DIFFUSION

OF HALOGENATED
HYDROCARBONS
IN HELIUM
3679

mations 01 << w s and w12 ~ 1 <<~ 1,2 these equations reduce

to

where N is the number of paramagnetic ions per unit

and volume; d is the average closest distance of approach
between a proton in the bulk and spin S; B3p, the
-1- - y12Ys2n2s(S + ’) [7J(0)+ 13J(ws)1 (A7)
Bessel function of order 3/2; and D = DI Ds,the +
TzwS 4 self-diffusion coefficient of the bulk protons plus that
of the spins S. Equation AS can be integrated in
The term J ( w ) is the spectral density function a t the closed form. The result, given by eq 33 of Pfeifer’s
frequency w and is given by the integral paper,6l was used to calculate Tlwsand Tz,’.

Diffusion of Halogenated Hydrocarbons in Helium.

The Effect of Structure on Collision Cross Sections

by Edward N. Fuller, Keith Ensley, and J. Calvin Giddings

Department of Chemistry, University of Utah, Salt Lake Citv, Utah 84112 (Received December 27, 1968)

Experimental binary gas-phase diffusion coefficients for some halogenated hydrocarbon compounds diffusing
into helium have been determined utilizing the gas chromatography (GC) peak-broadening technique. Certain
modifications are described which have improved the speed and accuracy of the technique. A critical discussion
of the validity of the method is given. Based mainly on the acquired data, additional parameters have been deter-
mined for use in connection with the Fuller-Schettler-Giddings diffusion volume method for estimating binary
diffusion coefficients. The relationship between effective collision cross sections and various molecular shape
and structural features is discussed.

Widespread efforts to characterize numerous dif-
fusion processes more exactly have resulted in an in-
creased need for accurate diffusion data and improved
methodology. Gradually more diffusion data are being
published, but very little data are presently available
for most organic vapors. The major objectives of this
investigation were to develop improved experimental
apparatus and techniques, to generate reliable data for
new systems, and to study the dependence of collision
cross sections on complex molecular structure. I n
contrast to simple systems where temperature vari-
ations are of utmost importance, the present complex
systems yield more information when studied as a
function of structural features such as chain length and
substituent position.
Experimental. binary diffusion coefficients were
determined for thirty-one halogenated hydrocarbons
diffusing into helium. From these data additional
“atomic diffusion volume” parameters were obtained
for use with the method developed by Fuller, Schettler,
and Giddingsl for predicting diffusion coefficients.
With the exception of two confirmatory determinations,
all diffusion coefficients reported here are for systems
not previously investigated.
A limited amount of related diffusion data have been
reported in the literature. Relatively recent work in-
cludes studies reported by Trautz and Ries2for H2-CC1,;
Trautz and Mueller3 for OrCC14; Gilliland4 for air-
chlorobenzene; Klotz and Millers for air-CNCI, air-
COCl2 and air-CClaNO2; Lee and Wilke6 for Freon-
(1) E.N.Fuller, P. D. Schettler, and J. C. Giddings, Ind. Eng. Chem.,
58, 18 (1966).
(2) M. Trautz and W. Ries, Ann. Phga., 8, 163 (1931).
(3) M. Trautz and W. Mueller, ibid., 22,363 (1936).
(4) E.R.Gilliland, Ind. Eng. Chem., 26,681 (1934).
(6) I. M. Klotz and D. K. Miller, J. Amer. Chem. SOC.,69, 2667
(1947).
(6) C.Y.Lee and C. R. Wilke, Ind. Eng. Chem., 46,2381 (1964).

