EPA Region 5 Records Ctr.

Environmental Forensic* (2002) 3. 131-143doi:10.1006 enfo.2002 0087, available online at http:;/www.idealibrary.com on

269204

Speciation and Characterization of Heavy Metal-ContaminatedSoils Using Computer-Controlled Scanning Electron Microscopy

Stephen K. Kennedy Ph. D*, William Walker Ph. D and Barbara Forslund

RJ Lee Group, Inc. 350 Hochberg Road, Monroeville, PA 15146. Walker and Associates, 2618 J Street, Suite I.Sacramento, CA 95816 and Advanced GeoServices Corp, Chadds Ford Business Campus, Chadds Ford, PA 19317

(Received 6 November 2001, Revised manuscript accepted 10 April 2002)

In the analysis of heavy metal-contaminated soils, particle-by-particle determination of elemental composition (i.e.species or mineral) as well as the characterization of particle size, external and internal morphology, can be used toassess bioavailability. remediation potential and source. The imaging capabilities of the scanning electron microscope(SEM) coupled wi th the elemental analysis provided by energy dispersive spectroscopy (EDS) can be used to acquireboth chemical and physical information. Heavy metal particles can be analyzed automatically using computer-controlled scanning electron microscopy (CCSEM). The analysis first recognizes a particle and determines is peripheryand various size measures. An EDS spectrum is obtained, the constituent elements identified and the occurrenceclassified into a compositional type. A TIFF microimage with spectrum is saved for later off-line review and locationcoordinates are saved for on-line review.

Two examples illustrate the use of this technique in source identification. In one example, elevated levels of arsenicwere attributed to a manufactured product rather than a fugitive source. In a second example, the lead-bearingconstituents observed, as well as an association with a vuggy aluminosilicate occurrence and unburned coal, identifiescoal ash as the probable source. © 2002 AEHS. Published by Elsevier Science Ltd. All rights reserved.

Keywords: lead; arsenic; source identification; environmental analysis; automated analysis.

Introduction

In general, regulatory agencies require soil andsediment to contain less than specified levels of heavymetals. To determine if soil metal contents are withinregulatory guidelines, bulk samples are dissolved andsubjected to an appropriate method of chemicalquantitation or are ground to a powder and analyzedby X-ray fluorescence to yield heavy metal assays.These techniques can determine the concentrationsvery accurately, but information related to the chemicaland physical (size, shape, internal particle structure)characteristics of the heavy metal occurrences and theircarrier particles is lost in the process. The compo-sitional and physical characteristics of heavy metaloccurrences have been shown to influence bioavail-ability (see for example, Da\ is (.'tul., 1993, Gasscr etal..1996. and Ruby ci <il.. 1999). The compositionalcharacteristics of heavy metal contaminated soils havebeen recognized as a factor in remediation (see forexample M;i a t//.. IWx l.aperche ci <//.. 1996, andEighmy et it/.. IW7). and it seems that physicalcharacteristics (size and enclosed liberated nature ofthe occurrence) would affect the need lor or the efficacyof remediation as wel l . Fur ther , chemical and physicalcharacteristics can he used lo assisl in the identificationof contamination sources

"Corresponding au thor . I >r S K Kemu\K RJ I.cc Oroup 350Hochbcrg Road !5Mh MonroeMlle. PA USA lei: +1 724 325 1776Kax. +1 72-4 733 I7W E - m a i l -.kenned) „ i|k com

*2cpRun is a registered i i . i den i .nk ol Aspev Ins t rumen t s'WEBSEM is a registered t r a d e m a r k of R.I l-ce Ciroup

Determination of the compositional associationspresent is referred to as speciation. Speciation can beperformed using powder X-ray diffraction, but thistechnique requires that the species of interest be presentat a relatively high concentration, typically 5% ormore. Historically, speciation has been attemptedindirectly through sequential extraction or partialdissolution techniques. These methods involve selec-tively dissolving components using a sequence ofincreasingly rigorous reagents thought to be specificfor certain minerals or mineral classes.

There are problems with this approach, includingnon-specificity of the reagents involved, creation ofartifacts by the addition of partial dissolution reagents,and extraction schemes that are inappropriate for themineral or mineral assemblages existing in the sample.Use of partial dissolution techniques may result insolublilization of the metal of interest that may thenadsorb to other mineral surfaces not previously coatedor containing significant adsorbed species. This tendslo suggest that more metal exists in the adsorbed s la tethan that actually occurs.

The speciation methods mentioned above provide noinformation on physical characteristics. In some cases,the heavy metal occurs as a single-component discretepanicle, and in other cases, the heavy metal occurswi lh in a complex particle of multiple components,some heavy metal-hearing and some not. The phys ica lcharacteristics include the size and shape of i l i c hea\\metal occurrence and its carrier part ic le , and iheassociation of Ihe heavy metal component u i i h otherconst i tuents if it occurs as a complex particle.

I527-S422 112 03H131 - r< S35 onoxi :i«C Ai:HS Published r-v i;ise\ier Science Ltd All nghK reseived

132 Stephen K. Kennedy el al.

The association can be described, for example, asliberated, enclosed, rim or cement.

The elemental association identification (speciation)as well as the physical description (size, shape, andassociation), can be performed on epoxy-impregnatedpolished mounts of soil samples by particle-by-particleanalysis in the scanning electron microscope (SEM) orelectron microprobe (EM) with their imaging capabili-ties and elemental analysis by energy or wavelengthdispersive spectroscopy (EDS or WDS). When per-formed manually, potential heavy metal occurrences areidentified by backscattered electron intensity (bright-ness) and composition determined by EDS or WDS(Davis et al.. 1993, Link ft al., 1994). The size (longestapparent axis) of heavy metal occurrences are estimatedand the association described. Analysis generally con-tinues for 5-8 h or until an established number of heavymetal occurrences have been described.

