EPA/600/ J-89/0?1
Atmotjkmc Vol 23. No. 1 pp. 1*7 471 IW
Pnaud m Grwt Bntaia
USE OF ELECTRON MICROSCOPY DATA IN RECEPTOR
MODELS FOR PM-10
T. G. Dzl ray
Atmospheric Sciences Research Laboratory, US. Environmental Protection Agency, Research Triangle
Park, NC 27711. USA.
and
Y. Mamane
Environmental Engineering, Technion, Haifa 32000, Israel
(first received 19 January 1988 ami in final form 6 July 1988)
Ahmet—Coarse particles (15-10 inn) were collected in dichotomous samplers and analyzed by scanning
electron microscopy. The resulting data for coal-fly ash and botanical matter were included with x-ray
fluorescence and neutron activation data in a chemical mass balance receptor model for fine and coarse
particles, and the mass concentration of particles with diameters < 10/im was apportioned into 10
components Mineral concentrations deduced by electron microscopy were in good agreement with soil
concentrations deduced from x-ray fluorescence data. A comparison of results 'or an incinerator component
indicated possible biases in results.
Key word index-. Coal fly ash, coarse particles, electron microscopy, fine particles, fly ash, incinerator,
minerals, PM-10, receptor models, spores.
INTRODUCTION
Receptor models, along with dispersion models, may
be used in polluted areas of the U.S. to develop plans
to meet the new national air quality standard for
particles with diameters <10/mi (PM-iO) (EPA,
1987). A receptor method that is a composite of
multiple linear regression and chemical mass balance
(CMB) models was applied to aerosol data from
Philadelphia, PA and Camden, NJ, and PM-10 was
resolved into several components (Dzubay ft al., 1988;
Dzubay, I98G). One component was crustal matter,
which included unresolved amounts of soil and coal-
fly ash.
A scanning electron microscope (SEM) with an
energy dispersive x-ray detector (EDX)can distinguish
individual coal-fly ash particles from soil minerals by
morphology. In studies conducted in Philadelphia
over 95*/. of coal-fly ash particles with diameters
< 10/mi appeared as smooth spheres whereas soil
minerals were irregular or crystal shaped (Mamane el
al.. 1986; Mamane and DzuU.y, 1988). Furthermore,
mass concentration for each particle type can be
deduced by estimating particle volume from SEM
data and assuming a value for density. Here we
expk"* 'lie feasibility of increasing 'he resolution of
theCM'» method by using SEM-EDX dati with A-ray
fluorescence (XP.F", data.
In .he Philadelphia study aeroso' samples were
collected or Teflor. f 'ters in di;hclnmr>tj samplers M
three sites fiowr in Fig. I and aiai; zul lor b ill
elemental composition (Dzubay et al., 1988). Addit-
ional samples were collected simultaneously on Nu-
clepore filters in dichotomous samplers at two of the
sites. Here we describe SEM-EDX analyses of coarse
particles (15-10 /im aerodynamic diameter) and pre-
sent estimates of mass concentrations of minerals,
coal-fly ash, botanical matter and incinerator . -lis-
sions. We show how the SEM-EDX data for botan.eal
matter and coal-fly ash can be included in CMB
calculations. We compare SEM-EDX and CMB re-
sults for incinerator emissions and discuss possible
biases for both methods.
Small spacing among fine particles (<15/im) col-
lected in dichotomous samplers prevented them from
being analyzed reliably by SEM. Elsewhere we de-
scribe a modified dichotomous sampler that collects
both fine and coarse particles suitable for analysis by
XRF and SEM (Mamane and Dzubay, 1989).
EXPERIMENTAL PROCEDURE
Aerosol sampies were collected on 0.3-/im pore size Nu-
e'epor* filters in Beckman automated dichotomous samplers
al sites 12 and 28|See Fi*. 11 between 14 July and 13 August
1982. Each sampler had a PM-10 inlet that was a prototype
of Siena Andersen mod J S.\ 246 inlet. The flow rate through
each inlet was 16.7 /min~'. Filters were changed twice daily
at 0600 and 1800 EDT. Additional samples were collected
simultaneously on Teflon Piters ,n collocated dichotomous
sa.nol'i Coirsc fraction Teflon t-vt Nuclepore filters were
coat- i vk)i "'S (igcm " * of r-.incral oil before ¦ jn >ling (ap'l
Wert tare weight determinations) to improve adhesion of
«£7
-------
468 T. G Dzumy and Y Mamane
w#i
PA TURNPIKE
STEEL
IRON
POLYMER
INCINERATOR
PHILADELPHIA
OPPER SMELTER/
GRAIN HANDLING
mineralU
CAMDEN
GYPSUM
REFINERY
GRAIN HANDLING
INTERNAT
HflASSy* TITANiUM ZINC
-ANTIMONY ^
A TYPE SHOWN ON MAP
CATALYST CRACKER
RESIDUAL OIL FtREO BOILER'
O COAL FIRED BOILER
O AIR SAMPLING SITE
..•••"FIBERGLASS
Fig. 1 Map of Philadelphia environ! showing sampling sites and emission sources.
