Icarus 233 (2014) 101–105

Contents lists available at ScienceDirect

Icarus
journal homepage: www.elsevier.com/locate/icarus

Temperature dependence of the sublimation rate of water ice: Influence

of impurities
Konrad J. Kossacki ⇑, Jacek Leliwa-Kopystynski
Institute of Geophysics of Warsaw University, Pasteura 7, 02-093 Warsaw, Poland

a r t i c l e i n f o a b s t r a c t

Article history: The sublimation rate of ice is commonly calculated using the simple Hertz–Knudsen formula. This for-
Received 10 September 2013 mula is derived from the kinetic theory of gases and ignores microphysical processes determining the
Revised 22 January 2014 actual sublimation rate. The microphysical processes can be accounted for by including the temperature
Accepted 22 January 2014
dependent sublimation coefficient (Kossacki, K.J., et al. [1999]. Planet. Space Sci. 47, 1521–1530;
Available online 3 February 2014
Gundlach, B., Skorov, Y.V., Blum, J. [2011]. Icarus 213, 710–719). Kossacki and Markiewicz (Kossacki,
K.J., Markiewicz, W.J. [2013]. Icarus 224, 172–177) discussed to what extent inaccuracy of the simple
Keywords:
Hertz–Knudsen equation affects the calculated temperature of Comet 67P/Churyumov-Gerasimenko.
Comets
Comets, nucleus
Numerical simulations presented in Kossacki and Markiewicz (Kossacki, K.J., Markiewicz, W.J. [2013]. Ica-
Comets, composition rus 224, 172–177) indicate, that derivation of the temperature below the dust mantle from the measured
Ices water production rate ignoring temperature dependence of the sublimation coefficient can lead to an
underestimate of the temperature by more than 10 K. Thus, it is important to know the dependence
on the sublimation coefficient of the composition of the real cometary ice, which can be far from purity.
We intended to check whether a small amount of dissolved minerals can affect the temperature depen-
dence of the sublimation coefficient of ice. According to our experiments the answer is positive.
Ó 2014 Elsevier Inc. All rights reserved.

1. Introduction (Skorov and Rickman, 1995). In its original form the HK

Sublimation processes are significant for many materials of sci-
entific and industrial importance. Typically, the flux of sublimating
molecules and the recession rate of a surface due to sublimation
are calculated using different forms of the classical Hertz–Knudsen
equation (also called Hertz–Knudsen–Langmuir equation). Farther
on we call it the HK equation. For the recession rate it is written
 
dz a l 0:5
ðTÞ ¼ ðpsat  pÞ;
dt . 2pRg T
where T is the temperature on the sublimation surface, . is the den-
sity of the sublimin material, p is the pressure over the subliming
surface, psat is the equilibrium saturation pressure at temperature
T, l is the molar mass of ice, Rg = 8.314 J mol1 K1 is the universal
gas constant, and alpha is a coefficient which we investigate exper-
imentally (for sublimation of the water ice) in this work.
The HK equation is derived from the kinetic theory of gases
assuming an equilibrium distribution of molecular velocity
equation contains the dimensionless coefficient a ¼ 1. This
assumption is not sufficient to explain correctly the sublima-
tion-condensation phenomena, even for the pure materials.
Moreover, materials encountered in nature are never the pure
elements or the pure chemical compounds but they always con-
tain different contaminants. Influence of the volatile impurities
and/or the non-subliming (solid) contaminants in the material
can be taken into account by replacing a ¼ 1 by the temperature
dependent sublimation coefficient as being also function of the
ð1Þ
contents and the properties of the impurities. The influence of
the collisions and backscatter of molecules can be taken into
account by adding an additional multiplicative coefficient. In
this case the coefficient is called the return flux coefficient B.
Thus, Eq. (1) becomes
 
dz Bas ðTÞ l 0:5
ðTÞ ¼ ðpsat  pÞ; ð2Þ
dt . 2pRg T
or for the flux of subliming molecules