The recognition and characterization of particles canbe performed by image analysis techniques. Krinslcyet al. (1998) give an extended coverage of imageanalysis pertaining to SEM data. In the methodspresented, images are acquired and then processed off-line to derive spatial attributes (size, shape, orientation,etc.). To derive compositional information, X-raymaps of multiple elements are acquired and stackedinto a single multilayer image. X-ray maps requireseveral minutes per field for acquisition.

A more efficient method of data acquisition is toprocess the "live" SEM image (see Henderson et al.,1989), a procedure known as computer-controlledscanning electron microscopy (CCSEM). In this way,particles are recognized, measured, their elementalcompositions and microimages acquired automatically.Before introduction of the Personal Computer (PC),these automated analyses were controlled by themicroprocessor in the EDS detector. In the early1980s, SEMs became more fully integrated with PCs(see for example Schamber. 1993) and CCSEManalysis became considerably more efficient. Someapplications of CCSEM are illustrated in Schwoebleet al. (1988, 1990), Casuccio <•/ al. (1983, 1989) andSong ct al. (1999). The methods as applied to theanalysis of heavy metals in soil is described herein andillustrated with two case studies.

differences in the soil geochemistry. Samples should bedried as soon as possible after collection to preventcontinued chemical reactions.

Sample preparation

A "strewn mount" can be prepared quickly for SEManalysis by sprinkling disaggregated soil onto an SEMstub with double sided tape. Although preliminaryinformation can be readily acquired, this preparationmethod does not allow internal particle structure(association) to be observed, and the EDS spectrumcan be somewhat degraded by surface topography,making compositional characterization somewhat moredifficult.

The preferred method of sample preparation is toproduce an epoxy impregnated polished mount. Soilcommonly displays a great range in particle size andshould be sieved. The polished mount is usually eitherone inch or one and a quarter inches in diameter, andobviously, pebbles or larger particles need to beremoved. This can be accomplished by sieving at1000 urn (an arbitrary value) to remove very coarsesand and coarser particles. If the issue is one ofingestion, sieving may be performed to obtain a samplefor analysis containing finer particles that are likely toadhere to surfaces that children place in their mouths.This size is not known with certainty, but 250 urn hasbeen suggested (Que Hee el al., 1985). Representativefractions of the whole sample and the various size splitsshould be saved for the determination of the heavymetal concentration in each size fraction using asuitable chemical technique.

To prepare the mount, a representative split ofapproximately one gram of the sieved dry sample isvacuum impregnated with epoxy in a ring and allowedto cure. A sample label is placed in the epoxy forpermanent identification. The "puck" is polished toremove at the minimum a thickness of material equalto the radius of the largest particle. Polishing occursusing kerosene or isopar as a lubricant to avoidchemical reactions that might occur w i t h water. Thislubricant is removed with ethanol. Finally, the sampleis given a thin coating of carbon to prevent chargingwhile under the electron beam in a vacuum.

Methods

Sample collection

Bulk soil or sediment samples are collected. As with anysample collection process, a strategy must be developedthat both maintains the in tegr i ty of the sample andrepresents the area from which the sample is collected.Samples are collected using in s t rumen t s such as plasticor Plexiglas that wi l l not introduce contamination,especially metals, to the sample. In some cases thesample should be collected and maintained in anitrogen atmosphere to l imit the introduction of oxygenthat may s ign i f ican t ly a l t e r speciation patterns. Ifhuman exposure is of i n t e r e s t , samples of residentialsoils close to the surface should he obtained in grassyareas, bare areas. ne;ir the d r i p - l i n e of the residence, andin any areas where chi ldren play. Composite samplesfrom deeper areas can be used to assess changes due to

Sample analysis

The SEM analyses for the case studies presented herewere performed on an RJ Lee Instruments (now AspexInstruments) PSEM 75 using the ZepRun* softwarefor CCSEM automation. Decisions related to theconduct of the automated analysis are established ina run parameter file, an element file and a rule file asdescribed below.

Ainil\sis set-up. The sample ring is scratched u i t h tworegistration marks and small strips of carbon andcopper tape arc tixed to the upper surface be I ore placingmul t ip l e samples into the SEM. The electron beam issaturated us ing an accelerating voltage of 20 KeV andimaging is set to the backscattered electron ( I H . ) modew h e r e the brightness of a component is propor t ional toits average a tomic number (;). The image br ightnessand contrast controls are adjusted such tha t the

Speciation and Characterization of Heavy Metal-Contaminated Soils 133

intensity of the carbon and copper tape are atestablished values thus assuring consistency in theanalysis among all samples. In addition, a brightnessthreshold value is selected to separate components ofpotential interest (bright particles) from the back-ground (less-bright particles and epoxy).

The epoxy and low atomic number occurrences suchas coal appear as a dark background. The mostcommon soil-forming components (e.g. quartz, feld-spars, calcite, dolomite, clay) are intermediate inaverage atomic number and appear as a mediumgray. Iron-bearing occurrences, not uncommon insoils, appear as a lighter gray, and heavy metaloccurrences are bright. In complete characterizationof all particles, the threshold value is chosen as justbrighter than the particle substrate. This is referred toas an "all particle" run. A particle of potential interestin the context of heavy metal characterization isdefined as one that is brighter than an appropriateintensity level chosen by the analyst. These brightparticles are candidates for a heavy metal occurrence.This is referred to as a "high-Z" analysis. If heavymetals are concentrated (e.g. anglesite (PbSO^ is 68%lead), the threshold level can be set quite higheliminating from consideration the vast majority ofthe particles. However, if heavy metals occur at lowlevels in a constituent (e.g. leaded iron oxide/hydroxidemay have lead levels at a few percent), the thresholdlevel must be set at a lower value resulting in moreheavy metal-free occurrences being evaluated by EDS.

The area for analysis of each sample is defined bydriving the stage to four edges of the circular puck andsetting their X and Y coordinates and the focus.Establishing the focus at these known locations allowsthe focus to be calculated for any position on thesample in the event that the sample surface is notexactly horizontal. The stage is also driven to the tworegistration marks whose locations are also saved. Theregistration marks define a new (local or sample)coordinate system to replace the stage coordinatesystem. The first mark defines the origin and is assignedthe X,Y coordinates of 0.0. The second mark definesthe alignment or orientation of the A'-axis which willpass through the identified point. Setting the origin andalignment to define a local coordinate system allowsparticles to be relocated even after the sample has beenremoved and replaced into the SEM days or years later.

