Atmos. Chem. Phys., 17, 6291–6303, 2017
www.atmos-chem-phys.net/17/6291/2017/
doi:10.5194/acp-17-6291-2017
© Author(s) 2017. CC Attribution 3.0 License.
Evaporating brine from frost flowers with electron microscopy
and implications for atmospheric chemistry and sea-salt
aerosol formation
Xin Yang
1
, Vilém Ned
ˇ
ela
2
, Ji
ˇ
rí Runštuk
2
, Gabriela Ondrušková
3,4
, Ján Krausko
3,4
, L’ubica Vetráková
3,4
, and
Dominik Heger
3,4
1
British Antarctic Survey, Natural Environment Research Council, Cambridge, UK
2
Environmental Electron Microscopy Group, Institute of Scientific Instruments of the CAS, Brno, Czech Republic
3
Department of Chemistry, Faculty of Science, Masaryk University, Kamenice 5/A8, 625 00 Brno, Czech Republic
4
Research Centre for Toxic Compounds in the Environment (RECETOX), Masaryk University, Kamenice 5/A29,
625 00 Brno, Czech Republic
Correspondence to: Xin Yang (xinyang55@bas.ac.uk) and Dominik Heger (hegerd@chemi.muni.cz)
Received: 13 January 2017 – Discussion started: 25 January 2017
Revised: 29 March 2017 – Accepted: 18 April 2017 – Published: 23 May 2017
Abstract. An environmental scanning electron microscope
(ESEM) was used for the first time to obtain well-resolved
images, in both temporal and spatial dimensions, of lab-
prepared frost flowers (FFs) under evaporation within the
chamber temperature range from −5 to −18
◦
C and pressures
above 500 Pa. Our scanning shows temperature-dependent
NaCl speciation: the brine covering the ice was observed
at all conditions, whereas the NaCl crystals were formed
at temperatures below −10
◦
C as the brine oversaturation
was achieved. Finger-like ice structures covered by the brine,
with a diameter of several micrometres and length of tens to
100 µm, are exposed to the ambient air. The brine-covered
fingers are highly flexible and cohesive. The exposure of the
liquid brine on the micrometric fingers indicates a significant
increase in the brine surface area compared to that of the flat
ice surface at high temperatures; the NaCl crystals formed
can become sites of heterogeneous reactivity at lower tem-
peratures. There is no evidence that, without external forces,
salty FFs could automatically fall apart to create a number of
sub-particles at the scale of micrometres as the exposed brine
fingers seem cohesive and hard to break in the middle. The
fingers tend to combine together to form large spheres and
then join back to the mother body, eventually forming a large
chunk of salt after complete dehydration. The present mi-
croscopic observation rationalizes several previously unex-
plained observations, namely, that FFs are not a direct source
of sea-salt aerosols and that saline ice crystals under evapora-
tion could accelerate the heterogeneous reactions of bromine
liberation.
1 Introduction
Ice and snow constitute an important reaction medium on
Earth and are known to accumulate and concentrate signif-
icant amounts of impurities that are stored, transformed, and
eventually released. The knowledge of the exact location and
speciation of these chemical impurities in ice and snow un-
der various environmental conditions is crucial for assessing
their reactivity (McNeill et al., 2012; Bartels-Rausch et al.,
2014; Gudipati et al., 2015) and further fate.
The ions originating from sea salt (including, for example,
Na
+
, Cl
−
, and Br
−
) have been widely observed in polar re-
gions in media such as aerosols, snow packs, and ice cores
(DeAngelis et al., 1997; Rankin and Wolff, 2003; Fischer et
al., 2007; Legrand et al., 2016). The sea salts trapped in snow
packs form a large chemical reservoir and therefore embody
a significant part of chemical reactions in the polar bound-
ary layer (Abbatt et al., 2012). Conversely, inactive ions such
as Na
+
recorded in ice cores could serve as a palaeoclimate
proxy for the past climate (Rankin and Wolff, 2003; Abram
et al., 2013). Although the sea spray and bubble bursting in
the open ocean surface dominate sea-salt aerosol (SSA) pro-
duction on most of Earth, the winter SSA peaks observed at
Published by Copernicus Publications on behalf of the European Geosciences Union.
