Heterogeneous freezing of super cooled water droplets in micrometre range-freezing on a chip

A new setup to analyse the freezing behaviour of ice nucleation particles (INPs) dispersed in aqueous droplets has been developed with the aim to analyse ensembles of droplets with sizes in the micrometre range, in which INPs are immersed. Major disadvantages of conventional drop-freezing experiments like varying drop sizes or interactions between the wateroil mixture and the INP, were solved by introducing a unique freezingchip consisting of an etched and sputtered 10 15x15x1 mm gold-plated silicon or pure gold film. Using this chip, isolated micrometre-sized droplets can be generated with sizes similar to droplets in real world clouds. The experimental set-up for drop-freezing experiments was revised and improved by establishing automated process control and image evaluation. The new set-up is economical, quick in handling the sample, precise in measurement and the results are more next to real conditions than former approaches. We were able to show the efficiency and accuracy of our setup by comparing measured freezing temperatures of different INPs (Snomax, K15 feldspar, birch pollen (Betula pendula) washing water, juniper pollen suspension (Juniperus communis) and ultrapure water) with already published results. The T50 values of ultrapure water (T50=-37,2°C), birch pollen washing water (T50=-18°C) and juniper pollen (T50=-22,7°C) match the data given in literature. Microcline shows higher freezing temperatures (T50= 16,4°C) than literature values from us and others, which can be explained by different preparing/milling parameters. The slightly lower freezing temperature of Snomax (T50=-8,9°C) received by using the freezing-chip, compared to 20 measurements already published, can be explained by different concentrations and droplet sizes. Our measurements and comparisons with the literature data show the important impact of droplet size, INP concentration and number of active sites on the T50 values. Here, the new set-up exhibits its strength in reproducibility and accuracy which is due to the defined and isolated droplets. Finally, it opens a temperature window down to -37°C for freezing experiments which was not accessible with many former approaches and allows determination of IN also with weak nucleation activity. 25 1 Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2017-31, 2017 Manuscript under review for journal Atmos. Chem. Phys. Published: 16 January 2017 c © Author(s) 2017. CC-BY 3.0 License.