⇑ Corresponding author.  0:5

l
E-mail addresses: kjkossac@igf.fuw.edu.pl (K.J. Kossacki), jkopyst@mimuw. F s ðTÞ ¼ Bas ðpsat  pÞ: ð3Þ
edu.pl (J. Leliwa-Kopystynski). 2pRg T

http://dx.doi.org/10.1016/j.icarus.2014.01.025
0019-1035/Ó 2014 Elsevier Inc. All rights reserved.
102 K.J. Kossacki, J. Leliwa-Kopystynski / Icarus 233 (2014) 101–105
The aim of this work is an experimental investigation of the
dependence of as , on temperature, taking into account presence
of dissolved impurities. We intended to check whether a small
amount of dissolved minerals, less than 0.1% by mass, may have
noticeable influence on the temperature dependence of the sub-
limation coefficient of ice. The samples were produced of the
water ice made of the demineralized water, and of a spring water
with dissolved minerals. We performed series of measurements
of the sublimation rate at temperature within the range 190–
260 K.
2. Sublimation coefficient
Before describing the coefficient a for the water ice we present a
short chronological review of representative papers related to
determination of as for different materials.
2.1. Sublimation of different solids
Hansen et al. (1969) studied sublimation from oriented single-
crystals and from powdered lead selenide. Analysis of the data
showed that the sublimation coefficient is unity over the experi-
mental range 887–1052 K.
Ishiyama and Fujikawa (2004) studied the evaporation and con-
densation at an interface of argon vapor and its condensed phase.
They investigated the validity of kinetic boundary condition for
the Boltzmann equation, which prescribes the velocity distribution
function of molecules outgoing from the interface, by the numeri-
cal method of molecular dynamics. They found that the condensa-
tion coefficient is close to unity below the triple-point temperature
and decreases gradually as the temperature rises.
Butman et al. (2008) studied the sublimation of praseodymium
tribromide PrBr3 under Knudsen and Langmuir conditions. The
high-temperature mass spectrometry over the temperature range
804–957 K were applied. During sublimation from the open sur-
face of a single crystal, the sublimation coefficient of PrBr3 changed
abruptly at the polymorphic transition point (UCl3 – type low-
temperature polymorph transforms to PrBr3 – type high-temperature
polymorph, T = 934 ± 3 K). The change was reproducible in the
heating–cooling cycles.
We note, that comparison of numerical results presented by dif-
ferent authors is rather difficults since it is not clear whether they
described the values of as , or Bas .
2.2. Water ice
This sublimation of water ice was investigated many times,
either with respect to the sublimation, or condensation of ice
(Kramers and Stemerding, 1951; Knake and Stranskii, 1959;
Davy and Samarjai, 1971; Beckmann and Lacmann, 1982;
Kossacki et al., 1999; Gundlach et al., 2011) and other authors.
Unfortunately, theoretical calculation of the sublimation coeffi-
cient (Knake and Stranskii, 1959) is too computationally inten-
sive to be included into a model describing long term
evolution of comet nuclei. Thus, experimental values need to
be used.
According to Gadsden (1998) and Winkler et al. (2012) ‘‘for the
water ice the value a = 0.83 is commonly used’’. Unfortunately, it
is not clear, whether the mentioned authors took into account the
return flux of subliming molecules. Thus, the values can be either
as , or Bas . According to the kinetic theory of the non-equilibrium
boundary layer, the relative return flux of the water molecules to-
ward such a sublimating surface is 0.25 (Cercignani, 1981; Skorov
and Rickman, 1998). Hence, B = 0.75.
Kossacki et al. (1999) and Gundlach et al. (2011) investigated
temperature dependence of the sublimation coefficient in wide
ranges of the temperature. Kossacki et al. (1999) considered the
pure ice samples, while Gundlach et al. (2011) investigated the
ice samples covered by different layers of dust.
According to Kossacki et al. (1999) the temperature dependence
of the sublimation coefficient can be described by
    
