220     CHEMICAL  ANALYSES  BY  SPECTRUM  OBSERVATIONS. 
the  prism  and  mirror.  A  small  telescope  placed  some  way  off 
is  directed  towards  the  mirror,  and  through  this  telescope  an 
image  of  a  horizontal  scale  fixed  at  some  distance  from  the  mir- 
ror is  observed.  By  turning  the  prism  round  every  color  of  the 
spectrum  may  be  made  to  move  past  the  vertical  wire  of  the 
telescope  C,  and  any  required  position  of  the  spectrum  thus 
brought  to  coincide  with  the  vertical  line.  Each  particular  por- 
tion of  the  spectrum  thus  corresponds  to  a  certain  point  on  the 
scale.  If  the  luminosity  of  the  spectrum  is  very  small,  the  wire 
of  the  telescope  C  may  be  illuminated  by  means  of  a  lens,  which 
throws  a  portion  of  the  rays  from  a  lamp  through  a  small  open- 
ing in  the  side  of  the  tube  of  the  telescope  C. 
From  a  long  series  of  preliminary  experiments  with  this  ap- 
paratus, the  authors  satisfied  themselves  that  the  appearance  of 
certain  bright  lines  in  the  spectra  may  be  regarded  as  absolute 
proof  of  the  presence  in  the  flame  of  certain  metals,  and  that 
they  serve  as  reactions,  by  means  of  which  these  bodies  may  be 
recognized  with  more  certainty,  greater  quickness,  and  in  far 
smaller  quantities,  than  can  be  done  by  help  of  any  other  known 
analytical  method,  no  matter  what  may  be  the  nature  of  the 
body  with  which  the  metals  are  combined. 
The  wonderful  delicacy  of  the  spectrum-reaction  of  sodium  is 
evinced  by  the  following  experiment,  which  the  writer  had  the 
good  fortune  to  witness  in  the  laboratory  of  Professor  Bunsen, 
in  Heidelberg.  In  a  far  corner  of  the  experiment  room,  the 
capacity  of  which  is  about  60  cubic  metres,  (one  cubic  metre  = 
35-3  cubic  feet,)  was  burnt  a  mixture  of  3  milligrammes  (0-0462 
gr.)  of  chlorate  of  sodium  with  milk  sugar,  whilst  the  non- 
luminous  flame  of  the  lamp  was  observed  through  the  slit  of  the 
telescope.  Within  a  few^  minutes  the  flame,  which  gradually 
became  pale  and  yellow,  gave  a  distinct  yellow  sodium  line,  co- 
incident in  the  solar  spectrum  with  Fraunhofer's  dark  line  D, 
lasting  for  about  ten  minutes,  and  then  entirely  disappearing. 
From  the  weight  of  the  sodium  salt  burnt,  and  the  capacity  of 
the  room,  it  was  calculated  that  in  one  part  by  weight  of  air, 
there  was  suspended  less  than   I —  of  a  part  of  soda  smoke. 
^  20000000  '■ 
As  the  reaction  can  be  quite  easily  observed  in  one  second,  and 
as  in  this  time  the  quantity  of  air  which  is  heated  to  ignition  by 
the  flame  could  be  calculated  from  the  rate  of  issue,  and  from 
