Sstrhsrtj

I have a file which looks like this

   18DMA      H 9996   0.886   5.687   5.320

   18DMA      H 9997   1.019   5.764   5.247

   18DMA     Np 9998   0.947   5.584   5.151

   18DMA      H 9999   1.033   5.541   5.113

   18DMA     Cn10000   0.880   5.674   5.050

   18DMA      H10001   0.831   5.616   4.971

   18DMA      H10002   0.814   5.751   5.091

   18DMA      H10003   0.957   5.735   5.003

   18DMA     Cn10004   0.837   5.486   5.185

The desire output is to delete column 3 however since from a certain row/line and next there is no a space between atom name and number I cannot make the deletion by column. Is there any way to make the deletion by selecting certain number of characters? The desire output should be

   18DMA      H    0.886   5.687   5.320

   18DMA      H    1.019   5.764   5.247

   18DMA     Np    0.947   5.584   5.151

   18DMA      H    1.033   5.541   5.113

   18DMA     Cn    0.880   5.674   5.050

   18DMA      H    0.831   5.616   4.971

   18DMA      H    0.814   5.751   5.091

   18DMA      H    0.957   5.735   5.003

   18DMA     Cn    0.837   5.486   5.185

Assuming you don't have <TAB>s but multiple spaces as field separators, and by looking at and counting your sample data, I came up with

  $ sed -E 's/^(.{15}).{5}/1/' file

   18DMA      H   0.886   5.687   5.320

   18DMA      H   1.019   5.764   5.247

   18DMA     Np   0.947   5.584   5.151

   18DMA      H   1.033   5.541   5.113

   18DMA     Cn   0.880   5.674   5.050

   18DMA      H   0.831   5.616   4.971

   18DMA      H   0.814   5.751   5.091

   18DMA      H   0.957   5.735   5.003

   18DMA     Cn   0.837   5.486   5.185

It's using a "back reference" for the first 15 characters to restore them using 1
in the replacement part of the substitute command.

Use cut in character mode:

cut -c1-15,21-

you may need to tweak the exact character numbers.
Again, this assumes the input doesn't use TABs (t characters) as delimiters (which it probably doesn't, since then you wouldn't have the problem of the joined fields in the first place).

$ awk -v OFS='t' 'NF == 5 { sub("[0-9]*$", "", $2) } NF == 6 { $0 = $1 OFS $2 OFS $4 OFS $5 OFS $6 } { print }' file

18DMA   H       0.886   5.687   5.320

18DMA   H       1.019   5.764   5.247

18DMA   Np      0.947   5.584   5.151

18DMA   H       1.033   5.541   5.113

18DMA   Cn      0.880   5.674   5.050

18DMA   H       0.831   5.616   4.971

18DMA   H       0.814   5.751   5.091

18DMA   H       0.957   5.735   5.003

18DMA   Cn      0.837   5.486   5.185

This short awk program will do different things to the input line depending on whether it contains 5 or 6 whitespace-delimited fields.

If it contains five fields, it removes all digits from the end of the second fields and leaves the rest as it is. If it contains six fields, it rewrites the line but omits the third field.

The output will be tab-delimited (or delimited by whatever you set OFS to on the command line).