Volume 75,Number 11 November 1860

3680
12-water, Freon-12-benzene and Freon-1Z-ethanol;
Bose and Chakraborty’ for OrCC14, Nz-CC14 and air-
CCl,; and Mueller and CahiW for CH4-CF4. Since
this work was completed, Byrne, Maguire, and Clarke9
reported values for four halogenated hydrocarbons in
Nz, Hz, and Ar. Earlier work is summarized in the
International Critical Tables.’O
Experimental diffusion coefficients were determined
using the gas chromatography (GC) peak-broadening
technique first developed by Giddings and Seager.”,lZ
With this technique which employes the experimental
sequence of gas chromatogaphy, a narrow pulse of a
trace component is introduced into a carrier gas which
is flowing a t a constant rate down an empty tube of
uniform circular cross section. As the trace component
is swept down the diffusion tube, diffusion and non-
equilibrium processes cause axial broadening of the
sample peak. At the far end of the tube, sample
passes through a detector producing a proportional
electrical signal. A strip-chart recorder traces out the
resulting Gaussian shaped concentration profile. From
the observed peak width the binary diffusion coefficient
can be deduced.”J2 The method is much like that pro-
posed by Tay10r’~J~and applied to liquid systems;
his equations are not valid in our flow range, but c o r
respond to the high-velocity limit mentioned below.
The diffusion coefficient is rather arbitrarily expressed
here in terms of the experimentally determined “plate
height,” the quantity used in gas chromatography to
characterize peak dispersion, by the theoretical ex-
pression
D~~ = @(H + HZ - roz/z) (1)
where ir is the gas velocity averaged over the tube cross
section, rothe tube radius, H the plate height determined
as LrZ/tZ, L the tube length, 9 the peak variance in
time units and t the tube residence time. The positive
mot of eq 1 is valid up to a certain critical velocity, %,
given byI2
gc = 4 6 DAB/ro (2)
At velocities greater than %I the negative root becomes
valid. When >> gothe Taylor limit becomes valid.
I n practice the proper choice of roots poses no serious
problem since one can make a reasonable estimate of
go beforehand or, alternately, make diffusion runs a t
several velocities, the data from which specify $.
Experimental Section
Apparatus. A diffusion apparatus with improved
characteristics was constructed. The basic compo-
nents are shown schematically in Figure 1.
Previously, using commercial GC apparatus with
excessive dead volume, it was necessary to obtain rather
complete data from a short correction tube for sub-
traction from the principal data in order to compensate
The J o u m l of Physical Chemiatru
K. ENSLEY,AND J. c. GIDDINQS
E. N. FULLER,
Recorder
lol
Electro-
meter
Vent ^Carrier
. .. II <nFlame
...-. Ionization
Sample
Stream
Gas i
Samplin
Volve
Figure 1. OC broadening apparatus for
diffusion measurements.
for end effects. The latter originate mainly in dead
volumes associated with sampling and detection devices.
The need for this cumbersome two-column procedure
has been eliminated through direct on-column sample
introduction and a rigorous reduction of detector dead
volume. In addition oven dimensions were increased
to accommodate the diffusion tube without tight coils
and bends which tend to distort tube cross sections and
otherwise alter peak dispersion.
Temperature in the electrically heated oven was con-
trolled by a Sim-Ply-Trol pyrometer manufactured by
Assembly Products, Inc. During any one run temper-
ature was controlled within about *0.2”; for any re-
peating set of runs used for a given diffusion coefficient,
the control level was about +0.4”. This temperature
regulation is adequat.e within the limits set by the ex-
perimental precision of the present technique.
A nominal 50-ft by 0.d. stainless steel beverage
tube with a bright interior finish was obtained from
Superior Tube Co. The length of the coiled tube was
determined by taping a wire to the tube’s side midway
between the inside and outside radius. Laid flat on the
floor, the length of wire (tube) was measured as 1521
cm. The volume of the diffusiontube was determined
(7) N. K. Bose and B. N. Chakraborty, T7ana. Ind. Inst. Chem. Eng..
8,67 (1955-56).
(8) C. R. Mueller and R.W. Cahill, J . Chem. Phys., 40.661 (1984).
(9) J. J. Byme, D. Mamire, and J. K. A. Clark. J. Phye. Chem., 71,
3051 (1967).
(10) “Internationd Critical Tables,” Vol. 5, McGraw-Hill Book Co..
Inc., New York, N. Y.,1926. P 62.
(11) 3. C. Giddings and S. L. Seager, J . Chem. Phya.. 35,2242 (1968).
(12) J. C. Giddings and 8.L. Seaper. I. and E. C. Fund.. 1,277 (1962).
(13) G. I. Taylor, Proc. Rou. Soc.. AZ19,186 (1953).
(14) G. I. Taylor. ibid., AZ55,473 (1954).
DIFFUSIONOF HALOGENATED IN HELIUM
HYDROCARBONS
from the weight of water required to fill it, giving 343.7
cm3. From these data the average tube radius was
