Maximum Evaporation Coefficient of Water

---

ENGINEERING, DESIGN, AND PROCESS DEVELOPMENT

K. C. D. HICKMAN' Research Iuborafories, Eastman Kodak Co., Rochesfer, N. Y.

A

VAST body of knowledge has accumulated ( 4 ) covering the over-all process of evaporation and condensation of water; yet, curiously enough, little is known of the actual mechanismhon- and a t what rate HzO molecules transfer across the interface between liquid and vapor. The consensus ( 7 ) , backed by different experiments ( 2 ) ,has been that not more than one molecule of water in 25, approaching the surface from either side, actually merges with the interfacial layer. This paper suggests that this opinion is in error and that it is unreasonable to expect a fixed or characteristic evaporation coefficient for water (or other liquid). The instantaneous coefficient is believed to depend on the circumstances-whether the surface is resting or disturbed, and to what, extent the equilibrium is displaced. The relevant' question, and one that does not yet appear to have been answered satisfactorily, is whether there are conditions a t which the evaporation coefficient of water is unity. This paper answers the question affirmatively and thus removes water from the class of anomalous liquids, as t'o evaporation. Coefficient Can Be Measured by Evaporation into Nearly Saturated Vapor or into High Vacuum

A valuable resume of research on the water problem, toget'her with some new experiments, has been given by Pruger ( 7 ) , working in Mache's laboratory in Gienna (1937 t'o 1940). By following in part Pruger's terminology, the rate of evaporat'ion is expressed in three interconvertible units,

k-,= weight evaporated per unit t,ime and area viith unit fall of Dressure L , = nuhber of complete niolecular layers evaporated per second quantity observed evaporating E = evaporationcoefficient = quantity calculated Experimental measurements of evaporation coefficient fall into two broad classes-those done with t'hrottling into a nearly saturated vapor and those done without restriction into high vacuum. The choice of the former has often been a compelling one-namely, to diminish loss of heat, a t the liquid surface and t,o avoid freezing and gross uncertainty concerning the temperat,ure of the outermost layers. The restricted methods comprise the following: 1. Evaporation or drying into a foreign gas, usually air. The quantitative problem is then to calculat'e the resistance to escape of vapor through the gas. The escape coefficient, f, is dominant and highly disproportionate to the evaporation coefficient. 2. Evaporat,ion into the nearly saturated vapor. The liquid evaporat,es a t pressure p' and t'emperature t into a vapor a t lesser pressure p . The optimum rate of escape of vapor is then assumed llf

to he 583(p - po) - > where ( p - p o ) subst,itutes for p (milliT meters of mercury) in the Langmuir equation (a),or for (pf) in Equation 1. Experimentally, this method involves the collection of vapor after it has passed the constricted opening of the evaporator and determination of the actual t,emperature of the evaporating surface to the limit, of accuracy wit'li a fine thermocouple (Baranaev, 5, and Pruger, 7 ) , or by the simultaneous measurements of surface tension (Blty, 1 ) . These investigators dernonstratcd a drop in temperature of as much as 3.5" C. before the thermocouple broke the surface of the water, but even allowing for this, the emission of vapor Iras depressed below the calculated optimum by 25 to 100 times. (According to 1 Present addreas, 13G Pelhain Road, Rooheaer 10, K.Y.

1442

Pruger and Baranaev, E = 0.01, and according to Alty, E = 0.04. Even by introducing all allon-able corrections, Pruger was unable to raise the value above E = 0.04.) Unrestricted Evaporation. The conditions most likely to realize the maximum evaporation coefficient exist when a clean random surface of water is continuously exposed to high vacuum and instantly removed. The device most suitable for producing these conditions without undue chilling or formation of ice is a rapidly floiring stream issuing from one orifice and immediately entering another. The flowing-stream tensimeter, in the form developed for phlegmatic liquids, is unsuitable, since it cannot be employed satisfactorily a t a saturation pressure above 200 microns. Water a t the freezing point exerts a pressure of about 4500 microns, and the latent heat of evaporation is many times that of most heavy liquids. Severtheless, with suitable alterations, a stream of water was evaporated into high vacuum for periods ranging from a few seconds to many minutes before ice appeared, and E was determined from the quantity of vater collected in a distant condenser. The experiment showed that the evaporation coefficient of water is a t least 25 times higher than previously found and probably approaches unity for clean, very new surfaces. Modified Flowing-Stream Tensimeter Is Used for Experimental Work

