Gas Chromatography of Cigarette Smoke, Part III.
Separation of the Overlap Region of Gas and
Particulate Phase by Capillary Columns*
by K. Grob, F. J. Burrus & Cie, Boncourt, Switzerland, and Department of Organic Chemistry,
University of Zurich, Switzerland
Abstract
Following the investigation of gas
phase constituents boiling within the
range of 20-100°C, as presented in Part
II, the section of less volatile sub-
stances with boiling points between 100
and 200°C has heen studied. The sep-
aration by capillary columns is demon-
strated, and 64 substances are indicated
as probable gas phase constituents.
The partition of certain substances be-
tween gas phase and particulate phase
is shown to be of great importance.
Some influences upon the partition
equilibrium, for instance charcoal fit-
ters, are discussed.
Cigarette smoke—as well as
smoke from other tobacco products
—is commonly described as con-
sisting of two main parts: the in-
visible part of the aerosol, the so-
called gas phase; and the visible
part, made up of approximately
three billions per ml of droplets
with diameters between 0.1 and lµ,
called the particulate phase.
Up to a few years ago the par-
ticular interest of chemical and
medical investigators was concen-
trated on the particulate phase.
52 J. of G. C.—February, 1965
This may be illustrated by the fact
that the efficiency of regular filter
tips was essentially described by
the amount of particulate phase re-
tained on the filter.
Hilding (1) was the first to show
that cigarette smoke was capable of
depressing the self-cleansing ac-
tivity of the respiratory tract by
inhibiting the action of the ciliated
epithelium. Further investigation
(2) proved this inhibition to be
caused by certain components of the
gas phase. Consequently it became
urgent to gain more knowledge of
gaseous smoke components.
In 1962 I reported on the gas
chromatographic analysis of the gas
phase, inclu^]ing components with
boiling points in the range of 20-
100°C. (3). Thirty-six substances
were identified and quantitatively
estimated. Since then I have tried to
extend the range to include smoke
components with boiling points up
to 200°C, still keeping, as far as
possible, the principle of analysing
fresh smoke, i.e. avoiding any quali-
tative or quantitative changes due
to trapping, concentrating or sepa-
rating procedures. On the other
hand, analysis of the fresh and com-
plete gas phase presents severe
problems caused by four inherent
characteristics:
—the labile partition equilibrium
of many substances between gas and
particulate phase.
—the great number of compo-
nents: approximately 200 constitu-
ents of interest within the men-
tioned range.
—the differing chemical character
of the components: all basic func-
tional groups of organic chemistry
are present.
—the high dilution: about 99%
of the gas phase is not of interest
in this study, i.e. nitrogen, oxygen,
noble gases, hydrogen, water, carbon
monoxide and dioxide.
The prevalent aim of the analyti-
cal work therefore is maximum res-
olution at maximum sensitivity,
which led to the use of capillary
columns.
*presented in part at Tobacco Chem-
ists' Research Conference, Raleigh,
N.C., October 22, 1964.
 Methods
Production of the Gas Phase
Cigarettes from tobacco without
additives and without filter tips were
used. According to my earlier obser-
vation (4) that the concentration of
the gas phase produced at moderate
humidity is less dependent on small
fluctuations of tobacco moisture, the
moisture content of the cigarettes
employed in these experiments was
standardized by allowing the cigar-
ettes to stand over a saturated solu-
tion of calcium nitrate, i.e. in an
ambient of approximately 50°J rela-
tive humidity, instead of 65% as
normally used.
The smoking was performed ac-
cording to the international stand-
ard: one puff of 40 ml volume and
2 seconds duration per minute.
From the cigarette the smoke im-
mediately passed through a special
glass fibre filter (Phipps and Bird,
Richmond Va.) which retains the
particulate phase by practically
100?1. Eventual changes in the
composition of the gas phase caused
by the filtration step are discussed
later.
The sampling of the gas phase
for GLC can be accomplished in
two ways:
As described earlier, the puff to
be analyzed can be smoked directly
on the gas sampling valve of the
chromatograph. The filtered smoke
immediately passes to the sample
loop from which it is introduced
into the carrier gas stream. The ad-
''Ii
2 ^^ y E
i iï
30 W?-, 30 40 50
Figure 1. Gas Phase of Cigarette Smoke Separated on 50m. x 0.01 in. Column (see text for details).
vantage of this procedure is a time
interval of less than one second be-
tween smoking and the start of the
separation. Secondary reactions of
sensitive smoke components are
brought to a minimum that can
hardly be further reduced. The
practical importance of this time in-
terval is not yet sufficiently eluci-
dated. Secondly, the gas phase from
one puff or from one or several en-
tire cigarettes can be collected in a
large syringe and pushed through
the sampling valve. This method
provides the advantage of analys-
ing samples representing the aver-
age composition of one or several
puffs.
In the following experiments both
methods are used, according to the
particular purpose of the work.
GLC Columns.
In accordance with my earlier ex-
perience (3) stationary liquids of
high polarity are required to obtain
satisfactory resolution. For three
reasons this requirement is the
source of many difficulties:
—Coating of capillary columns
becomes increasingly difficult with
increasing polarity of the coating
material.
—High polarity unfortunately
means a small practical temperature
range of the stationary phase, the
lower limit being fixed by the
viscosity, the upper limit by the
vapor pressure of the liquid.
—The higher the polarity the
Jil
1
ii
eo
lower the thermal stability of the
liquid.
The Jatter two problems are of
special importnce in this case
since, as a conzequence of the very
diluted gas phase, detectors have to
be operated at maximum sensitivity.
Much effort was made to avoid
these difficulties by using liquids
with moderate polarity. But all at-
tempts with poly propylene glycol,
poly phenyl ether, polyesters, cyano
and fluoro silicones failed com-
