United States
                   Environmental Protection
                   Agency
Hazardous Waste Engineering
Research Laboratory
Cincinnati OH 45268
                   Research and Development
EPA/600/S2-86/055   Sept. 1986
x°/EPA         Project Summary
                   Predicting the  Effectiveness  of
                   Chemical-Protective  Clothing:
                   Model  and  Test  Method
                   Development

                   A. S. Bhown, E. F. Philpot, D. P. Segers, G. D. Sides, and R. B. Spafford
                     A predictive model and test method
                   were developed for determining the
                   chemical resistance of protective poly-
                   meric gloves exposed to liquid organic
                   chemicals (solvents). The prediction of
                   permeation through protective gloves
                   by solvents was emphasized.
                     Several theoretical models and test
                   methods for estimating permeation re-
                   lated properties were identified during
                   a literature review and were evaluated
                   in comparison to performing direct per-
                   meation tests. The models and test
                   methods chosen were based on theo-
                   ries of the solution thermodynamics of
                   polymer/solvent systems and the diffu-
                   sion of  solvents in polymers (as op-
                   posed to being based on empirical ap-
                   proaches). These models and test
                   methods were further developed to es-
                   timate the solubility, S, and the diffu-
                   sion coefficient, D, for a solvent in  a
                   glove polymer. Given S and D, the per-
                   meation of a glove by a solvent can be
                   predicted for various exposure condi-
                   tions using analytical or numerical solu-
                   tions to  Pick's laws.
                     The model developed for estimating
                   solubility is based on Universal Quasi-
                   chemical Functional-group Activity Co-
                   efficients for Polymers (UNIFAP) the-
                   ory, which is an extension of the
                   Universal Quasichemical Functional-
                   group Activity Coefficient (UNIFAC)
                   method for predicting phase equilibria.
                   The model recommended for estimat-
                   ing diffusion coefficients versus con-
                   centration is the Paul model, which is
                   based on free-volume theory. The pre-
                   dictive test method developed is  a
liquid-immersion absorption/desorp-
tion method that provides estimates of
S and D.
  The models and test method chosen
were incorporated into an algorithm for
evaluating protective gloves recom-
mended for use with new chemicals. Fi-
nally, limited confirmation of the devel-
oped models  and test method was
performed by comparing estimated val-
ues of S  and D with reported experi-
mental data and by using the estimated
values to predict instantaneous perme-
ation rates, breakthrough times, and
steady-state permeation rates for com-
parison with experimental permeation
data.
  This Project Summary was devel-
oped by EPA's Hazardous Waste Engi-
neering Research Laboratory, Cincin-
nati, OH,  to announce key findings of
the research project that is fully docu-
mented in a separate report of the same
title (see Project Report ordering infor-
mation at back).

Introduction
  Section 5 of the Toxic Substances
Control Act (Public Law 94-469) requires
prospective mMhufacturers of new
chemicals to submit Premanufacture
Notifications (PMNs), which are re-
viewed by the  U.S. Environmental Pro-
tection Agency's Office of Toxic Sub-
stances (OTS). PMN submittals often
propose  specific chemical-protective
clothing to limit the dermal exposure of
workers to toxic chemicals. Because
OTS has only 90 days to complete each
PMN review, and because testing by the

-------
manufacturer must be kept to a mini-
mum, the development of reliable mod-
els for predicting the performance of
protective clothing is desirable. Thus
the first objective of this  study was to
develop predictive models for evaluat-
ing the chemical  resistance of protec-
tive clothing exposed to liquid organic
chemicals.
  No matter how sophisticated the
models developed here or in future ef-
forts, there may often be insufficient
data available to allow a given model to
make predictions as accurate as those
requested (i.e., ±50% for permeation
rate and ±20% for breakthrough time).
Thus predictive test methods are also
needed to allow estimates of the perme-
ation of  chemical-protective clothing
under expected exposure conditions.
The second objective of the study was
to develop such methods, either by rec-
ommending the use  of  existing test
methods  or by developing new ones.