6292 X. Yang et al.: Evaporating brine from frost flowers with electron microscopy
most near-coastal sites in polar regions (Wagenbach et al.,
1998; Rankin et al., 2004) are clearly out of phase with the
distance to the open water. Several lines of evidence suggest
that winter sea salt cannot derive only from the long-range
transport of the aerosol produced over the open ocean. The
winter maximum observed seems inconsistent with the fact
that the nearest open water is hundreds of kilometres further
away in the given season because of extended sea ice. In ice
cores, significantly higher concentrations of salts are found
in glacial periods, when sea ice was even more widespread
and furthermore when relevant models do not suggest any
greater transport (Mahowald et al., 2006). The most direct
evidence of the salt that should originate from zones covered
with sea ice arises from the composition of sea-salt aerosol
and ice cores. Frequent episodes when the sulfate / sodium
[SO
2−
4
/ Na
+
] ratio is below that of seawater, despite the
addition of the non-sea-salt sulfate resulting from the oxi-
dation of dimethlysulfide, are observed (Wagenbach et al.,
1998). This is believed to occur due to the effect of mirabilite
(Na
2
SO
4
.10H
2
O) precipitating from the brine when the tem-
perature drops below −6.4
◦
C (Wagenbach et al., 1998; Jour-
dain et al., 2008 ; Butler et al., 2016b; Marion et al., 1999), a
segregation inapplicable to sea spray particles.
The sea ice microstructure is permeated by brine chan-
nels and pockets that contain concentrated seawater-derived
brine. Cooling sea ice results in further formation of pure
ice within these pockets as thermal equilibrium is attained,
resulting in a smaller volume of increasingly concentrated
residual brine (Light et al., 2003; Butler et al., 2016b). A
fraction of such concentrated brine will be expelled upwards
to form a thin layer of brine on the sea ice surface, where frost
flower (FFs) can grow under a certain weather condition. The
formation of mirabilite results in removing the major portion
of the dissolved SO
2−
4
from the brine, with less effect on the
Na
+
due to its large abundance compared to the sulfate (e.g.
Butler et al., 2016b). The SSA produced from these resid-
ual brines consequently displays a depleted [SO
2−
4
/ Na
+
] ra-
tio. However, for sea spray particles, the Na
2
SO
4
will not be
fractionated in the atmosphere or the following deposition,
even when these particles are exposed to sub-zero tempera-
tures: the precipitated mirabilite remains within the body of
the aerosol and has no effective pathway to escape.
FFs are commonly observed on fresh sea ice and pref-
erentially grow on small-scale roughness nodules sticking
above the surface or out of the brine, which is typically
colder by 5
◦
C compared to bulk ice (Domine, 2005; Galley
et al., 2015); at these conditions, the supersaturation of wa-
ter vapour is frequently achieved (Style and Worster, 2009).
Frost flowers often consist of feather-like dendritic ice crys-
tal structures, and their surface can be covered by concen-
trated brine (Perovich and Richter-Menge, 1994; Barber et
al., 2014; Galley et al., 2015). A detailed chemical composi-
tion analysis was performed, finding, inter alia, that FFs can
reach the salinity of the concentrated brine of 120 practical
salinity units (Douglas et al., 2012), which is in the effective
range of the mirabilite precipitation (Butler et al., 2016b).
