The Snowball Chamber: Neutron-Induced Nucleation in Supercooled Water

Matthew Szydagis,∗ Corwin Knight, and Cecilia LevyUniversity at Albany

Department of Physics1400 Washington Av.

Albany, NY 12222-0100(Dated: July 25, 2018)

The cloud and bubble chambers have been used historically for particle detection, capitalizing onsupersaturation and superheating respectively. Here we present the snowball chamber, which utilizessupercooled liquid. In our prototype, an incoming particle triggers crystallization of purified water.We demonstrate water is supercooled for a significantly shorter time with respect to control data inthe presence of AmBe and 252Cf neutron sources. A greater number of multiple nucleation sites areobserved as well in neutron calibration data, as in a PICO-style bubble chamber. Similarly, gammacalibration data indicate a high degree of insensitivity to electron recoils inducing the phase tran-sition, making this detector potentially ideal for dark matter searches seeking nuclear recoil alone,while muon veto coincidence with crystallization indicates that at least the hadronic component ofcosmic-ray showers triggers nucleation. We explore the possibility of using this new technology forWIMP and low-mass dark matter searches, and conclude with a discussion of the interdisciplinaryimplications of radiation-induced freezing of water for chemistry, biology, and atmospheric sciences.

PACS numbers: 95.35.+d,29.40.V,29.40.-n,25.40.Dn

Keywords: dark matter, direct detection, sub-GeV, water, supercooling, low-mass WIMPs

I. INTRODUCTION

The nature of dark matter has remained an enduringenigma for over eight decades now, for both cosmologyand astroparticle physics. Continued lack of unambigu-ous evidence from direct detection experiments of thetraditional Weakly Interacting Massive Particle (WIMP)has led to an impetus to consider particle masses bothhigher and lower than before, driven by many hypothe-ses/models [1]. The goal of this work is inexpensive, scal-able detectors for low masses, but also multi-purpose.

Water has the advantages of hydrogen content, idealfor considering dark matter candidates O(1) GeV/c2 inmass due to the recoil kinematics, and the possibility of ahigh degree of purification, even en masse [2]. Thresholddetectors for dark matter, such as bubbles chambers em-ployed by COUPP [3] then PICO [4], while possessing noenergy reconstruction, do have the advantage of a highdegree of insensitivity to electron recoil backgrounds, ina search for nuclear recoil. The recoil energy thresholdcan remain low while the dE/dx threshold is high, bothset simply by temperature and pressure of operation.

Instead of using superheated water in a bubble cham-ber, implemented successfully in the past [5] (though athigher energy threshold, not for a dark matter search)we consider here supercooled water, oft-studied [6]. Thereason is that freezing is exothermic not endothermic likeboiling. This should naıvely imply near-0 energy thresh-old, as the phase transition will be entropically favorablein this case. The frontier of lower-mass dark matter be-comes within reach with lower-energy recoil threshold.

∗ Corresponding Author: mszydagis@albany.edu

II. EXPERIMENTAL SETUP

A cylindrical fused quartz vessel from Technical GlassProducts with hemispherical bottom and quartz flangeat top for sealing was prepared with 22±1 g of water anda partial vacuum on top, 8.5±0.5 psia of water vapor atroom temperature. The overall volume of water as activedetector was limited by the low throughput of the finalfilter used, described below, likely caused by particulatebuild-up. The quartz vial was fully submerged in a Huberministat circulator from Chemglass Life Sciences for ther-mal regulation, instrumented with three thermocouplesfor recording the temperatures, including the exothermicincrease [7]. These were located near the top (below theflange), middle (water line), and bottom (hemisphericaltip). A piece of plastic scintillator with an attached sili-con photomultiplier (SiPM) served as the muon veto, sit-uated below the thermo-regulating circulator, but alignedwith the central vertical axis of the quartz.

A. Water Purification

Ordinary tap water was passed through a commercialdeionizer and a 0.150-µm filter first, then boiled. Steampassed through multiple µm-scale filters and a final 20-nm NovaMem PVDF thin-film membrane filter similarto that used by [8], which remained in place above thequartz jar during operation. The quartz was prepared byultrasonic cleaning with an Alconox solution for 15 min-utes at 50◦C and 25 kHz, rinsed with deionized and pre-filtered (150-nm) water above, then dried before sealedin its flange assembly, in a Class-1000 cleanroom. A low-power vacuum pump reached ∼1 psia before steam flow.

arX

iv:1

807.

