Water frosts were deposited on a liquid nitrogen-cooled plate from a humid air stream. The plate formed
the bottom of a high-aspect wind tunnel. Reynolds numbers were varied from 3770 to 15,800, gas-phase
humidities from dew points of 14" to 58"F, and gas temperature from 34" to 93°F. Frost densities and
thermal conductivities were determined and the heat and mass fluxes measured. The data were correlated
b y a simple analytical model which emphasized the internal diffusion processes within the frost. Density
measurements were also made within the thin frost layer. These measurements indicated that within the
accuracy of the method used there are no significant density gradients in the frost. This result, though not
thoroughly explained, i s believed due to a combination of internal diffusion and nucleation with particle
transfer by thermal diffusion forces.
- V
- &i
w-
- C
Ln
0
- $ 0
- +
0 0
0 Tg =93OF, T,= 15OF
I ,
0
I I I I
0.05
I , ,
0.10
I I I I
0.15
0)
I @ T g = 7PF, Ts = 8OF
Distance from Cold Wall (inches) @ Tg =50°F, Ts = 5OF
Figure 2. End of run temperature profile within frost @ T, =34OF, Ts = 2OF
distances from the cold plate. For each set of system condi-
tions, the temperature profiles were obtained for the runs
with different deposition times. These profiles indicated that I .o I I I I I I I 1 I 1
during the transient state of the frosting process, the tem- 2x10~ io4 4x11
perature a t any depth within the frost decreases with an Reynolds Number
increase in the deposition time (Shah, 1968). For a few system
conditions, the temperature profiles were obtained for two Figure 3. Heat transfer coefficient vs. Reynolds number
identical repeat runs. These profiles were found to be in agree- (at various gas temperatures) under frosting and nonfrost-
ment within 10% deviation in the temperature values. ing conditions
Typical experimental results are shown below, when a
comparison with analytically predicted values is made. It Gas-phase mass transfer coefficients were calculated using the
can be noted in Figure 3, however, that using the measured Lewis analogy (Eckert and Drake, 1955) and the theoretical
frost surface temperature and experimental heat fluxes, values for the heat transfer coefficient are shown in Figure 3.
calculated gas-phase heat transfer coefficients agreed well This technique led to the predicted values of the mass transfer
(within 10%) with those predicted from the theoretical equa- coefficient shown in Figure 4. Experimental values of this
tion of Sleicher and Tribus (1957) for heat transfer in the coefficient were obtained from the mass deposition rate data
entrance region. The theory assumes a nonfrosted flat plate, and the water vapor partial pressure in the bulk gas and at
whereas in the experiment there is obviously an effect of a the frost surface, the latter being assumed to be the vapor
rough surface. Other experiments, not described here, wherein pressure of ice at the frost surface temperature. There is some
dry nitrogen gas was passed over the nonfrosted copper test Bcatter, which is believed to be due to slight inaccuracies in
plate yielded heat transfer coefficients within 3% of those the surface temperature measurements, but the agreement
predicted using the Sleicher and Tribus method noted above. between theory and experiment is satisfactory.
A microscopic, detailed examination of the frosting process Table 1. Empirical Correlation of Frost Thermal Con-
ductivity with Average Frost Density and Temperature
would reveal it to be extremely complex. The movement
of water vapor to the frost surface, if temperatures in the gas Equation Conditions
boundary layer are not too low, is believed to be similar to
other gas-phase transport processes. While not simple, such
IC,3.875
= x 10-6~1.441+ 4.08 x 10-9
- 0.025)Ta.0S5
(PI T
> 0.025
< 460
processes are describable and amenable to predictive calcula-
tions. Heat transfer and deposition rates will, however, be a
IC, +
= 8.49 X 10-'7T5~44 6.86 X p / > 0.025
(fit - 0.025)T4.84 T > 460
function of the frost surface temperature and therefore of the
properties of the frost layer. The frost is not a simple thermal
+ 8.17 X 10-6p,T'.74
k f = 6.58 X 10-6T1.272 < 0.025
T < 460
resistance. The thickness increases with deposition time, the p, in g/cma; IC, in Btu/hr-ft-OIt; T i n OR.