large particles. All filters were analyzed for element and mass
concentration! by XRF and beta gauge, and fine Teflon
filters were analyzed Tor element concentrations by instru-
mental neutron activatic-i analysis (1NAA) (Dzubay ft of,
1988).
Individual coarse particles on Nudepore filters were anal-
yzed by SEM-EDX for 12 sampling periods. Periods were
not selected randomly but were selected to provide a variety
of cases of high and low concentrations of incinerator
particle* A 4-mm by 4-mm piece of each Nudepore filter was
coated under vacuum with a thin carbon laye> and analyzed
m an Amray Model 1000 SEM Analysis conditions were 30-
keV electron energy, 100-jiA beam current. 26' sample tilt
and 12-rnin working distance. For each sample, all part ides
larger than 13 pm were analyzed manually in five to ten
randomly chosen fields. About 200-300 particles were anal-
ysed per sample For each particle two perpendicular dia-
meters were recorded, morphological characteristics were
noted and an x-ray spectrum was obtained for 30-to 60-s
counting time
RESULTS AND DISCUSSION
Scanning electron microscopy result*.
Cnlena shown in Table I were used to classify
coarse particles into the following categories' mineral,
coal-fly ash, botanical and inctneralor emissions To
Table I Criteria for classifying panides into categories
Density
(gem"')
Category
Morphology
Major i-ray peaks
Mineral
2.6 ±0 3
Nonsphencal
High peak to background ratio for Si or Ca and
variable anountf of Al, K, Ti and Fe
Coal-fly ash
2.6 ±03
Smooth sphere
Same as mineral
Botanical
1 1 ±02*
Characteristic of
spores or pollent
Low peak to background lalio for variable
amounts of P, S. K, and Ca
Incinerator
2.6 ±0 31
Non-specific
Spectrum lor mineral elemcnts§ or carbon} with
CI. K and Zn
Sea salt
2.2*02
Non-specific
Mainly CI with detectable Na
* Density of spores from Sussman and Halvorson (l°66(.
t Morphologies fur many spores and pollens are iHi>slra.ed by Kapp 11969)
t For Ihe few carbonaceous incinerator particles, a ilen .ily of 1 g cm ~' was assumed
{X-ray spectra (or incinerator particles are shown by Mamane (198%).
-------
Um at electron mcroaoopy data it receptor models
Fig. 2. Micrographs of coar* particles representing (a) resuspended toil from downtown
Philadelphia, (b) axl-ffy uh emiuioni and (c) airborne ipt e at lite 28.
-------
470
T. G. DzutAY and Y. Mamane
illustrate the use of those criteria. Fig. 2 shows micro-
graph* of coarse particle samples of (a) surface dust
from downtown Philadelphia that was resuspended
by an aerosol generator (Batterman el at„ 1988), (b)
coal-fly ash emissions from a coal fired power plant in
Philadelphia (CMmcz el 1989) and (c) a spore from
ambient air at Site 28; Fig. 3 shows typical x-ray
spectra. Both morphology and x-ray spectra are nee-
ded to distinguish among such particles. X-ray spectra
can distinguish between carbonaceous and cnistal
particles (Fig. 3). Morphology can distinguish be-
tween fly ash and minerals within the cnistal category
and between botanical particles and soot within a
carbon category. Figure 4 shows a micrograph with
several coarse particle types indicated. Although oil-
fired power plants in Philadelphia emitted spheres
with major x-ray peaks similar to those of coal-fly ash,
most oil-fly ash particles were emitted in the fine
fraction (Olmez el al., 1989). By using V and Ni as
indicators , we detected only a negligible amount of
oil-fly ash in ambient coarse particles.
