Max size reached in AugHess

[jevandezande]

When I attempt to run a ZCASSCF calculation on H₂O + F, after a few iterations, I get the error.

         3   0       -175.51133581     2.87e-06     14.11
         3   1       -175.51133581     0.00e+00     14.11
         3   2       -175.16510034     0.00e+00     14.11
         3   3       -175.16510034     0.00e+00     14.11
         3   4       -175.09321538     0.00e+00     14.11
         3   5       -175.09321538     0.00e+00     14.11

         res : 2.90e-04   lamb: 1.00e+00   eps : -9.10e-08   step: 5.88e-04    0.68
         res : 1.07e-04   lamb: 1.00e+00   eps : -1.14e-07   step: 8.27e-04    0.68
         res : 4.24e-05   lamb: 1.00e+00   eps : -1.23e-07   step: 9.94e-04    0.68
         res : 2.25e-05   lamb: 1.00e+00   eps : -1.24e-07   step: 9.92e-04    0.68
         res : 1.39e-05   lamb: 1.00e+00   eps : -1.25e-07   step: 9.86e-04    0.68
         res : 2.22e-05   lamb: 1.00e+00   eps : -1.25e-07   step: 9.82e-04    0.69
         res : 2.53e-05   lamb: 1.00e+00   eps : -1.25e-07   step: 9.89e-04    0.68
         res : 8.91e-06   lamb: 1.00e+00   eps : -1.25e-07   step: 9.99e-04    0.69
         res : 4.22e-06   lamb: 1.00e+00   eps : -1.25e-07   step: 1.00e-03    0.69
         res : 2.96e-06   lamb: 1.00e+00   eps : -1.25e-07   step: 1.00e-03    0.69
         res : 1.65e-06   lamb: 1.00e+00   eps : -1.25e-07   step: 1.00e-03    0.68
         res : 1.14e-06   lamb: 1.00e+00   eps : -1.25e-07   step: 1.00e-03    0.68
         res : 9.68e-07   lamb: 1.00e+00   eps : -1.25e-07   step: 1.00e-03    0.69
         res : 4.74e-07   lamb: 1.00e+00   eps : -1.25e-07   step: 1.00e-03    0.68
         res : 3.78e-07   lamb: 1.00e+00   eps : -1.25e-07   step: 1.00e-03    0.69
         res : 5.09e-07   lamb: 1.00e+00   eps : -1.25e-07   step: 1.00e-03    0.69
         res : 2.99e-07   lamb: 1.00e+00   eps : -1.25e-07   step: 1.00e-03    0.69
         res : 2.64e-07   lamb: 1.00e+00   eps : -1.25e-07   step: 1.00e-03    0.69
         res : 1.48e-07   lamb: 1.00e+00   eps : -1.25e-07   step: 1.00e-03    0.69
  ERROR ON RANK 0: EXCEPTION RAISED:max size reached in AugHess
terminate called after throwing an instance of 'std::runtime_error'
  what():  max size reached in AugHess
[n020:14245] *** Process received signal ***
[n020:14245] Signal: Aborted (6)
[n020:14245] Signal code:  (-6)
[n020:14245] [ 0] /lib64/libpthread.so.0(+0xf100)[0x2b6935b13100]
[n020:14245] [ 1] /lib64/libc.so.6(gsignal+0x37)[0x2b6936f0b5f7]
[n020:14245] [ 2] /lib64/libc.so.6(abort+0x148)[0x2b6936f0cce8]
[n020:14245] [ 3] /lib64/libstdc++.so.6(_ZN9__gnu_cxx27__verbose_terminate_handlerEv+0x165)[0x2b6936a179d5]
[n020:14245] [ 4] /lib64/libstdc++.so.6(+0x5e946)[0x2b6936a15946]
[n020:14245] [ 5] /lib64/libstdc++.so.6(+0x5e973)[0x2b6936a15973]
[n020:14245] [ 6] /lib64/libstdc++.so.6(__cxa_rethrow+0x49)[0x2b6936a15be9]
[n020:14245] [ 7] BAGEL[0x52ad3a]
[n020:14245] [ 8] /lib64/libc.so.6(__libc_start_main+0xf5)[0x2b6936ef7b15]
[n020:14245] [ 9] BAGEL[0x529669]
[n020:14245] *** End of error message ***

This error appears seemingly at random, with some similar geometries working fine and others that are only slightly displaced (0.05 Angstrom) having the same error. As I am attempting to optimize this complex using single points and an external optimizer, I need this to work for a region around the optimal structure.

Input File:

{ "bagel" : [
    {
        "title": "molecule", 
        "basis": "cc-pVTZ",
        "df_basis": "cc-pvtz-jkfit",
        "symmetry": "C1", 
        "angstrom": true, 
        "geometry": [
            { "atom": "F", "xyz": [   0.031,    1.021,    0.000] }, 
            { "atom": "O", "xyz": [   0.031,   -1.081,    0.000] }, 
            { "atom": "H", "xyz": [  -0.549,   -1.014,    0.762] }, 
            { "atom": "H", "xyz": [  -0.549,   -1.014,   -0.762] }
        ]
    }, 

    {
        "title": "zcasscf", 
        "state": [ 0, 3 ], 
        "nact_cas": 3, 
        "nclosed": 9, 
        "only_electrons": false, 
        "finite_nucleus": false
    }
]}

[shiozaki]

This happens when you specify a wrong set of states. Are you sure that there is no quartet state below the third doublet states? Note that in relativistic calculations you cannot only compute doublets because spin is not a good quantum number.

[shiozaki]

PS: Can you give me casscf.log if you think it is the case?

[RDReynolds]

It looks like it's just not converging fast enough; the default setting is to stop after 20 microiterations. The input file you provided works for me if I add the following line to the ZCASSCF input:

"maxiter_micro" : 50,

[jevandezande]

Increasing maxiter_micro fixes it.

Thanks