Microscopic Observations of Riming on an Ice Surface Using High Speed Video

Microscopic Observations of Riming on an Ice Surface Using High Speed Video

C. Emersic1 and P. J. Connolly2

1School of Electrical and Electronic Engineering, The University of Manchester, Manchester, UK

2School of Earth, Atmospheric and Environmental Sciences, The University of Manchester, Manchester, UK

Abstract

Microscopic droplet riming events on an ice surface have been observed using high speed video. Observations includedgreater propensity for droplet spreading at temperatures higher than -15°C on flatter ice surfaces, and subsequently, the formation of growing rime spires into the flow, allowing glancing droplet collisions and more spherical freezing of smaller droplets. Insight into differences between laboratory observations of the Hallett-Mossop processis offered, relating to the nature of droplet spreading associated with the structure of the rimer surface prior to impact. Observations of a difference between air speed and resulting droplet impact speed on an ice surface may affect interpretations of riming laboratory studies, and may explain recent observations of a high secondary ice production rate in supercooled layer clouds.

Keywords: Hallett-Mossop, Ice Multiplication, Riming, High Speed Camera, Ice Crystal

Corresponding Author Information: Dr Christopher Emersic, School of Electrical and Electronic Engineering, The University of Manchester, Manchester, UK. Tel: +44 (0)161 306 4679;

1Introduction

In-situ observations have been made of cloud regions in the temperature range 0 to -10°C showing ice crystal concentrations far in excess of the number of ice nuclei present (e.g. Koenig 1963; Crosier et al. 2011). A possible explanation for these observations was investigated in laboratory experiments showing the production of ice crystals as riming occurred on arotating target moving through a cloud of supercooled water droplets (Hallett and Mossop 1974; Mossop and Hallett 1974). This phenomenon, often referred to as the Hallett-Mossop process of secondary ice production, has since been studied a number of times in the laboratory to attempt to characterise the rates of splinter production under a range of conditions(e.g. Goldsmith et al. 1976; Mossop 1976, 1978; Foster and Hallett 1982; Heymsfield and Mossop 1984). Nevertheless, it remains unclear how ice splinters are produced by this process.

In the original experiments of Mossop and Hallett(1974), a cloud chamber 2m1.2m1.8m was used with a riming rod 30cm long, 0.2cm in diameter, mounted vertically and moving in a circular path of diameter 30cm at a velocity of approximately 2.6ms1 in a boiler-generated cloud of supercooled water droplets. The low number of ice crystals produced were counted by eye in a beam of light. They found one splinter was produced for approximately every 160 droplets of diameter greater than 23µm that were accreted (roughly 350 splinters per mg of rime), for target velocities ranging from 1.4 to 3ms1 at a temperature of -5°C. Mossop (1978, 1985) later found that the abundance of dropletssmaller than 13µm in addition to those larger than 23µm,increased the splinter production rate further.

Saunders and Hosseini (2001) extended the droplet impact velocity to 12ms1 and found maximum splinter production rate occurred at 6ms1, with 70 splinters produced for every mg of rime. They also provided a comprehensive review of the literature associated with the Hallett-Mossop process, including hypothesised mechanisms of splinter production. Briefly, Choularton (1978; 1980)suggested that larger droplets (larger than 24µm) can sometimes freeze quasi-spherically on smaller accreted droplets (smaller than 13µm) which can act as a narrow bridge to prevent heat flow to the graupel bulk ice. Subsequent heat loss leads to symmetric surface freezing of the larger dropletto form a shell. As the bulk liquid within the shell freezes, the expansion of remaining liquid can form a protuberance through the shell. This protuberance can later break off into one or more splinters. However, Dong and Hallett (1989) noted significant droplet spreading at all temperatures greater than -8°C which inhibited spherical droplet freezing, and proposed an alternative mechanism to explain splinter production. This relied on the surface of spreading droplets freezing and cooling more rapidly than the liquid nearer the droplet-graupel interface. The thermal contraction associated with cooling results in surface stress which ultimately leads to shattering and splinter production. Both Griggs and Choularton (1983) and Dong and Hallett (1989) provided still photographic evidence to support their proposed mechanisms.