Typical particle-count data are shown by size and
category in Table 2, and component mass concentra-
tions are shown in Table* 2 and 3. Mass concentre-
Fig. 3. X-ray spectra of (i) (oil mineral, fb) coal-fly ash p~r.iclc, and (c) spore.
-------
Ute of dec: ran microscopy data in receptor models
471
B r
\ I
MM B
\ I
i I
\
B
i -t »
i " - it '
• . v. v .£:i *¦ i
. * c ''%¦• c ' *
•'• „ JMj ®j»' • ;
'' -i v' y i f ?%¦
7—r
B
M
Fig 4. Micrograph of atmospheric coarse pa/tides collected at Site 28. Particle types are
botanical (B), mineral (M) and coal-fly uh (F>
Table 2. Coaise particle count and estimated mass concentration by category at Camden. NJ. Tor 12-h periods
beginning at 0600 and 1800 on 21 July 1982
Diameter range* Mass
Category £l.S $11 $3.1 $5.1 $7.1 $10 Oigm"')
Mineral
1
15
26
29
10
6
7.7 ± 3.1
Cl-rich mineral
1
0.01 ± 005
Coal-fly ash
1
1
0.06 ±0.06
Botanical
II
23
33
6
2
2.3 ± 0.9
Indnerator
1
3
1
0.1 ±01
Other
4
14
6
6
0.8 ± 0.3
Mineral
,
36
38
30
8
1
8.5 ± 3.4
Cl-rich mineral
2
2
6
1
1
2.1 ±08
Coal-fly ash
6
17
7
4
0.7 ± 0.3
Botanical .
4
II
9
4
1
2.0 ± 08
Indnerator
3
13
10
10
1
1.8 ±0.7
Zn-rich
1
1
0.41 ±0.2
Other
6
6
to
8
09 ± 0.9
* For the sequence of size ranges shown, the effective diameters were < 1.3. 1.8, 2.SS, 4.05, 6.3 and 8.8 pm.
1200 particles were analyzed in a scanned area of 47,300 pm1.
J 258 particles were analyzed in a scanned area of 90,000 jim'.
tions were estimated Tor each das* of particles from
the following data: effective diameter for each size
range, assumed density, fraction of filter area scanned,
and air volume sampled. Accuracy estimates were
based on uncertainties .1 density shown in Table 1,
40% uncertainty in estimating particle volume and
counting statistics.
Minerals rich in CI .were detected orly for the
sampling period that began at 1800 on 21 Jul;- when
the indnerator component had its largest value. We
assume that the Cl-rich minerals were associated with
indneratur emissions but did not know if they were
emitted by inditerators or if they were the produd of
reaction* between $as phase CI from indnerators and
minerals from other sources. Therefore, the Cl-rich
minerals component shown in "table 2 was divided
betweer. the incinerator .ind mineral components as a
footnote in Taolf: specifies.
-------
472
T G Dzubay and Y Mamane
Table 3 Mats concentrations in jig m "1 estimated by SEM-EDX for 12-h coarse particle tamplcs at
Sites 12 and 28 during 1982
Sample Mart
Date Time Botanical Coal-fly ash Mineral Incinerator
Site 12
3 August
6 August
6 August
II August
11 August
Site 28
21 July
21 July
3 August
6 August
6 August
II August
11 August
Unweighted mean
Standard deviation
1800
41 ± 16
000 ±002
38 ± 1 5
•
0600
13 ±0.3
016 ±006
38 ± 1 3
•
1800
09 ± 04
010 ±004
2.2 ±09
00±
0600
1 5 ± 06
ftl8±007
1 6 ± 06
00±
1800
2.1 ±08
008 ±003
04 ± 0 1
00±
0600
2.3 ±0.9
006 ±004
77 ± 3 1
0I±
1800
10 ±08
0 74 ± 0 30
9 6 ± 3 6t
2.9 ±
1800
40 ± 16
004 ±002
53 ± 2.1
•
0600
10 ± 04
0 32 ±013
38 ± 2.1
•
1800
04 ± 0J