Data acquisition. After the initial set-up, the SEManalysis begins under the control of the computer usinginstructions defined in the run parameter file. The stageis driven to the first field. The fields are selected eitherrandomly if the whole surface is not to be analyzed(high magnification defined by the ana lys t ) , or selectedto analyze all fields in order if the whole surface is to beanalyzed (low magnification defined by the analyst).The field is inspected for pixels above the thresholdbrightness level to iden t i fy high-Z particles. Eachparticle of interest is sized and the electron beam isplaced wi th in that particle ( the ana lys t can specifyeither the center of the par t ic le , or rastered in thelargest enclosed rectangle). I 'he l iberated X-rays arecollected lor a specified number of seconds using alight clement detector and parsed in to 2048 energychannels, each 10 cv wide. Each element produces

characteristic X-rays that are displayed as peaks at oneor more specific energy levels. As the spectrum is beingacquired, it is evaluated for elemental composition asdetermined by peaks in the EDS spectrum. Theevaluation is only for elements selected by the analystas indicated in the element file.

Due to the potential for overlapping peaks fromdifferent elements, the position and shape of the peaksare compared to spectra acquired on elemental stan-dards to identify the elements and determine thenumber of X-ray counts attributed to each element.(See section EDS Analysis and Peak Overlaps for adiscussion of issues related to overlapping peaks.) TheEDS peak area for each element is expressed as apercent of the total area of all elements analyzed. Ingeneral, the EDS peak area percent increases anddecreases as the element concentration increases ordecreases, but it should not be considered as a weightpercent. In the results presented herein, no correctionsfor atomic number, absorption, or fluorescence aremade. Classification as minerals (known stoichiometry)is done by comparison to spectra collected from knownstandards. Changes in non-stoichiometric componentscan be expressed as changes in the EDS peak area asopposed to weight percent.

Although the light element detector processesinformation for fluorine to carbon, the sensitivity forthese elements is less that that for the heavier elements.The lack of sensitivity makes the recognition andquantification of carbon and oxygen considerably lessaccurate. In addition, hydrogen is not recognized at all.These limitations make it difficult to separate oxidesfrom carbonates, and impossible to distinguish oxidesfrom hydroxides.

As the spectrum is being collected and the EDS areapercents calculated for the high-Z particles, eachoccurrence is assigned to a class defined by the analystin the rule file. At the minimum, this classification isused to separate particles of interest from particles thatare not of interest. In the present context, a brightzircon (ZrSiO4) particle would not be of interest, but abright lead-bearing particle would be of interest. If theparticle is of interest, the instructions can specify thatthe X-ray acquisition be extended (generally from 2 s to5-7 s), and that a tagged information file formal(TIFF) data object be saved. This data object containsthe image at a specified pixel resolution (minimum64 x 64 pixels, maximum 2048 x 2048 pixels), the 2048channel EDS spectrum, and additional informationrelated to instrument working conditions as well asparticle physical measures and elemental composition.In addition to the data stored in tags, data consisting ofthe physical measures and the elemental compositionand particle location are saved for all particle:, in tableform for later review and summary.

The classification rules can be simple such as "heavymetal-bearing" and "other", or can be more complex inorder to give a running total of the different com-ponents observed in the analysis. It should be notedtha t there is the opportunity to classify (and re-classify)particles af ter the analysis is complete so tha t the on-line classification may not be cr i t ica l .

The analyst continues \ \ i th the ncxl particle in thefield and addi t ional fields arc analyxetl mini a user-dclincd slopping cr i ter ion is met. The stopping cr i ter ia

134 Stephen K. Kennedy et al.

include completing the defined area or reaching apredetermined number of fields, total number ofparticles, number of a specified particle type, or elapsedtime. The stopping criteria used in these analyses was theentire I in. diameter mount for the low magnificationanalysis and 2 h for the higher magnification. Totalanalysis time was about 3 h. The analysis proceeds on thenext sample in the specimen chamber until all sampleshave been analyzed. Four one-inch diameter samples canbe placed in the standard sample chamber, but largerchambers are available.

Multiple magnifications. The fact that particles ofinterest in CCSEM analysis can occur in a wide varietyof sizes presents the following problem — How mucharea should be analyzed? As a general rule, the smallerthe particle, the more often particles of this type occur,but the less it contributes to the total volume or mass.After the analysis of a certain area, the number ofoccurrences for the smaller particles may be sufficient tostatistically characterize their distributions, but thesmall number of occurrences of the larger particlesmay not be sufficient to characterize their distributions.Each doubling of the area (which would also double theanalysis time) would provide additional informationneeded for the larger particles but would providestatistically redundant information for the smallerparticles. To provide the needed information moreefficiently, the analysis can be divided into two or threemagnifications — a large area could be covered at a lowmagnification for the larger particles and a smaller areacovered at a high magnification for the smaller particles.The area analyzed at each magnification is known andthe data for the smaller particles can be prorated ornormalized to the larger area.

In the case of multiple magnifications, there is a set ofstopping criteria for each magnification. In the currentcontext, the entire surface is inspected for particleslarger than 5 or 10 |im, and 500-1000 randomlyselected fields are inspected for particles down to1 urn in diameter.

Data review. In the simpler studies, the CCSEM datamay be totally satisfactory and not need much in theway of additional SEM review. When the particlemorphology is complex, when the observed elementshave severe peak overlap problems, or when it isimportant to understand the context in which the heavymetal occurrence is found, additional review may berequired. At the completion of the analysis, the data canbe reviewed on-line or off-line.

