56  .Mr.  L.  R.  Ingersoll  on  the  Faraday  and 
M^nd  M2,  the  polarizer  was  in  general  set  so  that  its  plane 
of  incidence  was  parallel  and  perpendicular  respectively  to 
the  planes  o£  incidence  on  these  two  mirrors.  It  should  be 
noted  that  the  elliptical  polarization  which  would  inevitably 
be  introduced  by  the  oblique  reflexion  at  the  silver  surface, 
immediately  after  passing  the  analyser,  as  well  as  the  polariza- 
tion which  would  occur  at  the  faces  of  the  prism,  could  have 
no  disturbing  effect,  since  after  the  radiation  had  once  passed 
the  analyser  its  state  of  polarization  was  of  no  moment. 
Purity  of  the  spectrum  was  tested  by  double  dispersion, 
and  although,  as  might  be  expected,  considerable  impurity 
due  to  overlapping  slit-images  was  found,  only  a  trace  of  stray 
radiation  could  be  discovered.  To  take  account  of  the  error 
due  to  impurity,  correction  was  made  for  the  width  of  slit 
and  bolometer-strip.  The  *5  mm.  strip  subtended  an  angle 
of  5',  corresponding  to  from  '02  /n  to  '±/jl  in  the  rock-salt 
spectrum.  The  slit  was  even  wider  as  a  rule,  and  it  might 
be  thought  that  such  an  arrangement,  while  necessary  to 
secure  sufficient  energy  for  the  measurements,  would  involve 
such  a  large  correction  as  would  seriously  prejudice  the 
accuracy  of  the  work.  But,  while  by  no  means  negligible, 
the  corrections  in  the  present  case  prove  to  be  much  smaller 
than  in  most  other  spectrobolometric  problems  involving  a 
like  amount  of  impurity,  because  they  are  of  a  differential 
nature;  for  the  rotations  were  determined,  not  by  a  single 
measurement  of  energy  at  any  point  of  the  spectrum,  but 
by  the  ratio  of  the  two  quantities  dl/I,  and  hence  only 
the  difference  of  the  separate  corrections  to  dl  and  I  need 
be  applied.  These  separate  corrections  were  determined 
for  each  of  the  curves  of  dl  and  I  plotted  with  wave- 
length (the  curve  of  dl,  while  not  an  energy-curve  in  the 
ordinary  sense,  is  exactly  analagous  to  one  and  may  be 
treated  in  a  similar  manner)  by  the  simple  method  of  correc- 
tion for  slit-width  due  to  Lord  Rayleigh,  and  outlined  in 
Preston's  '  Heat,'  p.  606  (last  edition).  The  difference,  or 
resultant  correction,  varied  from  2  per  cent,  at  \=1//,  to 
—  4  per  cent,  at  3  jn.  That  the  correction  should  be  small 
was  indicated  by  measurements  of  rotation  for  the  same 
wave-length,  with  different  slit- widths,  which  gave  results  in 
good  agreement. 
Results. 
Table  1.  is  a  synopsis  of  the  principal  results.  Potations 
are  given  for  forty-two  points  of  the  spectrum  (including 
duplicates),  as    produced    by  carbon    bisulphide   in    a    tube 