Task 1:  Developing Predictive
Models for the Permeation  of
Polymeric Membranes by
Organic Solvents
  Because of the expense of conducting
permeation tests, models that would al-
low  the  prediction  of  permeation
through polymeric materials  would be
useful for screening candidate protec-
tive materials. The work reported here
emphasized the prediction of perme-
ation through unsupported, natural-
rubber glove  formulations because of
the relative purity of these formulations
(typically 93% latex) compared with oth-
ers (e.g., PVC-based  gloves typically
contain >40% plasticizer). Also, experi-
mental solubilities and diffusion coeffi-
cients for various solvents in natural
rubber are available.

Applicable Mass  Transport
Theory
  If it is assumed that Pick's laws of dif-
fusion are valid, that the diffusion coeffi-
cient is independent of solvent concen-
tration in the polymer, and that swelling
of the polymer is negligible, then the
permeation  rate (J)  for a  solvent
through a glove sample  whose  outer
surface is in contact with the liquid sol-
vent is given by
     •exp  -mVDt/e2
where D is the solvent diffusion coeffi-
cient, S is the solubility of the solvent
in the polymer, and fis the sample thick-
ness. Thus if the diffusion coefficient
and solubility of the solvent in the poly-
mer can be predicted by a  model, then
permeation-rate curves and other quan-
tities of interest (e.g., breakthrough
times, lag  times, and steady-state per-
meation rates) can be predicted. This
simple model may also be modified to
account for phenomena such as the
generally  observed  concentration de-
pendence  of the diffusion coefficient.

Solubility Predictions
  Two approaches to predicting solubil-
ity were attempted: The Flory-Huggins
theory  (combined with the  use of solu-
bility parameters) and the Universal
Quasichemical Functional-group Activ-
ity Coefficient (UNIFAC-FV)  or Universal
Quasichemical Functional-group Activ-
ity Coefficient  for Polymers (UNIFAP)
theory. Both approaches assumed that
none of the polymer dissolved in the
liquid phase (that is, that the solvent re-
mained pure)  and that the standard
state for the solvent in  both the liquid
and the polymer phase was the pure
solvent (for which the activity, a1r  is
unity). Thus it was only necessary to de-
termine the solvent volume fraction  in
the polymer that yielded a^l to predict
the solubility of the solvent in the poly-
mer.
  If it is assumed that the molar volume
of the solvent, VT is much less than the
molar volume of the polymer, then the
Flory-Huggins equation may be written
as
1n a, = 1n
                                 (2)
  J -   DS/f   1  +   >  2 • cos(irm)   (1)
     \    /[    m = 1
where x is the Flory interaction parame-
ter for the polymer/solvent system and
Q>1 is the  solvent volume fraction. At
equilibrium, ai = 1 and Equation 2 may
be written as

  0=1n*, + (1-*l) + x(1-*i)2  (3)

It has been suggested that \ is related to
Hildebrand's solubility parameters for
nonpolar systems:
                                     where  6, and S2 are the solubility
                                     parameters  for the solvent and poly-
                                     mer, respectively.  In  this work, .
                                     Hansen's so-called three-component or \
                                     three-dimensional solubility parame-
                                     ters were used to calculate the separa-
                                     tion in  solubility space, A, which was
                                     substituted for the difference (&i-82) in
                                     Equation 4 to make the theory more ap-
                                     plicable to  polar systems. Three ap-
                                     proaches to the calculation of A have
                                     been reported:
    A = [(8g - S.J)
  A' = [4(85 -»d)2 + (8JJ-8J)2
                                                                      (5)
                                                                       (6)
  A" = [8g -
                                                    + 0.250(TP - Te>2]1/2  (7)
        x = (P!/RT» (8, - 82)2
                               (4)
where the p and I superscripts refer re-
spectively to the polymer, and the liquid
solvent, 8d, is for dispersion forces, 8p is
for polar effects, 8h is for hydrogen
bonding, and T = (8J; + 82,)1'2 .
  Solubilities calculated for a series of
solvents in natural rubber are listed in
Table 1. These  calculated solubilities
correlate poorly with  reported experi-
mental solubilities. For this reason and
because the UNIFAP approach to calcu-
lating solubilities appeared to be more
promising,  attempts to  use the Flory-
Huggins theory and solubility parame-
ters were discontinued.
  The UNIFAC theory is based on the
use of  functional-group interaction
parameters calculated from experimen-
tal phase-equilibrium data to determine
the activities of each component in  mix-
tures. This work was extended to in-
clude free-volume theory and thus to
make it applicable to polymer/solvent
systems. A software package (UNIFAP)
is available to conduct calculations that
enable the prediction of solvent solubil-
ity in polymers. The primary input  data
are the solvent and polymer densities at
the temperature of  interest  and the
identity  of the functional groups in the
solvent and the polymer.
  The data  in Table 2 compare solubili-
ties calculated  using UNIFAP theory
with experimental data for a series of
solvents in natural rubber and  four
other polymers. The typical agreement
is within a factor of five or better.
  The data in Table 2 reveal two prob-