At the SEM, the stored image and spectrum can bereviewed, either by paging through all particles, orselecting only particles thai have been assigned tocertain heavy metal-bearing classes. Upon review ofthe stored data collected automatical ly, it may bedesirable to reanalyze a particle to obtain a longer EDSacquisition, to assess the composition of associatedcomponents, or to obtain an image that puts the heavymetal component in the context of it 's surroundingmatr ix . To accomplish this , pressing the space bar onthe PSEM keyboard v. i l l mo\e the stage to the locationof th;it occurrence where the a d d i t i o n a l analysis can beperformed. At the press of a but ton, a composite imagecan be saved to a file. The composite image consists of a

lower magnification view for context, a high magnifi-cation view illustrating the feature, and a labeled EDSspectrum. When in the review and relocation mode, athumbnail image of the original image obtained by theCCSEM analysis along with the identifying particlenumber, classification and average size is superimposedon the composite image as well. The information savedin the TIFF image is illustrated in Figure 1.

In addition to the on-line review, all tabulated dataand the images and spectra obtained in the CCSEManalysis can be reviewed using off-line programs. Thedata can be pre-sorted by class, element, size, etc. andthe image and spectrum reviewed, either on a computermonitor, or as printouts consisting of multiple (2-30)image/spectrum pairs per page. In the analysesreported herein, manual review ranged from less than1 h to about 2 h.

Remote review. In many cases, it is desirable to sharethe results of the analysis with other experts or clientswho are not located at the SEM facility. The images andspectra are in a digital format and can be rapidlydistributed over the Internet or on a CD for review withthe appropriate software. However, questions may stillarise that cannot be addressed adequately or quickly byphotomicrographs. In this situation, manual review ofthe sample can be performed remotely using theWEBSEM® f , a fully functional remote control of theSEM over the Internet. The site with the PSEMfunctions as a host server and several remote viewerscan observe the images and spectra as they are beingcollected. SEM control can be passed from the host toany of the remote sites. The remote function has thepotential to significantly improve project coordination

I igurc I Composite SRM TIK image obtained d u r i n g manua l reviewot C'C'SliM-acuuircd data . The review program shows the liDSspec

buttdialpan•li.n

eonspec

rum a>m olctci .1ele wa11 l.> h•\l Mill

H! microho 'mug'' l'< pirelocate*plici K':

mage acquired during the CCSTM run in the. in ihis case lor panicle ? 'X . w i t h .inand assigned !•

1 tor additional n1 in ilie right imai:

eating that thi.-. particle warum was acquired and replaces th

ihe I'r.-rk.nM,,: > la.mual re\ icw . The pa

1 lie ten imau'i. p i < > \liberated An exteiuk

veraue11-1-

icle isle- thei I-:DS

automated spectrum

Speciation and Characterization of Heavy Metal-Contaminated Soils 135

and reduce miscommunication, important factorsespecially in the initial stages when the nature andscope of a project are being defined, and in the laterstages when the conclusions are being formulated.

Data summary. The particle-by-particle data gener-ated by CCSEM can be summarized by classifying(reclassifying) the occurrences by elemental compo-sition and tabulating various size distributions. Classesare defined by the user based on experience with knowncompositions and by compositional associationsobserved in the sample. In the present context, the

Table I. Example summary report for lead-hearing soilsdistributions

Summary Distributions

Classes Number (%) Area (urn2) Area (%)

others 14PbPCa 1PbMnOPbTiPbPPb/SPb(Zn.Kc)PbSiPb(AlSi)PPbOPbKcOTotals 18

3 76.21! 13.60'. 0.15i 364> 1.78! 0.15

0.890.890.890.890.89

) 100.00

13463.61219.2211.7133.9116.197.972.745.437.224.818.1

15440.7

87.207.901.370.870.750.630.470.290.240.160.12

100.00

number percent and the area percent distributions bybins of average particle diameter are presented bycompositional class. An example of the results of a twomagnification run is given in Tables 1 3. Table 1 showsthat 10 lead-bearing classes were defined and theremaining high-Z particles are assigned to "others".Approximately 19% of the high-Z particles were lead-bearing. After normalizing for the different areasanalyzed at each magnification, the Number % aswell as Area and Area % values are calculated. Tables 2and 3 show the size distributions by number percentand area percent respectively. Instead of or in additionto average diameter, bins can be defined by othermeasures such as maximum diameter, aspect ratio orperimeter-to-area ratios.

Estimation of weight fraction

Bulk chemical analysis can obtain very accurate heavymetal assays that cannot be acquired on the SEM. TheSEM obtains information regarding the proportions ofthe various components that cannot be acquired bybulk chemical analyses. A combination of the tech-niques can be used to estimate the amount of heavymetal contributed in each compositional class.

Knowing the relative volume of each heavy metal-bearing class, the density of each class, and the weightpercent of the heavy metal in each class, the proportionof the total heavy metal contents in each class can becalculated. It has been shown that, for random sectionsof paniculate that is itself in random orientation, the

Tohle 2. Example summary report for lead-bearing soils si;e distribution h\ mi-rage diameter (microns/

Classes Number (%) Mean Std Dev 1.0-2 5 2.5-5.0 5.0-10.0 100-20.0 20.0-50.0 »>

OthersPbPCaPbMnOPbTiPbPPb/SPb(Zn.hc)PbSiPb(AlSi)PPbOPbHcOTotals

76.213.60.13.61.8O.I0.9090.90.90.9

100.0

2.72.4

10.51.72.27.42.72.11.61.61.32.6

2.81.44.30.71.00.10.00.00.00.00.02.5

66.759.00.0

98.050.00.00.0

10001000100.01000669

26.93940.000

5000 0

100.00.00.00.00.0

27.7

4.91.1

50.0200.0

10000.000000.0004.2

0.90.5

50.00.00.00.000000.000000.8

0.60.00.00.00.00.00.00000000.00.4

0.00.00.00.00.00.00.0000.00.00.000

Table i. Example summary report for lead-bearing soils Area Distribution by average diameter <microns I

