that this estimate was made for the largest RFO considered and for

a high source pressure. Reduction in these two parameters would lower the initial jet velocity,

U0, and, therefore, lower the entrainment (see Eq. [1]).

Additional useful information can be obtained by considering the decay of flow properties, such

as centerline concentration, Xc, centerline velocity, Uc, and jet half-width, b1/2, as a function of

distance, x, from the jet origin. The following relationships are generally accepted to describe

these variations for the case of subsonic, isothermal jets expanding in quiescent surroundings

[7,8]:


35/98


36/98

28


overestimated. The low overpressures recorded in the demonstration tests in the gas cabinet

indicate that the amount of silane that actually reacted was even less than the value calculated.

7.4 Average Silane Concentration in Exhaust Flow

After the initial transient associated with the silane release in the pressurized line, the leak into

the ventilated enclosure continues at a rate determined by the size of the RFO and by the pressureupstream of it. The hazard associated with this phase of the process is defined by the

concentration of unreacted silane in the stream leaving the ventilated enclosure. This is a relevant

quantity to consider, because its value determines whether an ignition of the mixture in the

enclosure will cause the ensuing explosion to propagate through the exhaust ductwork. Silane

flows through different size orifices and for a range of source pressures are shown in Table 4,

based on the predictions of an equation recommended by ISA [6].

Calculated values of silane concentration, assuming the release mixed perfectly with the

ventilation flow, are shown in Figure 8 for the 0.020 in. (0.51 mm) RFO at four source pressures.

Concentrations for a 0.010 and 0.014 in. (0.25 and 0.36 mm) would be approximately 25 and

49% of the values indicated. The curves in the figure can be used to calculate the minimumventilation flow to keep the concentration below the LEL of silane (1.4%). Flows of 895, 540,

220, and 83 scfm (0.42, 0.25, 0.10 and 0.039 m3/s) would be needed for source pressures of

1500, 1000, 500, and 200 psig (104, 70, 35.5 and 14.8 bar), respectively. Coincidentally, the

minimum flow requirement for the 1500-psig source corresponds to a ventilation velocity of

200 lfpm (1.02 m/s) in a gas cabinet with a 4.5-ft2(0.42-m2) cross section. Ventilation velocity,

however, is not a meaningful parameter since it has been found to have a negligible impact on

the ignition characteristics of silane releases. In addition, test results have shown that the forced

jet produced by the release induces its own mixing pattern which, in the case of a leak from a

1/4 in. line, will drop the concentrations to values near the LEL within a few feet from the

source.

If the average concentration in the enclosure is above the LEL value, but does not exceed about

4%, then the silane/air mixture is flammable and stable. In this case, a flame should be expected

to propagate when the mixture ignites. This outcome is not guaranteed, however, as shown by

some of the RFO and cabinet tests performed with the large RFO (0.020 in. [0.51 mm]), at

elevated source pressures and with a low ventilation rate of 450 scfm (0.21 m3/s), corresponding

to 100 lfpm (0.51 m/s). In this respect, the most severe test condition was reproduced in RFO

Tests #33 and 34 (see Table 2) and in cabinet Test #49, where the source pressure was 1380 psig

(96 bar) in the first two tests and 1300 psig (91 bar) in the last.


37/98

29


Table 4 Silane Flow Rates Through Restricted Flow Orifices Based on the Use of the

ISA Equation

Silane Flow Rate [slpm] (@70F Source Temperature, 0 psig Downstream Pressure)

RFO

Dia. [in.] Source Pressure [psig]

1500 1200 1000 800 600 400 200 100 50

0.020 355 277 214 157 108.5 67.8 32.7 16.6 8.74

0.014 174 136 104.9 76.9 53.2 33.2 16.0 8.13 4.29

0.010 88.8 69.2 53.6 39.3 27.1 16.9 8.18 4.15 2.19

NOTE: The flows through the 0.014 and 0.010 in. RFOs are equal to 49 and 25% of the flow through the 0.020

in. diameter RFO.