Introduction
The influence of clouds on the Earth's climate system is well investigated (Solomon, 2007, S. 8).Cloud microphysics determines for example cloud lifetime and precipitation properties.Clouds cool the climate system by reflecting incoming solar radiation and warm it by trapping outgoing infrared radiation (Baker and Peter, 2008).In all these processes, aerosol particles play a crucial role by acting as cloud condensation nuclei (CCN) for liquid droplets or as ice nuclei (INPs) for the formation of ice particles.INPs can be as small as a few nanometres and range up to several micrometres.They can be produced by natural processes, such as forest emissions, volcanic emissions, suspension of mineral dusts, and anthropogenic processes, such as burning of wood and fuels (Petters et al., 2009).Desert dusts (Field et al., 2006), volcanic ash (Bingemer et al., 2012), as well as biological particles (Despres et al., 2012), including non-proteinaceous (Hiranuma et al., 2015;Krog et al., 1979) and proteinaceous (Maki et al., 1974) particles, are known as efficient INPs (Atkinson et al., 2013).
In the atmosphere, ice crystals form through heterogeneous and homogeneous ice nucleation processes.For homogeneous nucleation, a temperature below -38°C is needed.Freezing processes at higher temperatures occur heterogeneously (Pruppacher and Klett, 1997).For heterogeneous ice nucleation, INPs are essential to provide a specific surface area which reduces the energy barrier of nucleation kinetics.Therefore aerosol particles act like a catalyst triggering the freezing process.In mixed phase clouds, INPs can cause glaciation temperatures higher than -15°C under certain conditions.Furthermore, biological INPs like Snomax ® , which consists of fragments of the bacterium Pseudomonas syringae, can foster heterogeneous ice nucleation at temperatures higher than -10 °C (Maki et al., 1974).However, the most abundant INPs in the atmosphere (mineral dusts, volcanic ashes and soot) are active in the temperature window between -10 and -37 °C.Therefore laboratory investigations of ice nucleation activity in this temperature range are highly demanded.
Atmospheric aerosols can experience a variety of different freezing modes, depending on temperature and water vapour saturation.(i) Condensation freezing takes place when the particle acts as a CCN at temperatures below the melting point of ice and afterwards freezes at the same temperature.(ii) In contact freezing mode, the particle initiates the freezing when it collides with a supercooled droplet.(iii) The deposition mode involves the growth of ice directly from the vapour phase.(iv) If an INP has already been immersed in a droplet and causes freezing, the process is termed immersion freezing.These scenarios are model situations.The reality is often more complex.Immersion freezing can be considered as the dominating mechanism in the atmosphere.The experimental setup presented here focuses on immersion freezing only.
Homogeneous nucleation depends on the droplet volume (Vali, 1971).By increasing the volume, the chance of forming of a critical ice cluster via fluctuation of intermolecular hydrogen bonds is higher.Heterogeneous nucleation is dependent on the amount of active surface area at the interface between the INP and water (Murray et al., 2011) and therefore has no volume dependence (Hartmann et al., 2016).However, this is only true for a single particle immersed in a droplet.In our experiments, we worked with solutions and suspensions of defined concentration, where larger droplet volumes lead to a higher absolute numbers of INPs, increasing the probability of nucleation events.Therefore heterogeneous nucleation in droplets with a defined concentration has a volume dependence.
Heterogeneous nucleation can be explained by two approaches: a singular model and a stochastic model.These separated but interrelated approaches are combined in the VS66 model by Vali (2014) which appropriately describes the freezing behaviours of INPs.The singular model assigns a characteristic temperature to each nucleation site (Vali, 1971).The stochastic model is analogous to first order chemical kinetics.The main parameter is the rate of nucleation as a function of temperature (Murray et al., 2012).There are at least three random contributions to freezing experiments: molecular fluctuations of the ice embryos, the distribution of INPs in the sub-sample (e.g.samples in which INPs are suspended and subsequently divided into droplets) and the location of the nucleation sites on the surface of the INP.The molecular floating of embryos is a stochastic process, INP distribution and surface site location are random events which do not involve time sequences and therefore behave according to the singular model (Vali, 2014).Droplet freezing experiments are suitable to distinguish between the dominant mechanism for certain kinds of INPs.
Since the total surface area of INPs per droplet is important, as predicted by classical nucleation theory and confirmed by Edwards et al. (1962), the characteristic parameter to describe nucleation is the so called ice nucleation active surface site density n s .Equation (1) implies that n s has a deterministic quantity, fixed by the characteristics of the INPs (Vali, 2014).The singular model neglects the time dependence.N 0 is the total number of droplets in the experiment and N F the number of frozen droplets at the temperature T and A is the particle surface per droplet.The fraction of frozen droplets f(t) is given by: One often used technique to investigate ice nucleation activity (INA) is the water in oil freezing technique.The droplets can be placed on a surface in a gaseous environment (Whale et al., 2015), embedded in an oil matrix (Pummer et al., 2015;Zolles et al., 2015) or studied in free-fall (Wood et al., 2002).In case of free-fall experiments, droplets generated by pipetting or microfluidic devices (Riechers et al., 2013;Stan et al., 2009), the resulting droplet sizes are well above the range of droplets usually present in clouds, which is about 10µm (Pruppacher and Klett, 1997).