Classes Number (%) Mean Std Dcv 1.0 - 2 5 : s 5 n 511 mn 100 :oo 20.0-500 >»

othersPhPCaPhMnOPhTiPhPPh SPN/n.l-e)I'hS.I'hi. \ISnPPhOPhi el )Tot.ils

87 27 41 4II 9II X0.6

0 5(1 1II 20.2>l 1

100 (I

15.44 T

1 3 4T 7

2 l'7.4-" 7

: ii '.1.61 4

14 .1

14.03 73 .11.9070.10.00.00(11(0ill)

1 3 6

12.722 S0.0

75.91 4 40.00.0

1000100/1100.0inn.ii

14 5

Id (i

4 9

15 S24.1(Ml

100000IIIIII (III I)

12.4130S4.2

0.00000onI I I I(I I I00IIII

I? 0

33.30.00.00.0o no o00oo0 (>onn ii

29.0

136 Stephen K. Kennedy et al.

area percent observed on the surface of a sample ofparticulate is equal to the volume percent in the sampleas a whole (Chayes. 1956). Having determined the totalassay, the amount of a heavy metal contributedfrom each compositional type can be estimated. Forexample, assume that a sample containing 813 mg/kg(mg/kg) As and 656 mg/kg Pb comprised three heavymetal classes in the relative proportions indicated by thevolume percent column in Table 4. The weight fractionof the total mineral that is the heavy metal is shown inWt. Fract. column. If the total volume of heavy metal-bearing occurrences was 100 cm1, the total weight ofthe each class is the product of the volume and thespecific gravity. The total weight of the heavy-metal ineach class is the product of the component weight andthe heavy metal weight fraction. The relative proportionof the heavy metal in each class can be applied to thetotal heavy metal assay to determine the contribution inmg/kg from each class. In this case, 661 mg/kg of thearsenic would be attributed to arsenolite (As?O3) and162 mg/kg to lead arsenate (Pb1As-,O8). Similarly,421 mg/kg of the lead would be attributed to leadarsenate and 235 to anglesite (PbSO4).

EDS analysis ami peak overlaps

X-rays of characteristic energy are generated by thejump of electrons to lower energy states after electronvacancies are produced by the interaction of beamelectrons and specimen electrons. Higher atomicnumber atoms produce a more complex spectrumbecause these atoms have a more complex electronicconfiguration. The resulting peaks in the spectrum arenamed for the shell into which the electron jumps (e.g.K. L or M). Although the complete X-ray spectra ofelements do not overlap with each other, it is possiblefor individual peaks generated by two or moreelements to be closely spaced or even largely overlap-ping. Unfortunately, this is the case with some of theheavy metals. Lead has an M peak that has a severeoverlap with the sulfur K (Figure 2(a)) and molyb-denum L peaks. The same is true to a lesser extent withmercury, thallium and bismuth. Lead can be differ-entiated from sulfur by the presence of smaller peaks at10.5 and 12.6 KeV for lead (Figure 2(b)). The majorpeak (L) for arsenic has a severe overlap with themagnesium K. peak (Figure 2(c)) and a secondary

Table 4. Distribution of arsenic and lead in three compositional class in u simiplc- \ \ i t h an arsenic concentration of 813 nig.kg and a leadconcentration of65fi mg.'kg

(wt. fract.)

Class

ArsenoliteLead ArscnalcAnglesite

Kormula

As,O,Pb,As,O,PbS04

As

0.7570.1670000

Pb

0.0000.6910.683

Spec. Grav.

3.8657.36.2

Vol (%)

5113020

(wt. ' IOOcm')x2

As

146436.500

Pb

0.0151.3

84.7

(ppm)x2

As

651162

0

Pb

0421235

c

D

l - i i j i i rc 2. EDS energy spectrum shotting c i j nca l regions for potential oterlaps of lead s u l f u r 1.1. h) and for As Me (c. d). Hjiiirc 2b indicate-, t h a ilead i* present. inil the >hv:h i u l l>c t to the l e f t in I i ^n rc 2a indicate:- t h a t s i i l l m i- p ic- i .n i a- v \ c l l I he cxacl o t c ik ip of peak;- in 1 .^-.ir . : ^indicates tha t magnesium is nut picsciil. In I isuirc 2d. the presence of arsenic i^ conhrmcJ h\ the peak al I I.? KeV. The presence ot\i peak at12.6 KeV and Ihe predominance of the 10 5 KeV peak over ihc 11 7 KeV peak indica tes t h a i lead is present as «ell

Speciation and Characterization of Heavy Metal-Contaminated Soils 137

arsenic peak (K) overlaps with a lead L peak(Figure 2(d)). However, the overlapping peaks arenot in exactly the same position nor are they identicalin shape. When the heavy metal is present inmoderately high concentrations (greater than about10%) the peak curve fitting program in the automatedanalysis does a good job of unmixing the spectra.When the heavy metal is present at lower levels, amanual review and the collection of an extended countspectrum and more detailed curve comparison may beneeded to confirm or deny its presence.

Case StudiesTwo case studies are presented in which the source ofheavy metals in contaminated residential soils areinvestigated. In each case, the offending heavy elementis alleged to have originated from a nearby industrialsite. In one case study, the soil on the property of thealleged source was sampled to serve as a referencewhich assisted in interpreting the results. Case study #1illustrates CCSEM where the primary component ofinterest is an arsenic oxide. Its physical characteristicsare relatively simple and manual review was notrequired to characterize the association. However,both lead and magnesium were present in the samplemaking automated identification of arsenic difficultwhen present in minor amounts. Manual review andextended EDS acquisition was performed on manyparticles. In case study #2, lead was the primarycomponent of interest. It occurred in a complexassociation with other high-Z occurrences and requiredsubstantial manual review. In addition, manual SEMobservations of components that were not actualcarriers of the lead were required for the interpretationof source.

Case study #/. Arsenic in residential soils

Background. Residential soil samples within aboutone mile of an industrial metallurgical site were foundto contain elevated arsenic concentrations. High levelsof soil arsenic ranged from 500 to 1000 mg/kg with oneoccurrence of over 6000 mg/kg. Due to the proximityto an industry that was known to have arsenic by-product on its site, it was considered possible thatthe industry was the source of the residential arsenicdistributed through the process of airborne dispersion.