estimated to be 0.2682 cm (or 0.1056 in.). This agrees
well with the value 0.210 * 0.004 in. i.d. specified by
the vendor. The lead connecting the sampling valve
to the diffusion tube, made from the same size tubing,
was considered part of the diffusion tube making a total
length of 1536 cm.
A stainless steel gas-sampling valve obtained from
Wahl-Henius Institute was used for sample introduc-
tion. Carrier gas and sample stream may be passed
continuously through two loops, each of about 0.5
cm3 volume, attached to a plate at the back of the
valve. By rotating the back plate and interchanging
carrier and sample loops, sample is introduced directly
on the diffusion column as a narrow plug.
Connected to the opposite end of the diffusion tube is
the base of the hydrogen flame ionization detector with
internal mixing volume of 0.38 cm3. The electrical
signal from the flame ionization detector was fed to an
electrometer whose sensitivity could be varied widely
by selecting input resistances of lolo, lo9, los or lo’
ohms. The amplified signal was fed to a Fisher Labora-
tory Recorder Riodel PWF, manufactured by Texas
Instruments, Inc., with 0.5 sec full-scale response time,
adjustable damping and gain, and variable chart speeds
up to 8 in./min.
Procedure. Gas samples were introduced slowly into
the sample stream. Since 0.5 cm3 of pure gas would
yield a signal much too large, and the approach to
“infinite dilution” is theoretically desirable, the sample
stream was greatly diluted with carrier gas through a
‘T” joint. The ratio of carrier gas to sample was ad-
justed until peaks of convenient size were obtained.
Vapor samples mere introduced into the sample
stream by slowly bubbling carrier gas through a test
tube of the liquid inside a cylindrical steel chamber.
The resulting gas mixture passed out the side of the
chamber into the sample stream. When necessary the
mixture was diluted as described above.
A small pressure drop across the flame ionization
detector was determined using a mercury manometer in
the hydrogen stream. To specify correctly the experi-
mental pressure this small correction (usually 30-40
mm) was added to the observed laboratory pressure.
The pressure drop through the diffusion tube was
negligible.
The plate height, H , was determined by measuring
the width of the recorder peaks at half height, measured
from outside the line of the leading edge to inside of the
trailing edge. The peak width at half height (in time
units) multiplied by the factor 1/(22/21n2) gives r for
Gaussian peaks. The diffusion coefficient was cal-
culated according to eq l .
Gases and Vapors. Helium and nitrogen were em-
ployed as the carrier gases. Grade A helium with
3681
stated purity 99.995% was obtained from the U. S.
Bureau of Mines. Whitmore Oxygen Co. supplied the
nitrogen, stated purity 99.989%, and hydrogen for the
detector, stated purity 99.98%. The gases and vapors
used as trace components were obtained from a variety
of sources and were in the range 95-99+% purity.
Analysis of Accuracy. The largest single source of
experimental error associated with the GC peak dis-
persion technique appears to be the measurement of r .
This presently can be determined to about *l%.
This gives, since H depends on the square of T, an ex-
perimental uncertainty of roughly 2% in H and DAB.
The magnitude of other possible sources of experi-
mental error can be estimated and shown to be generally
small in comparison.
The effect of finite sample size can be calculated ap-
proximately by assuming that the sample is introduced
as a plug of concentration co. The initial sample
variance will be given by
where x0 is the length of the sample plug in the tube.
The standard 0.5 cm3sample will occupy 2.26 cm of tube
length, i.e., xo = 2.26 cm and go2= 0.425 cm2. Experi-
mentally observed u2’s were in the range 250-200 cm2;
thus finite sample size represents a contribution of about
0.1-0.2%,
Using chromatographic theory derived by Giddings,*s
spurious plate height contributions arising from the
detector volume, stagnant pockets, and from tube coil-
ing can be estimated. The plate-height contribution
due to detector volume will be given by the expression.
(4)
where v d is the detector volume, in our case -0.38
cm3, ‘Ir, the diffusion tube volume, -350 cm3, and N d
the number of theoretical plates ascribable to the detec-
-
tor alone, -10-20.
Hd 1.3 X
For these dimensions we have
cm. Since observed plate heights
were usually about 0.2 cm, the error contribution
from this source is -0.1%.
Small pockets at various connections are perhaps
unavoidable. The contribution to H from stagnant
pockets can be approximated by the relation
assumed here that B -
where 6 is the depth of the stagnant pocket. It is
10 cm/sec, DAB 0.5 cm2/sec -
(15) J. C. Giddings, J. Gas Chromatog., 1, 12 (1963).
Volume 78,Number 11 November 1960
 3682 E. N. FULLER, AND J. C. GIDDINGS
K. ENSLEY,