-------
TaWe 1.
Solubilities Calculated for Various Solvents in Natural Rubber8 (Based on Flory-
Huggins Theory and Solubility Parameters)
                      Literature solubilityb
                                        Calculated solubilityc-d
Solvent
Methanol
Ethanol
Isopropanol
n-Butanol
n-Pentanol
Benzyl alcohol
n-Propanol
Acetone
2-Ethyl-1-butanol
t-Pentanol
Diethyl carbonate
Methyl ethyl ketone
Ethyl acetate
n-Propyl acetate
n-Hexane
n-Heptane
Tetralin
Cyclohexane
Cyclohexanone
Toluene
Tetrachloroethylene
Carbon tetrachloride
Trichloroethylene
10*xCexp
4.92
8.59
52.8
125
130
142
146
169
259
437
636
713
766
1270
1540
1580
3330
3380
3410
3870
4240
5370
5690
10s x C,
3.27
5.37
18.7
14.3
34.5
35.0
9.33
64.2
28.1
2.23
#
69.9
7400
28.4
8.57
5.7t
266
39.7
489
299
136
94.8
#
705 x C',
7.88
3.27
9.27
6.99
73.7
35.0
5.25
26.7
8.48
0.770
858
30.8
757
7.74
7.75
7.00
757
25.9
429
284
727
88.9
*
70s x C]
392
394
604
475
724
2630
439
*
483
92.9
#
#
*
363
752
T35
*
7380
#
#
*
#
*
"The concentrations (solubilities) reported are in units of moles/cm3 of unswollen polymer at
 298 K.
b These values are calculated from the experimental volume-fraction data reported by D.R. Paul.
These values are calculated using Flory-Huggins theory. The term C» means that Equation 5
 was used to calculate the separation in solubility space; C\ corresponds to Equation 6; and C]
 corresponds to Equation 7.
dThe symbol "*" means that no solubility could be calculated, because these systems are
 predicted to be miscible in all proportions.
lems with UNIFAP. First, it yielded pre-
dictions of total miscibility for a number
of solvents, which experimentally
showed large but finite solubilities. This
prediction occurred because UNIFAP
does not consider crosslinking. Second,
the database of interaction parameters
is limited to a relatively small  set of
functional groups. Thus the interaction
parameters for  functional groups
needed to define given polymers or sol-
vents may not be available.

Diffusion Coefficient
Predictions
  This work  emphasized the prediction
of solubility.  However, C.W. Paul's
model based  on free-volume theory
was used to estimate the concentration-
dependence  of the diffusion coefficients
for benzene  and n-heptane in  natural
rubber. Paul's model requires fewer ex-
perimental data (that is, physical prop-
erties) than other models of free volume
theory.
  A comparison  of  the predicted
concentration-dependence of the diffu-
sion coefficient for benzene in  natural
rubber is shown in Figure 1. Although
the values of  the maximum in both
                             curves are in good agreement, the loca-
                             tions of the maximum  relative to the
                             solvent volume fraction are not. The ini-
                             tial work in the application of the Paul
                             model will  continue in  an attempt to
                             eliminate the discrepancy shown.

