﻿55S 
  Mr. 
  K. 
  T. 
  Compton 
  and 
  Prof. 
  0. 
  W. 
  Richardson 
  on 
  

  

  at 
  will 
  by 
  sliding 
  an 
  electromagnet 
  np 
  or 
  down 
  outside 
  of 
  

   the 
  apparatus. 
  A 
  suspending 
  spring 
  C 
  was 
  just 
  strong 
  enough 
  

   to 
  keep 
  the 
  strip 
  in 
  its 
  regular 
  position 
  when 
  the 
  electro- 
  

   magnet 
  was 
  removed. 
  A 
  bit 
  of 
  the 
  metal 
  to 
  be 
  investigated 
  

   was 
  placed 
  in 
  the 
  quartz 
  furnace 
  H, 
  which 
  was 
  wound 
  with 
  

   a 
  platinum 
  heating-coil. 
  After 
  the 
  requisite 
  vacuum 
  had 
  

   been 
  obtained, 
  the 
  strip 
  S 
  was 
  drawn 
  into 
  the 
  bulb 
  T 
  and 
  

   the 
  metal 
  distilled 
  from 
  the 
  furnace 
  over 
  on 
  to 
  the 
  strip, 
  

   where 
  a 
  thin 
  film 
  of 
  the 
  metal 
  to 
  be 
  investigated 
  was 
  deposited. 
  

   Then 
  the 
  strip 
  was 
  returned 
  to 
  its 
  position 
  in 
  the 
  bulb 
  B, 
  

   and 
  photoelectric 
  measurements 
  were 
  made 
  as 
  soon 
  as 
  pos- 
  

   sible. 
  This 
  method 
  worked 
  very 
  satisfactorily, 
  and 
  can 
  be 
  

   used 
  for 
  most 
  metals 
  which 
  melt 
  below 
  about 
  1000° 
  C. 
  

  

  This 
  was 
  the 
  apparatus 
  used 
  in 
  testing 
  sodium. 
  The 
  low 
  

   melting-point 
  of 
  caesium 
  made 
  a 
  variation 
  of 
  this 
  process 
  

   more 
  convenient. 
  The 
  caesium, 
  after 
  being 
  washed 
  in 
  an- 
  

   hydrous 
  ether, 
  was 
  placed 
  in 
  the 
  small 
  bulb 
  K, 
  which 
  was 
  at 
  

   once 
  sealed 
  on 
  to 
  the 
  tube 
  ML. 
  After 
  a 
  liquid-air 
  vacuum 
  

   had 
  been 
  maintained 
  for 
  several 
  hours 
  the 
  caesium 
  was 
  melted 
  

   so 
  that 
  it 
  flowed 
  successively 
  through 
  the 
  constrictions 
  at 
  M 
  

   and 
  L. 
  The 
  caesium 
  oxide 
  and 
  other 
  impurities 
  were 
  left 
  

   behind 
  at 
  L, 
  and 
  the 
  tube 
  was 
  sealed 
  off 
  at 
  M 
  with 
  a 
  blowpipe. 
  

   Then 
  the 
  caesium 
  was 
  distilled 
  on 
  to 
  the 
  strip 
  by 
  gently 
  

   heating 
  the 
  bulb 
  T. 
  This 
  method 
  proved 
  very 
  satisfactory, 
  

   for 
  the 
  caesium 
  gave 
  practically 
  constant 
  photoelectric 
  currents 
  

   during 
  the 
  time 
  necessary 
  to 
  complete 
  a 
  test. 
  

  

  These 
  precautions 
  were 
  not 
  taken 
  when 
  aluminium 
  and 
  

   platinum 
  were 
  tested, 
  although 
  in 
  the 
  case 
  of 
  aluminium 
  we 
  

   should 
  probably 
  have 
  obtained 
  better 
  results 
  had 
  we 
  taken 
  

   them. 
  In 
  these 
  cases 
  the 
  metal 
  strip 
  was 
  scraped 
  with 
  a 
  

   clean 
  knife-blade 
  in 
  air, 
  quickly 
  placed 
  in 
  position, 
  and 
  the 
  

   vacuum 
  pump 
  started. 
  Certainly 
  some 
  oxidation 
  of 
  the 
  

   aluminium 
  surface 
  occurred 
  before 
  the 
  vacuum 
  was 
  obtained. 
  

  

  As 
  soon 
  as 
  the 
  strip 
  S 
  was 
  ready 
  for 
  the 
  test 
  the 
  drum 
  W 
  

   was 
  turned 
  so 
  as 
  to 
  cause 
  light 
  of 
  different 
  wave-lengths 
  to 
  

   fall 
  successively 
  on 
  the 
  strip. 
  For 
  each 
  wave-length 
  we 
  

   measured 
  the 
  deflexion 
  of 
  the 
  electrometer 
  in 
  some 
  definite 
  

   interval 
  of 
  time, 
  say 
  ten 
  seconds. 
  Thus 
  data 
  were 
  obtained 
  

   giving, 
  in 
  arbitrary 
  units, 
  the 
  number 
  of 
  electrons 
  liberated 
  

   in 
  unit 
  time 
  from 
  the 
  metal 
  by 
  the 
  light 
  of 
  each 
  wave-length 
  

   represented 
  in 
  the 
  spectrum 
  of 
  the 
  mercury 
  arc. 
  But 
  in 
  

   order 
  to 
  apply 
  these 
  results 
  to 
  test 
  equation 
  (6) 
  it 
  was 
  neces- 
  

   sary 
  first 
  to 
  make 
  a 
  correction 
  for 
  the 
  fact 
  that 
  the 
  various 
  

   lines 
  in 
  the 
  arc 
  spectrum 
  were 
  not 
  all 
  of 
  equal 
  intensity. 
  

   This 
  necessitated 
  a 
  calibration 
  in 
  order 
  to 
  find 
  the 
  relative 
  

   energies 
  in 
  different 
  parts 
  of 
  the 
  spectrum. 
  Then, 
  when 
  the 
  

  

  