If the elevated soil arsenic was due to contaminationby airborne dispersion from the industrial site, then itwould be expected that: (1) the spatial distribution ofheavy metals and metalloids (arsenic in this case)would be consistent with that dispersion model: (2) theresidential soils would be contaminated by the otherheavy metals also observed in the industrial site and insimilar ratios: (3) the specific arsenic occurrences wouldbe consistent with those observed at the industrial site,and (4) no other significant source could be identified.Therefore, in addition to the arsenic analysis, thesamples were analyzed for other elements includingcadmium and lead.

Relat ive to the first two points , review of the spatiald i s t r ibu t ion pat terns of the various elements, and therelative proportions of the various elements determinedby bulk chemistry techniques indicated that the results

were nol consistent with the suspected industry source.However, it could be argued that contamination couldhave occurred long enough ago that subsequentchemical reactions in the soils could adjust the relativeproportions.

CCSEM analysis. A CCSEM analysis of the soilparticulate to characterize the arsenic occurrences wasundertaken on a total of 29 samples for the purpose ofaddressing points 3 and 4 above. Eleven of the sampleswere residential soils containing various amounts ofarsenic (11 -6754 mg/kg). Fifteen samples were derivedfrom four location settings on the industrial site (IS)where arsenic content ranged from 100 to over8000 mg/kg. Three samples were used as backgroundsamples (on site but less than 19 mg/kg in arseniccontents). The samples were sieved at 250 urn andprepared as polished pucks then analyzed at twomagnifications for occurrences of heavy metals. Sixsamples were analyzed twice to assess reproducibility.Due to the EDS overlap of arsenic peaks withmagnesium and lead, all potential arsenic occurrenceswere manually relocated and reviewed.

In addition to occurrences containing only arsenicand oxygen, arsenic was found associated with one ormore elements. By far the majority of the occurrenceswere associated with calcium, antimony, aluminum,cadmium, and lead, but minor occurrences withsilicon, iron and zinc were also observed. The totalarsenic contribution from all occurrences of eachcompositional class was estimated for each sample.The density of each compositional type was assumed tobe that of the closest stoichiometric mineral and arseniccontent in each compositional type was estimated usinga semiquant program. Those types contributing morethan 5% to the total arsenic load are presented inTable 5. Images collected at 64 x 64 pixel resolutionduring the automated analysis representative of thecombined groups are shown in Figure 3.

Results of IS ami background samples. In the ISsetting I. 29% of the arsenic was associated withcalcium. 21% was associated with Sb, 17% wasassociated with Cd, and 13% was associated with Pb.In the IS setting 2, 79% of the arsenic was associatedwith Pb with small amounts associated with Ca and Cdand none associated with Sb. In the IS setting 3, themost common arsenic association was with Ca (25%).Second in abundance was arsenic associated w i t ha luminum (40%). About 16% was associated wi th Cdand 9% with Pb. In the IS setting 4, most of the arsenic

hlf / l i u ' m i p t ' i i i ' i t t itllrihittctl to iirH'nif-hi'iiring r / ( / v . > < p s in tht :

/VI /i J Indian uif Souri c v/fc.v (W(/f /H' Rt'.tittfnlitit Suits, i ()nt\ N / N M I *i i / ' fwmi/M/ < / * / v \ ( , inf i l l ing i'"i, tnlilt A\ tir ninrc. ,'

IS 2 IS .1 IS -4 Soils

As C:iAs AlA> ShAs CdAs I'hAs ()

2') 25411

:iI" 16111 7>) '1

-

7S '•S4

138 Stephen K. Kennedy et al.

Kigurc 3. Hive examples each of arsenic associated with calcium (lop row), antimony, a luminum, cadmium, and lead (bottom row). These low-resolution (64 x 64 pixels) images were acquired during the automated analysis. In ull cases, the arsenic was associated with the brightestregions

1 i g u i e 4 I mn examples from three Dimples of the dominant arsenic oxide occurrences in the anomalous soils. These low -resolution (64p : \ e K i I M U ^ C S \ \ere acquired d u r i n g the automated analysis .

. 64

Speciation and Characterization of Heavy Metal-Contaminated Soils 139

was associated with lead (78%), whereas most of theremaining arsenic occurred as very low concentrations.Very low levels of arsenic were observed in thebackground soils, but most of that was associatedwith lead.

Results of residential soils. In the anomalous soils,over 84% of the arsenic occurred as a quite pure arsenicoxide, but occasionally containing a trace of Cd. Mostof the remaining arsenic (9%) was associated with Pb.In addition, review of the images indicate that theoccurrences are generally solid discrete particles(Figure 4), in contrast to the majority of occurrencesin the other settings where it occurred more oftenclosely associated with other constituents in compoundparticles (compare the images in Figures 3 and 4). Inaddition, the size distribution of the arsenic particleswas rather limited. Over 70% of the particles werebetween 10 and 20 urn in average diameter and 93%were less than 30 urn.

Conclusions. The spatial distribution of arsenic andother heavy metals as well as the ratios of the variousheavy metals is not consistent with the hypothesis thatthe indust r ia l site was the source of the arsenic inresidential soils. The specific arsenic-bearing occur-rences in samples derived from the site are notconsistent with this hypothesis either. The majority ofthe arsenic in residential soils occurs as particles ofarsenic oxide with small amounts of cadmium. Theparticles occur in a narrow size range and chemicalpuri ty of the occurrences of arsenic in the residentialsoils suggest that they were derived from a manufac-tured rather than a fugitive source. One candidate is aproduct that has been used for crabgrass control. It is acombination of arsenic trioxide (25%) and leadarsenate (8.25%). The proportion of arsenic from thearsenic trioxide is 82.4%. This is in the same generalproportion as that observed in the anomalous soils. Itis concluded that it was the use (or misuse) of a productlike this that resulted in the elevated arsenic content ofthe residential soils in question.