000 ±002
2.9 ± 1 2
0.0 ±
0600
16 ±06
0 13 ±006
33 ± 1 3
06 ±
1800
03 ±0.2
008 ±0.03
1 1 ±04
00 ±
1 8 ± 07
0 16 ±006
40± 16
03 ±
12
019
2.8
06
'Not analyzed by SEM-EDX
t Includes I 1 ± I I jigm"1 of the Cl-nch minerals shown in Table 1
Table 4 Chemical mass balance Tor 12-h sample beginning at 0600 on 21 July 1982 at Site 28 when wind direction was 0= 13°
±22° Data are in ngm~'
Calculated Obs Soil Flyash Botanical SOJ Incinerator Vehicles Sb roast
Al*
315 ± 134
319
249
11
0
0
52
2
0
Si*
1070 ± 344
1082
968
II
0
0
88
3
0
S*
89 ± 40
89
13
0
7
32
33
4
0
a
148 ±32
34
1
0
0
0
144
3
0
K*
125 ± 22
123
38
1
19
0
46
1
0
Ca
177 ±42
347
63
1
6
0
103
3
0
Ti
51 ± 12
34
39
1
0
0
10
0
0
Cr
5 ± 3
6
4
0
0
0
1
0
0
Mn
4 ± 2
7
3
0
0
0
1
0
0
Fe
263 ±42
341
239
5
1
0
17
3
0
Ni
J±1
2
1
0
0
0
1
0
0
Zn*
47 ±38
49
2
0
0
0
44
I
0
Br
8 ± 3
8
0
0
0
0
1
6
0
Cd*
l±2
1
0
0
0
0
1
0
0
Sb*
-1 ± 3
-1
0
0
0
0
0
0
1
Pb*
39 ±9
39
2
0
0
0
17
19
0
Bo< an*
2254 ± 800
2270
0
0
2254
0
0
0
0
F-ash*
60 ±30
60
0
60
0
0
0
0
0
Mass
7434 ±2866
9445
3712
60
2254
95
1026
288
1
4mass
1336
2768
1164
30
790
127
371
197
4
* Species included in least squares fit
Chemical mass balance results
SEM-EDX results for fly ash and botanical particles
were included with XRF data in CMB calculations in
which the concentration C, of species i is
= (1)
where AtJ is Ihe abundance of species i in component
and M, is the total mass concen.. ation for component
}. (Jung data on A,j, unknown values Mt were found
by a least-squares method that minimizes the ex-
pression,
if-lSC.-Of/E} (2)
where O, is the observed concentration of species i, and
Ef is the effective variance, which includes uncertain-
ties in 0, and A,t (Watson etal, 1984) For a species i to
be included in Equation (2), two criteria must be met
(a) all sources of the species must be included in
Equation (I), and
lb) the jpecie^' abundance At) must be k nown for all of
those sources
Tables 4 and 3 indicate the species and sources
included in our CMB as well as typical CMB results
The soil signature represents an average for the Phila-
delphia area (Batrerman el al. 1988) Signatures for
-------
U«e o> electron necroscopy data in receptor models
473
Table J. Chnucal mui botutee for 12-h (ample beginning al 1800 on 21 July 1982 at Site 28 when wind direction was
0»3I8''±2O°. Data are in ngm"1
Calculated
Obs
Soil
Flyaih
Botanical
soj-
Inane rator
Vehicles
Sb roast
Al*
70U±262
747
390
142
0
0
162
7
0
&•
1939 ± 581
1903
1515
138
0
0
273
13
0
S*
178 ± 79
178
20
4
6
28
102
17
0
a
459 ±106
244
2
I
0
0
445
11
0
K*
263 ±56
260
90
10
17
0
143
4
0
Ca
453 ±83
485
99
17
5
318
14
0
Ti
106 ±17
67
62
10
0
0
32
2
0
Cr
9±4
9
6
0
0
0
3
0
0
Mn
9±3
16
4
0
0
0
3
2
0
Fe
502 ±66
JI2
373
60
0
0
54
14
0
Ni
6±2
2
2
0
0
0
3
0
0
Zn*
143 ±119
154
3
0
0
0
137
2
0
Br
29 ±13
25
0
0
0
0
4
25
0
Cd'
3±3
3
0
0
0
0
3
0
0
Sb*
7±4
7
0
0
0
0
1
0
6
Pb*
134 ± 33
134
3
0
0
0
54
76
0
Bo tan*
194} ±700
1950
0
0
1945
0
0
0
0
F-aih*
746 ±250
740
0
746
0
0
0
0
0
Mass
12,911 ±2842
14,716
5808
746
194!
