I don't understand your meaning. You do not need to export shm file by Shermo, you can directly read thermal correction to G from output of Shermo.

Note that if you request Shermo to employ RRHO model (ilowfreq=0), then the thermal corrections printed by Shermo are identical to those by Gaussian.

Thank you sir, I was exporting shm file to read ZPE corrections as from text file of multiple entries Shermo don't print ZPE corrections. The exported shm file takes electronic energy value from Gausssian log file and not the value provided in text file input. Manually changing electronic energy value in shm file printed same results.

Note that if you request Shermo to employ RRHO model (ilowfreq=0), then the thermal corrections printed by Shermo are identical to those by Gaussian.

This is a very common and standard way of calculating accurate Gibbs free energy.

However, if the system contains many low frequency modes, it is also important to apply quasi-RRHO thermodynamic model, which is supported by Shermo code (http://www.shanxitv.org/soft/shermo/)

sir the result of G from text file is different from shm file generated by shermo.
I have given text file as input with content gaussian-freq-job.log;-2812.272184125518 and I got value of G as -2811.969688 au. then I gave shm file exported by shermo as input and I get value of G as -2812.4253519 a.u.

However, if the system contains many low frequency modes, it is also important to apply quasi-RRHO thermodynamic model, which is supported by Shermo code (http://www.shanxitv.org/soft/shermo/)

Not "change", you just need to compute Gibbs free energy for each configuation, and then directly compare the values.

To compute accurate Gibbs free energy, is this right approach to compute frequency at medium level and add thermal/entropy corrections from this medium level frequency calculation to the single point energy calculated at higher level? For example opt and freq job at pbe0/def2-tzvp and then add thermal/entropy corrections to single point energy computed at PWPB95 D3 def2-QZVPP level will yield better Gibbs free energy.?

For H-bond dimer system, you can also use electron density at bond critical point to evaluate the binding energy due to H-bond interaction, thus only step 1 is needed (used to generate wavefunction of the complex). This calculation is quite easy by using Multiwfn, see Section 4.2.1 of latest version of Multiwfn manual for introduction and example, as well as my work J. Comput. Chem., 40, 2868–2881 (2019).

I have read the following sentences:

(A further interesting comparison is that of the binding energies, which we have calculated with respect to the fragments, both charged and neutral. We find that the binding energy of carboplatin is higher in both cases, by AEc = 2.9 eV and AE=1.8 eV, respectively. These results reflect the higher instability of the dicarboxylate ligand in the gas phase, both in the charged and the neutral state.)

DOI: https://doi.org/10.1016/0009-2614(95)01099-2

My question is:

How I figured out the binding energies, and whether it is achievable with Multiwfn.

thanks you very much