1 i c u r c 5 Revie\\ images and spectra of leaded manganese (a and h) and leaded iu and d) i>ceinrcnees.

140 Stephen K. Kennedy et at.

Case stiuly #2 lead in residential soils

Background. Soils in the front, side and back yards ofresidential properties in the vicinity of an industrialoperation involving lead sulfate were found to containelevated levels of lead (up to 2000 mg/kg). Due to theproximity of the residences to the primary truck routefrom the industrial site, it was considered possible thatthe lead was derived from that site and distributed toproperties near roadways by truck traffic.

If the elevated lead were due to contamination bylead transported by trucks from the industrial site, thenit would be expected that: (1) the spatial distribution oflead would be consistent with that transport mechan-ism, (2) the lead would occur as an original sulfate ordisplay a re-precipitated morphology, and (3) no othersource could be identified. The spatial distribution oflead on the residential properties was consistent with atraffic source in that the lead concentrations werehigher in the front and side yards than the back yards.

The concentrations did not decrease with increasingdistance from the site, however, as would be expected ifthe lead source was from the industrial site raising thequestion of whether the lead was derived from analternate source.

SEM analysis. Eleven samples were prepared forSEM analysis for the purpose of addressing points 2and 3 above. Samples were sieved at 1000 urn andprepared as polished pucks for high-Z CCSEM analysisrelated to the heavy metal content. Due to themorphological complexity of the heavy metal-bearingcomponents, all occurrences were relocated andinspected manually. In addition, each sample wasexamined by manual SEM techniques to characterizeother anthropogenic constituents in the soil.

Results of the high-Z analysis. The results of the high-Z analysis indicated that there were four lead-bearingcompositional classes. The two dominant classes are a

tin id)6. Review images and spectra of spherical leaded phosphorus (a), enclosed leaded hante (h) . lead hromochloridc (c). and liberated leaded

Speciation and Characterization of Heavy Metal-Contaminated Soils 141

leaded manganese and a leaded iron. Iron was oftenobserved in the manganese occurrences, and manga-nese was often observed in the iron occurrences andthese are thought to be end members of a continuousmixture.

Over 97% of the leaded manganese occurrencescontained silicon and/or aluminum in proportionsranging from very small peaks, to the dominantpeaks. Calcium and potassium were commonlyobserved as well. The rare occurrence contained atrace of t i tanium. Most of these occurrences were notsingle-component, discrete particles. Instead, theyoccurred as an apparent cement or an otherwisecomplex association. As a result of this morphologyor mode of occurrence, it is generally not possible to saywith confidence that the aluminum and silicon are apart of the leaded manganese or was derived from anadjacent component. However, the occurrences thatwere larger than the beam interaction volume showedaluminum and/or silicon and it is assumed that this istrue for the smaller occurrences. Many of the leaded

iron occurrences showed similar aluminum and siliconpeaks. However, copper, antimony, t i tanium, or zincwere occasionally present also. Representative imagesand spectra acquired during manual review of theleaded manganese and iron are shown in Figure 5.

The leaded phosphorus occurrences were commonlyassociated with silicon and/or aluminum as well. Inaddition, approximately one quarter of the leadedphosphorus occurrences were spherical. Iron, and to alesser extent zinc, were found with the phosphorus aswell (Figure 6(a)).

All samples contained a barium and sulfur (probablebarite - PbSO4) component and 8 of the 11 samplescontained a leaded variety of barite (Figure 6(b)). Rareoccurrences included one lead bromochloride (Fig-ure 6(c)). one leaded glass, one probable paint and twoleaded tin (Figure 6(d)), presumably solder.

Other than a single occurrence, lead sulfate was notobserved. The one occurrence was a 5 urn inclusion ina t i tanium particle that contained spherical holes.

re 7 Images and specira i l l u s t r a t i n g composit ion and morphology of the phase interpreted as hotioMi ash

142 Stephen K. Kennedy et al.

Results of the manual SEM analysis. Manual analysisof the samples revealed two unusual components inaddition to the normal soil constituents of quartz,feldspars, calcium and magnesium carbonates, andclay. A component containing a major peak of siliconand aluminum at half to equal the silicon content, withtraces of potassium, t i t an ium, iron and. less commonly,calcium was also observed. These particles commonlycontained vesicles and the particles themselves wereoccasionally spherical. This particle type is interpretedto be bottom ash resulting from coal combustion.Representative occurrences are shown in Figure 7.

A second unusual occurrence was carbon-rich con-taining a trace of sulfur and was interpreted as coal(Figure 8(a)). Rims or internal layers of a lead-bearingcomposition were common (Figure 8(b)). The coalcommonly had inclusions, stringers or rims of thealumino-silicate composition, some of which havingspherical vesicles (Figure 8(c) and (cl)). The close asso-

ciation of the coal and the aluminosilicate componentsuggests that these are remnants of coal that wereheated but remained unburned.

Discussion and conclusions. The occurrence of ash andpartially burned coal suggests an alternate potentialsource of lead other than the industrial site. Thecomposition of representative coals and coal combus-tion products is reported in DOE (1996). These coalscommonly have lead concentrations on the order of10 mg/kg. This lead is not generally volatilized uponcombustion of the coal and would be concentrated inthe ash. Manganese and barium may be found in coal inhigher concentrations (measured in 10s of mg/kg).Phosphorus may be found on the order of 0.1 % andsulfur and iron greater than 1%. The oxides of siliconand aluminum may total in the range of 10%. Thecombustion of coal, then, could supply all the elementsfound associated with lead (leaded manganese and iron.

Heurc 8. Images and spectra i l l u s t r a t i n g the composition of coal particle (a) w i t h a lead-hearing rim (h) and associated . i luminosil icatc phase (cl.The aluminosil icate phase also oecur> enclosed in coal (d) .

Speciation and Characterization of Heavy Metal-Contaminated Soils 143

leaded phosphorus and leaded barite). In addition thisprocess would also provide a source for the sphericalnature of the leaded phosphorus.