0925

3v1

[ph

ysic

s.in

s-de

t] 2

4 Ju

l 201

8

2

FIG. 1. Photograph of the chiller used to house the snowballchamber prototype reported on in this work. Visible at topare the solenoid valve for filling with purified water, and onthe screen the down/up temperature swings of running canbe seen in yellow. At the left is a pressure transducer. Seatedon the flange is an AmBe source. The thick black conduit atright is the connection to the bore-scope camera for observingfreezing. (Left inset) The quartz itself containing the 22 mL(room temperature) of deionized and filtered water for super-cooling. (Right inset) An example exothermic spike detectedduring an event. The temperature does not rise all the wayto 0◦ due to the chiller actively cooling still.

B. Data Collection

The water was cooled in an ethanol bath to −35◦ at∼ 2◦/min, the best rate of the circulating chiller (Fig-ure 1). While introducing a lag in the water tempera-ture, which we account for in the systematic uncertainty,this had the advantage of reaching a low temperature andthus higher degree of supercool rapidly [9], in an effort toreach low energy threshold. The chiller sat on vibration-dampening pads, with a cut-out for the veto to sit di-rectly beneath it. In addition to the three thermometersmentioned earlier for recording the internal bath temper-ature, a fourth recorded the room temperature, whosesmall variation was not observed to affect the results.

All data (temperature, pressure, camera, veto) wasread in using National Instruments hardware and theirLabView software, and taken continuously day and night,alternating control and source runs to minimize system-atics (48-hour-long runs in 2017; 24-hour runs instead in2018). An effort was made to ensure equal numbers ofcontrol and source runs with no preference for day, night,weekday, or weekend. The increase in temperature fromthe exothermic reaction of supercooled water crystalliz-ing was used as the trigger for saving camera images,going back in the image buffer to ensure the moment ofinitial nucleation was captured, as there was a delay inthe temperature spike (Figure 1, right inset). The mid-dle thermocouple, closest to the water line (see water inbottom of quartz in Figure 1, left inset) was used as themost reliable trigger source. See Figure 2 for a sample ofcamera data.

FIG. 2. An example of a triple-nucleation event, from 2018AmBe data, suspected to be caused by multiple scattering.Red circles indicate the first frames in which a nucleation siteappears. The first two “snowballs” merge rapidly; the secondappears much later, implying it is from a different neutron.Unlike in a bubble chamber, there is no pressure increase ac-tivated after a trigger, so the unfrozen water volume remainssupercooled and thereby active during an ongoing event. Alsonot like in superheating, nucleation is a slow process here [10].For brevity/clarity, only every 3rd frame is pictured (150 ms).

3

One bore-scope camera, set to 480x234 resolution and20 FPS, recorded events. A cool-down took 27 min start-ing at 20◦, followed by 27 min to increase back to 20◦ seton the thermoregulator, to quickly melt the active massand reset for the next 54-min cycle. The time spent su-percooled by the water was logged for control, ∼200 n/sAmBe [11] nominally 90 µCi, Pb-shielded AmBe, and 10µCi 137Cs 662 keV gamma-ray source (2017 data-takingrun) and for control, AmBe, Pb-shielded AmBe, and Pb-shielded ∼ 3000 n/s 252Cf [12] nominally 1 µCi source(2018). Shielding for these neutron sources was intendedto prevent AmBe and Cf gammas from interfering withthe operation of the thermocouple readouts [13].

A thermocouple and not the one camera was used forthe trigger due to image artifacts making reliable pixelsubtraction for uncovering differences challenging, suchas glare from the LEDs lighting the chamber, reflectionsincluding off the thermocouples on the quartz wall, thewall itself, or the water line, or bubbles or dust in the cir-culating ethanol surrounding the quartz (see Figure 2).Nevertheless, a resulting image can be relatively clear,with a black background filling in with the color of theLED used, since the light is scattered more effectively byice crystals than in clear, pure liquid water. This is anal-ogous to bubbles in the 15 kg COUPP bubble chamberwith 90-degree lighting [14].

The entire setup was disassembled and rebuilt in 2018and data re-taken with a mixture of the same and dif-ferent sources. The 3 red LEDs, which failed to remainoperational at consistent lighting level long-term in theethanol with thermal cycling, were replaced with a singleblue LED. While image quality was somewhat enhanced,its different luminosity may have created a slightly dif-ferent thermal load: control data in 2018 showed slightlylonger supercool times. The calibration sources runs wereonce again alternated with control.

