J. Aerosol Sci., 1976, Vol. 7, pp. 167 to 173. Pergamon Press. Printed in Great Britain.

A GAS PYCNOMETER (MICROLITER) FOR

DETERMINING THE MEAN DENSITY OF
ATMOSPHERIC AEROSOL PARTICLES

JiARG THUDIUM
Institut fur Meteorologie,Johannes Gutenberg-UniversitY,t, Mainz, G.F.R.

(First received 27 M a y 1975; and in revised form 6 August 1975)

Abstract--The physical principle of a gas pycnometeris explained with fundamental consider-

ations about the accuracy of measurement. For the first time a pycnometer has been devel-
oped with which volumes of ~2.5 mm3 can be determined within an error of 10%. Mean
densities of samples of dry atmospheric aerosol particles from various locations vary between
1.8 and about 3 g/cm3.

1. THE USE OF THE MEAN D E N S I T Y OF A T M O S P H E R I C AEROSOL

PARTICLES IN M E T E O R O L O G Y
Aerosol particles influence the radiation balance of the atmosphere. Cloud droplets
and ice crystals originate on them and, afterwards, elements of precipitation also. In
order to be able to calculate these influences and processes, knowledge of the size
distribution of the aerosol particles in dependence on the relative humidity is necessary
(thermodynamic equilibrium between the particles and their environment). For the com-
putation of radiation characteristics of the particles their complex refractive index,
depending on the relative humidity, is also needed.
A model showing the humidity dependence of size distributions and of complex refrac-
tive indices was developed by Hanel (1968, 1970). The following parameters requisite
for that purpose are determined by measurements on samples of particles:
The size frequency distribution of the aerosol particles at a known humidity, the
mass of a sample of particles as a function of the relative humidity at thermodynamic
equilibrium, the mean complex refractive index and the mean density of the aerosol
sample in a dry state. The humidity dependence of the size distribution of the aerosol
particles and of their mean complex refractive index can be computed from these four
measurements.
In this model, physical properties measured for samples are transferred to populations
of airborne aerosol particles. This requires the assumption that single dry particles are
the same as the dry sample in chemical composition and in structure. This assumption
has a basis in the fact that the single aerosol particles which are important for cloud
and radiation physics are each conglomerates of substances of different composition.
When the size distribution of aerosol particles at a known humidity is determined
with an impactor, a knowledge of the density of the collected particles is also necessary.

2. THE PHYSICAL P R I N C I P L E OF A P Y C N O M E T E R
2.1 Theoretical basis
The mean density of aerosol particles can be determined by measuring the mass
and the volume of a sample of particles. The volume of a body with no well-defined
geometry, as is the case with aerosol particles, can be determined with a pycnometer:
The body displaces a liquid or a gas, the so-called pycnometer agent, and its volume
can be determined from this displacement either directly or--using gas pycnometers--
from physical laws.
167
168 JtJRG THUDIUM

Only a gas pycnometer is suitable for finding the volume of a sample of aerosol
particles. The sample is not changed in chemical composition and in structure and
remains available for further investigations. Using a liquid agent, capillary gas compres-
sion can occur at the surface of the sample.
Volume determination by a gas pycnometer is based on the ideal gas law. A closed
system containing a gas, in which no phase transitions occur, is considered. Let
Z o = the initial value of a state variable of the gas (pressure p, temperature rE
volume V),
AZ = the change of a state variable and
AZ* = the ratio AZ/Zo.
The ideal gas law can be written in the form:

poVo (Po + Ap)(Vo + AV)

.... (1)
To To + AT
Transformations yield
1 + AT* = (1 + Ap*)(1 + AV*) (2)
and
AV*(1 + Ap*) = AT* - Ap*. (3)
Accordingly the initial volume V0 of the system can be computed with equation [3):
1 + Ap*
Vo = AV (4)
AT* - Ap*"
When a sample of the volume X is brought in the system and the system is afterwards
shut off again, then the volume of the enclosed gas is only Vo' = Vo - X. Now equation
(3) gives, analogous to equation (4):
1 + Ap*'
V~ = Vo - X = AV' AT*' - Ap*" (5)

In principle the volume of a sample can be determined from equations (4 and 5). For
physical and technical reasons some changes in state (or variables containing changes
in state) are reproduced at all prevailing pycnometer models; in particular, the tempera-
ture should be kept constant during the measurements.