pletely, although coating of capil-
lary columns is much easier with
these substances than with highly
polar ones. Tetraethylene glycol
dimethylether (TED) that was
best suited for the gas phase com-
ponents boiling below 100°C. is
too volatile to be used for higher
boiling substances. Unfortunately a
longer chain homologue of TED is
not commercially available. I there-
fore methylated*, in the classical
manher with silver oxide and
methyl iodide, a polyethylene
glycol (PEG) of 600 molecular
weight—as indicated by the sup-
plier. The columns coated with this
material had to be baked for at
least three days at 120°C. to achieve
a reasonable base line stability at
*Methylation has the same aim as
acetylation or silanization mentioned
previously (7), but provides better
chemical stability.
^ ^ r
^^^^ c é ÉL t
c isd i. -+
J. of G. C.—February, 1965 53
maximum sensitivity. This rough
treatment at the upper temperature
limit of the liquid obviously
damaged the column by causing
considerable dislocation of the
liquid film. Replacement of PEG
600 by PEG 1000 did not help be-
cause of excessive viscosity of the
latter at lower temperatures. Gel
filtration of commercial PEG 600
on Sephadex G 25 yielded at least
10°J polymer in the molecular
weight range of 100-400. I therefore
collected and methylated the first
75% of the effluent from the Sepha-
dex column. Capillary columns
coated with this material in ether
solution were ready for use after
baking overnight at 110°C.; they
provided the best resolution of all
liquids tested. By varying the meth-
ë glans fiber filter 25* c
glans fiber filter so•c
no filter
whole smoke en column
column femp.25°c
Figure 2. Three Methods of Separating
the Gas Phase from the Particulate (Col-
umn 50m. x 0.02 in.).
ylation rate, the polarity of the
liquid may be slightly altered. This
influences the sequence of column
effluents, allowing a choice of col-
umns which have the minimum of
coincident substances.
For identification purposes a
second column was required which
resolved the gas mixture in a basi-
cally different way. Since this study
was mainly concerned with identi-
fication in the high boiling part of
the gas phase, viscosity of the liquid
was a less restrictive factor for the
alternative column. With much suc-
cess, Dehydrophene C (Dehydag,
54 J. of G. C.—February, 1965
Düsseldorf), a non ionogenic de-
tergent was employed.
Stainless steel columns (0.01 in.
i.d., 50 m. long) were used for
identification purposes. As deter-
mined with nonpolar test substances
they showed 100,000-150,000 theo-
reretical plates. This high efficiency,
however, could not be fully utilized
since, due to the small volume of
the capillary, too much (approxi-
mately 0.02 ml) of the diluted gas
had to be introduced to get reason-
able response, even at maximum
sensitivity. Unfortunately a good
stainless steel capillary of 0.015 in.
i.d., which would be ideal for this
purpose, is not available.
The reproducibility of the runs
on 0.01 in. columns is not sufficient
to allow quantitative estimations.
0.02 in. columns were used for this
purpose, although resolution was
poorer.
A
Apparatus and Operation.
The decisive requirement in the
choice of apparatus in this study
was the necessity to convey the gas
sample on the capillary column
without further dilution. Single col-
umn and dual column chromato-
graphs from Carlo Erba, Milano
were used.
C
Flame ionisation detectors were
best suited for this purpose. More
sensitive detectors were not suc-
cessful since the restrictive factor
was column stability rather than
detector sensitivity. 0.01 in. col-
umns were run with a nitrogen flow
of about 1 ml per minute at 1.3
atmosphere inlet pressure. For 0.02
in. columns the flow was 5 ml per
minute at 0.3 atmosphere. The
sample volume of 0.5 was split in
the ratio of 25:1 for 0.01 and in the
ratio of 20:1 for 0.02 in. columns.
Column temperature normally
was programmed from 30 to 100°C.
A program rate of 3-4° per minute
yielded the best average resolution
in the different parts of the chro-
matograms. Two identical columns
were used to get a nearly horizontal
base line.
Resuits
Composition of the Gas Phase
Philipp et al. (5) have recently
presented a comprehensive survey
of the most volatile part of the gas
phase. The chromatogram in Fig. 1
on which this part is compressed to
about half a minute, gives a picture
of the rest of the gas phase. The
chromatogram omits only a few sub-
stances present with fairly measur-
able concentrations and with higher
boiling points. The first half, treated
already (3) , appears here in a con-
siderably different sequence, due to
the different stationary phase. One
important component of this section
not yet definitely identified in (3)
has come out as isopentenone, a
strong irritant and homologue to
butenone and acrolein. On this par-
ticular column it coincides with
acetonitrile.
Many substances of the second
half of the chromatogram have
already been reported as smoke
components e.g. by Johnstone, Quan
and Carruthers (8) and by Sted-
man and coworkers (9). I was able
to confirm the presente of almost all
these substances. No important
component has been overlooked by
these authors. A few additional com-
ponents, present in minor amounts,
have been identified by the same
methods described in (3) , essen-
tially by cochromatography on dif-
ferent stationary phases.
The following remarks with re-
spect to substances not indicated in
Fig. 1 may be added.
—Some very reactive smoke com-
ponents, such as glyoxal, diacetyl
and furfurol do not pass through the
steel column.
—Acidic and basic components
do pass through the column but are
not reasonably resolved. The broad
bend in the base line in the area of
pseudocumene, for instance, repre-
sents the "peak" of acetic acid!
—Low molecular weight alcohols
and esters are often reported as im-
portant constituents of the gas
phase (6). Except for the lowest
ones, e.g. methanol, methyl formate
and methyl acetate, this class of
substances plays a very modest role.
Several members could not be de-
tected on any column.
Much more important than alco-
hols and esters are aldehydes and
nitriles. The predominant compo-
nents of the second half in Fig. 1,
besides certain terpenes, are aro-
matics.
Semiquantitative estimation was
performed by preparing a mixture