The flowing-stream tensimeter was compressed vertically to expose a short, rapidly moving stream for evaporation. The emitting and receiving jets were construct,ed in the form of a water aspirator, used in reverse, with the vapor from the exposed stream led by a short, wide pipe to a large condenser, to be immersed in dry ice and acetone. The system was evacuated through a second large freezing t'rap, xhich alternately returned water to the system or maintained a source of high vacuum (20 to 30 microns) during runs. The apparatus contained an adjustable vane for controlled throttling of the vapor between the mater jet and condenser. The vane was fully open in the present, experimenk

As shoxn in Figure 1, the ejector jet, A , is thrust through t,he rubber stopper, B , and plays into the receiving Bt, C, which is held in stopper D. The glass tubular housing, E, 5 em. in diameter, is constricted at. F to allow draining of spillage from the jets which occurs during start-up or after faulty runs. The space between the lower jet and the housing is nearly closed by a wire spiral. The tube above the ejector jet, A , contains a roll of stainless steel wire mesh which smooths out, the st,ream of water. This tube, in turn, is connected with an inverted flask which serves as a centrifugal trap for bubbles and other nuclei. The flask is supplied through the pipe, G, from a glass centrifugal pump, H , fitted with a four-bladed, stainless steel impeller on a Trevoy shaft. The pump is closed a t the base by a large rubber bung cont'aining the glass entrance pipe, J . The circulating water passes from the receiving jet, C, to a double reservoir system, which has one bulb, K , loosely packed wit,h absorbent cotton. The vapor from the housing, E, passes by the two tubes, L and M , t o the ice traps, N and P , which, in turn, are connected with the master freezing trap, Q, monitored by the Pirani gage, Q'. Primary vacuum is secured from the two rotary oil pumps, R and 8, while the system can be vented to the atmosphere or to a small flaak by the tube, 2'. A drainage and ventilating channel, U , connects the base of housing E and the larger reservoir bulb with the base of trap &. TFO paired glass thermometers take

I N D U S T R I A L A N D E N G I N E E R I N G C H E h.I I S T R Y

Vol. 46, No. 7

ENGINEERING. DESIGN, AND PROCESS DEVELOPMENT

Figure 1 .

High Velocity Jet Tensimeter

the temperature of the stream before and after passage through the jets. An oil manometer, V , which can be short-circuited by the stopcock, W , measures the pressure of steam in housing E. Reference vacuum i8 produced by a separate mechanical pump, W', not shown, and is checked by the Pirani gage, Ti'. While it is diagramatically correct, Figure 1 bears little resemblance to the actual apparatus, which was heavily supported by rods, clamps, and adjustment screws. Rubber stoppers and connections were retained in the interests of flexibility. There was no indication that harmful contamination was introduced with respect to the experiment in hand. The over-all vacuum and the pressure of water over jet A were meamred by the three-limb mercury barometer tubes, X. The steel shaft of the Trevoy (8) seal was lubricated variously with water (containing a trace of ammonia to delay rusting), triacetin, or castor oil. The apparatus held a vacuum excellently when it was shut o f f from the pumps. Operation. Vacuum was induced by the roughing pump, S, which could always be relied on to pull a vacuum of less than 1 mm. of mercury absolute. Ordinary laboratory-distilled water was admitted by tube Y until reservoirs K and K' were filled. Pump H was energized intermittently until a steady circulation of water was secured through jets A and C, at which time bubble trap G' was full and K and K' one third full. Vacuum was relieved and the water drained out and replaced two or three times until the glassware, rubber connections, and cotton filter were considered clean. After a working filling had been obtained, trap Q was half filled with acetone, and enough dry ice was added to leave a small excess undissolved. Until the actual moment when a reading was taken, the large Dewar vessel in the center of Figure I was, of course, removed and kept in readiness. July 1954