Understanding secondary ice multiplication is of high importance to numerical weather forecasts, which are sensitive to the number of ice crystals within clouds. Ice crystal concentrations have direct impact on cloud evolution, lifetime, and radiative properties(Ramanathan et al. 1989; Hogan et al. 2003b; Hogan et al. 2003a; Hogan et al. 2004), which is important for climate change model forecasts. Improving the accuracy of forecasts also has direct economic impact (Katz and Murphy 1997). The focus of this study was to investigate the fundamental riming process in the laboratory to see whether it can be better understood. Specifically, attempts were made to record microscopic riming processes on an ice surface in higher temporal detail than has been studied before,using a high speed camera,to try and identify possible factors which may contribute to secondary ice crystal production. It was intended that these observations would provide additional insight into differences observed between experimental researchersinvestigating the Hallett-Mossop process, and encourage further research on the mechanismsof secondary ice production. While not investigated in this study, solutes commonly present in cloud droplets may affect the freezing process of ice, as observed in studies of levitating droplets (e.g. Johnson and Hallett 1968; Leisner et al. 2014). Further investigations on the influence of solutes on freezing processes during riming would be beneficial.

2Experimental Setup and Procedure

Observations were made using a high speed video camera (PhotronFastCam MC-1) which allowed recording speeds up to 20,000 frames per second (0.05ms). A chest freezer measuring approximately2m1m1m was used at temperatures down to -20°C to simulate the general Hallett-Mossop temperature range and colder. Clouds of supercooled water droplets were produced using an atomiser and the droplet size distribution was measured using a Forward Scattering Spectrometer Probe (FSSP) (Fig. 2.1). Droplets of sizes historically found to be required for splinter production (less than 13µm and greater than 25µm) were present and were also directly observed in video data. Clouds were generated for approximately 1 minute before riming commencedto ensure homogeneity and droplet thermal equilibrium. Once the cloud was formed, a vacuum pump was used to draw the air through a tube at 6ms1and past a stationary target rod acting as the rimer, located laterally across and 1cm deep into the pipe (Fig. 2.2). The rod was 4mm in diameter and had a length which spanned the pipe tube of diameter 40mm. The rod was also made of ice to ensure most realistic riming conditions; initial attempts using plastic rods prevented droplets freezing readily on contact. Metal rods were also examined, but unlike transparent ice, did not allow for good lighting required for optical recording. Using a stationary rod introduces differences relative to a natural graupel. There is no tumbling or precession of the rimer in the experiments and there will therefore be relatively greater latent heating on the riming face for a given cloud liquid water content and droplet size distribution. However, the target rimer is considerably larger than a graupel particle and thus will have greater heat capacity, possibly offsetting this. The relative increase in rime accretion rate is likely within the range of what could be experienced by a graupel particle in a natural environment, particularly as wet growth was not observed. Calibrated type K thermocouples were used to measure the ambient temperature, and were carefully hooked on the ice rod to measure the temperature of the rimer surface. Lenses on the camera were used to zoom to microscopic scales and had a field of view of approximately 500 by 150µm and pixel resolution of approximately 2µm. The depth of field was approximately 50µm. Riming events were recorded while looking at the rime from top-down (viewing parallel to airflow) and,separately, from the side (nearer perpendicular to the airflow). Recording microscopic high speed video proved problematic; with increasing frame rates and shutter speeds, the amount of light captured per frame was reduced. This was exacerbated with increasing magnifications which reducedthefield of view and thus the area from which light could be gathered. Furthermore, it was not possible to increase light levels to ideal values to compensate, as this increased heat input to the ice sample resulting in its destruction. A careful balance was found to optimise lighting while avoiding observable heating of the system.

Fig. 2.1. Percentage of total cloud produced by atomiser droplets measured by FSSP within specified size ranges.

Fig. 2.2. Experimental setup: An atomiser-produced cloud in a cooled chest freezer which was drawn past an ice target rod using a vacuum pump and piping system. A high speed camera filmed microscopic riming events on the rod’s surface; thermocouples measured the air and rod surface temperature (T1 and T2 respectively).