1 1 T  a2 p
as ¼ 1  þ tanh a3 tan p  ; ð4Þ
a1 a1 273  a2 2
where a1 ¼ 2:342, a2 ¼ 150:5, and a3 ¼ 4:353.
Gundlach et al. (2011) investigated sublimation of a dust cov-
ered ice samples. The main reason was investigation of the diffu-
sion of subliming molecules through the layers of dust of
different properties. The authors have found, that the diffusivity
of dust is insufficient to describe decrease of the vapor flux when
compare to that predicted by HK equation for clean ice. Thus, they
proposed the new formula for the sublimation coefficient of ice
b1
as ðTÞ ¼ þ b4 ; ð5Þ
1:0 þ exp½b2 ðT 1  b3 Þ
where b1 ¼ 0:854, b2 ¼ 11:58  103 , b3 ¼ 4:99  103 , and
b4 ¼ 0:146.
Both formulas predict as ¼ 1 at temperatures lower than 195 K,
and as ¼ 0:15, at temperatures higher than 230 K. However, at
moderate temperatures where the sublimation coefficient features
the strongest temperature dependence the formulas predict sub-
stantially different values of as . The discrepancy between formulae
(4) and (5) can be the consequence of limited number of data avail-
able when the formula (4) was developed. It should also be noted,
that the qualitative temperature dependence of the sublimation
coefficient predicted by Kossacki et al. (1999) and Gundlach et al.
(2011) is in qualitative agreement with the theoretical predictions
for subliming argon (Ishiyama and Fujikawa, 2004).
3. Measurements
3.1. Procedure
Sublimation of ice from a sample can be measured in different
ways. Gundlach et al. (2011) determined the flux of water mole-
cules escaping the ice samples, using measurements performed
by the mass spectrometer. This method allowed building of auto-
matic measurement system, but produced very large scatter of
readings when ice slowly sublimed at low temperatures.
We decided to use another approach, direct non-contact mea-
surements of the position of the sample that recedes due to subli-
mation. Knowledge of the surface position versus time, when the
surface temperature is known makes possible calculation of the
sublimation rate versus temperature.
3.2. Experimental set-up
The surface temperature of the investigated samples is recorded
in two or three different points using resistance thermometers. The
temperatures are measured every minute. For each of the sensors
we create the separate profile of the temperature versus time.
We use a cylindrical vacuum chamber of 50 cm diameter and
50 cm height. A cooler in the form of an open cylinder, 20 cm high,
is located inside the chamber. Liquid nitrogen is pumped through
the cooler. Similar cooled chambers are used by Kossacki et al.
(1999). On the flat bottom of the cooler we placed samples within
plastic, or metal cylindrical holders up to 10 cm in diameter. Ther-
mal isolation of the samples at the sides is not needed because we
do not warm up the samples.
 K.J. Kossacki, J. Leliwa-Kopystynski / Icarus 233 (2014) 101–105
Above the cooler is installed a device to measure evolving posi-
tions of the surfaces of the samples. The distance to the sample
needs to be measured with accuracy of the order of 0.1 mm. Thus,
we used laser triangulation method. The laser beam is projected
onto the target surface, and the diffuse element of the reflection
of the light spot is imaged by a CCD receiver located at some dis-
tance to the optical axis of the laser beam. The incidence angle of
the received light depends on the distance to the scattering target.
From the output signal of the CCD element a digital signal proces-
sor (DSP) calculates the distance between the light spot on the ob-
ject being measured and the sensor. We use integrated device
Micro Epsilon ILD1700-200. It contains semiconductor laser with
power less than 1 mW, 670 nm (red), of the spot diameter
1300 lm. Accuracy of the measurements is 220 lm, and resolution
12 lm.
The surface temperature of the samples is measured by preci-
sion resistance thermometers, 100 X and 500 X connected to the
datalogger Agilent 34972A. We use 2, or 3 thermometers, located
close to the wall of a sample container. The spot of laser light is
roughly in the center of the samples. The distance between the
spot and the thermometers was 1–3 cm, depending on the size of
the sample.
3.3. Samples
We intended to answer the question, whether a small mass
fraction of the dissolved minerals can significantly affect the tem-
perature profile of the sublimation coefficient. Thus, we investi-
