1260 
is filled with polygonal convection cells, as described 
earlier. 
2. When the top plate is moved at a rate not exceed- 
ing a certain low limiting value w, the original convec- 
tion cells are distorted mto a horseshoe pattern, as 
shown in Fig. 9. 
3. When the top plate is moved at a rate exceeding 
tm, but not exceeding another limiting value w, the 
chamber is filled with transverse rolls. 
4. When the top plate is moved at a rate exceeding 
Us, the chamber is filled with longitudinal rolls. 
For a given depth of chamber, w and w. will both 
increase as the difference of temperature between top 
and bottom of the chamber increases. It is not possible 
to state the values of uw and we for different depths of 
chamber and for different temperature gradients. Only 
one value, v2 for depth 10 mm and temperature differ- 
ence 40C, is known, wu. then being 0.5 cm sec7, ap- 
proximately. 
Depth of Chamber Less Than 7 mm. In the absence of 
shear the chamber is filled with structures such as 
are shown in Fig. 5. When a large shearing motion is 
imposed, the chamber is filled with spindlelike struc- 
tures, and never shows the long continuous rolls which 
occur in deeper chambers. As the rate of shear is de- 
creased the spindlelike structures become shorter and 
more irregular, but at no stage is the chamber filled 
with transverse rolls. Some idea of the completely 
different structures obtained in different depths of fluid 
may be gathered from Fig. 11, which was taken in a 
Fre. 11—Change of cell pattern with depth. The picture 
represents a chamber of depth 8 mm at the top and 6 mm at 
the bottom of the picture. The temperature difference between 
top and bottom of the chamber was about 42C, and the rate 
of motion of top plate was 7 em sec—. 
chamber in which the depth varied from 6 mm at one 
side to 8 mm at the opposite side, with a temperature 
difference of 42C. The photograph in Fig. 11 was 
taken shortly after the top plate had been stopped. 
A Comparison of the Phenomena in Air and in Carbon 
Dioxide 
Dassanayake [5] carried out an investigation of the 
forms of convection in carbon dioxide, the gas being 
bubbled through hydrochloric acid and a solution of 
ammonium carbonate in Dreschel bottles, and finally 
LABORATORY INVESTIGATIONS 
through concentrated sulphuric acid to remove all water 
drops. The amount of fogging was kept as low as was 
consistent with ease of observation of the patterns of 
flow, in order to avoid seriously altering the character of 
the medium. 
Carbon dioxide was chosen as a medium for this 
investigation because the values of « and » are consider- 
ably lower for carbon dioxide than for air. Thus at 15C 
xv 18 0.030 for air, and 0.0076 for carbon dioxide. In 
Rayleigh’s criterion quoted earlier, the critical differ- 
ence of relative density is proportional to cv/h?. Thus 
for a given depth of chamber the critical condition 
established by the criterion is satisfied by a much lower 
gradient of relative density in carbon dioxide than in 
air. Thus it was found possible to form polygonal cells 
in carbon dioxide in a chamber of 4.5 mm depth, 
whereas in air no polygonal cells could be formed in a 
chamber of less than 7 mm depth. It should be noted 
that (4.5/7)' is equal to 0.265, which is in reasonable 
agreement with the theoretical value (0.25) of the 
ratio of xv for the two gases at 15C. 
CONVECTION OF 
TYPE | (CELLULAR) 
CONVECTION 
OF TYPE II 
OEPTH IN MILLIMETERS 
0.04 0.06 
AY: 
Fig. 12—A summary of Dassanayake’s observations of limiting 
conditions in carbon dioxide (cf. Fig. 7). 
0.02 0.08 0.10 0.12 
Figure 12 shows the limiting curves derived by Das- 
sanayake for carbon dioxide. The results are similar 
in character to those of Chandra shown in Fig. 7. 
Applications To the Atmosphere 
In the atmosphere, motion in the convection cell is 
normally the reverse of that observed in the laboratory 
experiments, a central ascending current being sur- 
rounded by a descending current of much smaller speed. 
Clouds form within the ascending currents when these 
extend to a sufficient height to produce condensation. 
In considering the atmospheric analogues of the lab- 
oratory phenomena, we should therefore look for cloud 
formation within the regions corresponding to those 
in which descending motion is found in the laboratory 
experiments, bearing in mind that instability is an 
essential condition for the occurrence of these phe- 
nomena. The precise nature of the clouds will depend 
on the rate of shear, that is, on the rate of variation 
of wind with height. The main classes of convection 
cloud are classified in Table I. This table includes both 
cloud forms directly formed by condensation, and those 