of pure test substances yielding a Propionitrile 30 Pseudocumene 6

similar chromatogram to that ob- Butenone 28 Isobutanol <6

tained from the gas phase. Methylfuran 20 n-Butanol 5

Measured amounts of this mixture Isavateraldehyde 20 Anisole 5

were injected into a large syringe by Furan 18 n-Butyronitrile 5

a microsyringe and the mixture di- Dimethylfuran 16 n-Propanol 4

luted to 100 ml with nitrogen. The Crotonaldehyde 14 sec-Butanol <4

method provided a gaseous mix- Methyl propyl ketone 12 /3-Pinene 3

ture that could be handled in an Isobutyraldehyde 12 Hemimellitene 3

identical manner to the gas phase Diethyl ketone 12 3-Ethyl toluene 3

from smoke. Thus the chromato- n-Butyraldehyde 10 4-Ethyl toluene 2

gram from this mixture served as Methylacetate 8 n-Valeronitrile 2

calibration. Obviously safe results Methacrolein 8 m-Methylstyrene 2

are obtained only from peaks repre- n-Valeraldehyde 8 o-Methyl styrene 1

senting pure smoke components. Methyl isopropyl ketone 6 n-Butylacetate 1
Results are indicated in table 1. Ethylcapronate 1

Pentadiene 20 n-Capronitrile 1

Acrylonitrile 10 Mesitylene 1

Tetrahydrofuran 8 2-Ethyl toluene 1

Isobutyronitrile 8

Pivalaldehyde 4
Table I. Semiquantitative
Methylacrylate 3

Estimation of Gas Phase Methacrylonitrile 3

Cyclohexane 3
Components in Micrograms per
Ethylacetate
2 Distribution between Gas Phase
Cigarette. (Temp. of Glass Fibre Ethanol 2 and Particulate Phase.