density increases (whether or not deposition occurs) as it
matures, and the resulting average thermal conductivity is a
function of time as well as deposition conditions. As noted
I I I I I I I I I I I
earlier, however, all studies of heat transfer across a frost
layer have shown that the heat flux soon becomes almost lxperimental Data :
constant with time, even though the thickness continues to Gas Cold Surface
increase. With constant gas-phase conditions, this would lead Tempe rat u re Te mp e r atu re
one to predict that during the period of quasi-steady heat 0 93OF + 15OF
flux, the frost surface temperature also becomes essentially 0 75°F + 8OF
time-independent. This assertion was shown to be true in A 5OoF t 5OF
earlier work (Brian et al., 1969) and also in the present studies. 0 34OF + 2OF
A very simple analytical model may be formulated during
this quasi-steady-state heat transfer portion of a frosting - Theoretical Predictions from
process. The development of the model is summarized else- Lewis Analogy
where (Brian et al., 1969). A key assumption in this model
is that water vapor diffuses away from the frost-gas surface
into the colder frost interior by virtue of the equilibrium
partial pressure gradient corresponding to the temperature
gradient existing a t the surface. I n some "undefined" way,
this water penetrates the frost to produce a frost of uniform
density, though this density is time-variant. This assumption
is examined in detail later and some experiments are described
to support the model.
The three basic equations which must be solved are as
i
follows: 0
Energy Balance
@ Tg=93OF; Ts=15OF
@ Tg=75OF, T s = 8OF
9 = h ( T , - T,) + - P , ) A ~ T= ICf9s(dT/dz)s= O Tg =5OoF, Ts = 5°F
i , V S- T,)/6 (1) @ T, = 34OF, Ts= 2?
Mass Balance
'"2x IO 3 ~ 1 0 ~
N Kg(pg - PA) = Pi(ds/d) + 6(dp,/de) (2)
103
Internal Diffusion
(dp,/W = -PVA[(I - P , / P ~ ~ ~ ) / ~ I ( ~ T(3)
/~&
I n addition, one must have, for water frost, a relationship
of the form k , = f(T, p,). For water frost deposited on a flat
plate, such a correlation was developed from the present
experimental data (Shah, 1968) and the data of Brazinsky
(Brazinsky, 1967; Brian et al., 1969). Expressed analytically,
this correlation yields the equations shown in Table I.
Also, to solve Equations 1 to 3, initial values of 6 and pf
are required. To test the applicability of the model, initial
values of 6 and p , were chosen equal to the first experimental
values taken in a frosting test. The equations were then
solved by finite difference techniques, using a digital computer
to predict the time variation of frost thickness, total mass de-
posited, average frost density, and heat flux. The predicted
values are shown and compared with experimental values in
Figures 5 and 6 for several widely differing deposition condi-
tions. The agreement is excellent. The conclusion one can
draw from comparisons such as those illustrated in Figures
5 and 6 is that, given only initial values of p , and 6, with a
- 0.40
System Conditions
R e = 14,700
G 0.30
HumidityZGas dew point
"'
e
t 0.20
.-
:0.10 0.15
at 29.2"F
Tg= 7 O o F
I C - 0 - 0 -v
0.14
> 0,13
0
N
-
*
A
.-
1,000
t
/ A'
0
Y
LL
$ 0.04
40 80 120
Time (rnins.) Time (rnins.)
Figure 5. Experimental facts and theoretical predictions
of the effect of gas temperature on time variation of frost
properties and heat and mass fluxes 0,05- einitioi 3 0 mins.
0 Re = 5 7 9 0 ; humidity = gas dew point a t 27.2 OF; T, = 3 4 O F I I I I I 1
A Re = 5 1 0 0 ; humidity = gas dew point a t 27.4 OF; T, = 9 3 OF
-- Theoretical curve
Experimental Conditions
''I7- -
Re = 15,800
0.16 - H u m i d i t y = G a s dew point at 29.8v
30 35 40 45 50 -
-- Tg = 7 l 0 F
Time ( m i n s )
0,20 I -1 0.14
m 0.15
U
VI
v) Svstem Conditions 1:. 0.13 -- A / 9 4 .
bIy
c
a)
c l-
X n) - -
n 0.12
V Q
0.11 - -
.- $ c.10-
c
ln-
0
0,14
0
30
. 1
35
2
40
H u m i d i t y = G a s dew p o i n t
L
45
-
a t 29.2OF
A
50
f
. ~
0.09 -
0.08
- :/: -
-
-
Time ( m i n s . ) 0.07 - -
1 om - -
0,08 vr i F i g u r e 9.