3Resultsand Discussion

3.1Observations from above the rimer

When observing riming events by viewing down onto the surface from above, lighting difficulties meant that the surface had to be initially freeof rime; rime accumulation meant increased scattering of light and opaqueness of the iceand therefore decreased lighting through the surface. Droplets landing onto a pristine, flat ice surface were observed to spread considerably at all temperatures greater than -10°C (Fig. 3.1A). Only at temperatures less than approximately -15°C did droplets retain a defined hemispherical shapeafter freezing on the surface (Fig. 3.1B). Once riming had increased the pristine surface depth by approximately 20µm, such that sufficient curvature from the frozen droplets became a dominant surface feature, the spreading of subsequent impacting droplets tended to become less likely. Despite the rapid droplet spreading, however, a varying degree of spreading could still occur, even between similar droplets impacting within ~1s of each other(Fig. 3.2), and in some cases, appeared to behave in a liquid-like manner and spread considerably (Fig. 3.2 A1). In Fig. 3.2A2, the droplet landed on the region in the final panel of Fig. 3.2A1 and froze in a distinct manner, showing a freezing wavefront and relatively solid structure.

Fig. 3.1. Droplets landing on a flatter ice surface tend to spread more at temperatures associated with the Hallett-Mossop process. Below approximately -15°C, droplets tended to retain hemispherical shape. Viewed from above and perpendicular to the ice surface plane.

Fig. 3.2. The amount by which droplets spread on flatter ice surfaces can vary even at similar temperatures and droplet sizes. Viewed from above, the droplet in A2 lands on the previously frozen region from A1 but spreads less. Similarly, the droplet in B2 lands very close to the droplet in B1 and again spreads less. The impact velocity of thedroplets was approximately 1ms1 and the second droplet impacting in each case occurred within 1s of the first droplet impact.

As reported by Griggs and Choularton (1983), we also observed freezing wavefronts propagate symmetrically across the droplet from the edges to the centre (Fig. 3.3). Here however, this was observed even in cases where spreading was substantial (Fig. 3.2 A1). In the majority of instances where a freezing wavefront was observed, the wavefront would form a very small central dark spot. This may suggest an increase in ice depth formed by a possible small protrusion beyond the surface, although it was not possible to measure the protrusion depth. These dark spots were observed at temperatures as low as -15°C,which isconsiderably colder than the Hallett-Mossop lower temperature limit of -8°C. However, the higher rate of freezing at these lower temperatures appeared to produce reduced darkness of the spot. Furthermore, this central dark spot was observed for all frozen droplet shapes—including those that were irregular and considerably flatter than circular frozen droplets.

Fig. 3.3. Time series showing freezing wavefronts observed on many freezing droplets viewed from above. The bottom row of images replicates the top row but with white dotted lines added to help emphasise the freezing wavefront. These result in a dark central spot at the end of the freezing process.

3.2Observations from the side of the rimer

As riming continued from an initially flat ice surface, rime thickness increased and was viewed from a side angle. The presence of the rod deflected airflow and droplets around it, with video footage showing many droplets moving into and out of the focal plane. As rime grew away from the rod’s surface into the deflected flow, droplets were observed to be accreted from glancing collisionsin addition to direct head-on impact. More spherical droplet freezing was observed as a result of the glancing collisions, and eventually, this formed fibrous rime spire structures (Fig. 3.4). Approximately 50 riming events viewed from the side were observed in detail at a range of temperatures between -3 and -8°C.

While freezing wavefronts could sometimes be seen from this side-view on more spherical riming events, the central dark spot left by the freezing wavefront that was apparent when observing from above was not convincingly seen as a protrusion when viewed from the side. This may suggest they are perhaps very broad in profile with a small physical dimension lower than the resolution of the camera and optical setup (less than 2µm). It is also possible that they could instead be an optical artefact and not result in a protrusion.

The larger apparent protuberances photographed by Griggs and Choularton (1983, figure 3, panel 3) were not observed here, nor was any evidence of liquid ejection in any of the video footageof this physical riming situation. The droplets here were not as large as those used in other studies e.g. Griggs and Choularton (1983)(125µm diameter droplets), which may account for this, but droplets were greater than the 25µm diameter found to be required for the Hallett-Mossop process to occur (Fig. 2.1). While similar larger protuberances were not observed, structures of comparable size were formed frequently in the rime spiresthat grew (Fig. 3.4). These however were caused by the accumulation of several smaller droplets (approximately 10µm)freezing on top of one another in succession. Breakoff of these structures could possibly account for the droplet size dependence of the Hallett-Mossop process, even though this in principal could occur at lower temperatures. It is possible if an ever larger-growing spire was hinged by a small droplet, the torque during tumbling of the riming particle could cause breakoff and possibly account for increased splinter production rates observed in the presence of smaller droplets. This effect has not been observed directly here, however, and the rimer was stationary.