III. RESULTS

When an AmBe or Cf source is present during a data-taking run of multiple cool-downs, the water does not re-main in a metastable, supercooled state as long, freezingalso at correspondingly higher temperature, as expectedusing the fixed cool-down rate. This time is “nominal,” inthe sense it is an overestimate, although the same acrossall runs with different sources, so it is used as a rela-tive measure. There is thermal lag between the exteriorquartz wall and water inside. In Figure 3, a simple defini-tion is used: the interval between when the water is below0◦C and when it reaches its minimum temperature beforethe sudden sharp increase from freezing. No fit was per-formed, and so minute fluctuations are ignored, averagedover. An arithmetic mean of the supercool times as de-fined with all 3 thermocouples on the quartz is reported.Although no blinding was performed, bias mitigation wasaccomplished through a lack of any cuts in the analysis.The data are presented as is, in Figure 3.

FIG. 3. Averages for the durations spent supercooled by thewater volume for control and sources runs in 2017 (top) and2018 (bottom). Neutron sources lead to reduction, althoughstatistical significance is not reached until Pb shielding is in-troduced, likely due at least in part to the secondary neutronproduction. Columns chronological, and errors strictly statis-tical, based on multiple runs, of which control most common.

TABLE I. Temperature minima achieved in all cases, priorto exothermic shock, from both data sets recorded. Note for2017 data are Cs, but Cf instead in 2018. All measurementshave a ±2.5◦C systematic uncertainty caused by being able toonly measure external temperature, to avoid nucleation sites.

Calibration Type Tmin (◦C) 2017 Tmin (◦C) 2018Control (no source) -20.31 ±0.05 -20.07 ±0.07AmBe w/o Pb -20.70 ±0.10 -20.53 ±0.11AmBe w/ Pb -20.00 ±0.11 -19.69 ±0.14137Cs OR 252Cf -20.40 ±0.12 -19.30 ±0.09

A. Interpreting Supercooled Time

The supercool time can be interpreted as the inverse ofevent rate, so that lower than control implies sensitivityto radiation. This is a more natural unit due to long re-set times. Future work will explore more rapid methods

4

FIG. 4. Histograms of supercool times, normalized to the con-trol average, 2017 and 2018 separately normalized and com-bined. Control in blue, repeated; vertical error bars are statis-tical, while horizontal are bin widths. Scenarios are as labeledin legend. Tails to lower supercooled times are interpretableas stochastic heterogeneous nucleation [16] induced e.g. bymobile residual particulates of order of the critical radius forrunaway crystallization (versus re-collapse) [17]. Gaussian fitsare the color-matched solid lines. The χ2/d.o.f. for control,AmBe, AmBe with lead shielding, Cs gamma, and shieldedCf respectively are 7.51, 1.56, 1.50, 1.21, and 1.09. The moststriking difference at the multi-sigma level is between controland Cf at lower right, with means from fits being 0.957±0.003and 1.006±0.004. The result is similar with different binning.

of heating, a modular detector setup, or a supercooleddroplet detector, in the vein of PICASSO for superheat-ing [15], to increase livetime. The control and gamma-rayresults are nearly identical, as evidenced further by Fig-ure 4, despite the gamma source, albeit weak, possessinga rate over 3 orders of magnitude the neutron rate in theAmBe case, and 2 for Cf. This result implies that our de-tector possesses a “blindness” to electronic recoils similarto that of a bubble chamber, at least when compared tothe type used for dark matter searches recently, operatedat a lower degree of superheat than HEP chambers [4].

A lower temperature may mean lower energy thresh-old, assuming our device acts like a reverse bubble cham-ber essentially: for the case of superheating it is well es-tablished that a higher temperature (and/or lower pres-sure) implies lower energy threshold [3, 4, 15]. For ourearly proof-of-concept initial prototype we have not as yetquantified energy threshold as a function of temperature,but the fact that all supercool times are at least ∼700seconds, demonstrating a low effective event rate, evenwhen a relatively hot neutron source is near includingduring cool-down, suggests that supercooled water doesnot become a sensitive radiation detector at least until athreshold temperature is passed (see also Table I). The

systematic uncertainty in the temperature minimum wasestimated through camera observations of melting, andextrapolating the temperature as a function of time dur-ing cool-down to the moment of zero-crossing recorded bythe chiller, compared to the thermocouple thermometers.

The quantitative results are suggestive of low neutrondetection efficiency, O(1) MeV energy threshold still,so this is not yet a dark matter detector, at least forWIMPs. That being said, if 1 keV threshold is achiev-able for nuclear recoils (possible with superheated wa-ter at least [18]), then 10,000 kg-days of exposure (only100 kg of water underground for ∼3 months, or 10 kgfor ∼3 years) would have an estimated spin-independentsensitivity of O(10−7-10−8) pb in the mass range of ∼5-10 GeV. That assumes zero background, but the Cs re-sults are suggestive of this being possible. Improvementof the water purity and container cleanliness [19], hy-drophobicity [9, 20], and smoothness [21], should elim-inate wall and surface backgrounds, which can also befiducialized out with position reconstruction. The low-energy “shoulder” of the neutrino floor may be withinreach, with our projection 2 orders of magnitude lowerin cross-section than the most up-to-date world-leadingresults in this mass range from DarkSide [22]. For 500MeV we project a world-best O(10−40) cm2, using [23].