gated two types of samples: (a) frozen demineralized water, and
(b) frozen natural spring water Cisowianka. The latter was selected
due to relatively high mineralization, mass of dissolved compo-
nents 756 mg/l, and lack of added CO2 (see Table 1).
This selection is sufficient for our purpose, but not to investigate
the influence of particular components.
3.4. Data processing
From each of the experimental runs we get two types of data:
the profiles of the surface temperature versus time, and the profile
of the surface position versus time.
The thermal contact between the sensors and the ice is not al-
ways perfect and the sensors sometimes record different tempera-
tures. Cooling of the ice samples from the bottom means, that a
sensor which has poor contact with the sample is warmer than
the other. Thus, after each experimental run we compare the tem-
perature profiles and take for interpretation the coldest one.
Experimental runs take few hours and correspond to the
changes of the surface temperature up to few tens of Kelvins. Thus,
the recorded profiles of the surface position versus time are split
for sections satisfying two conditions:
(a) duration 30 min, and
(b) changes of temperature smaller than 10 K.
Table 1
Composition of the mineral water used to produce samples.
Component Content (mg/l) Component Content (mg/l)
Ca 128.3 HCO3 557.2
Mg 26.1 Cl <5
Na 9.6 SO4 <5
K <5 F <0.5
SiO2 (undissociated) 18.38
103
For each of the sections we calculate recession of the ice surface.
The dependence of the surface position z versus time is expressed
as
Z
F s ðTðtÞÞ
zðtÞ ¼ z0 þ dt; ð6Þ
.
where T is the measured surface temperature, and F s is the flux of
subliming molecules. We were able to reduce pressure p in the
cooled chamber with the sample inside to about 5  105 mbar.
The corresponding ratio p=psat ðTÞ was of the order 0.01. However,
when the chamber was not cooled to keep the sample at a temper-
ature about 260 K the pressure was about 1 mbar. The sublimation
coefficient as was considered, due to moderate changes of the tem-
perature in a section, as constant with the value of as being the
parameter. The calculations are repeated for different values of as .
The section n the value of as corresponding to the best fit of the cal-
culated zðtÞ to the data is considered as as ðT n Þ, where
T n ¼ 0:5ðT n1 þ T n2 Þ. The temperatures T n1 and T n2 correspond to
the beginning and the end of the section. In Fig. 1 we shown the
example set of data together with calculated profiles. It is shown,
how the quality of the fit can depend on the value of as . In most
cases we determined as with the accuracy 0.05–0.1.
4. Results
In Fig. 2 we show values of the sublimation coefficient derived
from the measured recession rates of the ice surface. The results
are for two types of the ice samples: (a) frozen demineralized
water, and (b) frozen natural spring water of high content of dis-
solved minerals. We have added fits to the data: curves calculated
according to the equation (Eq. (4)), proposed in Kossacki et al.
(1999), but with new coefficients. It can be seen, that the results
for the investigated ices are different. This is important, because
the amount of admixtures dissolved in the used water was very
low, less than 1 g/l. It can be also seen, that the curve for pure
ice is steeper than for impure one.
In Fig. 3 we shown results obtained for ice produced of demin-
eralized water, and in Fig. 4 for ice made of mineral water. In Fig. 3,
in addition to the data points are plotted curves as ðTÞ: (i) described
in the literature (Kossacki et al., 1999; Gundlach et al., 2011), and
(ii) calculated according to the equations proposed in (Kossacki
et al., 1999; Gundlach et al., 2011) (Eqs. (4) and (5)), but with
the new coefficients. The coefficients of Eqs. (4) and (5), corre-
sponding to the best fit to the data are presented in Table 2. The
2 270
Position of the surface
Measured
Calculated, αs = 0.7, p = 0.01
Calculated, αs = 0.8, p = 0.01
Calculated, αs = 0.9, p = 0.01
Position of the surface [mm]
1.5 T102
Temperature [K]
240
1
210
0.5
0 180
16.6 16.8 17 17.2 17.4
Time [hours]
Fig. 1. The example section of the data: the surface temperature TðtÞ, and position
of the surface zðtÞ. Added are the theoretical profiles zðtÞ calculated for different
values of as .
104 K.J. Kossacki, J. Leliwa-Kopystynski / Icarus 233 (2014) 101–105