Filter, 40° C.) Thiophene 2
The less volatile constituents of
Tetrahydropyran 2
the gas phase also appear in con-
Earlier Findings
densed state in the particulate
copied from (3) phase. Thus the composition of the
New Estimations
Acetone 360 gas phase is not a definite one but
Methanol 180 Isopentenone 45 depends on a partition equilibrium
Acetonitrile 140 Limonene 24 that can easily be influenced.
Toluene 80 m-Xylene 16 The most important influence is
Methyl ethyl ketone 80 Ethyl benzene < 14 temperature. I filtered the smoke
Acrolein 45 Capronaldehyde 12 from similar cigarettes through
Propionaldehyde 40 Styrene 10 filters of 25 and 60 °C. respectively.
Benzene 30 p-Xylene < 10 Chromatograms A and B in Fig. 2
Methylformate 30 o-Xylene 6 show that the temperature rise ap-

o , upper curve : 7th puff

luwer curve: 1 st puff

b
1

30 50 70 90 110C
average commercial charcoal filter E
Figure 3. Gas Phase of the First and Last Puff of a
Cigarette (Column 50m. x 0.02 in.).

Figure 4. Effect of a Commercial Charcoal Filter of Average

Efficiency (Column 50m x 0.02 in.).

J. of G. C.—February, 1965 55
proximately doubled the concentra-
tion of high boiling substances
whereas the low boiling ones were
not markedly increased. As an ex-
ception the behaviour of methanol
may be mentioned. The lower the
temperature, the lower the ratio be-
tween methanol and substances of
similar volatility.
Another influence may be con-
densation of the particulate phase.
It is conceivable that the condensed
particulate phase covering the glass
fibers of the filter does not act in
the same way as does the free par-
ticulate phase in the aerosol. This
influence, however, is difficult to
study. I introduced a sample of
whole, unfiltered smoke on a cold
column, thinking that this would
be the most direct separation of the
2 phases. In fact the result (Chro-
matogram C in Fig. 2) was quite dif-
ferent from that obtained by filter-
ing the smoke through a filter of the
same temperature. I am not able to
explain the effect in a simple way.
The result demonstrates the prob-
lem of dealing with the gas phase in
the same composition as it exists in
the free aerosol.
First and Last Puff of a Cigarette
The gas phases from the first and
the seventh pull (Fig. 3) show
similar differences to the gas phases
filtered by a cold and a heated
filter. In this case both influences
mentioned above may have func-
tioned.
The corresponding particular ob-
servation to that in Fig. 2 may be
56 J. of G. C.—February, 1965
stressed. The shorter the remain-
ing unburned cigarette, the higher
the relative concentrations of hy-
drophilic substances, such as meth-
anol and acetonitrile.
The Effect of a Charcoal Filter
The smoke from a cigarette of a
commercial brand with charcoal
filter was compared with that from a
cigarette of the same brand, from
which the filter tip was removed.
From both cigarettes 40 mm were
smoked and the smoke filtered
through a glass fibre filter of 25°C.
The effect of the charcoal filter as
shown by Fig. 4 does not need much
comment. The efficiency of the filter
used for this experiment is typical
of those available today. A real
specifity of absorption does not
exist. There is an increasing absorp-
tion with increasing boiling point.
Again typical hydrophilic sub-
stances show a particular behaviour.
Methanol and acetonitrile that are
strongly retained on unburned to-
bacco show the weakest absorption
by the charcoal filter.
It may be useful to add that the
gas phase cannot only be influenced
by filters, but depends qualitatively
and quantitatively on the tobacco it-
self.
Acidic and Basic Smoke Compo-
nents.
Undoubtedly the gas phase con-
tains acidic and basic substances.
For their separation on capillary
columns it is necessary to have sta-
tionary phases that possess either
strongly acidic or basic groupings.
These liquids, of course, must not
come into contact with substances
of the opposite group. This means
that the smoke has to be filtered
through a filter that has previously
been rendered acidic or basic. Un-
fortunately this treatment of the
filter brings about quantitative
changes in the substances that
should pass through. At present I
therefore doubt whether acidic and
basic smoke components can be
separated from fresh smoke, as is
possible for neutral components.
Literature Cited
1. Hilding, A. C., New Engl. J. Med. 254,
1155, (1956).
2. Kensler, C. J., New Engl. J. Med. 269,
1161, (1963).
3. Grob, K., Beitráge zur Tabakfor-
schung, 9, 315, (1962).
4. Grob, K., BeitrSge zur Tabakfor-
schung, 7, 285, (1962).
5. Philippe, R. J. et al. Anal. Chem. 36,
859, (1964)
6. "Smoking and Health," Report of
U.S. Dept. of Health, p. 52, (1964).
7. Grob, K., J. Gas Chromatog., 2, 80,
(1964).
8. Johnstone, R. A. W. et al. Nature,
195, 1267, (1962).
9. Stedman, R. L. CORESTA information
bulletin, 1963, No. 4, p. 11.
Manuscript received Nov. 16, 1964.