85
3179
1139
8
Anass
2707
2768
2148
248
699
257
1318
598
5
'Species included in least squares fit
stationary source emissions wrre from Olmez et al
(1989). The composition of pollen (Stanley and Lin-
skens, 1974) was assumed to represent botanical par-
tides. In Tables 4 and 5, the 'calculated' uncertainties
are Eb and imass is the uncertainty in mass derived in
the least squares analysis. (Any investigator may ob-
tain the source signatures, ambient data and CMB
code on a 360 kbyte IBM-PC readable diskette by
sending such a diskette to the principal author.)
To determine if SEM data were essential for deter-
mining botanical matter and coal-fly ash, CMB ana-
lyses were made using only XRF data. Uncertainties
increa**] by factors of 9 for botanical matter and 20
for coal-fly ash, and those components were undetect-
able.
CMB computations were madr for each of 12
samples using the sources and fitting species listed in
Table 4. FigureS compares CMB results for soil with
SEM-EDX results foi minerals The former are based
mainly on XRF data for Al and Si whereas the latter
are based only on SEM-EDX data. There is agreement
within the estimated uncertainties, which suggests that
both methods are equivalent for soil and minerals.
Thus, there was no need to include SEM data for
minerals
-------
474
T G Dzubay and Y Mamane
10
1 1
• SOIL ICMB)
O MINERAL (SEM EDXI
i r
n 2
E
5 0
$ «
4
6
1
_L
_L
_L
SITE 28
h
7/21
0600
7/21
1600
8/5
1800
8/6 8/6
0600 1800
DATE AND TIME
8/11
0600
8/11
1800
Fig. 5
Companion of SEMEDX results for mine rati and CMB result) for toil in the coarse
fraction. CMB computation were made u in Table* 4 and 5
Table 6. Average chemical mass balance results in jigm"' based on XRF, INAA, and SEM-EDX analyses
of the 12 samples listed in Table 3*t
Fine
Coarse
PMIO
Sulfate and related ions
II1±I2
03±03
II4±I2
Motor vehicle exhaust
2.4 ±07
0S±02
29±07
Soil
06101
J8±I0
44± 10
Coal-fly ash
:
0 16 ±006
02 ±01
Oil-fired power plants
01±0I
t
04 ± 0 1
Ti-nch paini pigment
0061004
t
01 ±01
Fluidued ealalyiic cracker
0 05 1 0 05
:
01 ±01
Municipal inciiKiaiortS
M10I
08 ±04
2 2 ± 0 4
Antimony router
0001001
001 ±001
001 ±001
Botanical matter (sports)
1
1 7±06
1 7 ±06
Calculated toun
1601 1 }
7 3 1 i 4
23 3 ± 2 1
Measuied total
1521 ) 1
67± 28
21 9± 4 2
* Became the sample* were nul randomly choien. the averages do not represent a particular season
tXRF and INAA were applied to line particlei, XRF and SEM EDX were applied to coarse particles
t Not determined
}As explained n the lent, reiutis shown here represent an upper limit for incinerator emimnni
iimilnr to those described previously (Dzuba, et at, bias described above, our results for incinerafcn are
1988) except th?t custrl i.ialier was ^presented by a upper limits
sig.^ture for soil rather .ban oust At i consequence of Sums for the fine and coarse fractions are presented
-------
IJsc of electron microscopy data in receptor models
475
• CMS
O SEM COX
SITE 12
«n o
E
2
8«
<
cc
¦5-
SITE 28
1} i
l_S.
7/12
0600
7/12
1800
-*5-
-5—1
a/6
1800
DATE AND TIME
8/11
0600
8/11
1800
Fig. 6 Companion of CM Band SEM-EDX results for incinerator emissions in the coarse
fraction. CMB computations were made as in Tables 4 and 5
as PM-10 results in Table 6 Although coal-fly ash and
botanical matter were determined only for the coarse
fraction, the fine fractions of those components is
expected to be relatively small (Olmez el al. 1989.
