﻿Bands 
  in 
  the 
  Positive 
  Band- 
  Spectrum 
  of 
  JSitrogen. 
  351 
  

  

  The 
  differences 
  between 
  these 
  numbers 
  form 
  three 
  rough 
  

   arithmetical 
  progressions, 
  which 
  are 
  shown 
  divided 
  by 
  lines 
  ; 
  

   and 
  these 
  I 
  believe 
  to 
  be 
  the 
  three 
  series 
  of 
  which 
  Deslandres 
  

   speaks. 
  But 
  the 
  divergencies 
  from 
  strict 
  regularity 
  are 
  

   too 
  large 
  to 
  be 
  accidental, 
  and 
  the 
  breadth 
  of 
  the 
  bands 
  is 
  

   only 
  about 
  one-tenth 
  of 
  that 
  of 
  the 
  bands 
  of 
  the 
  second 
  group. 
  

   It 
  seemed, 
  however, 
  improbable 
  that 
  the 
  law 
  which 
  is 
  so 
  

   abundantly 
  evident 
  in 
  the 
  second 
  group 
  should 
  not 
  be 
  re- 
  

   presented 
  by 
  some 
  similar 
  rule 
  in 
  the 
  adjoining 
  portion 
  of 
  the 
  

   spectrum. 
  

  

  I 
  searched, 
  therefore, 
  for 
  such 
  a 
  law, 
  and 
  found 
  that 
  the 
  

   frequencies 
  of 
  the 
  heads 
  of 
  bands 
  could 
  be 
  arranged 
  in 
  the 
  

   following 
  order 
  on 
  the 
  model 
  of 
  Deslandres' 
  series 
  (see 
  p. 
  352). 
  

  

  The 
  figures, 
  thus 
  distributed, 
  fall 
  into 
  thirteen 
  series, 
  in 
  

   some 
  of 
  which 
  only 
  two 
  members 
  are 
  present, 
  and 
  the 
  first 
  differ- 
  

   ences 
  decrease 
  in 
  arithmetical 
  progression 
  from 
  153*1 
  to 
  112*9.. 
  

  

  This 
  arrangement 
  differs 
  so 
  much 
  from 
  that 
  of 
  the 
  second 
  

   group 
  made 
  by 
  Deslandres 
  that 
  its 
  reality 
  may 
  appear 
  open 
  

   to 
  doubt. 
  The 
  number 
  of 
  bands 
  is 
  so 
  great, 
  and 
  the 
  differ- 
  

   ences 
  between 
  the 
  frequencies 
  of 
  the 
  heads 
  of 
  consecutive 
  

   bands 
  so 
  regular, 
  that 
  it 
  might 
  be 
  surmised 
  that 
  almost 
  any 
  

   law 
  could 
  be 
  " 
  fudged 
  " 
  out 
  of 
  the 
  figures. 
  But 
  the 
  following 
  

  

  © 
  © 
  © 
  

  

  reasons 
  tend 
  to 
  show 
  that 
  this 
  arrangement 
  is 
  a 
  true 
  one, 
  and 
  

   not 
  accidental 
  : 
  — 
  

  

  1. 
  It 
  accounts 
  for 
  every 
  one 
  of 
  the 
  bands 
  mapped 
  by 
  

   Angstrom 
  and 
  Thalen, 
  and 
  all, 
  with 
  two 
  exceptions 
  (16319 
  

   and 
  16474) 
  conform 
  to 
  the 
  law 
  with 
  very 
  considerable 
  

   accuracy. 
  Only 
  one 
  (18373) 
  is 
  used 
  twice. 
  

  

  2. 
  In 
  addition, 
  it 
  accounts 
  for 
  many 
  of 
  the 
  subsidiary 
  

   strong 
  lines 
  interspersed 
  in 
  some 
  bands, 
  and, 
  in 
  particular, 
  

  

  for 
  the 
  overlapping 
  which 
  occurs 
  at 
  about 
  - 
  = 
  18000 
  (t, 
  u 
  

  

  and 
  v 
  of 
  Angstrom 
  and 
  Thalen) 
  . 
  Thus, 
  the 
  strong 
  lines 
  at 
  

   18093, 
  18232, 
  and 
  19782 
  become 
  the 
  heads 
  of 
  bands, 
  and 
  

   the 
  curious 
  line 
  at 
  19872 
  is 
  accounted 
  for. 
  So 
  also 
  the 
  line 
  

   at 
  21225 
  is 
  seen 
  to 
  be 
  the 
  head 
  of 
  a 
  band 
  in 
  the 
  Series 
  IV. 
  

  

  3. 
  The 
  distribution 
  was 
  arrived 
  at 
  by 
  tabulating 
  differ- 
  

   ences 
  between 
  the 
  frequencies 
  of 
  every 
  pair 
  of 
  bands 
  in 
  the 
  

   group. 
  On 
  examining 
  the 
  table 
  it 
  was 
  seen 
  that 
  the 
  numbers 
  

   lying 
  in 
  a 
  certain 
  line 
  transversely 
  across 
  the 
  page 
  were 
  

   connected 
  in 
  the 
  manner 
  shown 
  above. 
  No 
  other 
  series, 
  so 
  

   far 
  as 
  I 
  can 
  discover, 
  can 
  be 
  picked 
  out 
  of 
  the 
  table 
  so 
  as 
  to 
  

   give 
  anything 
  like 
  the 
  same 
  regularity. 
  

  

  4. 
  The 
  series 
  die 
  out 
  after 
  the 
  thirteenth. 
  And 
  it 
  is 
  

   remarkable 
  that 
  the 
  last 
  member 
  of 
  that 
  series, 
  19502, 
  is 
  

   consecutive 
  to 
  the 
  last 
  member 
  of 
  the 
  first 
  series, 
  19607. 
  

   The 
  chances 
  against 
  this 
  occurring 
  in 
  an 
  accidental 
  arrange- 
  

   ment 
  are 
  high. 
  

  

  