The significance of the supercool times difference wasdiminished for the with- and without-source (AmBe) sce-narios in the second set of runs, but a likely explanationis differing ethanol levels changing the amount of neu-tron moderation in such a hydrogen-rich fluid. The liq-uid level in the thermal bath was perhaps slightly higherin 2018. We were unable to simply maintain this as lowas possible to avoid neutron moderation, as it was neces-sary to keep the entire water volume under the ethanolline to avoid a thermal gradient. To make a more defini-tive determination of the neutron detection ability weadded the 252Cf source, which, while possessing a slightlysofter spectrum, had the advantages of not only produc-ing more neutrons, but also being a more neutron-richsource (compared to its gamma production rate, vis-a-vis AmBe) [11, 12]. Further evidence of neutrons beingresponsible for freezing was accidentally observed whenthe first attempt at new data in 2018 showed no differ-ence from control: it was found the bath level was ∼3 cmhigher than in 2017. The source was on top as usual.

The low nucleation rate compared to source strengthsis indicative of a clearly non-zero energy threshold [24].While good for background rejection, the data are thusin tension with the original hypothesis of effectively nominimum energy. In retrospect, given that supercoolingis even possible, it is sensible a threshold must exist [25].Taken together the neutron and gamma-ray data suggesttwo thresholds, in energy and in stopping power, as in abubble chamber. In spite of the freezing water outputtingenergy, in the form of heat, a positive energy input cantrigger its nucleation with a local disturbance. The exis-tence of a homogeneous nucleation limit for water hintsat the threshold energy approaching zero close to it [26].

5

B. Image Analysis

Given the potentially controversial nature of this work,with a discovery claim of a new property of water, andthe reverse of what one would expect from the directionof the phase transition (for ionizing vs. non-ionizing ra-diation) at least when compared to the cloud chamberas a closer analogue than the bubble chamber, a furthercheck was deemed necessary. To this end, the number ofnucleation sites was counted to verify more in the pres-ence of neutrons, which should be moderated in water,scattering multiple times visibly in the camera images.

This validation was performed manually in the styleof decades-old bubble chamber photos, given the imagequality issues discussed earlier making an automated pro-cess challenging, but due to camera failures it could onlybe reliably performed on the 2017 data. Bias mitigationtook the form of mislabeled folders, initially resulting inan opposite conclusion to what is reported in this section.

The image analysis helped name this new particle de-tector type as the snowball chamber. A bulk event i.e.homogeneous nucleation (often clearly not edge or watersurface or hemisphere heterogeneous nucleations throughobservation of the crystal “plume” initial shape and sub-sequent development) often seems spherically symmetricin nature prior to collision with a side or second nucle-ation [27]. Furthermore, the process, unlike bubble for-mation, cannot be halted via pressure change [28], andtemperature increase is not rapid enough, leading to a“snowball” effect where the entire volume freezes in time.The physical consistency (i.e. texture) of frozen super-cooled water is also closer to snow than typical ice [29].

Events occurring on the quartz wall or at the interfacebetween the liquid and partial vacuum appeared to bemore common especially for control runs. This is promis-ing as a water purity indicator, but cannot be concludedreliably lacking 3D information. Therefore, this analysiswas restricted to a counting of interaction sites withoutfurther interpretation. Table II shows a clear increase inmultiple scattering, including 3 or more vertices, with anAmBe source 10 cm above the water level. The fractionof singles decreases further when 3 mm of lead shieldingis used. Neutrons, but not gammas, make a difference.

TABLE II. Numbers of events with varying snowball vertexcounts, listed as fractions of total events with usable images.This is consistent with the analyses of the differences in boththe supercooled times and the lowest temperatures achieved.The uncertainties come from counting statistics alone, whilesystematics are minimized using painstaking hand-scanning.