Kapp, 1969), and little mass was missed by not
applying SEM to fine particles Likewise, emissions
from oil-fired power plants and fluitfi/td crackers were
determined only in the fine fraction, bul coarse par-
ticle emissions from such sources were relatively small
at the source (Olmez et al, 1989) Because samples
were not chosen randomly, the average PM-10 >r ss
concentration shown in Table 6 docs not represent a
particular season and is. in fact, 40% lower than the
average for all samples collected between 14 July and
13 August 1982
Additional CM tt analyses were made with a coarse
particle marine component included On 6 August,
when winds were from the east, the magnitude of this
component was I 6 and 3 5 pg m ~ 1 at Sites 12 and 28.
respectively The average for all sites and samples was
03±02;/gm"J This component was determined
mainly by CI. a volatili/able element, that caused
overall accuracy to be poor (D/ubay, 1988)
CONC LLS.UNS
By supplementing *-rav fluorescence data with
microscopy data in our mass balance we weie able to
determine 'wo additional component" coal-r.y ash
and botanical Tinner Roiannal matter •~->n,rihi>t^d
-------
4T6
T C DzutAY and Y Mamane
B'/.toPM-IO, and coal-fly ashcontributed nearly 1%
White coal-fly ash was a smalt pari or PM-IO at sites
we studied, knowledge of that contribution can be
valuable in developing ccwt-cfTectivr strategies for
meeting air qual it_v standards at any location Coal-fly
ash and botanical truttcr could not be determined l>y
XRF data alone, and SEM data were essentrjl
For an event of known hig't impact from municipal
incinerators, there was good agreement between re
suits obtained by CMB and SEM-EDX. but at other
times there were significant discrepancies Loss or CI
dunng transport, would cause SEM-EDX to have
negative bias because that method depended on CI
along with Zn to identify incinerator emissions On
the other hand, the CMB method applied tu our data
could have positive bias because Z-i from source,
other than incinerators #as treated jy our model as if
it were from incinerators
Samples in our study were nol selected randomly
but were selected 'ogive insight into the application or
SEM and CMB Thus, the averages do nut represent a
particular season Ideally in future studies all samples
would be analyzed by alt methods, but the much
higher cost of electron microscopy analysis suggests
that only a subset of samples can be aiu'.icd by that
method Random selection of samples would enable
the subset to be representative of ihe entire popu-
lation Use of automated deviocs for control o( the
scanning electron microscope (Johnson. 1983, Casuc-
cioel at, 1983)could improve the speed and precision
of analyses that we performed manually
Achmnvlr/fgemtmi We lhark Robert Slevens and John
Miller of US EPA, David Buiney and Don Stone of Duke
University and John Cooper of NEA, tnc Tor valuable
discussions and assistance Although research described in
this article has been conducted at U S EPA, it has not been
subjected to Agency review and therefore does nol necessari-
ly reflect the views of the Agency, and no official endorsement
should he bjferrtd Mention of "»"""¦¦¦! product* and
compam name* does not constitute endorsement the
Agency
r
RFFMENCES
SatlcrmanS A .Dzubs) T G and Bajnv,at'J-rr R I (,#s»j
Development nf crusial profiles for recptor m-j-kli-t
Armovphcrtc £r>'trorvnenr 23, 1825-18?$
OaauccinG'S lai-ocko11 B .LccK I t,DaUnj;rS
1. and MgcbroT J SJ190) The i/k^mp'iicr uintm-
KeJ scsiming electron mKi^uopv in cn.;r.ini'rnla' ULd-
KS J At* Pailut ConsroI As* J?, QiT-94}
Dzubay T C tlW) Development ol coippom'.c teeepioi
method; L'S EPA, Office of Research Jnj Developmeni
Report Nn EPA/&00/>-88,fl36. Research Tnangle Psik.
nc jr» i
D?utw <> Km.m Apores und Pndfti Wm C
Brown. Dubuque, IA
Mumane Y tl9S8| F.surr.'.te uf municipal refuse inemeruior
conlnbulion to Philadelphia aerosol Air-otphrra t n-
rtrumvnl 22, 241! 2at S
Mamane Y . Miller > L and D/uhaj T C 11986) Character-
ization of individual fly ash particle* emitted from coal-
nnd oil-fired power plants Atmisoheru tntironnrni 10.
2121-2135
Mamane V andDzubayT Gurees in Phila-
delphia for receptor modeling applications J Ait Pnllui
Control Sac 38, (in presi)
Stanley R G and Linikms H F (I974| Pulirn, pp 111-129-
Spnrger-Verlag, Berlin
Sussman A S and Halvorson H O (|9«i| Spom Thrir
Dormamy and Cerminailnn Harper & ROw»>»a. unnwtft
rvprodcc# d* • J » ' *
utt to 4>w *4 ha m«ii *i c
------- |