Negative excitation energies with active-space CR-EOMCCSD


Just Got Here
Dear all,

A computation of benzene excitations with creom(t)ac with this input:

MEMORY total 48000 mb stack 2000 mb heap 2000 mb
START tce_benzene_MeOH_eomcc
GEOMETRY UNITS ANGSTROM
C    0.759749176   1.031275058   0.373772394
H    1.435016257   1.875664269   0.374704616
C    1.266617787  -0.267362341   0.421273076
H    2.334915967  -0.429180190   0.459432338
C    0.395320544  -1.355991164   0.424905109
H    0.788661934  -2.362492586   0.463035490
C   -0.982205637  -1.146654408   0.381270236
H   -1.657656316  -1.991140192   0.385121003
C   -1.489346115   0.151149791   0.337572337
H   -2.557947039   0.313750489   0.307719001
C   -0.618775155   1.240331207   0.333883729
H   -1.011761615   2.247106897   0.304369215
END
BASIS spherical
 * library maug-cc-pVTZ
END
SCF
 SINGLET
 RHF
END
TCE
creom(t)ac
emin_act -1.0
emax_act 1.0
  2eorb
2emet 13
  split 8
 TILESIZE 15
 FREEZE CORE ATOMIC
 NROOTS 2
 TARGET 1
IO GA
END
TASK TCE ENERGY
set tce:thresheom 1.0d-4


produces negative excitation energies:

CR-EOMCCSD(T)-ACT(II)  total energy / hartree       =      -231.241750851019901
 CR-EOMCCSD(T)-ACT(II)  excitation energy (eV)       =     -2.39452

corresponding excitation energy = 0
corresponding excitation energy = 0
corresponding excitation energy = 0
corresponding excitation energy = 0

CR-EOMCCSD(T)-act Intermediates cpu wall time            61.9           67.3
 cr_N i1_1 nr of boxes                    152
 cr_N i1_2 nr of boxes                   4150
 creom_N i2_1 nr of boxes                    152
 creom_N i2_2 nr of boxes                   4150
 creom_N i2_3 nr of boxes                    152
 creom_N i2_4 nr of boxes                   4150


CR-EOMCCSD(T)-ACT(II) triples loop cpu wall          1528.3         1527.3

 CR-EOMCCSD(T)-ACT(II)  total energy / hartree       =      -231.244410865723324

Now, I might be choosing a wrong active space, but already the end of EOM-CCSD is wrong. 2 last iterations read:

 Iteration    65
   0.0000001659096  -0.0000000016228   -0.00000    13.6    13.7
   0.0000000133072   0.0000000000255    0.00000    13.6    13.7

 Iteration    66
   0.0000000920125  -0.0000000006372   -0.00000    13.8    13.9
   0.0000000078340   0.0000000000164    0.00000    13.8    13.9


while if I change the window of active space to [-0.5,1.0] not only I obtain reasonable energies with CR-EOMCCSD(T)-ACT:
CR-EOMCCSD(T)-ACT(II) triples loop cpu wall           232.6          233.0

 CR-EOMCCSD(T)-ACT(II)  total energy / hartree       =      -230.652869292841928
 CR-EOMCCSD(T)-ACT(II)  excitation energy (eV)       =      5.81811


CR-EOMCCSD(T)-act Intermediates cpu wall time            83.5           86.6
 cr_N i1_1 nr of boxes                    152
 cr_N i1_2 nr of boxes                   4150
 creom_N i2_1 nr of boxes                    152
 creom_N i2_2 nr of boxes                   4150
 creom_N i2_3 nr of boxes                    152
 creom_N i2_4 nr of boxes                   4150


CR-EOMCCSD(T)-ACT(II) triples loop cpu wall           232.6          232.5

 CR-EOMCCSD(T)-ACT(II)  total energy / hartree       =      -230.623758973731810
 CR-EOMCCSD(T)-ACT(II)  excitation energy (eV)       =      6.61024


but also EOMCCSD converges to non-zero values:
 Iteration    38
   0.0000000000437   0.2234464413987    6.08029    20.8    20.9
   0.0000001117331   0.2495576796759    6.79081    20.8    20.9

 Iteration    39
   0.0000000000813   0.2234464413984    6.08029    20.8    20.9
   0.0000000675230   0.2495576799136    6.79081    20.8    20.9


Shouldn't at least EOM-CCSD give me the same answer (unless it's not really EOM-CCSD)?

Best,

Ewa Pastorczak


Forum >> NWChem's corner >> Running NWChem