Fatal error in newfileman.c

Hello, I am trying to run some DFT calculations with a double hybrid functional. And there is an error that reads like this:

" Q-Chem fatal error occurred in module libmdc/newfileman.C, line 384:

FileMan error: End of file reached prematurely reading (9014544) bytes in file FILE_VIAQ_INTS
Path: /home/***/scratch/qchem49588/617.0"

My $rem key looks like this:
$rem
AO2MO_DISK = 500000
MEM_STATIC = 10000
MEM_TOTAL = 120000
exchange = gen
omega = 300
lrc_dft = 1
hfk_sr_coef = 53000000
hfk_lr_coef = 100000000
correlation = rimp2
basis = aug-cc-pvtz
aux_basis = rimp2-aug-cc-pvtz
scf_convergence = 8
cis_singlets = 1
cis_triplets = 0
cis_n_roots = 20
MAX_SCF_CYCLES = 500
$end

The machine has more than 1 Tb of free disk space. So it is strange seeing failure with only 9014544 bytes

Could you kindly help?

Thanks

It’s not an out-of-space problem but rather than less was written to that file than Q-Chem was expecting to read later, usually a sign of either a bug or incompatible input options. It would be helpful if you could (a) post the whole input, including $molecule, and perhaps (b) find the minimal example that fails in this way.

Thank you so much for replying. The failure point is somewhere between 7-10 heavy atoms, but I don’t have the exact spot of failure.

$molecule
0 1
C 0.000000000 0.000000000 -1.433507970
C 1.208042190 0.000000000 -0.705394710
C -1.208042190 0.000000000 -0.705394710
C 0.000000000 0.000000000 1.361077520
C 1.205579080 0.000000000 0.676281000
C -1.205579080 0.000000000 0.676281000
C 1.251526350 0.000000000 -3.518737600
C -1.251526350 0.000000000 -3.518737600
N 0.000000000 0.000000000 -2.795722460
N 0.000000000 0.000000000 2.814439520
O 1.084589050 0.000000000 3.383352190
O -1.084589050 0.000000000 3.383352190
H 2.152075480 0.000000000 -1.219077130
H -2.152075480 0.000000000 -1.219077130
H 2.126799020 0.000000000 1.230607820
H -2.126799020 0.000000000 1.230607820
H 1.041017520 0.000000000 -4.582744420
H -1.041017520 0.000000000 -4.582744420
H 1.849465400 -0.884350720 -3.290102550
H 1.849465400 0.884350720 -3.290102550
H -1.849465400 0.884350720 -3.290102550
H -1.849465400 -0.884350720 -3.290102550
$end

$rem
MEM_STATIC = 10000
MEM_TOTAL = 120000
exchange = gen
omega = 300
lrc_dft = 1
hfk_sr_coef = 53000000
hfk_lr_coef = 100000000
correlation = rimp2
basis = aug-cc-pvtz
aux_basis = rimp2-aug-cc-pvtz
scf_convergence = 8
cis_singlets = 1
cis_triplets = 0
cis_n_roots = 20
MAX_SCF_CYCLES = 500
$end

$xc_functional
k 0.53
x mub88 0.47
c lyp 0.73
$end

A couple of things. First, having tried your job I can tell you that the SCF converges much faster if you pay attention to the warnings about S2THRESH and THRESH. Second, I failed to notice that you were doing excited states with double hybrids, there may actually be a memory problem. With the current version of Q-Chem (6.0.1), my job fails following the CIS calculation with a slightly different error:

 Q-Chem fatal error occurred in module libmdc/newfileman.C, line 376:

 FileMan error: Error reading (2147483632) bytes in file FILE_OIAQa_INTS
 Path: /tmp/slurmtmp.13028107/qchem98379/616.0

Clearly both problems are related to the RI-CIS(D) calculation that is required to compute the MP2 contribution to the excitation energies.

Update: I have tried this by requesting one root rather than 20, and a somewhat different (though probably related) error is encountered:

 Q-Chem fatal error occurred in module libmdc/newfileman.C, line 376:

 FileMan error: End of file reached prematurely reading (10170304) bytes in file FILE_VIAQ_INTS
 Path: /tmp/slurmtmp.13031738/qchem181222/617.0

I think I will have to leave this one for Q-Chem support. Does Q-Chem support excited-state calculations for hybrid functionals? This was not clear to me from the manual and I could not find an example.

Thanks immensely for this.

I am using Q-Chem (5.3.1) and with 5 states, the RI-CIS(D) portion completes successfully. This is enough to capture the charge-transfer state I was interested in.

Also, Q-Chem can support excited-state calculations for DHs, with some care. I was helped with this sometime ago: Manual specification of double hybrids

And, looking at the excited-state computations with DH from Q-Chem and other software, there is quite good agreement.

Hi John,

Thank you for the investigation. We’ll take it from here.

Kuan-Yu

Thanks Kuan-Yu.

I’d really appreciate some help/insights into this. Basically, we can’t use Q-Chem with triple-zeta basis sets for excited state properties with any Double-Hybrid DFA as soon as we ask for more than one state, especially once there are more than 10 heavy atoms. The RI-MP2 and RI-CIS(D) steps always fail. Although there are some alternative software, there are so many advantages to running these calculations with Q-Chem.