FFs have the specific surface area of 185 (+80–50) cm
2
g
−1
,
measured by methane adsorption; such a specific surface area
is about 5 times lower than that of freshly fallen snow. The
surface area of FFs is estimated to be 1.4 m
2
per m
2
of ice
surface (Domine, 2005). The fragile structure plus extremely
high brine salinity (Rankin, 2002) make FFs the likely cause
of chemical reactions (e.g. heterogenous, photochemical, and
redox; Perovich and Richter-Menge, 1994; Kaleschke et al.,
2004; Simpson et al., 2007) and source for SSA (Wagen-
bach et al., 1998; Wolff et al., 2003). However, recent stud-
ies propose that FFs are not as important as assumed pre-
viously (Obbard et al., 2009; Roscoe et al., 2011; Abbatt et
al., 2012). In particular, a recent wind tunnel experiment in-
dicated that FFs are not a direct source of SSA (Roscoe et
al., 2011). Apart from saline FFs, the snow lying on sea ice
can be contaminated by seawater (or saline) through various
pathways (Domine et al., 2004). These contaminated salty
snows have been hypothesized to act as an efficient source
of SSA (via blowing snow) and bromine (Yang et al., 2008;
Legrand et al., 2016; Zhao et al., 2016; Levine et al., 2014).
The relative importance of these two sea-ice-sourced SSA
to the polar winter sea-salt budget is still under debate (e.g.
Huang and Jaeglé, 2017; Xu et al., 2016; Rhodes et al., 2017).
In any case (FFs or salty snow), the formation of SSA from
salty ice particles requires its size to be reduced via the loss
of water through either the evaporation or the sublimation
processes, depending on the temperature. Until now, there
was no detailed image at the microphysical scale to indicate
what happens to saline ice under evaporation or sublimation.
Moreover, current atmospheric chemical models consider the
solutes’ impurities on ice to be present in a diluted liquid
solution on the ice surface (Domine et al., 2013). Such a
model is generally unsatisfactory in describing the real sit-
uation, and thus more realistic parameters for modelling are
needed. Some of us previously showed that the concentration
increase of nonpolar (Heger et al., 2011; Kania et al., 2014;
Krausko et al., 2015a, b) and polar compounds (Heger et al.,
2005, 2006; Heger and Klan, 2007; Krausková et al., 2016)
can even lead to their crystallization under certain conditions.
In this study, we grew FFs in a laboratory and inspected
them using an environmental scanning electron microscope
(ESEM) to obtain some information about the state of im-
purities in/on the ice. The preparation of the FF samples to
mimic the FFs naturally produced on sea ice is detailed in
Sect. 2 together with the related information on the ESEM.
The scanning results are presented in Sect. 3, the atmospheric
implications are discussed in Sect. 4, and the conclusions are
available in Sect. 5.
Atmos. Chem. Phys., 17, 6291–
6303, 2017 www.atmos-chem-phys.net/17/6291/2017/
X. Yang et al.: Evaporating brine from frost flowers with electron microscopy 6293
Figure 1. A frost flower (FF) grown in a polystyrene-isolated beaker
in a walk-in cold room, at the temperature of −30
◦
C. Both pure wa-
ter FFs and saline FFs were prepared for further microscopic scan-
ning (see the text for details).
2 Methods
2.1 Growth of the frost flowers and preparation of the
samples
The FFs were prepared in a custom-built 2 m × 2 m walk-
in cold chamber. Inspired by the natural condition at which
FFs grow (Style and Worster, 2009) and exploiting previ-
ous methods of preparation (Roscoe et al., 2011), we cooled
the walk-in cold chamber down to −30
◦
C and inserted ves-
sels containing pure water or an aqueous solution of NaCl
(3.5 % w/w, similar to that of seawater) at 20
◦
C. The ves-
sels were isolated with styrofoam to minimize the contact
cooling of the solution by the floor of the walk-in chamber
and to promote cooling by the air. We typically observed
the following course of events: first, hoarfrost appeared on
the sides of the beaker; then, an ice crust formed on the wa-
ter level; subsequently, dendrite-shaped icy features (consid-
ered to be FFs) grew gradually, as shown in Fig. 1. After
the ice reached a certain thickness, the FFs stopped grow-
ing and were collected into a pre-cooled vial to be stored at
the temperature of liquid nitrogen. Care was taken to collect
only the FFs from the ice surface, avoiding the hoarfrost con-
densed on the walls of the beaker. The FFs were fragile and
fragmented during the manipulation. The FFs grown on the
surface of pure water were powdery; however, those grown
from the brine were sticky, and therefore two spatulas were
needed to place them into the vials. We attempted to follow
growth conditions similar to the natural ones; our sampling
Figure 2. The ANSYS Inc. fluent-based simulation of the water
vapour velocity distribution and the direction of the vapour flow in
the vicinity of the sample surface in the specimen chamber of the
applied ESEM AQUASEM II.