#Vertices Control AmBe w/ Pb 137Cs γ1 0.9532 0.8539 0.8000 0.94742 0.0468 0.1348 0.1846 0.05263 0.0000 0.0000 0.0154 0.00004 0.0000 0.0112 0.0000 0.0000(uncertainty) 3 x 10−5 5 x 10−4 9 x 10−4 2 x 10−4

FIG. 5. Histograms of the time of freeze as determined visu-ally minus the time of closest muon veto pulse. Bin width is50 ms (1 camera frame) and the double-wide peak indicates a1-frame uncertainty in determining freeze time. Colors are aselsewhere: control blue, unshielded AmBe is the red dashed,shielded AmBe in solid red, and 137Cs in green. Surprisingly,only the non-Pb-shielded AmBe data exhibits a strong peak.Possible explanations are discussed in text. Given this oddity,the analysis was cross-checked with the time of the exothermicspike, which agrees. Different particles appear in the “veto,”including MeV-scale neutrons, as it is a simple scintillator, soa peak is not necessarily interpretable as muons (or their sec-ondaries). Errors are statistical, while the black dotted line(horizontal) indicates the accidental coincidence probability,of ∼3%, the same across all data sets, so direct scintillation inthe veto from the sources was not observed above background.

C. Cosmic Ray Muon Veto

We compared the timestamp assigned to the first savedcamera frame in which at least 1 nucleation is visible asbrighter pixels over a mostly-dark background to the clos-est muon veto pulse in time, forward or back, to deter-mine if freezing could occur immediately after an SiPMpulse. This analysis was performed en masse after firstcollecting all freezing times before looking at veto data.

When these times were subtracted, a significant coinci-dence peak appears, even considering the non-negligibleaccidental coincidence rate from all the background ra-diation passing though the scintillator (see Figure 5).The statement that radiation can freeze water is thusstrengthened with the veto analysis, changing from amerely statistical statement over many events to a deeperevent-level understanding. However, it is statistically sig-nificant only for the unshielded AmBe data, which is puz-zling, especially given the fact that particular data barelyshows any reduction in supercool time compared to con-trol, unlike in the lead-shielded AmBe and Cf data sets.

6

Given the apparent electron recoil rejection, it is morelikely that muon-induced neutrons and not muons them-selves are generating events, as is often the case in di-rect WIMP detection experiments focusing on nucleonscattering. The sole peak may be caused by neutrons di-rectly entering the scintillator via the quartz vertically,but lowered in energy and changed in direction (secon-daries) with shielding. Future work will investigate thiswith full simulations. Another interpretation is the coin-cidence comes from the AmBe gammas, higher in energythan Cs, and the rejection power is a function of energy.This possibility also provides an alternative explanationof homogeneous nucleation of supercooled water droplets,which has been shown to scale with their size [20], andmay be related to the “baked Alaska” model of cosmicrays inducing a phase transition in superfluid 3He [30].From both the theoretical as well as existing experimentalperspectives, cosmic-ray-induced phase-transitions simi-lar to that in this work are not unknown [31].

In a future dark matter experiment deployed deep un-derground it may be possible to compare to existing dataon the homogeneous nucleation threshold as a functionof droplet volume in water [20, 26], to see if it occursat a lower temperature underground, where the cosmic-ray flux is considerably lower [32]. This has never beenattempted previously to the best of our knowledge.

This may be the solution to the technical and thermo-dynamic challenge of maintaining a large volume super-cooled, while lowering the temperature in search of lowerenergy threshold. A large target is necessary for a com-petitive dark matter search, while a continuous volume isexcellent for multiple-scatter characterization as we seehere. There is yet a good deal of room to go below -20◦Ctoward the min possible of below -40◦C for water [33],and we have not even studied other liquids yet.

IV. CONCLUSION

We have documented here statistically significant evi-dence that non-ionizing radiation in the form of neutrons,but not ionizing radiation in the form of gamma-rays,can initiate the solidification process in sufficiently pu-rified and supercooled water, a world first, to the bestof our knowledge. This was accomplished by logging theamount of time a volume of water spent supercooled ina quartz vessel both with and without the presence of aradioactive calibration source. We have also shown thatelastic neutron multiple scattering leads to multiple nu-cleation sites forming in a single volume of supercooledwater in a short period of time, likely from nuclear re-coils within the water molecules, another new discovery.We dub our new device the snowball chamber, echoingthe names of the cloud chamber and the bubble cham-ber from particle physics history, based on the nature ofphase transition used.

The combination of properties of this new type ofdetector are suggestive of being useful for direct dark

matter detection. The high rate of coincidence betweenthe timing of crystallization and pulses registered in themuon veto underneath the active volume of water im-pact a different field as well, namely, atmospheric sci-ence, wherein supercooled water is studied, being highlygermane [34]. Our result implies that an impurity suchas dust is not necessary for nucleation of supercooledwater in the atmosphere, but only radiation itself, ofwhich there is naturally a greater amount in the formof cosmic rays in the uppermost reaches of the atmo-sphere. This work is complementary to that of CERN’sCLOUD [35], and may also intersect tangentially withastronomy. Planets like those in the Trappist-1 systemclose to a red dwarf star producing more radiation thanour sun, if possessing (supercooled) water in their atmo-spheres, may experience different cloud formation ratesand climate than currently modelable [34, 36].