2.2 Accuracy
The measuring accuracy of a gas pycnometer will now be discussed fundamentally.
Within a gas pycnometer, for each change of state there will be measured a change
of pressure or a change of volume or a measure for one of them, depending on the
volume of the sample brought in. The errors in these values have essentially the following
causes:
I. Changes in temperature,
2. Variations in those properties being reproduced at each measurement,
3. Error in the internal volume of the system due to opening and shutting the system
in a non-reproducible manner, and
4. Errors in reading.
For X ~ Vo, all these errors are practically independent of the volume of the sample.
In order to keep small the resulting error in the volume of the sample, it is necessary
to make the initial volume Vo of the system as small, and the changes in state during
the measurements as large as possible. This is common to all gas pycnometers. In
the latter case it must be considered that adsorption and desorption of gas molecules
at walls and at the sample increase with increasing changes in state. On the other
 Determiningthe meandensityof atmosphericaerosolparticles 169

hand, the ratio X/Vo, down to which the volume of a sample can still be determined
within a certain limit of error, is a measure for the technical quality of the apparatus.
Two essential phenomena originating in the measuring system have fundamental in-
fluence on the accuracy of measurement. They are independent of the actual conditions
of measuring technique. (1) In those instruments in which the absolute volume change
AV (Baranowski, 1973; Juda, 1966) or a variable containing AV (Krutzsch, 1954) is
reproduced, no exact linearity exists between the measured values and the volume of
the sample; (2) in those instruments consisting of one chamber only (Baranowski, 1973;
Juda, 1966), or having a second chamber as a pressure reference only (Krutzsch, 1954;
H~inel, 1972b), every fluctuation of temperature in the measuring chamber gives an
error of measurement; however in the systems with two chambers (Miiller, 1964) only
changes in the difference between the temperatures of the two chambers give an error
of measurement.

3. DISCUSSION OF A GAS PYCNOMETER FOR VOLUMES OF LESS THAN

FIVE MICROLITERS
According to statements by the users of previous gas pycnometers, the smallest
volumes could be determined by the apparatus of H~inel, some 20-40 mm 3 with an
error up to 10~o. However, shortening of the collecting time for better interpretation
of the results and the limitations of the flow rate through a collection apparatus which
should be portable for expeditions, the available samples of aerosol particles are very
small, especially when they are collected from rather clean air masses. Thus an apparatus
for volumes of less than 5 mm 3 has been developed. Its fundamental physical principle
is equal to that of the apparatus of Beckman (Miiller, 1964): the ratio Ap/po = Ap*
is reproduced at each measurement. But the changes in state are now caused by expan-
sion and not by compression of the chambers. Indeed it is inexplicable how in Beckman's
apparatus the changes in state are caused by compression. At the large volume changes
realized in this apparatus (AV = Vo/2), the error of measurement could be reduced
by half if an increase instead of a decrease of the volumes of the chambers by the
same amount were realized (then Vo would be half as large). Moreover, sorption effects
would be smaller.
Problems had to be solved for the development of this pycnometer which could
be disregarded in all previous models. These problems will be discussed at the end
of that chapter.
The development of a prototype (Thudium, 1973) led to an apparatus with which
some measurements on samples of aerosol particles were made (Fig. 1).
The apparatus consists of two systems, the sample system and the balancing system,
the latter having the initial volume Vr. These two systems can be shut off reproducibly.
Thus the initial volumes Vo of the sample system and VK of the balancing system are

vatves ~ [ ~ z ~ pressure sensor

/

sample ~ ('"") ~

\ ~ 'v,- / \ ~/ /
sample system IVo ) balancing system (VK)

Fig. 1.
170 ,It;RG Fl~:tq)it \~

constant. The sample is brought into the sample chamber. With the balancing chamber
the volume VK is approximated to V0 for reasons of measuring technique. A pressure
sensor measures the pressure difference between the systems. The volume of each of
the systems can be changed with a syringe whose piston is moved by a micrometer.
At the start of each measurement the pressures in the two systems are equal. The
volumes of both systems are increased, the volume of the balancing system always
by the same amount, AI/k, the volume of the sample system in such a way that after
the expansion the pressure difference between the systems again equals zero.
According to the ideal gas law

Po + Ap |~,
l'o l'k + At}< !0)
Hence it follows
A~
Ap* . . ~k. + . A~k constant. {7t

whence, using equations (4 and 5) and remembering that A T = 0:

Vo = cAVand I/i)' = cAV' (c = constant). (8)
Thus the following conditional equation for the volume of the sample results:
X = c A V - cAV' (AV' = AV'(X)), (9)
The apparatus is calibrated with the help of the linear equation (9).
As a result of volume changes, changes in temperature occur within the systems.
They are balanced by heat conduction. But this balancing proceeds somewhat differently
in the two systems, leading to a difference in temperature between them. On this account
the amount of the volume change of the sample system deviates from the value tbr
constant temperature. So far as this deviation proceeds from internal temperature in-
fluences described above, it is a linear function of the volume of the sample (nearly
constant for X ~ Vo), and is, consequently, independent of temperature and pressure.
It is included implicitly in the calibration line if the volume changes are always made
in the same manner.
Experience has shown that series of measurements have the smallest scatter if a com-
pression according to the same principle is accomplished after the expansion, and the
mean of these two measurements is taken. This compression is the reverse process
of the preceding expansion and does not change the initial volumes Vo and VK. Distur-
bances caused by external influences of temperature can be compensated most nearly
in that way.
Pressure changes realized in this apparatus come to approx. 10 mb only. Occurrence
of desorption or adsorption has no effect on the error of measurement.
The pressure balancing method has the advantages that
1. The measurements are independent of the pressure of the external air,
2. Only changes in the difference between the temperatures of the systems, but not
changes in temperature common to both of the systems influence the measuring error,
and
3. Regarding the measurement of pressure, it is a technically advantageous null method.
The following main problems had to be solved to obtain an accuracy of measurement
sufficient for volumes of less than 5 mm3:
l. The initial volumes of the systems must be as small as possible; in fact they are
very large considering the small volumes of the samples (some mmS); that is essentially
caused by the internal volumes of the pressure sensor. But on the other hand, the
high accuracy of the pressure sensor requires a certain dimension of its internal
volumes for technical reasons; it must be remembered that small changes in state
 Determining the mean density of atmospheric aerosol particles 171

within this apparatus are only practicable because of the very high accuracy of the
pressure sensor. Volumes down to the ratio X/Vo ~ 1:1300 can be measured with
an error in the volume of the sample up to 10~o.
2. Fluctuations of temperature inside the apparatus must be kept very small; fluctuations
of 10 . 4 °C are detectable. This requirement could be fulfilled with the help of a
number of alternating insulating and conducting layers. The apparatus stands in a
box in which temperature is kept constant within 0.1°C.
3. With strongly hygroscopic samples, above all, the desorption and adsorption of water
vapor caused important errors. They were eliminated by keeping the whole instrument
virtually free from water vapor (r.h. <0.0002%) and by drying out the samples with
the help of phosphorus pentoxide.
Absolute errors in volume of only 0.2-0.25 mm 3 are obtained, so that volumes of
2-2.5 mm 3 can be determined within 10%.

4. RESULTS
The calibration of the apparatus was accomplished with a substance which is not
hygroscopic and which has a similar surface structure to atmospheric particles: pulver-
ized elementary silicon. The result of the calibration was tested by measurements with
strongly hygroscopic potassium chloride. Measured densities of potassium chloride and
of several samples of dry atmospheric aerosol particles are shown in the following Table
1.

Table 1. Densities of KCI and of six samples of dry atmospheric aerosol particles
Error in p
No. Sample Mass (mg) Density p(g/cm3) (%)
1 KCI 7.28 2.00 +_0.13 6.6
2 Israel I 6.95 2.59 + 0.20 7.6
3 Israel II 2.25 2.72 _ 0.74 27
4 Mainz I 15.97 1.867 __+0,066 3.6
5 Mainz II 5.74 1.99 + 0,14 6.9
6 Jungfraujoch 3.37 2.87 + 0,59 21
7 Mace Head 18.67 1.93z + 0,039 2.0

R e m a r k s on Table 1.
No. 1. KC1 has a density of 1.984g/cm 3. The ascertained value of 2"004g/cm a is
only 1~o too large whereas the calculated error is nearly 7~.
Nos. 2 and 3. These samples were collected in the Negev Desert, "Israel I" during
15-24 May 1973 and "Israel lI" during 14-15 May 1973. "Israel II" was collected in
very hazy air masses containing particles of non-local origin (during the so-called
"Sharav'). During the collection of ~'Israel I', the air turbidity was very low. The absorp-
tion spectra for these samples (Fischer, 1975) indicated in both cases large parts of
silicates (with densities of about 2.5-3 g/cm3). For "Israel II" the absorption spectrum
was similar to that of quartz. The measurements of absorption of airborne aerosol
particles, made by Grassl at the same time and place, have yielded values which differ
for a density of 2.6 g/cm a at most by only 30% from Fischer's values for deposited
particles; that is a very good agreement in this case (Fischer and Grassl, 1975). A
significant difference between the densities of these two samples can not be ascertained
because of the small quantity of "Israel II'.
Nos. 4 and 5. These samples were collected at Mainz (University): "Mainz I" during
11-12 January 1970 in a persistent fog with weak circulation, a situation which devel-
oped in the night from 10th to 11 January 1970; "Mainz II" during 19-23 January
1974 in varying weather situations with frequent showers and moderate winds, mainly
westerly. For three days before collecting "Mainz II", strong north-westerly winds were
prevalent in Central Europe, and there were storm gusts on 16 and 17 January 1974.
172 Jrm(; THul)lt M