method guarantees that the features were grown on the ice
surface, and thus the examined samples are believed to be
very similar to natural FFs.
2.2 Environmental scanning electron microscope
The ESEM (AQUASEM II) is unique in the observation of
nonconductive, wet, or liquid samples, with the specimen
chamber pressure as high as 2000 Pa and temperatures rang-
ing from 0 to −30
◦
C (Tihlarikova et al., 2013). The indicated
temperature is measured on the sample holder. The tempera-
ture of the ice surface is estimated to differ by no more than
2
◦
C from that of the holder on which the temperature is mea-
sured. This estimate is based on the observation of the ice
surface melting. The major source of the heat is the energy
from the electrons used for scanning.
The conditions inside the chamber allow for the obser-
vation of ice samples in conditions similar to those under
which ice and snow occur naturally. No conductive coating
of the sample is needed, because the positive ions resulting
from the electron–gas ionization in high-gas-pressure condi-
tions of the ESEM discharge the accumulated charge. The
strength of this apparatus lies in the delicate control of the
dynamic conditions in the specimen chamber via an origi-
nally designed hydration system enhanced with temperature
and vapour flow control and an advanced cooling system in-
tegrated in the sample holder. The specimen chamber can be
evacuated very slowly, with the possibility of reaching high-
humidity conditions in the sample vicinity without purge–
flood cycles (Ned
ˇ
ela et al., 2015). The water vapour temper-
ature is estimated to be around 10
◦
C. Care was taken to di-
rect the steam away from the sample to prevent any heat-up.
The regulation of the temperature in the vicinity of the sam-
ple allows us to study ice in precisely controlled conditions
(Krausko et al., 2014). The temperature, pressure, and rela-
tive humidity in the chamber of the ESEM can be set close
www.atmos-chem-phys.net/17/6291/2017/ Atmos. Chem. Phys., 17, 6291–
6303, 2017
6294 X. Yang et al.: Evaporating brine from frost flowers with electron microscopy
Figure 3. The dynamical in situ images of the formation of brine
fingers during slow evaporation of water from the frost flower. The
individual fingers bending and flapping around are highlighted in
circles. The width of the seven indicated necks in Fig. 3f is mea-
sured as d
I
= (2.23 ± 0.43) µm, mean ± standard error of the mean.
Imaged with the applied ESEM AQUASEM II; beam energy at
20 keV, ionization detector, water vapour pressure of 348 Pa, sample
holder temperature of −5.2
◦
C, and sample-to-aperture distance of
2 mm. Scale bar: 100 µm. A video of this case is attached (S1).
to the frost point to cause ice sublimation or gradual growth.
The ESEM is equipped with a tungsten hairpin cathode as a
source of electrons and also with two custom-built detectors
(Ned
ˇ
ela et al., 2011): an ionization detector for secondary
electrons (surface sensitive to provide information about the
morphology of the ice surface) and a highly material sen-
sitive detector of backscattered electrons. A comparison of
these two modes on identical samples yields complemen-
tary information on the morphology of the ice surface and
ice grain boundaries contaminated by impurities.