Small amounts of bromomethane and chloro-methane can be released from the combustion ofsome coals and this may be the origin of the leadbromochloride occurrence. The only occurrence of alead sulfur component was observed as an enclosedparticle in an ash fragment.

If the lead is derived from the coal burning process, amechanism is also needed to transport the lead to theresidential soil. The residential properties in this studyare old and were historically heated by coal. It isknown that some communities would collect the coalash to use as traction material in the winter, and thatsome individual home owners would spread the ash onsidewalks and driveways for traction or on gardens andlawns as a fertilizer or conditioner. We suggest this asthe origin for the majority of the lead in the residentialsoil of this study. We note that some of the lead in thesamples was not described by CCSEM; the low levelsin coal were not sufficiently bright to be detected, andhad to be found in manual review. The associationwith the other non-leaded components was the key tounderstanding this origin.

Discussion and ConclusionsThe description of heavy metal-bearing species, size,and association was shown to be critical in thedetermination or confirmation of source of the heavymetal contamination in the two settings illustrated.This information is only available using e-beaminstruments — the scanning electron microscope orelectron microprobe. The manual analysis of polishedsamples is time consuming, especially considering thatconcentrations may occur at levels down to several tensof mg/kg. However, automated analysis considerablyreduces analysis time, even if considerable manualreview is required, and should be considered as anancillary tool in the analysis of heavy metal-contami-nated soils and sediment.

ReferencesCasuceio. G.S., Janocko, P.B., Lee. R.J.. Kclley, J.F., Daltncr, S.L.

and MgcbrorT. J.S. 1983. The use of computer controlled scanningelectron microscopy in environmental studies. APCA J. 33,937-943.

Casuceio, G S., Schwocblc, A.J., Henderson, B.C., Lee, R.J., Hopke,PK and Svcrdrup, G.M. 1989. The use of CCSEM and

microimaging to study source/receptor relationships. In: Trans-ai tionx Refi-pior Models in Air Resources Management, pp. 39-58.(Watson. J.G.. Ed.). Air & Waste Management Association.Pittsburgh, PA .

Chaycs. F 1956 In: Petrographic Modal Analysis. New York, JohnWiley and Sons. 113p.

Davis. A . Drcxlcr. J.W., Ruby. M.V. and Nicholson, A. 1993Micrommcralogy of mine wastes in relation to lead bioavailability,Buttc, Monlana. Environ. Sci. Technol. 27, 1415-1425.

DOE (Department of Energy). 1996. A comprehensive assessment oftoxic emissions from coal-fired power plants. Pittsburgh EnergyTechnology Center. September 1996.

Eighmy. T.T., Crannell. B.S., Butler, L.G . Cartlcdgc. F.K., Emery.E.F.. Oblas, D . Krzanowski, J.E., Eusdcn, JrJ D . Shaw, E.L andFrancis. C A. 1997. Heavy metal stabilization in municipal solidwaste combustion dry scrubber residue using soluble phosphate.Environ Sci. Technol. 31. 3330-3338.

Gasscr. U G . Walker. W.J., Dahlgren, R.A . Borch. R S. and Burau,R G 1996 Lead release from smeller and mine waste impactedmaterials under simulated gastric conditions and relation tospcciation Environ. Sci. Technol. 30, 761-769.

Henderson, B.C , Stewart, l.M. and Casuceio, G.S 1989. Rapidacquisition storage of electron microscope images. Am. Lah.November 49-52.

Krinscly. D H , Pyc, K , Bogs. Jr., S. and Tovcy, N.K. 1998. Chapter9 Image analysis. In: Backsca/tered Scanning ElectronMicroscopy and Image Analysis of Sediments and SedimentaryRocks, pp. 145-172. Cambridge University Press

Lapcrchc. V, Traina, S.J., Gaddam, P. and Logan. T.J. 1996.Chemical and mineralogical characterizations of Pb in a contami-nated soil reactions with synthetic apatite. Environ. Sci. Technol.30. 3321-3326

Link. TE , Ruby, M . V , Davis, A. and Nicholson, A.D. 1994. Soillead mineralogy by microprobe: an interlaboratory comparison.Environ Sci. Technol. 28, 985-988.

Ma, Q.Y , Logan. T J. and Traina, S.J. 1995. Lead immobilizationfrom aqueous solutions using phosphate rocks Environ. Sci.Technol. 29, 1118-1126.

Que Hec, S.S., Peace, B., Clark, C.S., Boyle, J R . , Bornshein, R.L.and Hammond. P.B. 1985. Evolution of efficient methods tosample load sources, such as house dust and hand dust, in thehomes of children. Environ. Rex. 38, 77-95.

Ruby. M V.. Schoof, R., Brattin. W.. Goldade, M.. Post, G.,Harnois. M., Mosby, D.E., Castccl, S.W., Berli, W. Carpenter,M , Edwards, D , Cragin, D. and Chappcll, W. 1999. Advances inevaluating the oral bioavailability of inorganics in soil for use inhuman health risk assessment. Environ. Sci. Technol. 33,3697-3705

Schamhcr, F.H. 1993. SEM for nontraditional users. Am. Lah..Junc34-38.

Schwocblc. A J., Dallcy, A.M., Henderson, B.C. and Casuceio. G.S.1988. Computer-controlled SEM and microimaging of finepanicles J. Metals. Aug. 11-14.

Schowcble. A J., Lcntz. HP. , Mcrshon. W J. and Casuceio, G.S1990. Microimaging and off-line microscopy of fine particles ininclusions. Mater. Sci. Eng A124. 49-54.

Song, X -H . Hadjiiski. L., Hopkc. P.K., Ashbaugh, L.L . Carvacho,O.. Casuceio, G.S. and Schlacglc. S 1999 Source apportionmentof soil samples by the combination of two neural networks basedon computer-controlled scanning electron microscopy. J. Air &Waste Mfiti t . Assoc 49, 772-783.