                             Prediction of a Permeation
                             Rate Curve
                              The solubility predicted for benzene
                             in natural rubber using UNIFAP and the
                             diffusion-coefficient curve given in Fig-
                             ure 1 (despite its disagreement with ex-
                             perimental data) were used to predict
                             the permeation rate as a function of
                             time  for a 0.046-cm-thick glove (Fig-
                             ure 2). This curve was calculated using
                             the Crank-Nicholson implicit finite  dif-
                             ferences approach to solving Fick's sec-
                             ond law of diffusion with a reference
                             table for the diffusion coefficient. The
                             agreement between the predicted curve
                             and that derived from data reported by
                             Weeks and McLeod should be consid-
                             ered somewhat fortuitous but nonethe-
                             less encouraging.
                            Task 2: Developing Predictive
                            Test Methods for the
Permeation of Polymeric
Membranes by Organic
Solvents
  Permeation tests are often conducted
to scree^ elastomeric materials that
may be suitable for the formulation of
chemical-protective clothing.  Perme-
ation tests are time-consuming and ex-
pensive, and they often require the de-
velopment of analytical methods for the
permeant.  These analytical methods
usually are not universal,  and thus
skilled technicians may be required to
conduct permeation tests. This work in-
cluded the development of a simple test
method that allows the prediction of the
permeation  of an organic solvent
through a  polymeric membrane. The
test method, based on liquid-immersion
absorption and desorption  measure-
ments, is simple, universal, and can be
carried out by technicians with minimal
training.

Experimental
  For each experiment, a  polymeric
sample with known dimensions and
weight was immersed in the solvent of
interest at t=0. The sample was re-
moved  periodically from the solvent,
blotted  to remove excess solvent, and
then weighed. The  sample was  reim-
mersed in the solvent after each weigh-
ing.  This test was  referred  to as the
liquid-immersion absorption test.
  After the weight of the  polymeric
sample reached a constant value (indi-
cating equilibrium), it was removed
from the solvent and blotted to remove
excess solvent. The sample  was  then
suspended from a hook on an analytical
balance, and its weight was recorded as
a function of time. This test was referred
to as the liquid-immersion desorption
experiment. Desorption  experiments
were conducted only for those samples
that  exhibited appreciable absorption.
The materials tested were circular sam-
ples  of  unsupported gloves that were
3.5 cm in diameter. The tests were con-
ducted with the samples and solvents in
the temperature range of 73 ± 6°F.

Data Analysis
  The absorption (or desorption)  of a
solvent  by a planar sample (if edge ef-
fects are negligible)  is given by
  Nit/Mo, = 1 - (8/ir2)     (2m + 1)~2  (8)
                 m=0

-------
Table 2.
Polymer
Natural
rubber
















Butyl
rubber



Neoprene
rubber


Nitrite
rubber

Poly
(vinyl
chloride)

Comparison of Solubilities Calculated Using the UNIFAP Model or Obtained Exper-
imentally with Manufacturers' Chemical-Resistance Guidelines"
Degradation Permeation
Solvent W5 x Cuni 105 x Cmo ratingb ratingc
Methanol
Ethanol
Isopropanol
n-Butanol
n-Pentanol
Benzyl alcohol
n-Propanol
Acetone
2-Ethyl-1-butanol
t-Butanol
t-Pentanol
Diethyl carbonate
Methyl ethyl ketone
Ethyl acetate
n-Propyl acetate
n-Hexane
n-Heptane
T&trftlln
i cm GUI i
Cyclohexane
Cyclohexanone
Toluene
Tetrachloroethylene
Carbon tetrachloride
Trichloroethylene

Acetone
Cyclohexane
Isopropanol
Toluene

Acetone
Cyclohexane
Isopropanol
Toluene
Acetone
Cyclohexane
Isopropanol
Toluene

Acetone
Cyclohexane
Isopropanol
Toluene
26.5
42.4
81.6
93.3
103
27.5
100
185
114
81.5
96.9
202
277
477
587
«e