As shown in Fig. 2, water vapour flows around the sample
and through the detector’s aperture during the scanning of
the sample. The flow speed varies from 2 m s
−1
on the sam-
ple surface to 16 m s
−1
at the distance of 0.7 mm above the
sample surface (simulated for the experimental pressure of
300 Pa in the specimen chamber of the ESEM AQUASEM
II and for the spherical shape of the sample). The flow is
influenced by the shape of the sample, pumping speed, and
ESEM aperture diameter. The flow speed was simulated as
described previously (Maxa, 2011, 2016).
3 Results and discussions
3.1 FFs at a high temperature: brine fingers formation
The FFs were scanned at the chamber temperature of
−5.2
◦
C. Figure 3 shows many spikes sticking out from the
main ice body; here, these will be referred to as fingers. The
smooth texture is indicative of surfaces covered with a layer
of a solution in contrast to the dry ice crystal surface observed
at temperatures below −30
◦
C and pressures below 50 Pa
(McCarthy et al., 2007; Blackford, 2007; Pfalzgraff et al.,
2010; Bartels-Rausch et al., 2014). The image differs from
that of a water drop also in the irregular and non-spherical
features. Thus, we are of the opinion that the exposed finger-
like spikes consist of ice covered with brine. More arguments
to support this interpretation will be proposed in the follow-
ing parts of the text. The brine is expected to become more
concentrated as a result of the loss of water during progres-
sive evaporation. The exposed thin fingers can be as much as
100 µm long, still remaining quite cohesive and hard to break.
In Fig. 3f, we estimate the thickness of the fingers’ necks
at their most narrow points to be d = (2.23 ± 0.43) µm; the
given values are mean ± standard error of the mean. In some
cases, a rounded sphere appeared on the top of a finger during
evaporation, as encircled in Fig. 3. At temperatures exceed-
ing ∼ −10
◦
C, which is well above the eutectic point temper-
ature (T
Eutectic
= −21.21
◦
C; Brady, 2009), the concentrated
brine was always observed as liquid, and no NaCl crystals
were perceived. The viscosity of the concentrated brine at
the 20
◦
C is not even 2 times higher than that of pure water
(Weast et al., 1987). Although we did not find any reference
to the values of the brine viscosity at sub-zero temperatures,
the viscosity of seawater at zero temperature is only slightly
higher than that of pure water (1.3 times; Sharqawy et al.,
2010), and the viscosity of supercooled water at −17
◦
C is
only 3.8 times larger compared to that at 20
◦
C (Dehaoui et
al., 2015). Therefore, we do not assume that the viscosity of
the brine will increase significantly enough to be the only ex-
planation for the formation of the fingers. The fingers were
observed to easily bend and flap following the airflow in the
chamber (Fig. 3, oval, and Supplement S1). When these fin-
gers are close enough to one another, they may tangle to-
gether to join into a larger one.
The relative humidity in our experiments was set to be
slightly below the frost point, and therefore slow loss of the
water from the sample could be observed. Thus, the micro-
graphs obtained already at the beginning of the observations
are not fully undisturbed; we assume that the water evapo-
rates faster from the brine of a lower concentration compared
to the more concentrated one (in accordance with Raoult’s
law). The vapour pressures above the water, ice, and satu-
rated brine (8.3 % w/w) at −5
◦
C are 422, 402, and 403 Pa,
respectively. These values were calculated from the applied
equations for the vapour pressure above the water and ice as
adopted from Buck (1981); for the brine, the relevant formu-
lae are proposed within the article by Perovich and Richter-
Atmos. Chem. Phys., 17, 6291–
6303, 2017 www.atmos-chem-phys.net/17/6291/2017/
X. Yang et al.: Evaporating brine from frost flowers with electron microscopy 6295
Figure 4. The phase diagram for a water–NaCl system. Indicated
(red and green arrows) are our experimental conditions at about
−5 and −17
◦
C. HH stands for hydrohalite (NaCl.2H
2
O). Based
on equations from Brady (2009).