Time (min. )
Density distribution within frost as function of t i m e
0,O7 ~ ~ ~ ~ ~
30 35 40 45 50
Time (mins.) the end of the run varied between 0.091 aiid 0.256 inch. T o
F i g u r e 8. C o m p a r i s o n of t h e o r e t i c a l p r e d i c t i o n s of s i m p l e
sample the frost at the end of each run, a coring tube was
and v a r i a b l e density models constructed by gluing together 0.0625-inch polystyrene rings.
When a frost saniple was removed with this tube, the 0.0625-
iuch rings (with frost) could easily be separated with a cold
sponding to a dew point of 29.2'F, and a gas temperature of razor blade in a -20'F cold room. The density of each seg-
70'F. Results for other conditions were qualitatively similar. ment was determined by measuring the frost volume and
For the results in Figure 7 , the calculation was started a t a mass. The results of these tests are shown in Figure 9. Though
time of 30 minutes with a frost layer 0.14 inch thick of uniform the density varied for runs of different length, the frost den-
density equal to 0.076 gram per cc. This layer was then sity on the bottom 0.0625-inch layer was always essentially
exposed to the 70°F gas with a dew point of 29.2'F. As addi- the same as the average density for the entire layer. With the
tional frost deposited, internal diffusion of water vapor caused help of repeat runs, the data shown in Figure 9 were estimated
densification of the frost layer, but to an appreciable extent to be accurate within a inaxinium error of 10%.
only in the outer regions of the layer. Apparently, in the These results and others of n similar nature force one to
inner regions of the frost layer the temperature is so low and reject a n internal densification model based on the transport
the ice vapor pressure so small that frost densification occurs of water b y vapor-phase diffusion driven by a n equilibrium
a t a n insignificant rate. Therefore, as shown in Figure 7 , large partial pressure gradient, a t least in the colder regions of the
variations of density with position were predkted. frost interior. Indeed, vapor-phase diffusion is unlikely even
Over a short period of time, this model predicted heat and with a rionequilibrium partial pressure profile. Using the
mass fluxes and frost thicknesses in good agreement with experimentally determined rate of densification of t'he frost
the simple model and with experimental data. Figure 8 in- layer for a given run, the partial pressure profile required to
dicates the type of agreement for one set of conditioiis. B u t produce a uniform densification of the frost layer by vapor-
the two niodels disagree strongly in the predicted density phase diffusion can be calculated. Such calculations (Shah,
distribution within the frost. The only previous experiinental 1968) resulted in water vapor partial pressure profiles such as
data on density gradients within a frost were taken by Braein- that shown in Figure 10. The slope of the required partial
sky (Braeinsky 1967; Brian et al., 1969) who found no varia- pressure profiles must, at each point within the frost, be equal
tion of density with depth. As these data were few and the to the rat,e of densification of the frost layer between t'he
result was somewhat unexpected, additional experimental point in question and the cold wall. Thus the slope a t the cold
confirmation was also sought in this work. wall is zero, and the slope a t the frost surface is that required
to densify the entire frost layer. R u t while the slope of the
Experimental Measurements of Density required profile is fixed a t every point', the absolute pressure
Gradients in a Frost layer
level is not. Thus the required profile is represented by a
A series of frosting runs was made a t a gas-phase Reynolds family of curves obtained by sliding the curve in Figure 10
number of 15,800, a humidity corresponding t o a 30'F dew up or down. B u t since negative partial pressures must. surely
point, arid a gas temperature of 71'F. These conditions be excluded, the curve presented in Figure 10 represents the
closely approximate those chosen for the theoretical results lowest possible curve in the family of partial pressure profiles
in Figure 7 . The deposition times varied from 26 to 144 which could account for a uniform frost densification a t the
minutes; corresponding to these times the frost thickness a t experimentally observed rate.
A
S
Rn:I,yo
d
':i;tn
ies = 14,700 diffusion velocity corresponding to the temperature gradient
H u m i d i t y = G a s dew point at various points within the frost, as given in Figure 2. The
calculated velocity varies from about 40 feet per hour near the
* e o T4= 7 O o F frost surface to about 170 feet per hour near the cold wall.
The water flux required to explain frost densification is of the
I .o
order of 0.05 lb,/hi ft2 near the fiost surface, but it decreases
3.0 to zero at the cold wall. I n order to get a flux of 0.02 lb,/hr
ft2with a particle velocity of 100 feet per hour, the concentra-
tion of particles would have to be 2 X lb,/ft3, which
corresponds to a volume fraction of 4 X 10-6. This does not
appear to be an unreasonable nuclei concentration.
