THE FORMATION OF ICE CRYSTALS 
Two examples showing the process of formation are 
illustrated, together with the condition of formation, in 
Figs. 21-24. Figure 22 shows the mode of formation of a 
stellar erystal with plates at the ends of the branches 
(Pih in the general classification). The conditions of 
formation are given in Fig. 21. The lettered arrows 
on the 7, curve show the moments at which the photo- 
micrographs of Fig. 22 were taken. During the stages 
between a and d the air temperature 7, was kept at 
nearly —15C and 7’, at about +12C; that is, the condi- 
tion was within the region for dendritic development, 
although the degree of supersaturation was compara- 
tively low. The crystal became a broad-branched type 
Pid, owing to the small amount of supersaturation. 
After this stage had been reached, 7, was gradually 
lowered, while 7, was kept at —16C. This change in 
condition made the rate of growth gradually smaller. 
The condition changed to that for plate development, as 
shown in Fig. 21. Accordingly, the tips of the branches 
began to widen (e), and finally the expected crystal 
was obtained as shown in f. 
Figures 23 and 24 show the course of formation of a 
capped column. This crystal was obtained by first 
making a column under the condition of low tempera- 
ture and humidity, and then changing the condition 
rapidly to that favorable for dendritic development. 
In the initial stage the column was slender in form. 
This column thickened without marked longitudinal 
elongation. In this example it took nearly five hours to 
attain the dimension observable in the central part of 
the natural capped column (Fig. 24c). Then the tem- 
perature of the water in the reservoir was raised rather 
quickly in order to increase the degree of supersatura- 
tion. In this condition a plate began to develop at each 
end of the column (d). The crystal finally developed 
into a completed, capped column as shown in f, through 
the stage e. 
These two examples show that any crystal belonging 
to a certain type or a combination of several types can 
easily be produced artificially. The method is simple, 
the only thing to do is to establish the required condi- 
tions one after another. 
APPLICATION OF SNOW CRYSTAL STUDIES TO 
METEOROLOGY 
Upper-Air Conditions and Snow Crystal Forms. As 
described in the preceding paragraph, a snow-crystal 
germ is born in the upper atmosphere and develops to a 
snow crystal of complicated shape as it is subjected to 
different conditions while falling through the various 
layers of the atmosphere. The final form of the snow 
erystal observed at the earth’s surface is an accumula- 
tion of the elements produced in the various strata. 
Shedd [14], Wegener [15, p. 284], Findeisen [6], and 
others inferred the mechanism of snow crystal forma- 
tion by comparing many photographs of snow crystals. 
Of these theories, Wegener’s will be chosen as repre- 
sentative. According to Wegener, a primitive hexagonal 
crystal is first formed in a sufficiently supersaturated 
zone of the atmosphere and then the space between the 
branches is filled with ice while the crystal is falling 
217 
through a less supersaturated layer, transforming the 
dendritic crystal into plate form. The next stratum of 
sufficient supersaturation will add other dendritic ap- 
pendages. When the crystal comes down through a 
subjacent layer of less supersaturation, it becomes a 
larger plate. By the repetition of processes similar to 
these, many variations occur in the form of the snow 
erystals. 
Wegener’s theory was examined in an artificial snow 
experiment, and the transformation of a dendritic crys- 
tal into a plate in a less supersaturated atmosphere was 
found not to be the actual case, as described in the ex- 
periment of Fig. 22. However, the general idea of his 
theory is correct. Our experiments show that a com- 
plicated form of snow crystal is made by adding the 
characteristic features, corresponding to variations in 
the external conditions. The simplest example is a 
plate with dendritic extensions. It is produced when a 
plate is made in the upper atmosphere and the den- 
dritic branches grow from the corners of the plate while 
the crystal is falling through lower layers suitable for 
dendritic development. 
By examining the natural snow crystals from this 
point of view, we can infer to some extent the structure 
of the upper atmosphere, because the external condi- 
tions controlling the form of snow crystals are now made 
clear. 
Comparison of Natural and Artificial Snow Crystals. 
By comparing natural and artificial snow crystals, the 
upper-air conditions can be estimated for each of various 
types of crystals, using the (T., 7)-diagram. Four 
examples will be described. 
Fern-Like Hexagonal Crystal. Natural and artificial 
crystals of fern-like type are reproduced in Fig. 25a—b. 
The form and structure of the branches are very similar 
in both cases, although the central part is somewhat 
different. The course of formation of this fern-like 
crystal is shown in Fig. 25c. During the primitive stage 
the temperature is lower and the supersaturation is less 
than the critical value for dendritic formation. Then 
the initial stage of this crystal is the irregular assem- 
blage of small sectors. From this stage both 7, and T., 
are increased so that the condition is most favorable 
for dendritic development. The crystal rapidly grows 
in a fern-like form and reaches the stage shown in Fig. 
25b in about 15 min. The rate of growth is very large 
in this case. The rate of fall is about 1 km hr™ for 
crystals of this type. Thus we can infer that there must 
be an atmospheric layer with ample moisture and a 
temperature of about —15C near the earth’s surface, 
when the crystal of the type shown in Fig. 25a is ob- 
served at the earth’s surface. The thickness of this 
layer will be about 14 km. 
Stellar Crystal with Plates at the Ends of the Branches. 
The crystal shown in Fig. 26a is frequently observed in 
nature, and is designated as Pih in the general classifi- 
cation. The artificially produced crystal shown in Fig. 
26b certainly belongs in this group. The fine strips in 
the plate portion are less marked in the case of natural 
snow, which is probably due to sublimation in the 