Menge (1994):
e
w
=
h
1.0007 + 3.46 × 10
−6
p
i
×
6.1121 × e
17.966×t
247.15+t
e
i
=
h
1.0003 + 4.18 × 10
−6
p
i
×
6.1115 × e
22.452×t
272.55+t
e
b
= e
w
(1 − 0.000537 × S
b
),
where e
w
is the saturation vapour pressure above the water, e
i
is the saturation vapour pressure above the ice, e
b
is the satu-
ration vapour pressure above the brine, p is the atmospheric
pressure in millibars, S
b
is the brine salinity in parts of mass
per thousand, and t is the temperature in
◦
C. In an additional
experiment with pure water FFs (not shown here), we found
out that they sublimate markedly faster than brine-covered
FFs.
At the temperature of −5.2
◦
C and concentration of NaCl
lower than 8.3 % (w/w), the phase diagram (Fig. 4) indicates
the presence of a liquid solution of NaCl and ice. Therefore,
if the equilibrium conditions are established, there will be ice
and ca. 8.3 % NaCl solution covering its surface. As the water
is gradually evaporated from the brine, the ice must melt to
maintain the equilibrium concentration. This process is rep-
resented with the red arrow in the phase diagram of Fig. 4.
This rationalizes well our observations: the evaporation of
the water from the brine on the fingers causes its concentra-
tion to increase above the equilibrium concentration; there-
fore, the water must be supplied from the ice body towards
the brine fingers to dilute the brine. This process results in
gradual melting of the ice body until all the ice is melted.
An examination of the sequences of the micrographs sug-
gests that the evaporation proceeds faster from the main ice
body than from the fingers. This can be seen in the video
Figure 5. The dynamical in situ micrographs of a large (∼ 100 µm)
brine-covered piece of ice formation and breakaway during slow
evaporation of water from the frost flower. Imaged with the ESEM
AQUASEM II; beam energy at 20 keV, ionization detector, water
vapour pressure of 348 Pa, sample holder temperature of −5.2
◦
C,
and sample-to-aperture distance of 2 mm. Scale bars: 100 µm. A
video of this case is attached (Supplement S2).
of Supplement S1 as the fingers exhibit a relatively stable
shape even if the main ice body gradually abates. Thus, the
concentration of the brine in the surface layer of the fingers
is deemed to be higher than that on the main body. We can
speculate that the higher concentration of NaCl on the fin-
gers is a result of previous water vapour evaporation from
the brine on the fingers. Possibly, the most concentrated so-
lution is found on the tips of the fingers, where small spheres
are sometimes formed. The increased local concentration of
salt would effectively lower the water vapour evaporation
and hence reduce further melting of the ice forming the fin-
gers’ interior, thus not allowing its breakaway from the main
body. For example, if the NaCl saturation concentration of
25 % (w/w) is reached at −5
◦
C, the water partial pressure
drops to 365 Pa from the 403 Pa at the brine equilibrium con-
centration (8.3 %).
A particular consequence of a higher rate of water evapo-
ration from the side wall of a finger and the main ice body
compared to the fingertip is the formation and propagation
of gulfs. This is well exemplified in Fig. 5, where the process
resulted in the breakaway of large pieces of ice (> 100 µm)
from the mother body. First, a very deep gulf was formed
which later separated the two pieces by a very thin neck,
eventually leading to the breaking off of the two parts (Sup-
plement S2). This case indicates that the evaporation or sub-
limation process indeed could cause a large ice particle to
fall aside, but this phenomenon is not common, as we no-
ticed it only once in all our observations (20 experiments).
Moreover, there is no evidence that the brine fingers can fall
apart to form a number of micrometre-sized particles.
The understanding of the structure of FFs is still far from
complete. The 3-D X-ray micro computer tomography ex-
periments suggest that salt impurities are present mostly on
the ice surface (Hutterli et al., 2008). Such a finding is con-
sistent with our observation and can be well understood, tak-
www.atmos-chem-phys.net/17/6291/2017/ Atmos. Chem. Phys., 17, 6291–
6303, 2017