The temperature of the water in the system fell rapidly to below 8' C. and the liquid jet, which had been splattering violent,ly, quieted down to a clear transparent stream. It was useful to stop pump H occasionally during the cooling period to be sure that all air bubbles had been removed from G, G', and A. With circulation re-established, trap Q was allowed to warm up and return ice water to the system, after which it was again supplied with dry ice. The charge of water was allowed to circulate a t least 3 hours to remove the last traces of air, suspended matter, and nuclei for condensation. To make t'he final run, trap Q was thawed and then refrozen so that it was covered with only a thin lager of ice. Pump S was replaced by pump R (the Pirani gage, Q', should read less than 0.1 mm.) after which stopcock TV was closed, isolating the reference vacuum in the right limb of the oil manometer, V . Louvers LL and M M remained fully open. Lamp Z flooded housing E with light and u-armth, and it was ascertained that no trace of splashing was coming from the jet. The upper and lower housing, E, the connecting tubes, and traps were completely free from ice and water. The t'emperature of the water a t both thermometers should be 7 " to 8" C., according to the author's experience. If it was warmer, the vapor stream became unmanageable during the run period; if it was cooler. ice always appeared. At this point, the t,hermocouple, Th, registered 8 " to 10" C. (when light 2 was off) and the oil manometer recorded 4 to 5 mm. of mercury. The manometer, X , should register a ivater pressure behind jet A of a t least 50 cm. of mercury. The large Dewar vessel, loaded with acetone and dry ice, was readied by the operator who wore protective gauntlets. h second operator undamped the long rubber tube, T , att'ached to air-filled flask, Z", and held it closed by pinching in a loop. Holding t,he pinched tube in one hand and a timer in the other, the second operator watched the stream and noted both thermocouple and oil manometer minimum readings, while the first operator raised the Dewar vessel to immerse traps P and N . The timer was started a t the moment of immersion. If, and as soon as, ice appeared a t the jet, tube T was released, the timer was stopped, and stopcock W was opened, after which the mechanical pumps and water circulator were shut down as rapidly as possible. Admission of the limited volume of air from 2' effectively stopped all movement of water vapor without causing enough pressure to drive splashings into the traps. The first operator had already lowered the large Dewar vessel. It was now found convenient t'o keep a small region, N ' , cold with a wad of cotton soaked in acetone-dry ice, while the rest of the ice in t'rap N thawed and fell to the measuring section. Any small droplets left on the upper portions rapidly distilled to t'he cold spot, N', and were thawed to join the main yield. During the effective portion of the run, the temperature of the water fell about 1 C. in 5 minutes, but the two thermometers showed only a fraction of a degree difference a t any one moment. The majority of the runs-perhaps a hundred, performed over a period of 3 months-lasted less than 5 seconds before ice appeared. These were discarded without admission of air, and t,he apparatus was given time to ready for another trial. The successful runs were almost invariably preceded by a long uneventful period of degassing. Origin of Ice. Ice was not first generat,ed on the exposed portion of the stream, but appeared to come from one of three sources-from inside the receiving jet, whence a layer of ice suddenly started bo climb up the wall; from the freezing of a minute drop on the rim of the lower jet; or from inoculation from the vapor phase. The vapor, after emission by the liquid, suffered a t least a 14 to 1 isentropic expansion as it flowed to the trap, N . It was this expansion, rather than abstraction of latent heat from the surface of the stream, that produced the low reading of - 15' C. on the thermocouple, and the true vapor temperature '1vas probably much lower. The vapor was now highly supercooled with respect t o ice and availed itself of any

INDUSTRIAL AND ENGINEERING CHEMISTRY

1443

ENGINEERING, DESIGN. AND PROCESS DEVELOPMENT

W Figure 2.