Figure 8 Calculated Silane Concentration in the Case of Perfect Mixing of a Release

with a Ventilation Flow

The average silane concentration resulting from these releases is around 2.5%, well within the

range that would support flame propagation into the exhaust duct. Propagation was actually not

observed in the tests because the release did not ignite at shut-off (in Tests #33 and 34) and, in

the one case of ignition (Test #49), this occurred after a 1-second delay, allowing for some

additional dilution of the mixture by the ventilation flow. If these tests were to be repeated with

changes in the internal geometry of the release (by the addition of obstacles in the path of the jet,

for example), the outcome could be different.


38/98

30


If the ventilation flow were low enough to allow for the average concentration to exceed about 4

4.5%, then the resulting mixture would be metastable and should be expected to spontaneously

ignite. This condition would clearly be most hazardous and should be carefully avoided.

8 RECOMMENDATIONS

The results from this test program, in combination with the data on stability of silane/air mixtures

obtained by FMRC, provide the basis for a revised approach to the design of ventilation for

silane storage areas. Since ventilation velocity (in lfpm or m/s) does not appear to have a

measurable effect on ignition characteristics, its use as a meaningful design parameter should be

discontinued. Instead, focus should be shifted to ventilation requirements (in scfm or m3/s),

which should be set on the basis of dilution calculations aimed at ensuring that the average

concentration from a leak will remain below the LEL of silane (i.e., less than 1.4% by volume).

In addition, the ventilation should be designed to eliminate dead zones near the leakage point to

minimize the likelihood of pockets with high concentrations of silane. Finally, a limit should be

imposed on the maximum amount of silane initially released as a percentage of the volume of the

enclosure to avoid a large pressure increase should the release promptly ignite. Note that thecabinet tests, where ignition pressures of 0.1-0.3 psig (7-20 mbarg) were measured, involved

initial silane inventories with standard volumes of about 0.2% of that of the cabinet.

More specifically, the following criteria are suggested for gas cabinets or other enclosed areas

that might experience leaks of 100% silane :

1. The ratio between the volume of the enclosed space and the estimated volume at

standard conditions of the released silane inventory should be large enough to

reduce the potential for developing unacceptable pressures leading to prompt

ignition. The silane inventory should be estimated by including the line between

the leakage point and the nearest flow-limiting device. The limited data obtained

during the program indicate that a value of 500:1 for this volume ratio willgenerally limit the pressure rise to the range of 0.10.3 psig (720 mbarg).

2. The possibility of explosive concentrations developing in the exhaust duct

should be eliminated by providing a minimum ventilation rate through the

enclosure equal to 75 times the standard volume flow rate of silane through the

flow-limiting device at the maximum pressure available from the source.

Additional ventilation may be provided to add a margin of safety to this

minimum requirement.

3. The ventilation should be arranged so that dead zones cannot form near the sites

where leaks might occur.


39/98

31


9 CONCLUSIONS

This study of 100% silane releases into ventilated enclosures has provided new data for designing

safety measures for storage cabinets and other areas where silane leaks might occur. Testing

focused on the behavior of accidental leaks from a 1/4 in. (6.4 mm) outside diameter line filled

with silane at a regulated process pressure or at the full pressure of the supply. In all cases, a

flow-limiting device was present upstream of the leakage point.

The main conclusion of the study is that the ventilation velocity does not have a measurable

effect on the prompt vs. delayed ignition characteristics of the silane. Discharges of silane

initially at high pressure (>200 psig [14.8 bar]) were generally found to form non-reacting jets,

whereas leaks from low-pressure lines (30 psig [3.1 bar]) tended to promptly ignite. Analysis of

the results has pointed to the importance of the dilution of the leak by the ventilation flow. This

is a relevant factor to prevent average concentrations in the exhaust stream from exceeding safe

values. A set of recommendations has been developed based on a dilution criterion.

Further research will be necessary to satisfactorily address some unresolved issues. For example,

the question

ignition of silane

Documents