Table I : Experimental Results for Nz-Ethane and Nz-Butane Compared with Literature Valuesa

No. of DABf atd dev, Lit. values and ref

System Runs T, OK om2/sec (corrected to exptl temp)

NS-ethene 10 302.6 0.170 i 0.0034 0.174 (0.163 a t 291.2’K)1*

Nz-butane
4 Reduced to 1 atm pressure.
0.174 (0.170at298.2’K)8
0.167 (0.163 a t 298.0°K)18
10 302.4 0.100 f 0.0016 0.0985 (0.0960 a t 298. O0K)l!J
0.1006 (0.0930 a t 293.2”K)ZO

-
and 6 1 mm. The effect of two or three such pockets
would be entirely negligible. Table 11: Experimental Diffusion Coefficients for Some
He-Halogenated Hydrocarbon Systemsa
Another spurious contribution to H comes from the
coiled configuration of the diffusion tube. The effect
No.
arises in the unequal path lengths and pressure gradi- of DABi std dev,
ents along “inside” and “outside” streampaths. The System Runs TIOK cml/sec
contribution from this effect can be approximated by16

H, - 0.1% rO4
~ -
DABRO~
1 . 4 X 10-6cm (6)
He-difluoromethane
He-1, 1-difluoroethane
He-1-fluorohexane
He-fluorobenzene
5
9
10
8
430.8
429.6
431.6
429.7
0.874 f 0.030
0.754 f 0.015
0.492 f 0 . 0 0 6
0.566 i 0.008
where Ro is the coil radius -30 cm, with fi and DAB as He-hexafluorobenzene 11 428 7
I 0.453 f 0.008
He-4-fluorotoluene 10 431.6 0.508 f 0.006
above. This contribution to H is presently negligible He-dichloromethane 10 427.5 0.750 i 0.009
but could become important for diffusion tubes with He-trichloromethane 8 429,l 0.624 xk 0.012
tighter coils and larger cross sections. He-l,2-dichloroethane 8 427.1 0.683 f 0.006
Coiling also leads to secondary flows,16 which may He-1-chloropropane 11 427.5 0.631 f 0.009
be important in extreme circumstances. However, in He-1-chlorobutane 12 429.2 0.555 f 0.010
He-2-chlorobutane 9 429.1 0.561 f 0.008
the vicinity of velocity dc (our velocity was generally He-1-chloropentane 8 428.2 0.518 f 0.004
< fit), a displacement across the tube radius occurs about He-chlorobenzene 10 430.9 0.542 f 0.006
equally fast by diffusion and by the primary flow.17 He-dibromomethane 8 427.7 0.665 f 0.007
Secondary flows are, of course, of much smaller mag- He-bromoethane 15 427.7 0.740 i 0.011
nitude. No effect was observed in earlier work12 a t He-1-bromopropane 20 428.2 0.592 i 0.009
He-2-bromopropane 10 428.0 0.606 f 0.012
velocities up to nearly 3 d, using a tighter coil; we there- He-1-bromobutane 11 426.6 0.545 f 0.006
fore assume the effect to be negligible here. He-2-bromobutane 7 427.2 0.553 f 0.013
He-1-bromohexane 6 427.5 0.461 f 0.008
Experimental Results He-2-bromohexane 7 427.9 0.470 f 0.012
The accuracy of the present results was checked using He-3-bromohexane 8 428.5 0.469 f 0.004
He-bromobenzene 9 427.1 0.543 f 0.010
nitrogen-ethene and nit,rogen-butane systems since He-2-bromo-1-chloro-
consistent literature values8*18-20
were available. Both propane 6 427.2 0.570 & 0.016
the experimental and literature values, reduced to one He-iodomethane 7 431.2 0.783 f 0.016
atmosphere pressure, are shown in Table I. The latter He-iodoethane 9 428.4 0.648 f 0.013
were corrected over short temperature ranges to our He-1-iodopropane 10 430.0 0.579 4 0.007
He-2iodopropane 11 430.2 0.579 f 0.012
experimental temperature assuming a T1.75 dependence. He-1-iodobutane 8 428.1 0.524 f 0.007
The comparison shows that present results differ, in He-2-iodobutane 6 427.1 0.545 f 0.013
each case, from the average literature value by less
than 1.5% and fall between extreme values. Reduced to 1 at.m pressure.
Results of binary gas-phase diffusion measurements
for thirty-one helium-halogenated hydrocarbon sys-
tems are shown in Table 11. All results have been re-
(16) R. J. Adier and J. A. Koutsky, Can. J. Chem. Eng., 42, 239
duced to one atmosphere pressure assuming a p-l (1964).
dependence. Each diffusion coefficient represents the (17) J. C. Giddings, “Dynamics of Chromatography, Part I. Princi-
average value from a t least five diffusion runs made at ples and Theory,” ;Marcel Dekker, Inc., New York, N. Y . ,1965.
two or more different gas-flow velocities. Precision is (18) J. H. Knox and L. -McLaren, Anal. Chem., 36,1477 (1964).
indicated by the standard deviation of the individual (19) C. A. Boyd, N. Stein, V. Steingrimsson, and W. IF. Rumpel,
J . Chem. Phys., 19,548 (1951).
determinations, in most cases about 2% of the observed (20) L. McLaren, Ph.D. Thesis, University of Edinburgh, Edinburgh,
diffusion coefficient. 1965.