322
#
29.7
#

*
*
69.8
	 f
— -

*s
*
49.5
*
4.92
8.59
52.8 (81.4)d
125
130
142
146
169 (263)
259
414
437
636
713
766
1270
1540
1580
3330
3380 (31907
3410
3870 (3480)
4240
5370
5690

(78.7)
(3240)
(6.8)
(1920)

(696)
(1040)
(95.5)
(3410)
(3130)
(117)
(389)
(1560)

(—)"
(—)
(—)
(—)
E
E
E
E
NA
NA
E
E
NA
NA
NA
NA
G
G
F
NR
NA
NA
NA
NA
NR
NA
NR
NA

NA
NA
NA
NA

G
NA
E
NR
NR
NA
E
F

NR
NA
G
NR
E,NN
VG,NN
E,NN
G
NA
NA
VG
F, NN
NA
NA
NA
NA
P,NN
G,NN
F,NN
NA
NN
NA
NN
NA
NN
NN
NN
NN

NA
RR
NA
NA

F,NN
NN
E
NN
NN
RR
E,RR
F, NN

NN
NN
E
NN
where t is the time, M, is the cumulative
amount of sofvent absorbed (or de-
sorbed), IVL is the cumulative amount
absorbed at infinite time (that is, at long 1
times), DO is the diffusion coefficient for
the solvent in the polymer, and i is the
thickness of the sample. This equation
assumes that the diffusion observed is
Fickian, that the swelling of the polymer
sample is negligible, and that the diffu-
sion coefficient is independent of sol-
vent concentration in the polymer.
Absorption or desorption data may
be fit to Equation 8 using nonlinear
techniques to obtain an estimate of D0
and Mo, (the solubility, S, of the solvent
in the polymer is M Jv where V is the
volume of the unswollen sample). IVL is
simply the maximum weight gain of the
polymeric sample immersed in the sol-
vent. If the quantity (M,/MJ is plotted as
a function of t1/2/Ł the so-called reduced
sorption curve is obtained. The initial
slope of this curve (up to M^M«, = 0.6)
may be used to estimate an apparent
diffusion coefficient by using the follow-
ing equation:

D — (ir/tRH2 (Q)
— \Ttf lw/1 \J/

where D is the apparent diffusion coeffi-
cient and 1 is the initial slope of the re-
duced sorption curve. Reduced sorption
curves may be plotted for absorption or
desorption data. Only one of these re-
duced sorption curves is necessary to
calculate D; however, both were used in
this study when available.
Once estimates of D and S are ob-
tained as described above, the solvent
permeation rate (J) as a function of time
(through a glove sample covered on
one side by an organic solvent) may be
predicted using the equation
8 The units of the calculated equilibrium solubilities, Con/, and the experimental solubilities, Cexp,
 are moles/cm3.
Degradation  ratings are  as follows: E = Excellent,  G = Good, F = Fair, NR = Not Recom-
 mended, NA = Not Available.
cPermeation ratings  are  as  follows:  E = Excellent (permeation  rate <0.15 mg/m2/sec),
 VG - Very Good (permeation rate < 1.5 mg/m2/sec), G = Good (permeation rate < 15 mg/m2/
 sec), F = Fair (permeation rate < 150 mg/m2/sec), P = Poor (permeation rate < 7500 mg/m2/
 sec), ND = None Detected. The double letters refer to ratings using by ADL (4f; RR (recom-
 mended) means that a large amount of test data indicates excellent chemical resistance; NN
 (not recommended) means that a large amount of test data indicates poor chemical resis-
 tance; NA means rating not available (or conflicting ratings are reported by ADL).
dThe  values in parentheses were determined during immersion tests conducted under the
 current effort; the other data in this column were calculated from data reported by D. R. Paul.
eThe symbol "*" in this column means that an activity of one was achieved only for a solvent
 volume fraction equal to one. (That is, the solvent and the polymer are predicted to be miscible
 in all proportions.)
 fDash indicates that no interaction parameters were available for the polymer/solvent pair
 indicated.
BThe UNIFAP solubilities for solvents in poly (vinyl chloride) were based on the polymer struc-
 ture alone; the presence  of plasticizer was not considered.
hDash in parentheses  indicates that weight loss was observed in the immersion tests  with
 poly (vinyl chloride).
  J = (DS/O-1 +      2 • cos(irm)   (10)
            I    m = 1
      •exp(-m2ir2Dt/e2)[