Description of the new setup
The experimental setup consists of four main parts: (i) the light microscope including a HD camera to observe the freezing experiment, (ii) the freezing-cell to cool down the sample, (iii) the freezing-chip carrying an ensemble of droplets and (iv) a computer to control cell temperature and cooling rate as well as to record and evaluate pictures of the freezing droplets.
The light microscope is equipped with a 20 fold LWD Objective and a 5.0 megapixel USB 3.0 camera (microQ L3CMOS) which is directly connected to the computer.The custom-built freezing cell is embedded in a hollow Polytetrafluorethylene (PTFE, Teflon®) cylinder with a diameter of 68 mm and a height of 25 mm, which can be sealed hermetically (see Figure 3).
Cooling is performed by a thermoelectric cooler (TEC; a Peltier element Quick-cool QC-31-1.4-3.7M)connected to a computer-controlled power supply.A water-ice mixture (~5°C) is pumped out of a storage tank into a heat exchanger attached to the warm side of the Peltier element with a water pump (EHEIM universal pump).With this set-up, we are able to cool our samples down to -40°C and below by regulating the electrical current through the TEC.A K-type thermocouple is attached directly on the cold surface of the TEC with a conductive adhesive to monitor the temperature.The thermocouple is connected to the computer via a thermocouple measurement device (NI USB-TC01).Two gas connectors on the shell of the freezing cell allow flushing with dry nitrogen.This is done before every experiment to remove humidity and establish a neutral atmosphere.Additional slots are available in the shell to insert the thermocouple and the electric connectors for the TEC into the cell.The top cover of the cell is removable to introduce the sample.The cover includes a glass window enabling the observation of the sample via light microscopy.The cell is placed on a stage directly under the objective of the light microscope.
As the common water-in-oil freezing technique has severe disadvantages, a new unique method was developed to generate whole sets of well-defined isolated micrometre-sized droplets directly.A novel freezing chip was designed from a (15 x 15 x 1) mm silicon plate, by etching a pattern of hemispherical cavities which allows us to create isolated droplets in the size range of 20-80µm.Reactive ion etching (RIE) was carried out with an OXFORD Plasmalab 80 with 10 cm³/min Argon and 20 cm³/min SF 6 as etching gas.The pattern consists of hemispherical cavities with defined diameters at defined distances from each other (see Figure 4).After the etching process, a gold layer (thickness 500 nm) was sputtered on top of the pattern, leading to an ice nucleation neutral surface.Tests of an uncoated silicon plate indicated that the silicon itself is IN active.A shift of freezing temperatures of ultrapure water from -38°C to approx.-20°C was found for droplets on the uncoated plate.
As an alternative to a gold sputtered silicon plate, a pure gold chip of similar dimensions was ion milled with a Focused Ion Beam (FIB) to introduce the same kinds of pattern.Since the gold chip is not ice nucleation active itself, further treatments of its surface are not necessary.If the surface of the gold sputtered silicon plate is scratched accidentally, the silicon gets revealed and the chip becomes ice nucleation active again.Small scratches on the surface of the pure gold chip were not found to have any influence on INA.
Once the chip is loaded with droplets (see below), it is directly placed on the TEC inside the freezing cell.The freezing process can be monitored on the computer screen and is recorded and saved as a video file automatically.All recorded videos include a time and temperature stamp.Freezing videos are provided under Supporting Information.The temperature ramp can be adjusted via the control software.The freezing videos are evaluated automatically by a LabVIEW VI (virtual instrument).During the freezing process the droplets turn dark, because ice shows a different light scattering behaviour than liquid water.The first step for evaluating the videos is to manually mark the droplets.Afterwards the software analyses the video and determines the time when the droplets turn dark respectively freeze.A contrast graph is generated for each droplet, linking the brightness of the droplet to the time respectively temperature, which enables to follow the freezing process (further information is given in the supplementary material).b) The birch pollen sample originated from the Czech Republic and was obtained from Pharmallerga ® .The preparation was carried out as described by Augustin et al. (2013). 1 g of birch pollen was suspended in 20 mL ultrapure water and placed for 12 hours in a refrigerator.Afterwards the suspension was filtered (Macherey-Nagel 640m) and the pollen washing water was diluted 1:2 with ultrapure water.c) Juniper pollen were obtained from Pharmallerga ® (Juniperus communis JUNU.0111).64 mg of juniper pollen were suspended in ultrapure water at a concentration of 50 g/L.After 20h at room temperature the suspension was directly used for the freezing experiment.d) Snomax ® was obtained from SMI Snow Makers AG.It consists of shredded Pseudomonas syringae, an ice nucleation active bacterium.About 1 mg of Snomax ® was suspended in ultrapure water to a concentration of 0,5 g/L.
The freezing behaviour of all these INPs is well described in the literature, rendering those substances suitable standards to test the efficiency and accuracy of our new setup.Their freezing temperatures cover a broad range between -5°C (Snomax ® ) and the homogeneous freezing temperature of water (-38°C).

Preparation of the freezing-chip
For sample preparation we applied a thin film of the dispersion on the freezing chip by placing 2 µL of the sample with a pipette (INPs dispersed in ultrapure water) on the chip and spreading it all over the surface with a rubber spatula.Parts of the suspensions are automatically collected in the cavities by capillary forces, while the rest is wiped off the surface with the spatula.This leads to droplets in the size of the etched cavities, with defined radii and defined distances between the droplets given by the etched pattern.Different droplet sizes can be achieved with different cavity sizes.After the cavities are filled, the surface gets covered by paraffin oil to prevent the Wegener-Bergeron-Findeisen effect, which occurs when water is present in both liquid and solid phases and would lead to continuous condensation of water vapour on ice, while at the same time liquid water evaporates until the liquid phase is entirely consumed (Korolev, 2007).A small droplet of paraffin oil is placed on the centre of the plate and spread by putting a microscope plate on top.Using this method we get a thin and evenly distributed oil film on top of our chip.
Contrary to the problems occurring when using oil during the water-in-oil freezing technique (cloudiness due to changes in viscosity of the oil, drawing of hydrophobic particles into the oil phase), the oil cover on the chip does not show any of the mentioned disadvantages.Since the emulsion is not vigorously mixed, hydrophobic particles remain in the droplets and the cloudiness of the oil film is negligible due to the thinness of the oil layer.
Due to the covering of the droplets and the distance between the cavities/droplets, completely separated and isolated droplets are prepared (Figure 5 and Figure 6).