The freezing of a supercooled liquid via incident radi-ation is also an undiscovered chemistry process, hereto-fore unknown. In chemistry, radiation such as x-rayshave been used to study the microscopic properties ofsupercooled water, but to the best of the authors’ knowl-edge, supercooled water has never been bombarded byparticles (at least not neutrons) in an effort to triggerthe process by which the water crystallizes [37–39]. Inbiochemistry, our new result may affect the study of an-imals which capitalize on supercooling of their blood inaddition to or in place of more ordinary means of freez-ing point depressing like salt content adjustment, suchas the arctic ground squirrel potentially [40]. Naturallyoccurring radiation may disrupt supercooled blood.

V. ACKNOWLEDGMENTS

This work was supported by the University at Albany,State University of New York (SUNY), under new fac-ulty startup funding provided for Prof. Szydagis for 2014-18, and under the generous Presidential Innovation Fundfor Research and Scholarship (PIFRS) grant awarded toProf. Szydagis for 2017-18 by the Office of the Vice Pres-ident for Research, Division of Research, the Universityat Albany, SUNY.

We also gratefully acknowledge the advice, encourage-ment, and support of the following individuals, who pro-vided their knowledge and expertise: Albany colleaguesProf. Mohammad Alam and Prof. William Lanford, engi-neer Dr. Kevin Kenny of General Electric (GE), chemistDr. Cornelius Haase of SI Group Inc., and Profs. JamesSchwab, Fangqun Yu, and Kara Sulia of the UAlbanyDepartment of Atmospheric and Environmental Sciences(DAES). We acknowledge Prof. James Buckley of Wash-ington University St. Louis for bringing the hibernationof the arctic ground squirrel to our attention. We thankthe UAlbany machine shop chief Brian Smith for the de-sign and production of the flanges used to seal the quartz.We thank MIT Prof. Janet Conrad and Spencer Axaniof Cosmic Watch (http://cosmicwatch.lns.mit.edu). We

7

thank Prof. Kathy Dunn of the SUNY Polytechnic Insti-tute for electron microscope imaging of the nano-filter,showing non-straight, cavernous pores. We thank Profs.Keith Earle and Carolyn MacDonald of the Departmentof Physics for bringing to our attention previous publica-tions on relevant experimental results and on nucleationtheories. We thank Dr. Marc Bergevin of LLNL for anextraordinarily helpful discussion on the relative meritsof AmBe and Cf sources and Pb shielding. We also thank

Dr. Ernst Rehm and Claudio Ugalde of Argonne NationalLaboratory for useful discussions regarding their waterbubble chamber for nuclear astrophysics.

Lastly, Prof. Szydagis wishes to personally thank hiswife, Mrs. Kel Szydagis, a linguist not physicist by train-ing, for suggesting the type of filter to use to purify thewater, and for coming up with the name “snowball cham-ber” for the detector.

[1] M. Battaglieri et al., “US Cosmic Visions: New Ideas inDark Matter 2017: Community Report,” 2017.

[2] S. Fukuda et al., “The super-kamiokande detector,” Nu-clear Instruments and Methods in Physics Research Sec-tion A: Accelerators, Spectrometers, Detectors and Asso-ciated Equipment, vol. 501, no. 2, pp. 418 – 462, 2003.

[3] E. Behnke et al., “First Dark Matter Search Results froma 4-kg CF3I Bubble Chamber Operated in a Deep Un-derground Site,” Phys. Rev., vol. D86, no. 5, p. 052001,2012. [Erratum: Phys. Rev.D90,no.7,079902(2014)].

[4] C. Amole et al., “Dark Matter Search Results fromthe PICO-60 C3F8 Bubble Chamber,” Phys. Rev. Lett.,vol. 118, no. 25, p. 251301, 2017.

[5] L. W. Deitrich and T. J. Connolly, “A study of fission-fragment-induced nucleation of bubbles in superheatedwater,” Nuclear Science and Engineering, vol. 50, no. 3,pp. 273–282, 1973.

[6] V. Holten, J. V. Sengers, and M. A. Anisimov, “Equationof State for Supercooled Water at Pressures up to 400MPa,” Journal of Physical and Chemical Reference Data,vol. 43, p. 043101, Dec. 2014.

[7] A. N. Nevzorov, “Some properties of metastable statesof water,” Physics of Wave Phenomena 14(1), p. 4557.

[8] D. T. Limmer and D. Chandler, “Phase diagram of su-percooled water confined to hydrophilic nanopores,” TheJournal of Chemical Physics, vol. 137, no. 4, p. 044509,2012.