The steady-state permeation  rate (JJ
may be estimated using
             J«, = DS/e
(11)
The breakthrough time (fc) based on a
given permeation rate (Jb) may be esti-
mated using
                                      4

-------
           0.75
           0.65-
         «  0.55

        I
        o

        "o  0.45
        §  0.35
        o
        o

        I
        Q  0.25
           O.tS
           0.05
                                        T
                           T
                                    Approximate Predictive
                                     Range 0.1 «Di <0.9
Predicted Using
the Paul Model
          -0.05'
               0.00    0.20      0.40      0.60      0.80     1.0O
                        Benzene Volume Fraction, *i

Figure 1.    Diffusion coefficient of benzene in natural rubber as  a  function of volume
           fraction.

300
E
^i
Permeation Rate, ,
§
c
\ \
—
I I I
—
Permeation Data
(Extrapolated from
Weeks and McLeodt
,'"'
- f
J I I
3 5 10


Predicted Using
Theoretical Model
I I I
15 20 25
Time, min
                                                         Jb = (DS/P) 1 +     2
                                                                        m=1
                                                                                                     cos(irm)  (12)
F/gun 2.    Comparison of predicted and experimental permeation rate curves for benzene
           through natural rubber.
      •exp(-m2ir2Dtj/('2)[


 Test Results
  Typical absorption and desorption
 curves are shown in Figure 3. Average
 solubilities and diffusion coefficients
 determined are given for four solvents
 and four polymers in Table 3.

 Predictions
  Predicted steady-state permeation
 rates and breakthrough times for the
 solvent/polymer combinations studied
 in this work are given in Table 4. This
 table also includes steady-state perme-
 ation rates and breakthrough times de-
 termined in permeation tests using the
 same polymeric glove samples and sol-
 vents. Figure 4 shows a predicted per-
 meation rate  curve (calculated  using
 Equation 10) and measured permeation
 rate curves for acetone through  nitrile
 rubber  and acetone through natural
 rubber.

 Conclusions and
 Recommendations
  Models that have been developed for
 predicting the permeation of organic
 compounds through protective  glove
 polymers have often been based on em-
 pirical approaches with little emphasis
 on  diffusion theory. The current work
 has demonstrated that predictive mod-
 els and test methods that yield diffusion
 coefficients and solubilities may be
 used to  estimate permeation data such
 as breakthrough times (for a given defi-
 nition),  steady-state permeation  rates,
 and permeation rate curves. These fun-
 damental parameters may be estimated
 using the theoretical models or the sim-
 ple test methods described in this re-
 port. Only relatively simple diffusion
theory and mathematical methods are
 required to calculate permeation data
from diffusion coefficients and solubili-
ties. These permeation data  may then
be  used to estimate the protection af-
forded  by polymeric gloves recom-
mended in PMN submittals.
  Most of the limited confirmation work
conducted here was performed manu-
ally. That is, although computers were
used to model  polymer solvent sys-
tems, to perform curve  fits of experi
mental data, and to predict permeation-

-------
      0   1234   S   67   8  9   10  11  12  13  14 15  16  17 18
Figure 3.   Absorption and desorption curves for toluene in natural rubber.
rate data,  the computer programs
written were not integrated into a single
software package, and they were not
made  user  friendly. Thus work under
the confirmation task was tedious and
time-consuming. For this reason, it was
not possible to evaluate all of the exper-
imental  immersion  and  permeation
data obtained, and the Paul model was
used to predict diffusion coefficients for
only two polymer/solvent systems.
However, any continued research effort
based on  the work described here
should begin  with consolidation of the
programs written into a single software
package. Thus future confirmation work
should largely begin to resemble the
preparation and evaluation of PMN sub-
mirtal and review software.
  Future efforts  to continue the work
described here should include the iden-
tification and use of computerized data
bases of physicochemical data that are
needed to predict diffusion coefficients
and solubilities (using, for example, the
Paul model or the UNIFAP theory). Also,
work should include addressing the lim-
itations of the UNIFAP  theory and the
Table 3.  Average Solubilities and Diffusion Coefficients Calculated from Liquid-Immersion Absorption and Desorption Test Data
Glove
Butyl rubber



