724 
Prof.  H.  Becquerel  on  some  Properties  of 
for  the    same  magnetic   field,   the  trajectories   of   the    rays 
which  had  merely  traversed  an  air  space  (fig.  2) . 
Fijr.  2. 
Aluminium  Sheet.  Air  only. 
A.  17/11/1905. 
It  was  directly  found  that  the  rays  which  have  traversed 
aluminium  give  a  shorter  image  than  the  others. 
Numerous  points,  whose  positions  were  determined  with  a 
micrometric  apparatus  reading  to  q^-q  mm.,  have  given  the 
mean  results  shown  below.  Values  of  the  distance  apart  2x 
of  the  divergent  curves  on  the  plates  were  obtained  for 
distances  from  the  lower  edge  corresponding  to  integral 
numbers  n  of  turns  of  the  screw  used  in  measuring  the  vertical 
coordinates,  the  pitch  of  the  screw  being  0"94684  mm.  The 
corresponding  heights  y  were  calculated  from  the  inclination 
of  the  plate  in  each  experiment. 
If  the  results  of  the  observations  are  plotted  on  a  large 
scale,  a  sinuous  curve  is  obtained  whose  mean  position  gives 
the  true  trajectory.  Theoretically,  in  a  uniform  magnetic 
field  and  in  vacuo,  this  trajectory  should  be  circular,.  I  had 
previously  realized  that,  for  increasing  values  of  ?/,  the  circles 
normal  to  the  field  which  pass  through  the  source,  the  upper 
portion  of  the  slit  and  a  point  of  the  trajectory,  have 
progressively  increasing  radii,  and  the  actual  experiments 
confirm  this  result.  But,  as  will  be  seen,  this  conclusion 
does  not  apply  to  the  radius  of  curvature  of  the  trajectory, 
and  is  due  to  an  inexact  interpretation  founded  on  imperfect 
knowledge  of  the  actual  distance  of  the  slit  from  different 
points  of  the  photographic  plate. 
If  the  origin  of  coordinates  is  taken  at  the  intersection  of 
the  edge  of  the  plate  with  the  bisector  of  the  two  pencils,  it 
is  observed  that  the  curves  do  not  pass  through  this  origin, 
but  cut  the  y-axis  below  the  edge  of  the  plate.  Different 
trials  have  shown  that  it  is  possible  to  draw  through  the 
mean  position  of  the  points  representing  the  observations  a 