F

E Icing Sequence of Tensimeter l e t

Time from Start, Sec.

A B

Interval between Pictures, Sec.

65 67 67.375 68.250 70.310

C W E F

2.0

0.375 0.850 2.06

95.3T0

suitable nucleus. A microscopic ice crystal may be su-ept backward by the turbulent vapor stream and may inoculate the fllm of water on the rim of the receiving jet. The author has observed these vapor-borne ice crystals in dark ground illumination. An essential function of the long degassing period is to remove from the vapor the potential nuclei for ice. Production of ice from inside the receiving jet is shoivn i n Figure 2. These photographs, enlarged from a 35. mm. film, reveal that transition from clear to iced condition takes place in a fraction of a second. Under Certain Conditions, Evaporation Coefficient of Water Approximates Unity

The evaporation coefficient is calculated from the formula

25.00

TT'

6 = -

iF;Aft \/ 279

0.0583~

(1)

The problem is the estimation of probable values for .f and T . hiany objections have been raised to employing the Langmuir formula for rate calculations. It has been argued that the formula m s originally intended to define accommodation coefficients under reversihlc conditions a t near equilibrium and it should not be used to predict rates, because rate processes are inherently different from equilibrium processes. The author does not agree and maintains that an equilibrium becomes a rate process as soon as it is disturbed, and a rate becomes an equilibrium process if it is condwted for a short enough time-that is, the two procewes conirerge asymptotically to the same point under limiting

Table I. Calculation of Evaporation Coefficient Dimensions of Water Stream Surface Head, Length, a r e a ( A ) , e m , Velocity, R u n S o . Diam.,crn. em. sq o m . Hg. cm./seo. 1.87

1,65 1.65 1.39 1.19 1.19 1.1s

*

21 24 35 10 56 50

56

798 798 962 514 1216 11.50 1210

Time of ExVol., posure in mi /see. Jet, See. 76 6 76.6 92.2 49.3 94.5 89 , 2 96.5

0.00216 0.00188 0.001% 0.00228 0,00099 0.00104 0.00093

Cal Extracted/ Mi. Passing 0.348 0.389 0.298 0.555

0.223 0.194 0.195

Temg, oi Water, Enter

Leave

7.5 7.0 7.0 7.0 6.6 6.4 6.0

7.16 6.6 6.7 6.4 6.3 6.2 5.8

c, Av. 7.3 6.8 6.8 6.7 6.4 6.3 5.9

Outer 0.1 Mrn. of Shell Max. Av. temp. Av. temp, temp. in2hell", of surdropa, O C. C. faced, C. 3.06 3.46 2.66 4.956 1.79 1.56 1.68

6.0

5.3 5.5 4.5 5.6 5.6

5.2

4.4 3.5 4.1 2.0 4.7 4.8 4.4

8 5 X (cal. extracted/ml. passing). passing). Temp. of n'ater entering minus ( m a l . temp. drop/?).

b 8.05 X (cal. extraoted!ml. C

1444

INDUSTRIAL AND ENGINEERING CHEMISTRY

Vol. 46, No. '7

ENGINEERING. DESIGN. AND PROCESS DEVELOPMENT conditions. In the present instance, the time element, 0.001 second, is sufficiently small that an equilibrium equation can be applied. The area of the evaporating stream varied from 1.87 to 1.19 square cm. in the Eeven runs, with an average of 1.40 square em. The area of the tube leading to the trap was 20.4 square cm. and the length to the condensing region was 20 cm., from which may be assumed a net admittance of 14 square cm. If the temperature of the evaporating surface of the water was 6" C., the saturated pressure would be 7 mm. and this pressure, expanding to the limit set by the admittance of the tube, would drop to 7 X

