Kerr  Effects  in  the  Infra-red  Spectrum.  63 
time  to  bring  the  beam  out  parallel  to  its  original  direction, 
and  hence  permit  the  use  o£  the  already  existing  apparatus 
with  the  smallest  possible  change,  double  reflexion  was 
resorted  to,  the  two  magnetic  mirrors  being  held  in  a  small 
brass  frame  mounted  between  the  pole-pieces  of  the  magnet. 
These  had  an  air-gap  of  about  3  cms.,  and  were  slightly 
shifted  out  of  line  to  allow  the  passage  of  the  beam  through 
them  as  before.  To  avoid  as  much  as  possible  the  many 
complications  of  the  problem  of  both  magnetic  and  optical 
character,  such  as  elliptical  polarization,  the  incidence  was 
made  as  nearly  normal  as  possible,  or  about  8°  at  each 
mirror.  While  this  is  somewhat  larger  than  that  used  by 
du  Bois  in  his  work  in  the  visible  spectrum,  it  is  well  within 
the  limit  of  15°  for  which  the  effect,  as  shown  by  Kighi  *,  is 
practically  the  same  as  for  normal  incidence. 
The  method  of  making  observations  and  their  reduction 
was  essentially  the  same  as  that  already  described,  and  the 
sources  of  error  were  also  largely  identical,  although  their 
relative  importance  was  very  different.  Thus,  errors  of 
adjustment,  of  impurity  of  spectrum,  and  the  like,  the  effect 
of  which  was  previously  shown  to  be  a  matter  of  only  a  few 
per  cent.,  were  not  worth  considering  here  in  view  of  the 
necessarily  large  errors  of  observation,  while  the  question  of 
deformation  of  apparatus,  due  to  magnetic  traction,  which 
was  entirely  negligible  before,  became  of  capital  importance, 
for  it  was  almost  impossible  to  mount  the  mirrors  so  that  they 
would  not  be  warped  or  shifted  by  the  magnet-field.  How- 
ever, since  the  mirrors  possessed  no  magnetic  polarit}r,  this 
effect,  which  in  general  gave  rise  to  an  increase  or  decrease 
of  intensity  of  the  radiation  measured,  did  not  change  sign 
on  reversing  the  magnet,  and  hence  readily  separated  itself 
from  the  true  Kerr  effect.  There  is  a  certain  amount  of 
elliptical  polarization  connected  with  the  rotation  in  the  Kerr 
phenomenon,  even  for  small  angles  of  incidence,  and  it  might 
be  thought  that  this  would  give  rise  to  an  error,  but  it  can 
readily  be  shown  that  such  error  would  be  only  of  the  second 
order,  and  since  direct  visual  tests  showed  the  amount  of 
elliptical  polarization  to  be  actually  very  small,  its  effect  was 
considered  entirely  negligible. 
The  apparatus  was  that  used  for  the  Faraday  tests,  with 
the  modifications  already  mentioned  and  some  minor  changes, 
such  as  the  increase  of  the  number  of  polarizer  and  analyser 
plates  from  six  to  twelve.  The  substances  tested  were  steel 
(hardened),  cobalt,  nickel,  magnetite,  Heusler's  metal,  and 
silver,  in  the  form  of  little  polished  plates  each  6  by  lb'  rams. 
*  Ann.  Chim.  Phys.  [6]  ix.  p.  132  (1886). 