Results
To investigate the efficiency of the new setup, the freezing behaviour of different kinds of INPs was analysed and compared with the existing literature.To describe the freezing behaviour of INPs freezing spectra and T 50 values are used.Freezing spectra show the fraction of frozen droplets f ice at a given temperature.The T 50 value describes the temperature at which 50% of the observed droplets are frozen.
In the Supporting Information detailed descriptions for the freezing behaviour of all analysed samples is provided.Results for ultrapure water and Snomax ® are given below.

Ultrapure water
The homogeneous freezing temperature of water, according to Pruppacher and Klett ( 1997), lies at -38°C or below.Ten measurements were performed with ultrapure water, with a resulting average T 50 value of -37,2 ± 0,4°C (Figure 9).The freezing spectra (Figure 7) are well comparable to the extended singular model VS66 as described by Vali (2014).

Snomax ®
The ice nucleation activity of Pseudomonas syringae is well documented and were recently published by Pummer et al.To clarify the reason for this shift in respect to the data by Pummer et al. (2012) we investigated the concentration dependence of the freezing behaviour of Snomax ® in a concentration range of 0,5 to 6,5 g/L using the water in oil technique.
We found a concentration dependence with lower concentration leading to lower freezing temperatures.This effect was also found by Wex et al. (2015) and explains the lower temperatures we received using the freezing chip, as the Snomax ® suspension we applied had a concentration of 0,5 g/L while Pummer et al. ( 2012) used concentrations of 24 g/L.Wex et al. (2015) worked with Snomax ® concentrations of about 0,5 g/L (the same as here), but they generated droplets with diameters about 1200µm, which was more than the hundredfold of the droplet radii generated on the freezing chip.This leads to higher freezing probabilities and therefore to higher T 50 values compared to our measurements.A summary of all freezing spectra using the freezing chip is given in Figure 10.

Conclusion
In order to study heterogeneous nucleation it is necessary to generate droplets with a defined size/volume.With the freezing chip it is possible to produce an array of droplets with the same well defined volume/size.Therefore investigations focused on size dependence can be done more easily and accurately.
In The accuracy of the light microscope videos can be significantly improved by the new technique.Using an oil-water emulsion with a thick oil phase can lead to cloudiness of oil at lower temperatures.Using the freezing chip allows a reduction of oil and therefore leads to higher contrast and focus in the recorded freezing videos.Another advantage over the classic emulsion method is the prevention of multi-layered droplets which complicates the evaluation.
The cavity matrix on the freezing chip provides clearly defined distances between the droplets.Droplets have no contact with each other, so frozen droplets cannot act as INP themselves for as yet unfrozen ones (infectious freezing).
The new controlling system of the temperature via LabVIEW increases the accuracy.Temperature programmes can be easily adjusted and repeated.The automatic evaluation via LabVIEW saves time and eliminates the problem of different results arising from varying personal interpretations of manually evaluated microscope pictures.The recording of the freezing process on video files makes reruns and additional evaluation possible.
A defined arrangement of droplets on the freezing-chip enables observation of a higher number of droplets per experiment.
In consequence random errors can be eliminated via statistical calculations.Furthermore droplets can be evaluated in a shorter period of time via automated evaluation.
We were able to show the efficiency and accuracy of our setup by comparing the measurements of freezing temperatures of different INPs with already published results (Figure 9).
Whale et al. (2015) have shown that ultrapure water droplets with a diameter of approx. 1 mm are often so strongly contaminated with INPs that the homogeneous freezing temperature is not reached and freezing occurs at around -25°C heterogeneously.Smaller droplets with smaller volumes open a wider temperature window and make investigations of INPs active at lower temperatures -even close to -38°Cpossible.By embedding droplets in an oil matrix, smaller radii down to 20µm can be realised.To generate the embedded droplets, a water (including the INP of interest) -oil emulsion in a test tube is shaken until homogeneity and then transferred on a sample carrier.Homogeneity is reached, when the emulsion turns white and opaque due to intense Mie scattering caused by the micrometre sized droplets.In this procedure five main problems occur: a) Droplets show a rather broad distribution of radii.b) Contact between droplets leads to possible droplet-droplet interactions during the freezing process (infectious freezing).c) The interface between droplets and the oil matrix can interfere with the nucleation process.If a hydrophobic INP (e.g.soot) is immersed in 5 an oil-water mixture as described by e.g.Pummer et al. (2012), it is very likely that the hydrophobic sample gets drawn into the oil phase.This significantly lowers the concentration of the INPs in the water droplet, yet it is difficult to quantify this effect.d) Partial crystallisation and changes in viscosity of the oil, due to decreasing temperatures, causes a cloudiness of the sample.All these influences can lead to a falsification of the results (Hauptmann et al., 2016).Our new experimental approach for droplet-freezing experiments eliminates this set of problems.In this paper we present 10 the advantages of a new setup using a defined freezing-chip over conventional drop-freezing experiments.Tests of the efficiency and accuracy of the new setup are performed by comparing results of INP measurements using well-known INPs (birch pollen washing-water, juniper pollen, K-feldspar, and Snomax ® ) as well as ultrapure water.Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2017-31,2017 Manuscript under review for journal Atmos.Chem.Phys.Published: 16 January 2017 c Author(s) 2017.CC-BY 3.0 License.