%

= 0.7 mm., which coincides with the observed values from

the oil manometer, v , of 0.5 to 1 0 mm. However, although the water is evaporating into an eventual vacuum of 1 mm., it is in contact with a density of vapor just outside the surface, corresponding to a pressure of 7.0 mm. The majority of the molecules are moving rapidly outward but, because of their random velocity and directions, are slowed by mutual collision. Only one calculation ( 5 ) and no experimental attempts have been made to determine the effective back pressure of such a freely emergent stream of vapor. Fraser predicted that the feedback of molecules to the evaporating surface could not exceed 50% and would generally be less. An extrapolation of the data of Trevoy (8) for 2 ethylhexyl phthalate esters on linear-log paper crobses the 60% line a t 10,000 microns. There is, of course, a great specie difference betxeen water and heavy ester molecules, and it IS not clear whether an evaporation coefficient or an escape coefficient, or both, was dealt with in the ester experiments. An escape coefficient of f = 0.65 =t 20, which combines retardations due to apparatus and vapor cloud, is used in the following calculations. Lack of information regarding the temperature, T, of the evaporating surface of the jet leads to uncertainties in the corresponding steam table vapor pressure, p , of about the same magnitude ( +30%). Temperature and, hence, vapor pressure diminish as the mater pasFes through the gap, and the average effective values are estimated with sufficient accuracy to be halfway betneen the extremes By using the formula, u = 42gh,in appropriate units, the stream velocity for a head of water of 56 em of mercury i*

v = 1216 cm. per second The time of exposure in a gap 1.2 cm. long is 0.001 second. Table I records that in run 5, for example, 5.5 X 592 calories were extracted from the 17,750 ml. of water that flowed through the gap during the 188 seconds of the experiment; hence, the total heat dr3p in the liquid was 0.183" C. If, and as an admitted guess, it is assumed that the heat drop was limited to the first 0 1 mm. ~ i i t h i nthe surface of the jet, the average temperature of this layer vould be 1.51" C. lower a t the end of its travel through the gap Its average temperature throughout the period of travel would be 0 75" C. lower than the incoming \vater. It

Table 1. Time of

Run

(0,

Sea.

90 20 27

72 183 188 49 6

4.0 1. 0 1.25 1.75 6.5 5.5 15.5

could be assumed, further, that the inner face of the first 0.1-mm. shell remained a t the same temperature as the bulk of the stream, while the evaporating face would be 1.51' C. lower, average throughout the gap. If it were assumed that the cooled shell of water was 0.05 or 0.025 mm. thick, the temperature of the outermost layer would be lowered twice or four times as much. In the absence of any experimental data, 0.1 mm. has been adopted as the shell thickness of thermal gradient. Two minor corrections have been neglected entirely in the calculations-the volume of vapor in the housing and trap before refrigerant was applied and leakage of vapor through the annulus between the water stream and the receiving jet. The former condensed to less than 0.05 ml. of water, and the latter amounted to less than 5% of the total vapor emitted by the exposed stream. If the surface temperature within the jet was actually below 0" C., this correction would be less than 2%. Of the many possible combinations of corrections, three have been chosen:

No corrections, temperatures as observed, f = 1 Temperature corrected to average of surface of outer shell, *f = -l Temperature corrected to average of surface of outer shell, f = 0.65 The experimental data and calculations are listed in Table

I. All the runs gave substantially the same values for evaporation coefficient. Runs 1, 5, 6, and 7, which lasted more than 60 seconds, provided the following conclusions : 1. The crude experimental evaporation coefficient, e = Poha Pcslcd. without any corrections, e averaged 0.243. 2. With conservative allowances for temperature and pressure gradients, e averaged 0.424. 3. If the temperature of the outer evaporating skin had actually fallen to an average of -6' C., which the freezing within the jet could well justify, and if a vapor escape coefficient of 0.65 is allowed, e is unity. 4. The rate of evaporation, expressed as K , X 10-6, ranges from 3800 t o 7800, compared with 150 t o 270 reviously reported. 5 . These high rates of evaporation, whic! reveal water as a normal liquid, were determined on a surface 0.001 second old and which lost 400 to 1100 complete molecular layers during the period of exposure. Conclusion

h clean, new surface of water exhibits an evaporation coefficient not less than B = 0.25 and the coefficient probably approximates unity. These values are 25 to 100 times greater than previously reported. Nomenclature

A = area of evaporating surface, sq. cm. d

= liquid density, gram/ml. = 1 for water (sufficiently

E

= escape coefficient for vapor = ratio of number

exact)

of molecules reaching condenser to number actually evaporating = ( p - p,,) in restricted vapor experiments

Calculation of Evaporation Coefficient (Confd.)