[9] E. K. Bigg, “The supercooling of water,” Proceedings ofthe Physical Society. Section B, vol. 66, no. 8, p. 688,1953.

[10] J. D. Atkinson, B. J. Murray, and D. OSullivan, “Rateof homogenous nucleation of ice in supercooled water,”The Journal of Physical Chemistry A, vol. 120, no. 33,pp. 6513–6520, 2016. PMID: 27410458.

[11] E. Aprile, J. Angle, F. Arneodo, L. Baudis, A. Bern-stein, A. Bolozdynya, P. Brusov, L. C. C. Coelho, C. E.Dahl, L. DeViveiros, A. D. Ferella, L. M. P. Fernandes,S. Fiorucci, R. J. Gaitskell, K. L. Giboni, R. Gomez,R. Hasty, L. Kastens, J. Kwong, J. A. M. Lopes, N. Mad-den, A. Manalaysay, A. Manzur, D. N. McKinsey, M. E.Monzani, K. Ni, U. Oberlack, J. Orboeck, D. Orlandi,G. Plante, R. Santorelli, J. M. F. dos Santos, P. Sha-gin, T. Shutt, P. Sorensen, S. Schulte, E. Tatananni,C. Winant, and M. Yamashita, “Design and performanceof the XENON10 dark matter experiment,” AstroparticlePhysics, vol. 34, pp. 679–698, Apr. 2011.

[12] C. E. Dahl, “The physics of background discrimination inliquid xenon, and first results from xenon10 in the huntfor wimp dark matter,” PhD Thesis, 2009.

[13] G. J. Dau, R. R. Bourassa, and S. C. Keeton, “Nuclearradiation dose rate influence on thermocouple calibra-tion,” Nuclear Applications, vol. 5, no. 5, pp. 322–328,1968.

[14] M. Szydagis, “Dark matter limits from a 15 kg window-less bubble chamber,” PhD Thesis, 2011.

[15] E. Behnke et al., “Final Results of the PICASSO DarkMatter Search Experiment,” Astropart. Phys., vol. 90,pp. 85–92, 2017.

[16] A. F. Heneghan, P. W. Wilson, and A. D. J. Haymet,“Heterogeneous nucleation of supercooled water, and theeffect of an added catalyst,” Proceedings of the NationalAcademy of Sciences, vol. 99, no. 15, pp. 9631–9634,2002.

[17] E. B Moore and V. Molinero, “Structural transformationin supercooled water controls the crystallization rate ofice,” vol. 479, pp. 506–8, 11 2011.

[18] F. Seitz, “On the Theory of the Bubble Chamber,”Physics of Fluids, vol. 1, pp. 2–13, Jan. 1958.

[19] N. E. Dorsey, “The freezing of supercooled water,” Trans-actions of the American Philosophical Society, vol. 38,no. 3, pp. 247–328, 1948.

[20] S. C. Mossop, “The freezing of supercooled water,” Pro-ceedings of the Physical Society. Section B, vol. 68, no. 4,p. 193, 1955.

[21] V. D. Aleksandrov, N. N. Golodenko, V. V. Dremov,V. A. Postnikov, O. V. Sobol’, M. V. Stasevich, and N. V.Shchebetovskaya, “Using numerical simulation methodsfor the analysis of heating and cooling thermograms char-acterizing the melting and crystallization of water,” Tech-nical Physics Letters, vol. 35, pp. 415–417, May 2009.

[22] P. Agnes et al., “Low-Mass Dark Matter Search with theDarkSide-50 Experiment,” 2018.

[23] C. McCabe, “The Astrophysical Uncertainties Of DarkMatter Direct Detection Experiments,” Phys. Rev.,vol. D82, p. 023530, 2010.

[24] D. Barahona, “Thermodynamic derivation of the activa-tion energy for ice nucleation,” Atmospheric Chemistry& Physics, vol. 15, pp. 13819–13831, Dec. 2015.

[25] G. G. Goyer, T. C. Bhadra, and S. Gitlin, “Shock in-duced freezing of supercooled water,” Journal of AppliedMeteorology, vol. 4, no. 1, pp. 156–160, 1965.

[26] D. E. Hare and C. M. Sorensen, “The density of super-cooled water. ii. bulk samples cooled to the homogeneousnucleation limit,” The Journal of Chemical Physics,vol. 87, no. 8, pp. 4840–4845, 1987.

[27] J.-I. Yano and V. T. J. Phillips, “Iceice collisions: An icemultiplication process in atmospheric clouds,” Journalof the Atmospheric Sciences, vol. 68, no. 2, pp. 322–333,

8

2011.[28] F. Aliotta, P. Giaquinta, R. Ponterio, S. Prestipino,

F. Saija, and C. Vasi, “Supercooled water escapingmetastability. watereurope meeting, zaragoza, june 12-14 (2014),” 06 2014.