According to Heintzenberg the collecting station was temporarily in the emission region
of the heating-installation of the university; for that reason the mean density of "Mainz
II" has probably somewhat decreased (density of soot: less than 2 g/cm31.
No. 6. The sample of the Jungfraujoch (3600 m above MSL, Swiss Alps) was collected
during 4 March- 5 June 1972. The rocky outcrop of this region shows densities between
2.5 and 3.4g/cm 3. Absorption spectra of this sample~ measured by Fischer (1975), point
to mineral contents.
No. 7. This sample was collected in Mace Head on the west coast of Eire during
13 30 November 1971. It is mainly of maritime origin. There are two reasons for the
mean density being lower than that of NaCI and also of KCI: 1. The organic compounds
(according to Ketseridis (1975) a fraction of 10-15°~,, of the total mass for maritime
aerosol) are relatively light: the influence of a compound on the mean density is deter-
mined by the volume fraction. 2. Aerosol particles of continental origin were present
at least from time to time.
A measurement on a sample from Tsumeb, south-western Africa, has resulted in a
density about 2.5 g/cm 3. This value is uncertain because sorption effects were possibly
not yet eliminated.
Densities of samples of atmospheric aerosol particles have been determined by H~inel
(1968, 1972a, b) by an indirect method, on the basis of refractometer measurements,
and with the help of a gas pycnometer. In the first case the densities have errors of
15-,0J~,,
3 ° and the samples were destroyed after the measurements; in the second case
samples of more than 60 mg were needed. In the following these results ate summarized:
1. F r o m January till September 1966, nine samples of aerosol particles were collected
at Mainz (University) at different weather conditions. The densities of them were
determined by refractometer measurements. In six cases these samples showed
densities between 1.8 and 2.8g/cm 3 with errors between + 0.3 and + 0.9g/cm 3,
in three cases the densities amounted to 3.3 and 3.4 g/cm 3 with errors between ± 0.7
and ± 1.0g/cm 3. On the average, the largest densities showed the largest errors.
2. A sample was collected over the Atlantic in air masses containing Sahara dust during
16-25 April 1969. Its density, determined by refractometer measurements, was
2.6 ___0.2 g/cm 3.
3. During 13 16 January 1970, a further sample was collected at Mainz (University)
in a persistent fog with weak circulation (s. above: "Mainz I"). The density of that
sample, determined by a gas pycnometer, was 2.8 ± 0.3 g/cm 3. This value was con-
firmed by refractometer measurements.
4. The density (determined by a gas pycnometer) of a sample from the Hohenpeissenberg
(1000m above MSL, G e r m a n lower Alps), which was collected during 18 June 27
August 1970, was 1.8 ± 0.2g/cm 3.
The new and the old values for the mean densities of dry atmospheric aerosol particles
are rather consistent on the whole. They are spread over a wide range from 1.8 to
about 3g/cm 3. It seems that the higher densities are caused by silicates and other
minerals, the smaller ones by light salts (ammonium sulfate above all) and by organic
compounds.

Acknowledqements- I wish to express my sincere thanks to Mr. Bullrich and Mr. H~inet for the help accorded
to me and for the numerous discussions, as also to the German Research Association (Deutsche Forschungsge-
meinschaft Sonderforschungsbereich "Atmosph~irischeSpurenstoffe") which has sponsored this work.

REFERENCES
Baranowski, J. (19731 Determinations of the densities of small quantities of powder with the help of an
air pycnometer (English translation of the title) Ochrona Powietrza 2, 38.
Fischer, K. (1975) Strahlungsabsorption durch Aerosol in den Spektralbereichen atmosphiirischer Transparenz.
Dissertation, Universitiit Mainz.
Fischer, K. and Grassl, H. (1975) Tellus 27, (5), 522.
Hiinel, G. (1968) Tellus 20, 371.
 Determining the mean density of atmospheric aerosol particles 173

H~inel, G. (1970) Beitr. Phys. Atmos. 43, 119.

H~inel, G. (1971) Beitr. Phys. Atmos. 44, 137.
H~inel, G. (1972a) J. Aerosol Sci. 3, 377.
H~.nel, G. (1972b) Bestimmung physikalischer Eigenschaften atmosph~irischer Schwebeteilchen als Funktion
der relativen Luftfeuchtigkeit. Diss., Universit~it Mainz.
Juda, J. (1966) Staub 26, 197.
Ketseridis, G. (1975) Private communication, Max Planck-Institut, Mainz.
Krutzsch, J. (1954) Chemiker-Zeitum3 78, 49.
Mi~ller, G. (1964) E. Schweitzerbart'sehe Verlagsbuchhandlung, Stuttgart. Methoden der Sedimentuntersuchung,
pp. 258-259.
Thudium, J. (1973) Entwicklung eines Gaspyknometers zur Volumenmessung im Kubikmillimeter-Bereich
zwecks Dichtebestimmung an Proben atmosph~irischer Aerosolteilchen. Diplomarbeit, Universit~it Mainz.