S a t d . Vapor Pressure, M m . Hg, a t Av. Temp. of Surface Shell Bulk (PI) (P*) (P3) 6.27 7.0 7.7 5.9 6.7 7.4 6.8 7.4 6 14 6.3 5.29 z.35 6.8 6.4 , 2 6 5 6.8 7.15 6 27 6.0 6.9

Evaporation Coefficient Calcd for

f - 1

L v , Mol. layers/

Q'

P2

p3

f = 0.65, p'

880.

0.254 0.346 0.308 0,226 0.314 0.258 0.284

0,227 0,309 0,278 0 187 0,294 0,244 0.267

0 207 0 275 0 265 n 161 0 277 0 214 0 255

0.391 0.532 0.478 0.347 0,467 0.392 0.446

119.5 1170 900 676 417 379 386

R a t e of Evaporation Ku x 10-6 Calcd. for f 1, f = 0.65, P' P'

-

3810 5070 4570 3270 4560 3780 4200

5860 7810 7040 5020 7010 5s40 6460

Other workers 270 (7) 1.50 (7) 6 1 3 (7)f

... .. .. .. ...

d Temp. of water entering minus max. temp. drop. 8 RIache, glass water. f Theory.

July 1954

INDUSTRIAL AND ENGINEERING CHEMISTRY

1445

ENGINEERING, DESIGN, AND PROCESS DEVELOPMENT = gravitational constant, cm./sec.P = velocity head, em. of Hg K , = rate of evaporation (Priiger, 7 ) - grams/(sec.) (sq. em.) ( p - p,) L, = rate of evaporation, molecular layers/sec., a t temperature T and saturated vapor pressure p - K , , ( M ~ ) ~x. Q10-7 ‘TI = molecular weight = 18 for wat’er P = saturated vapor pressure of water, mm. of Hg, at temture T Po = pressure (vacuum), mm. of Hg, into which water evaporates with pressure p ‘ 1 = total weight of water evaporated experimentally, grams t = duration of evaporation, see. T = temperature of evaporating surface, C. liquid evaporating, gmms,/sec. e = evaporation coefficient, W E = ___9

h

w=

583p

44

e

0.0583

-

4$\ q

=

pAft

References

(1) Alty, T., Phil. M a g . , 15, 82 (1933). (2) Alty, T., and Mackay, C. .4., Proc. Rou. Snc. (London),149A, 104 (1953). (3) Baranaev, J. R., J . Phve. Chem. (U.S.S.R.), 13, 1635 (1939). (4) Doraey, N. E., “Properties of Ordinary Water Substance,” New York, Reinhold Publishing Co., 1940. ( 5 ) Fraser, R., personal communication. (6) Hickman, K., IND.ENG.CHEM.,44, 1892 (1952). (7) Priiger, W., Z . Physik, 115, 202 (1940). (8) Trevoy, D. J., IND. ENG.CEEM., 44, 1888 (1952). RECEIVED for review September 38, 1953. ACCEPTEDAIarch 1 , 1954. Communication N o . 1612 from the Kodak Research Laboratories.

K. C. D. HlCKMANl

AND

W. A. TORPEY

Research Laboratories, Eosfmon Kodak Co., Rochester, N. Y.