Table 3
This work: The coefficients of Eq. (4) for the sublimation coefficient of H2O with minerals.
1 H2O, demineralized
fit to the data
H2O, mineral a1 ¼ 2:17  0:16
fit to the data
Sublimation coefficient α(T)

a2 ¼ 145:60  2:33
0.8 a3 ¼ 4:40  1:34

0.6

1.2
0.4 This work: data
1.1 fit, eq. 1

Sublimation coefficient α(T)

0.2 0.9

0.8

0.7
0
190 200 210 220 230 240 250 260 270
0.6
Temperature [K]
0.5
Fig. 2. Values of the sublimation coefficient derived from the measured recession 0.4
rate of the ice surface: (a) frozen demineralized water, and (b) frozen natural spring
0.3
water with 756 mg/l of dissolved minerals.
0.2

0.1
1.2
Kossacki et al. 1999 0
1.1 Gundlach et al. 2010 180 190 200 210 220 230 240 250 260 270
1 This work: data Temperature [K]
fit 1
Sublimation coefficient α(T)

fit 2
0.9
Fig. 4. Experimental results for the samples made of mineral water. In addition to
0.8 the values of as are presented curves as ðTÞ fitted to the data (Eq. (4) with new
coefficients). The coefficients are in Table 3.
0.7

0.6

0.5 Table 3. The coefficients a1 , and a2 are similar to these for pure
0.4 ice, but the coefficient a3 is two times smaller.
0.3