3
Materials and PreparationSeveral INPs of different types were used to investigate the efficiency of the setup: Microcline, birch pollen, juniper pollen and Snomax ® .The freezing experiments were carried out in ultrapure water type 1 generated by the MilliQ water purification system Merck Simplicity ® 2012.a) Microcline (K-feldspar, KAlSi 3 O 8 ) is a naturally occurring mineral and was supplied by Alfa Aesar GmbH & Co KG.The mineral was freshly milled with a swing mill (Retsch MM400) for 4 minutes and 30 swings per second immediately before the experiments.This led to a grain size between 1 and 10µm.Microcline was suspended in ultrapure water (concentration 20 g/L).
the water-in-oil emulsion technique, interactions between INPs and the oil matrix can have a substantial impact on the concentration of the INP depending on the hydrophobicity of the INP.By using the freezing chip, the water-INP suspension gets directly placed on the freezing chip without the necessity of suspending INPs in an oil-water mixture.Therefore hydrophobic INPs can also be analysed without loss of concentration.
The measured T 50 values of ultrapure water, birch pollen washing water and juniper pollen match the values found in the literature within ±2°C.Microcline (T 50 =-16,4°C) shows a trend to higher freezing temperatures than found by Zolles et al. (2015) (T 50 =-25°C) and Atkinson et al. (2013) (T 50 =-22,5°C).As mentioned, the volume and therefore the concentration, as well as the number of active sites play an important role for the freezing process.Small differences in these factors can lead to a significant shift of up to 7°C in T 50 values.The slightly lower freezing temperature of Snomax ® (T 50 =-8,9°C) compared with the measurements done by Pummer et al. (2012) (T 50 =-5°C) and Wex et al. (2015) (T 50 =-4°C) can be explained by different concentrations and droplet sizes.The freezing chip solves major disadvantages of the well-known and frequently used water-in-oil freezing technique and enables a new approach to determine freezing behaviours.Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2017-31,2017 Manuscript under review for journal Atmos.Chem.Phys.Published: 16 January 2017 c Author(s) 2017.CC-BY 3.0 License.

Figure 1
Figure 1 Schematic description of heterogeneous and homogeneous ice nucleation.Homogeneous nucleation takes place at -38°C and does not involve an INP.Heterogeneous nucleation takes place at higher temperatures at the interface between the immersed INP and water.

Figure 2 Figure 3
Figure 2 Schematic depiction of the measurement setup.Shown left is the light microscope equipped with a camera and the cryocell directly beneath.The K-type thermocouple to measure the temperature of the sample, the thermoelectric cooler used to cool the cell and the camera are linked to the computer.The temperature and the camera are controlled and regulated by a program based on LabVIEW.

Figure 4 Figure 5
Figure 4 Different electron micrographs of the freezing chip.Figure 4 (a): cavities of different size (ca.20 to 80µm) arranged on the freezing chip.Figure 4 (b): more detailed picture of the arrangement of the ca.20µm cavities.Figure 4 (c): a 40µm cavity sputtered in gold by ion milling and Figure 4 (d): a 45µm cavity etched in the gold plate via Ar-SF6 plasma.

Figure 6
Figure 6 Detail of freezing mask.Comparison of cavities filled with unfrozen and frozen ultrapure water.Differences in light 5

Figure 7
Figure 7 Fraction of frozen droplets fice against sample temperature of ten different freezing experiments using ultrapure water recorded with the new setup (M1 to M10).

Figure 8
Figure 8 Fraction of frozen droplets fice against sample temperature for nine different Snomax ® freezing experiments (S1 to S9)

Figure 10
Figure 10 Ice nucleation spectra of freezing experiments using the freezing chip with different INP: juniper pollen, birch pollen washing water, microcline and Snomax ® , as well as ultrapure water.