A

KY demonstration that a a t e r can distil a t the high calculated

rate ( 6 ) ,emphasizes its failure to do so (1. b) in everyday circumstances. Previous measurements of evaporatioii coefficients have reached only 1 to 4% of theory ( S ) , and n ater boils or bumps because it cannot emit sufficient steam from its existing surface. To suppose that there is a repressive film raises many new questions. Is it mechanical-due to gradients of heat and escaping vapor which disturb equilibrium? Is it caused by orientation or polymerization of the polar natcr molecule7 Is it a layer of chemical impurity, comparable in effect 771th a monolayer of insoluble fatty acid? Or is it a combination of all these and perhaps other factors, and if so, does one factor dominate the rest? It is difficult to understand how surface polymerization and orientation of the water substance can reduce evaporation a hundredfold or more, since all coniponents of liquid water are believed to be rapidly interconvertible. If, on the other hand, contamination is responsible, the degiee of obstruction should vary with the nature and quantity of impurity. Thus, a positive test for contamination should be variations in the vapor emission of different eamples of water, examined under identical physical conditions. A confirmatory test nould be the generation of a schizoid surface pattern, m here a freely evaporating area could be seen pushing an obstructive film to one side. This paper examines the questions of the variable emission of n ater and the torpidity of its surface (4). Vacuum Still Is Used to Demonstrate Torpidity of Water

To secure a working crater in a torpid fluid, the pressure must be low enough for vapor recoil to deprrss the surface. Water near the freezing point has a saturation pressure of 5 mm. of mercury, so that a vacuum of 3 to 4 mm. or lor~erwould be necessary for the formation of patterns. If water evaporated without obstruction, it nould freeze instantly nhen exposed to such a vacuum, and the vacuum itself could be maintained only if there waq a coextensive vapor path between the evaporator and condenser. If, on the other hand, the surface was substantially 1

Present address, 186 Palham Road, Roclicster 10 h-

1446

T

blocked, there is just the chance that water could be evposed to vacuum for an observational period before freezing and that vacuum could be maintained through a constricted vapor path, such as the neck of a flask. Apparatus. This reasoning dictated the construction of the simple pot still shown in Figure 1. Coil A , under the flask in Figure I, represents a commercial radiant heater, run a t 750 watts. Coil B, immersed in the water, consists of a few 1-cni. turns of Nichrome wire and can dissipate 5 t o 50 watts. Acting as a “torpidity tester” (41, it serves to open working craters in resistant samples of water and to maintain working areas in other samples. (All the patterns produced by this coil or by eyternal heating of the flaak could be obtained by prev, arming the water. The coil n-as a convenience, not an epaential.) A camera, C, is indicated on the right, and a Fresnel lens, D , about I5 square cm , is shown between the lamp and the flask, on the left. The flask is kept free from dew by air blasts. Operation. The apparatus was cleaned with acetone, follon ed by a large quantity of distilled water from the laboratory’s aluminum supply lines. The manometer mas charged nith mercury and sealed in position: the flask was half filled with distilled water, and vacuum was induced by a mechanical oil pump. Heaters *4 and B and the air blasts were energized, and watei soon began to distil into the trap v i t h many explosive bursts a, degassing proceeded. Trichlorethylene was placcd in the trap, and sniall pieces of dry ice were added to maintain a pressuie of S to 10 mm. in the assembly and prevent undue loss of n ater vapor to the pump. The pump oil r a s changed frequently and regenerated by blowing with air overnight. First Experiment. After half an hour, all bumping ceased and the trap was filled with dry ice. -4few minutes later, the manometer read less than 2 mm. and a crater appeared on the water, surrounded by symmetrically moving pits or “cratercts.” The photograph, Figure 2, was taken and a t once the surface froAe solid. The trap was allowed to warm up, and soon the ice began to melt in the flask, presenting a beautiful pattern of cratercts, just above the internal heater, as shoan in Figure 3. Figure 3 and the following photographs Rere taken v i t h the camera aimed a t the underside of the surface, as shown in Figure 4, because clearer pictures “ere obtained that nay. The craterets now appeared a9 humps or pimples.

I N D U S T R I A L A N D ENGINEERING C H E M I S T R Y

Vol. 46,No. 7