While nucleation within the frost appears likely and while
Brock's theory predicts thermal diffusion velocities high
enough to transport the nuclei within the fiost, there ale a t
this time no experimental observations to corroborate the
.--
+
0
5.0
postulate that theimal diffusion of nuclei is the mechanism
0
4.0 of densification of the cold regions of the frost. Furthermore,
a it is by no means clear how this water transport mechanism
3 .O would result in a densification rate which is uniform over the
2 .o frost layer. But no other mechanism of densification of the
cold layers appears likely, and so this postulate is currently
I .o being pursued.
0 Nomenclature
0 0.04 0.08 0.12 0.16 0.20
Distance f r o m Cold W a l l (inches)
D = molecular diffusivity of water vapor in air, ft2/hr
h = heat transfer coefficient, Btu/hr-ft2-OF
Figure 10. Required pressure profile compared to equi- AH = heat of sublimation (negative) of water vapor,
librium profile within frost Btu/lb,
k = thermal conductivity, Btu/hr-ft-OF
K, = mass transfer coefficient, ft/hr
Shown for comparison in Figure 10 is the profile of the ice M = molecular weight of water, lb,/lb mole
vapor pressure in equilibrium with the local frost temperature. AT = flux of water vapor to surface, lb,/hr-ft2
At the frost surface, the slope of the equilibrium profile is p = partial pressure of water vapor, mm of Hg
= heat flux, Btu/hr-ft2
nearly equal to the slope of the required profile because the
former agrees with the over-all rate of densification of the
kT = gas constant, lbi-ft3/ftz-lb mole-OF
= temperature, O F
frost layer. But the required profile lies substantially above the X = distance from the cold wall, ft
equilibrium profile, indicating supersaturation ratios over GREEKLETTERS
1000 within the frost layer. Even a t the frost surface, a
supersaturation ratio of about 5 is indicated. Such high e time, hr
=
P frost density, lb,/ft3
=
supersaturation ratios would appear to be unattainable in P(T,) DLlf2AHpg/R2Tsar,lb,/'F-ft-hr
=
such circumstances. But even more convincing is the fact 7 = tortuosity
that the required partial pressure a t the frost surface, 14 m m 6 = frost thickness, inches
of Hg, is substantially higher than the water vapor partial SUBSCRIPTS
pressure in the bulk gas stream, which was only 6.3 mm of f = frost
Hg. This contradiction indicates that vapor-phase diffusion 9 = gas phase
cannot account for a uniform densification of the frost layer. ice = solid ice
Therefore while the success of the simple model indicates S = frost surface
that water is transported from the frost surface to the interior W = wall
by vapor diffusion driven by an equilibrium partial pressure A bar over a symbol indicates an integrated average value.
gradient, the transport of water to the cold regions deeper
within the frost would appear t o occur by other mechanisms. Literature Cited
Biguria, Gabriel, Ph.D. thesis, Lehigh University, Bethlehem,
Water Transport as Nuclei Pa., 1968.
Biguria, Gabriel, Wenxel, L. A,, IND.ENG.CHEM.FUNDAM. 9,
Since the densification of the cold regions of the frost i 2 9 (i970). '
layer cannot be explained by vapor diffusion, it is postulated Brazinsky, I., Sc.D. thesis, Massachusetts Icstitute of Tech-
that very small ice particles are nucleated in the vapor within nology, Cambridge, hIass., 1967.
Brian, P. L. T., Reid, R. C., Braainsky, I., Cryog. Technol. 5 , 205
the frost and that these nuclei are transported to the cold 11969).
regions by thermal diffusion. Thus the water transport into Brock, J. R., J . Colloid Sci. 17,768 (1962).
the frost is visualized as occurring by vapor-phase diffusion Eckert, E. R . G., Drake, R. I f . , ('Heat and Mass Transfer,"
2nd ed., p. 7, McGraw-Hill, New York, 1955.
in the warm region near the frost surface but by thermal Reid, R. C., Brian, P. L.T., Weber, M.E., A.I.Ch.E. J . 1 2 , 1190
diffusion of nuclei in the cold region. Intermediate between (1966).
these two would lie a region in which both vapor-phase diffu- Shah, Y. T., Sc.D. thesis, Massachusetts Institute of Technology,
Cambridge, Mass., 1968.
sion and thermal diffusion of ncclei might be important. Sleicher, C. A,, Tribus, X., Trans. ASME 79, 789, 1957.
Brock (1962) has presented a theory for the movement of RECEIVED for review July 7, 1969
colloidal-sized solid particles in a gas with a temperature ACCEPTEDJune 1, 1970