The Journal of Physical Chemistry

DIFFUSIONOF HALOGENATED
HYDROCARBONS
IN HELIUM 3 683

Atomic Diffusion Volumes determined by Lee and WilkeS6 If these points are
The preceding data have been used to determine included, the parameters are -F = 11.9, -C1 = 22.5,
previously unavailable parameters for use in predicting -Br = 21.8 and -I = 29.8 cma. Since it seems un-
binary diffusion coefficients in the absence of experi- reasonable that addition of -C1 should increase the
mental data. collision cross section more than the addition of -Br,
Fuller, Schettler, and Giddings have developed a the listed values were for the present assumed more
correlation equation based on special atomic digusion reliable.
volumes.’ The equation is I n the latest least-squares analysis the temperature
exponent was found to be 1.776. For simplicity it was
decided to retain the previously determined 1.75. The
diffusion volumes of Table I11 were obtained with the
temperature exponent fixed a t this value. (Average
where T i s the temperature, OK; M A , MBthe molecular absolute error a t the least-squares minimum 4.1%;
weight, g/mol; p the pressure, atm; and the Vi’s are a t the constrained minimum = 4.2%.) The resulting
the atomic diffusion volumes, om3, to be summed over increase in average error is insignificant.
the atoms, groups, and structural features of each
diffusing species. Numerous vi’s and the 1.75 exponent Effective Collision Cross Sections of
to T were determined from a nonlinear least-squares Complex Molecules
analysis of over 300 experimental points. At that time The nature and magnitude of collision cross sections
the necessary data were lacking for -F, -Br, and -I, but are of considerable importance in the study of transport
values can now be established. More generally the and rate phenomenaas3 Efforts have naturally focused
data listed in Table I and I1 and additional points from on simple geometrical models (“hard” and “soft”
the literature (references 8, 20, 21-32) were used to spheres, spherocylinders, etc.) . Relatively little work
expand our data collection to 512 points. (Some points has been directed at the study of collision cross sections
from a few references were not included in order to keep for complex molecules and the effect of chemical and
the quantity of data within reasonable limits.) Another structural parameters. This neglect has been rein-
least-squares analysis was carried out as before. The forced by the long recognized difficulty in extending
resulting optimized values of the atomic diffusion rigorous theory beyond models of utmost simplicity,