[29] E. Ilotoviz, A. P. Khain, N. Benmoshe, V. T. J. Phillips,and A. V. Ryzhkov, “Effect of aerosols on freezing drops,hail, and precipitation in a midlatitude storm,” Journalof the Atmospheric Sciences, vol. 73, no. 1, pp. 109–144,2016.

[30] P. Schiffer and D. D. Osheroff, “Nucleation of the ABtransition in superfluid 3He: Surface effects and bakedalaska,” Rev. Mod. Phys., vol. 67, pp. 491–501, Apr 1995.

[31] S. Veretenenko and M. Pudovkin, “Effects of the galacticcosmic ray variations on the solar radiation input in thelower atmosphere,” Journal of Atmospheric and Solar-Terrestrial Physics, vol. 59, no. 14, pp. 1739 – 1746, 1997.

[32] D.-M. Mei, C. Zhang, K. Thomas, and F. Gray, “Earlyresults on radioactive background characterization forsanford laboratory and dusel experiments,” Astroparti-cle Physics, vol. 34, no. 1, pp. 33 – 39, 2010.

[33] C. Goy, M. A. C. Potenza, S. Dedera, M. Tomut,E. Guillerm, A. Kalinin, K.-O. Voss, A. Schottelius,N. Petridis, A. Prosvetov, G. Tejeda, J. M. Fernandez,C. Trautmann, F. Caupin, U. Glasmacher, and R. E.Grisenti, “Shrinking of rapidly evaporating water micro-droplets reveals their extreme supercooling,” Phys. Rev.Lett., vol. 120, p. 015501, Jan 2018.

[34] F. Yu, “Formation of large nat particles and denitrifica-tion in polar stratosphere: possible role of cosmic raysand effect of solar activity,” Atmospheric Chemistry andPhysics, vol. 4, no. 9/10, pp. 2273–2283, 2004.

[35] E. M. Dunne, H. Gordon, A. Kurten, J. Almeida, J. Du-

plissy, C. Williamson, I. K. Ortega, K. J. Pringle,A. Adamov, U. Baltensperger, P. Barmet, F. Benduhn,F. Bianchi, M. Breitenlechner, A. Clarke, J. Curtius,J. Dommen, N. M. Donahue, S. Ehrhart, R. C. Flagan,A. Franchin, R. Guida, J. Hakala, A. Hansel, M. Hein-ritzi, T. Jokinen, J. Kangasluoma, J. Kirkby, M. Kul-mala, A. Kupc, M. J. Lawler, K. Lehtipalo, V. Makhmu-tov, G. Mann, S. Mathot, J. Merikanto, P. Mietti-nen, A. Nenes, A. Onnela, A. Rap, C. L. S. Redding-ton, F. Riccobono, N. A. D. Richards, M. P. Rissanen,L. Rondo, N. Sarnela, S. Schobesberger, K. Sengupta,M. Simon, M. Sipila, J. N. Smith, Y. Stozkhov, A. Tome,J. Trostl, P. E. Wagner, D. Wimmer, P. M. Winkler,D. R. Worsnop, and K. S. Carslaw, “Global atmosphericparticle formation from cern cloud measurements,” Sci-ence, vol. 354, no. 6316, pp. 1119–1124, 2016.

[36] M. Gillon, E. Jehin, S. M. Lederer, L. Delrez, J. deWit, A. Burdanov, V. Van Grootel, A. J. Burgasser,A. H. M. J. Triaud, C. Opitom, B.-O. Demory, D. K.Sahu, D. Bardalez Gagliuffi, P. Magain, and D. Queloz,“Temperate Earth-sized planets transiting a nearby ul-tracool dwarf star,” Nature (London), vol. 533, pp. 221–224, May 2016.

[37] R. Schiller and I. Kules, “Radiation chemistry of su-percooled water,” The Journal of Physical Chemistry,vol. 75, no. 19, pp. 2997–2999, 1971.

[38] D. E. Hare and C. M. Sorensen, “Raman spectroscopicstudy of bulk water supercooled to 33c,” The Journal ofChemical Physics, vol. 93, no. 1, pp. 25–33, 1990.

[39] A. Belitzky, E. Mishuk, D. Ehre, M. Lahav, andI. Lubomirsky, “The source of electro-freezing of super-cooled water by polar crystals,” vol. 7, 12 2015.

[40] “What the supercool arctic ground squirrel teaches usabout the brain’s resilience,” Scientific American.