0.2
5. Summary and discussion
0.1

0 The main icy component of the cometary nuclei is the H2O ice.
180 190 200 210 220 230 240 250 260 270
However, it can be far from purity. Sunshine (2006) argued that the
Temperature [K] ice present on the surface of Comet Tempel 1 is thermally and
physically decoupled from the dust. However, this does not mean,
Fig. 3. Experimental results for ice made of demineralized water. In addition to the
values of as are presented curves as ðTÞ fitted to the data: fit 1 (Eq. (4) with new that this ice is clean. Spectral observations of the cometary comae
coefficients), and fit 2 (Eq. (5) with new coefficients). The latter is at low indicate presence of various gas molecules. Unfortunately, directly
temperatures nonphysically high, above 1. For comparison are drawn curves: observed are ions. The most abundant parent molecules are: CO,
as ðTÞ according to Kossacki et al. (1999), as ðTÞ according to Gundlach et al. (2011). CO2, CH4 and H2S, H2CO, CH3OH and NH3 (Bockelee-Morvan,
2011). Some of the molecules can be trapped and form clathrates,
but some (e.g. NH3) do not form clathrates. We investigated the
both formulas for the temperature dependence of the sublimation
temperature dependence of the sublimation coefficient for ice pro-
coefficient can fit to the data. However, at very low temperatures
duced of water: clean, and with dissolved minerals. We performed
only Eq. (4) predicts expected as ¼ 1. This suggests, that Eq. (5)
measurements of the sublimation rate at different temperatures,
can be used only when the available data cover the whole range
using new experimental approach.
of temperature. In the case of limited data set it seems better to
We intended to check, whether a small admixture of dissolved
fit Eq. (4). Another effect is, that the curves calculated using new
minerals can significantly affect the sublimation coefficient. Thus,
coefficients are significantly more steep than described in the ori-
we investigated samples of two compositions: pure ice, and frozen
ginal papers (Kossacki et al., 1999; Gundlach et al., 2011). The coef-
spring water with small concentrations of admixtures, about 0.1%
ficients of Eq. (4), corresponding to the best fit to the data are in
by mass. However, the temperature dependencies of the sublima-
tion coefficient are clearly different. At temperatures below 190 K
the sublimation coefficient as is in all cases equal to unity. At about
Table 2
The coefficients of Eqs. (4) and (5) for the sublimation coefficient of pure H2O, original 200 K the sublimation coefficient decreases versus temperature.
and fitted to the new data. This decreasing is less steep for the ice with impurities than for
the pure one. At high temperatures the values of as are: about
Eq. (4) Eq. (5)
0.14 for pure ice, and about 0.08 for ice with admixtures. However
Original New Original New it should also be noted, that our experimental setup is most suit-
a1 ¼ 2:342 a1 ¼ 2:33  0:08 b1 ¼ 0:854 b1 ¼ 0:13  0:03 able for measurements at temperatures about 200 K and we have
a2 ¼ 150:5 a2 ¼ 138:72  0:94 b2 ¼ 11; 580 b2 ¼ 18; 757  5926 very few data points at high temperatures, above 230 K. This
a3 ¼ 4:353 a3 ¼ 9:30  2:05 b3 ¼ 0:00499 b3 ¼ 0:00486  0:00027
makes our measurements at these temperatures less accurate than
b4 ¼ 0:146 b4 ¼ 0:88  0:18
at lower temperatures.
 K.J. Kossacki, J. Leliwa-Kopystynski / Icarus 233 (2014) 101–105
Comparison of the temperature profiles of the sublimation coef-
ficient with these results presented in Gundlach et al. (2011) indi-
cates, that: (a) our experimental technique gives the same results,
and (b) the mineral contaminants have significant influence on the
temperature dependence of the sublimation coefficient. So, the
more detailed searches concerning dependence of that coefficient
versus temperature and versus concentration of the impurities
(the mineral ones as well as the volatiles) are needed.
The main components of comets are dust and H2O ice. The dust/
ice mass ratio is commonly assumed to be 1. However, this is a
simplification due to lack of knowledge. Numerous observations
indicate presence of non-volatile dust particles on the surfaces of
comet nuclei, and in the comae (e.g. sample return mission Star-
dust). Observations dealing with volatiles indicate, that the pro-
duction rates of molecules CO, CO2, CH3OH, CH4, H2S, and NH3
relative to H2O are grater than 1% (Bockelee-Morvan, 2011). These
volatiles could be present as separate ices, or as impurities in H2O
ice. Unfortunately, it is not known, how impure H2O ice really is.
Numerical simulations indicate, that derivation of the temperature
below the dust mantle from the measured water production rate
ignoring temperature dependence of the sublimation coefficient
can lead to a significant underestimate of the temperature
(Kossacki and Markiewicz, 2013). Thus, further experiments
dealing with different concentrations of different minerals are very
important.
Acknowledgments
This work was partially supported by the Polish Ministry of
Education and Science (Grant 4036/B/H03/2010/39). We thank also
the Reviewers for their valuable suggestions.
105
References
Beckmann, W., Lacmann, R., 1982. Interface kinetics of the growth and evaporation
of ice single crystals from the vapour phase II. Measurements in a pure water
vapour environment. J. Crystal Growth 58, 433–442.
Bockelee-Morvan, D., 2011. An overview of comet composition. In: Cernicharo, J.,
Bachiller, R. (Eds.), IAU Symposium Proceedings Series, vol. 280, pp. 261–274.
Butman, M.F., Motalov, V.B., Kudin, L.S., Kryuchkov, A.S., Grishin, A.E., Krmer, K.W.,
2008. A jump change in the sublimation coefficient of the PrBr3 single crystal at
the polymorphic transition point. Russ. J. Phys. Chem. A Focus Chem. 82, 1972–
1974.
Cercignani, C., 1981. Strong evaporation of a polyatomic gas. Progr. Astron. Aeron.
74, 305–320.
Davy, J.G., Samarjai, G.A., 1971. Studies of the vaporization mechanism of ice single
crystals. J. Chem. Phys. 55, 3624–3630.
Gadsden, M., 1998. Noctilucent clouds seen at 60N: Origin and development. J.
Atmos. Solar-Terr. Phys. 60, 1763–1772.
Gundlach, B., Skorov, Y.V., Blum, J., 2011. Outgassing of icy bodies in the Solar
System – I. The sublimation of hexagonal water ice through dust layers. Icarus
213, 710–719.
Hansen, E.E., Munir, A.A., Mitchell, M.J., 1969. Sublimation pressure and sublimation
coefficient of single-crystal lead selenide. J. Am. Ceram. Soc. 52, 610–612.
Ishiyama, T.Y., Fujikawa, S., 2004. Molecular dynamics study of kinetic boundary
condition at an interface between argon vapor and its condensed phase. Phys.
Fluids 16, 2899–2906.
Knake, O., Stranskii, I.N., 1959. Mechanism of evaporating. Adv. Phys. Sci. 261, 6–88.
Kossacki, K.J., Markiewicz, W.J., Skorov, Y., Koemle, N.I., 1999. Sublimation
coefficient of water ice under simulated cometary-like conditions. Planet.
Space Sci. 47, 1521–1530.
Kossacki, K.J., Markiewicz, W.J., 2013. Comet 67P/CG: Influence of the sublimation
coefficient on the temperature and outgassing. Icarus 224, 172–177.
Kramers, H., Stemerding, S., 1951. The sublimation of ice in vacuum. Appl. Sci. Res.
A3, 73–82.
Skorov, Yu.V., Rickman, H., 1995. A kinetic model of gas flow in a porous cometary
mantle. Planet. Space Sci. 43, 1587–1594.
Skorov, Y., Rickman, H., 1998. Gas flow and dust acceleration in a cometary Knudsen
layer. Planet. Space Sci. 47, 935–949.
Sunshine, J.M. et al., 2006. Exposed water ice deposits on the surface of Comet 9P/
Tempel. Science 311, 1453–1455.
Winkler, R., Wang, L.Y., Lin, Y.H., Chu, C.S., 2012. Robust level coincidences in the
subband structure of quasi-2D systems. Solid State Commun. 152, 2096–2099.