Natural rubber



Neoprene rubber



Nitrite rubber



Solvent
Acetone
Cyclohexane
Isopropanol
Toluene
Acetone
Cyclohexane
Isopropanol
Toluene
Acetone
Cyclohexane
Isopropanol
Toluene
Acetone
Cyclohexaned
lsopropanold
Toluene
Solubility,
g/cm3
0.0457 ± 0.0010
2.7261 ±0.1088
0.0041 ± 0.0010
1.7690 ± 0.0372
0. 7527 ± 0.0099
2.6877 ± 0.0635
0.0489 ± 0.0079
3.2067 ±0.7524
0.4044 ± 0.0262
0.8787 ± 0.0739
0.0574 ± 0.0079
3.1388 ± 0.0974
7.8204 ±0.7236
0.0982 ± 0.0795
0.2340 ± 0.0234
1.4377 ± 0.0463

709 x Da
5.6 ± 7.4
750 ± 3.4
7.7 ± 73
380 ± 73
400 ±200
250 ± 30
42 ± 6.4
530 ± 67
250 ± 200
37 ± 6.7
7.3 ± 0.67
350 ± 38
420 ± 59
0.25 ± 0.067
0.42 ± 0.20
59 ± 5.9

709 x Dd
—
330 ± 7.3
—
280 ± 720
380 ± 20
490 ± 750
—
260 ± 35
260 ± 36
54 ± 4.4
—
200 ± 33
450 ± 82
—
—
200 ± 40

709 x Da
5.0
180
0.43
320
690
260
40
580
520
40
1.2
0.79
19
0.23
110
230
12
20
51
64
2.3
0.31
410 7.8
440 ±160
0.30 ± 0.080
0.72 ± 0.23
98 ± 15

109 x D'd
—
280 ± 82
—
270 ± 770
450 ± 55
480 ± 150
—
290 ± 25
377 ± 50
43 ± 6.5
—
270 ± 46
480+ 61
—
—
750 ± 78
aThe term Da is the apparent diffusion coefficient estimated from the initial slope of the reduced absorption curve. Dd is the apparent diffusion
 coefficient estimated from the initial slope of the reduced desorption curve. All data points [including (0,0)] for which  MJM«, <0.6 were used
 in these calculations; if no data points meeting this criterion except (0,0) existed, then the calculation was based on (0,0) and the first data point
 above M/Mx = 0.6.
bThe term De is the apparent diffusion coefficient calculated from a curve-fit of data obtained in an absorption test.
 Dd is the apparent diffusion coefficient calculated from a curve-fit of data obtained in a desorption test.
°The symbol "—" means that the experiment indicated was not conducted.
dSome of the nitrite-rubber samples immersed in these solvents continued to gain weight even after 530 hr.

-------
Table '4.   Comparison of Measured Breakthrough Times and Steady-State Permeation Rates with those Predicted from Immersion Test Data8


                                            Breakthrough time,b min
Glove
Natural
  rubber
Neoprene
  rubber
        Solvent
      Acetone
      Cyclohexane
      Isopropanol
      Toluene

      Acetone
      Cyclohexane
      Isopropanol
      Toluene
                                         Measured
                                                Predicted
17 ± 1.2
19 ± 7.7
> 750, < 7400
9.7+ 1.5

22 ± 2.0
>60

12
  9.9 ±  5.3
  4.7 ±  0.8
130  ±35
  4.0 ±  0.1

  8.6 ±  6.0
 28  ±  3.8
     on
  3.2 ±  0.2
                                                                             Steady-state permeation
                                                                               rate, ]Lg/(cm2-min)
                                             Measured
 34 ±   3.5
520 ±  48
  1
780 ±  27