volumes are listed in Table I11 in comparison with the As Watson and Goldberger point out, the collision of
previous values. The list includes parameters for bound systems is an aspect of the intractable many-
those halogen substituents not previously determined body problem. 34 Understanding is further complicated
and gives improved values for the others. by the uncertain statistical populations of rotamers in
The diffusion volumes in Table 111 were obtained gas-phase, flexible-chain molecules.
excluding three data points for freon-12 systems Transport coefficients, particularly viscosity and
more recently crossed molecular beam scattering mea-
surements have provided cross sectional data for many
Table I11 : Special Atomic Diffusion Volumes (cmJ systems.35 Hirschfelder, Curtiss, and Bird point out
A. Atomic and structural diffusion volume increments, vi’s
(21) R. E. Bunde, University of Wisconsin Naval Research Labora-
C 15.9 L16.51 (F) 14.7 ... tory Report C M 850, 1955.
H 2.31 [1.98] (Cl) 21.0 [19.51 (22) J. P. Kohn and N. Romero, J. Chem. Eng. Data, 10, 125 (1965).
0 6.11 [5.48] (Brj 21.9 ... (23) A. P. Malinauskas, J. Chem. Phys., 42, 156 (1965).
(N) 4.54 [5.691 (1) 29.8 *.. (24) R. Paul and I. B. Srivastava, ihid.,35, 1621 (1961).
(S) 22.9 [ 17.01 Aromatic or Heterocyclic (25) R. Paul and I. B. Srivastava, Ind. J. Phys., 35,465 (1961).
Ring-18.3 [-20.21 (26) R. Paul and I. B. Srivastava, ibid., 35,523 (1961).
(27) W. F. Rumpel, University of Wisconsin Naval Research Labora-
B. Diffusion volumes of atom and simple molecules, Zui’s tory Report CM 851, 1955.
He 2.67 [2.88] 18.0 [18.9] (28) S. C. Saxena and E. A. Mason, Mol. Phys., 2,379 (1959).
Ne 5.98 [5.59] 26.7 [26.9] (29) K. P. Srivastava, Physiea, 25, 571 (1959).
Ar 16.2 [16.1] 35.9 [35.9] (30) R. E. Walker and A. A. Westenberg, J. Chem. Phys., 32, 436
Kr 24.5 L22.81 20.7 [14.9] (1960).
Xe 32.7 r37.91 13.1 [12.7] (31) A. A. Westenberg and G. Frazier, ihid.,36,3499 (1962).
Hz 6.12 [7 071
I 71.3 [69.7] (32) S. Weissman, S. C. Saxena, and E. A. Mason, Phys. Fluids, 4,
Dz 6.84 [6.701 38.4 i37.71 643 (1961).
Nz 18.5 [17.9] 69.0 j67.21 (33) J. 0. Hirschfelder, C. F. Curtiss, and R. B. Bird, ‘‘Molecular
16.3 [16.6] 41.8 [41.1] Theory of Gases and Liquids,” John Wiley & Sons, Inc., New York,
0 2
N. Y., 1954.
Air 19.7 L20.11
(34) M. L. Goldberger and K. M.Watson, “Collision Theory,” John
a Note: ( ) indicates the listed value is based on only a few Wiley & Sons, Inc., New York, N. Y . ,1964, Chapter 11.
d a t a points. [ ] indicates previous values, ref 1. (35) E. W. Rothe and R. B. Bernstein, J. Chem. Phys., 31, 1619
(1959); R. W. Landorf and C. R. Mueller, ibid., 45,240 (1966).

