644  Mr.  Inglis  on  the  Isothermal  Distillation  of 
argon  used,  which  was  kindly  supplied  to  me  by  Sir  Wm. 
Ramsay,  was  purified  by  means  of  a  hot  mixture  of  quick- 
lime and  magnesium,  and  was  fractionated  with  liquid  air  to 
remove  traces  of  helium  and  neon.  Its  spectrum  showed 
that  it  was  extremely  pure. 
The  analyses  in  each  case  were  carried  out  by  measuring 
oft'  10-13  c.c.  of  the  gas,  and  then  removing  the  oxygen  by 
means  of  a  pellet  of  yellow  phosphorus.  A  measurement  of 
the  volume  of  the  residual  gas  then  gave  the  molecular 
composition.  Tests  showed  that  analyses  carried  out  in  this 
way  did  not  have  a  greater  error  than  O'l  per  cent.,  which 
was  sufficiently  accurate. 
Two  parts  of  the  apparatus  needed  exhaustive  testing 
before  one  could  be  sure  of  the  results.  The  method  of 
taking  the  sample  of  the  liquid  was  based  on  the  assumption 
that  by  taking  a  fine  enough  capillary,  the  liquid  would 
evaporate  as  a  whole  and  would  not  fractionate  itself.  The 
first  capillaries  employed  were  found  to  give  very  variable, 
and  therefore  untrustworthy  samples ;  but  by  inserting  a 
fine  drawn-out  capillary,  as  drawn  in  fig.  1,  concordant 
results  were  obtained.  These  tests  were  carried  out  as 
follows  : — About  5  litres  of  dry  air  were  condensed  in  the 
bulb  B,  and  samples  of  the  liquid,  which  half-filled  the  bulb, 
were  then  taken  and  analysed.  The  percentages  of  nitrogen 
found  were  78-77,  78'52,  78-74,  78'66,  and  78'84,  which 
results,  omitting  the  second  one,  give  78' 75  per  cent,  as  a 
mean,  the  deviation  of  the  second  one  from  this  mean  being 
0*2  per  cent.  But  all  the  results  are  lower  than  the  true 
percentage  of  nitrogen  in  the  air,  which  was  found  to  be 
79"06,  78*99  per  cent,  in  two  consecutive  experiments.  This 
difference  can  be  explained  as  follows  : — When  the  air  was 
condensed,  the  spiral  4  and  the  part  of  the  bulb  B  above  the 
liquid  were  filled  with  vapour  which  was  in  equilibrium  with 
the  liquid,  and  which  therefore  did  not  have  the  same  com- 
position as  the  air.  The  vapour,  in  fact,  contained  an  excess 
of  nitrogen,  so  that  the  liquid  contained  too  high  a  proportion 
of  oxygen.  The  vapour  also  being  at  a  temperature  of  less 
than  80°  Abs.,  the  dead  space  contained  nearly  4  times  as 
much  gas  as  the  same  space  would  at  ordinary  temperatures. 
Thus  25  c.c.  of  dead  space  represented  100  c.c.  of  vapour 
measured  at  ordinary  temperatures.  This  vapour  contained 
about  5 "5  per  cent,  of  oxygen,  so  that  the  dead  space  con- 
tained 94-5  — 4x  5'5  =  72'5  c.c.  of  nitrogen  above  the  proper 
amount  for  the  oxygen.  This  amount  of  nitrogen  is  to  be 
deducted  from  5  litres  of  air  ;  and  this  brings  down  the  true 