150 ±  20
 35 ±  27

640
                                               Predicted
Butyl
rubber


Acetone
Cyclohexane
Isopropanol
Toluene
>1100
55 ± 6.3
—
24
CO
5.5 ± 0.4
CO
4.7 ± 0.5
<0.47
440 ± 110
—
400
0.25
660
0.028
560
0.05
53
0.05
91
  51
 900
   2.2
1200

 120
  49
   0.10
1100
   22
  160
    0.7
    2.5

±  68
±   5.5
    0.05
   45
Nitrile
rubber


Acetone
Cyclohexane
Isopropanol
Toluene
10 ± 2.0
>1400
—
52
3.2 ± 0.2
CO
00
72 ± 1.4
960 ± 26
<0.12
—
2JO
870
0.028
0.77
790
87
0.07
0.06
19
"The symbol "—" means that no permeation test was conducted with this glove/solvent combination. The symbol "°°" means that the
 permeation rate would always be less than that used to define the breakthrough time.
bThe predicted breakthrough times are based on the minimum permeation rates that could be detected in permeation tests: 0.47 \t.g/(cm2-min)
 for acetone, 0.12 \Lg/(cm2-min) for Cyclohexane, and 0.11 \Lg/(cm2-min) for toluene, and 0.31 \ig/(cm2-min) for isopropanol.
       1200
       1000
        800
        600
        400
        200 -
                                                  Permeation Data
                                        Predicted from
                                        Immersion Data
                  10
                         20
                                30     40
                                   Time, min
                                 50
                                        60
                                               70
                                                      80
Figure 4.
Comparison of predicted and experimental permeation rate curves for
acetone through nitrite rubber.
                                           modified Paul  model described in this
                                           report. In addition, the consideration of
                                           other theoretical models should be en-
                                           couraged.
                                            Perhaps the  major recommendation
                                           to result from the current study is that
                                           there should continue to be a strong
                                           emphasis on the development of pre-
                                           dictive algorithms that are  based  as
                                           much as possible on the rigorous  inter-
                                           pretation of diffusion theory. This ap-
                                           proach will allow permeation rate and
                                           cumulative permeation curves to be cal-
                                           culated as desired;  other permeation-
                                           related parameters (for example, break-
                                           through times) can be determined from
                                           these curves. Note that rigorous predic-
                                           tive algorithms can always be modified
                                           to yield  simple correlations;  however,
                                           the extension of an algorithm based on
                                           empirical correlations to the calculation
                                           of quantitative data is often difficult if
                                           not impossible.
                                            The full report was submitted in fulfill-
                                           ment of Contract No. 68-03-3113 by
                                           Southern Research Institute under the
                                           sponsorship of the U.S. Environmental
                                           Protection Agency.

-------
     A. S. Bhown, E. F. Philpot, D. P. Segers, G. D. Sides, and R. B. Spafford are with
       Southern Research Institute, Birmingham, AL 35255.
     Michael D. Royer is the EPA Project Officer (see below).
     The complete report, entitled "Predicting the Effectiveness of Chemical-Protective
       Clothing: Model and Test Method Development,"  (Order No. PB 86-209
       087'/AS; Cost: $16.95, subject to change) will be available only from:
            National Technical Information Service
            5285 Port Royal Road
            Springfield,  VA 22161
             Telephone: 703-487-4650
     The EPA Project Officer can be contacted at:
            Releases Control Branch
            Hazardous Waste Engineering Research Laboratory—Cincinnati
            U.S. Environmental Protection Agency
            Edison, NJ 08837-3679
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Official Business
Penalty for'Private Use $300

EPA/600/S2-86/055
                       0000329    PS
                                                       *GENCY
                       230  S  DEARBORN  STREET
                       CHICAGO                IL    60604
                                                                            * U.S GOVERNMENT PRINTING OFFICE. 1986 — 646-017/47157

-------