Volume 78. Number 11 November 1969

3684
the advantage of binary diffusion, as opposed to vis-
cosity and thermal conductivity, in that it reflects
directly the interaction between dissimilar molecules.33
Adding to this advantage, the GC peak broadening
method, with the extreme sensitivity of existing GC
detectors, can be used for components in exceedingly
small concentrations (e.g., mole ratio). Hence
large and complex molecules, which may decompose
before reaching their boiling point, can be studied a t
high vapor dilutions well below this temperature.
Gas chromatography itself has been applied to high
boiling compounds 200-250” below their boiling
points;3e such species would be difficult to study by
viscosity, molecular beam or most diffusion techniques.
The statistics of collisions between like molecules be-
come exceedingly complicated if the molecules are com-
plex and especially if they are flexible. By our ap-
proach the cross sectional probe is a helium atom or
other inert species. This “atomic probe” has the ad-
vantage of simple geometry, low mass, smallness, and
inertness. It is little affected by attractive forces,
thus yielding an approximation to the repulsive shell
independent of dipole and chemical factors. By virtue
of its small size and low mass, DABwill reflect a degree
of structural fine graining.
For simple systems, temperature is a key variable,
providing evidence on the depth and shape of the poten-
tial well. For complex systems, particularly where one
is probing for the relatively hard shell of repulsion,
chemical structure is presently a more relevant param-
eter. The fine nuances of temperature variability,
complicated by structural change, would, to be useful,
require theoretical elaboration much beyond present
knowledge. This explains our present emphasis on
structural factors.
Liquid droplet models. E q 7 has its roots in hard-
sphere kinetic theory. The denominator is essentially
a replacement for p ( a A + a d 2 , where a’s are collision
diameters. Arnold replaced the a’s by the sum of the
atomic LeBas volumes raised to the -*/3 power.37
This has proven an effective way for estimating the
cross sections of complex molecules using bulk volu-
metric data, Equation 7 relates to the Arnold ap-
proach except that the cross sections are provided by the
diffusion data themselves. Therefore eq 7, with allied
atomic diffusion volume parameters, provide an ap-
proximate measure of relative collision cross sections.
The above are essentially “liquid droplet” models,
in which the group volumes add together and somehow
assume spherical shape. The success of these ap-
proaches is somewhat surprising; one would expect
shape (nonsphericity) factors to have more influence On
cross section. There is presently no tractable theoreti-
cal approach to this question for truly complex mole-
cules, although significant advances have been made in
basic theory. 38 some semiquantitativeconsiderations
are given below based on collision “shielding.”
The Journal of Physical Chemistry
E. N. FULLER,
K. ENSLEY,
AND J. C. GIDDINGS
Groups within complex molecules are partially
shielded from collision by other groups which exempt
certain collision paths. When a group is completely
shielded it contributes nothing to the cross section of
the target molecule. Extended molecules, with ex-
posed (minimally shielded) groups, have relatively large
total cross sections.
The contribution of group i to cross section may be
written as at = ato (1 - St) where Siis the shielding
probability. If other groups were positioned randomly
around group i we would have
St = 1 - (1 - Sl5)
j#i
where st5 is the probability that group j intercepts the
collision path. However the structural correlation
introduced by fixed bond lengths, angles and, for rigid
molecules, rotational states, will alter the result some-
what.
The effect of molecular dimensions can be shown by
assuming a probe molecule of diameter up and a shield-
ing group of diameter u5 a t distance d, from group i.
Providing dt5 is large compared to the a’s, orientation
averaging yields
(9)
i.e., the shielding probability falls off with the inverse
square of group separation distance. One would there-
fore expect larger st5’s and smaller cross sections for
molecules in a droplet form than in an extended chain
configuration. The contradiction of this with the
widespread success of the “liquid droplet” model in-
vites closer scrutiny.
Experimental Evidence. The experimental data in
Table I1 show that smaller diffusion coefficients are
associated with halogens in the l-position than in the
2- or 3-position on a straight chain. The latter are