Not all riming events on such fibrous rime spire structures led to spherical droplet freezing. As observed when viewed from above, on some occasions under very similar environmental conditions (successive riming events within approximately 100ms), some droplets of comparable size would freeze and retain hemispherical structure, whereas some would experience spreading. No causal relationship could be identified for this observation. On two occasions at a temperature in the Hallett-Mossop range (-3to -8°C), a larger approximately 30µm droplet was observed to land on a frozen approximately 11µm droplet—conditions which form a requirement for the shell fracture hypothesis splinter production(Choularton et al. 1978). However, in these two observed cases, spreading of the larger droplet occurred which resulted in its complete envelopmentof the smaller droplet before freezing (Fig. 3.5). No evidence was found for larger droplets retaining their structure and freezing sphericallyon smaller frozen droplets in the 50 observed riming events.

Fig. 3.4. As riming continues, spires develop away from the surface. As these protrude into the airflow, video footage showed increasing numbers of glancing collisions which expanded the spires laterally to produce the observed structures. Increased glancing collisions at this stage appeared to increase the likelihood of dropletsfreezing spherically. In these images, airflow is generally from top to bottom with a component towards the page.

Fig. 3.5. Two observations of larger 30+µm droplets landing on pre-frozen smaller approximately 11µm droplets at approximately 1ms1at the colder end of the Hallett-Mossop temperature range, and enveloping the smaller droplet instead of freezing spherically. The incoming droplet in panel 1 of B spreads sufficientlyon impact to result in no larger spherical droplet structure after freezing.

3.3Interpreting earlier studies

3.4These observations may help interpret some of the differences between observations of past laboratory research. Choularton et al. (1978) hypothesised a shell fracture mechanism based on observations of more spherical droplet freezing, whereas Dong and Hallett (1989) proposed a temperature gradient stress mechanism instead to account for the droplet spreading they observed at temperatures considered relevant to Hallett-Mossop ice multiplication. The observations of Dong and Hallett were made by viewing from above on what appears from photographs to be flatter, more pristine ice surfaces, whereas the photographic evidence provided by Griggs and Choulartonsuggested side-viewing and images of structures indicative of rime spire growth. The new observations in this study revealed that flatter, more pristine ice surfaces tended to encourage less spherical drop freezing and greater spreading, whereas accretion on rime spires resulted in more spherical droplet freezing events occurring from glancing collisions. These observations are therefore consistent with past studies and can account for the range of observations and behaviours. No observations made here preclude either proposed splintering hypothesis from being responsible for ice multiplication via the Hallett-Mossop process. Conversely, in approximately 1300observed droplet freezing events viewed from above on more pristine, flat ice surfaces, and ~50 droplet freezing events viewed from the side, no direct evidence of secondary ice production consistent with either hypothesis was observed.Observations of air speed changes

While the air speed at the location of the riming rod was measured to be 6ms1, when the rod was in position, video footage showed that droplets were only impacting at approximately 1 to 2ms1. The approaching droplets are likely decelerating due to the rod’s presence in the air flow. Simulations of the potential flow around a cylinder for an air speed of 6ms1 revealed the velocity field profile (Fig. 3.6) showing reduced flow speed on the front face of the target down to 1ms1 (Fig. 3.6 A) and increased air speed about the lateral edges by up to approximately a factor of 2 (Fig. 3.6 B). Trajectories of water droplets of varying sizes for different initial air speeds were subsequently calculated and their velocity on approach to the rod surface is shown in Fig. 3.7. Droplets 10µm in diameter impacted the rod surface at approximately 1 to 2ms1in the computations only when the air speed was reduced to approximately 5ms1 rather than 6ms1 observed here. This discrepancy can be accounted for in part by noting that the impact speeds measured are projected 2D trajectories and thus the speed of 1 to 2ms1 observed represents a minimumimpact speed. Furthermore, only 2 or 3 frames of footage provided droplet position from which to estimatethe projected speed. Both observations and simulations reveal that droplet impact speed at the front face of a riming target is lower than the air speed.