presumably more globular. The helium-l-iodopropane
and helium-2-iodopropane systems, with nearly equal
diffusion coefficients, were the only exceptions. Sta-
tistical tests39 indicate that these small differences are
significant; for l-bromohexane and 2-bromohexane a t
the 75% confidence level and for the other pairs a t the
95% level.
Early diffusion data for halogenated hydrocarbons
diffusing in air, summarized in the International Critical
Tables,l0 show the same effect. Similar results for iso-
meric butanols diffusing in air have been observed by
(36) B. L. Karger and W. D. Cooke in “Advances in Chromatog-
raphy,??Val. 1, J. C, Giddings and R.A. Keller, Marcel Dekker, Inc.,
New York, Ed., N. Y., 1965,Chapter 9.
(37) J. H.Arnold, Ind. Eng. Chem., 22,1091 (1930).
(38) C. F. Curtiss and J. S. Dahler, J . Chem. Phua., 28, 2352 (1963).
(39) C. A. Bennett and N. L. Franklin, “Statistical Analysis in Chem-
istry and the Chemical Industry,” John Wiley & Sons, Ino., New
York, N. Y.,1954.
DIFFUSION
OF HALOGENATED
HYDROCARBONS
IN HELIUM
Gilliland4and for propanols in helium by Seager, Geert-
son, and Giddings.40
While many data for unsubstituted hydrocarbons
show a similar trend, diffusion coefficients for branched
hydrocarbons are often equal or less than values for the
straight chain compounds. 1@,20,41 Such results have
been used to infer a crumpled form for straight-chain
paraffin^.^
Additional evidence comes from the comparison of
diffusion volumes (added from Table 111) with LeBas
boiling points volumes (-C = 14.8 and -H = 3.7 cma)
for straight-chain hydrocarbons. With increasing car-
bon number beginning at CIthe ratio is 0.848, 0.883,
0.896, 0.901, 0.906, 0.908 . . . . 0.923; the latter being
the limiting long-chain value. It appears that the
diffusion data may have “distorted” the carbon-hydro-
gen diffusion volume ratio (responsible for the above
variation) so that slightly reduced diffusion among long
chains is obtained. This broad trend reflects results
with over one hundred binary pairs. Some contradic-
tions arise among aromatic compounds; e.g., benzene,
compact in structure, has a ratio of 0.948. This may
result from incorrectly lumping heterocyclic and other
aromatic rings into a single parameter.
We conclude that the liquid-droplet model works
surprisingly well, which may relate to the following con-
siderations. Few chain compounds are expected to
actually assume near-spherical configurations. How-
ever cross sections, when averaged over all external
rotations, apparently have roughly similar asymmetric
contributions. (The abnormal sphericity of methane
is allowed for in the adjustable C-H volume ratio;
also the calculated cross sections are usually too high.)
The long-chain molecules, whose deviations alone
3685
could be extreme, may have rotational-isomer popula-
tions weighted toward intramolecular associated forms.
This would contrast with liquid-state, and especially
solid-state, populations where the trans-trans confor-
mation about each single bond is favored by roughly
500 ~ a l .However
~ ~ the net attractive interactions be-
tween segments in solvent-free (i.e., gas-phase) long-
chain molecules may decrease mean dimensions. This
is not unlike Langmuir’s arguments for coiling due to
excess surface energy in unlike media, and has been
confirmed at intervals for gas-phase paraffins based on
transport cross ~ e c t i b n s . The
~ ~ ~effect
~ ~ would likely
increase with chain length, but increasing nonvolatility
hinders investigation. Perhaps the present technique,
applicable at high dilutions, could produce cross section
data bearing on the matter.
More theoretical work is needed with these complex
systems in order to ascertain the effects of rotameric and
external-rotation averaging, increasingly important
secondary collisions (successive collisions of the “probe”
with the same “target” molecule) and realistic inter-
group and intermolecular potentials.
Acknowledgment. This investigation was supported
by Public Health Service Research Grant GM 10851-12
from the National Institutes of Health.
(40) S. L. Seager, L. R. Geertson, and J. C. Giddings, J. Chem. Eng.
Data, 8,168 (1963).
(41) G. A. McD. Cummings and A. R. Ubbelohde, J . Chem. SOC.,
3751 (1953).
(42) R. A. Scott and H. A. Scheraga, J. Chem. Phya., 44, 3054
(1966); 9. Mizushima, “Structure of Molecules and Internal Rota-
tion,” Academic Press, New York, N. Y.,1954.
(43) R. M. Melavin and E. Mack, J . Amer. Chem. Soc., 54, 888
(1932); J. C. McCoubrey, J. N. McCrea, and A. R. Ubbelohde, J.
Chem. Soc., 1961 (1961).
Volume 73,Number 11 November 1960