Molcas Forum

arshadmehmood118

Member

Registered: 2026-05-28

Posts: 2

Dyson RASSI biorthonormalization failure for core-ionization

Dear developers and users,
I am computing XPS Dyson spectra from a valence-excited S1 state.

My protocol is neutral RASSCF + XMS-CASPT2 (S0, S1), then core-ionized RASSCF with HEXS + XMS-CASPT2 (20 doublet roots), then RASSI with DYSOn.  The core orbital sits in RAS1 (1 orbital), with 3 valence orbitals in RAS2 and 3 pi* in RAS3.
Input attached below.

  Basis=cc-pvdz
  Group=C1

&SEWARD
&RASSCF
  Title    =  N1s neutral reference, S0/S1, RAS(8,7)
  File     =  RefFC_RasOrb
  Symmetry =  1
  Spin     =  1
  Inactive =  28
  Ras1     =  1
  Ras2     =  3
  Ras3     =  3
  Nactel   =  8 1 6
  SupSym   =  1; 1 2
  Alter    =  1; 1 2 29
  Ciroot   =  2 2 1

&CASPT2
  IPEA      = 0.00
  IMAGinary = 0.20
  XMultiState = All
  MaxIter   = 2000
  FROZen    = 0
  PROPerties

>> COPY $Project.JobMix JobMix_Neutral_S1_N1sRAS
>> COPY $Project.JobIph JobIph_Neutral_S1_N1sRAS
>> COPY $Project.RasOrb RasOrb_Neutral_S1_N1sRAS

&RASSCF
  Title    =  N1s core-ionized final states, RAS(7,7)
  File     =  RasOrb_Neutral_S1_N1sRAS
  Symmetry =  1
  Charge   =  1
  Spin     =  2
  Inactive =  28
  Ras1     =  1
  Ras2     =  3
  Ras3     =  3
  Nactel   =  7 1 6
  SupSym   =  1; 1 29
  HExs     =  1; 1
  Ciroot   =  20 20 1

&CASPT2
  IPEA      = 0.00
  IMAGinary = 0.20
  XMultiState = All
  MaxIter   = 2000
  FROZen    = 0
  PROPerties

>> COPY $Project.JobMix JobMix_N1s_Ionized
>> COPY $Project.JobIph JobIph_N1s_Ionized
>> COPY $Project.RasOrb RasOrb_N1s_Ionized

>> COPY JobMix_Neutral_S1_N1sRAS JOB001
>> COPY JobMix_N1s_Ionized       JOB002

&RASSI
  NrOfJobIphs = 2 1 20	        			
  2						
  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
  SUBSets = 2
  EJob
  DYSOn
  DYSExport
  1 0
  CIPRint

With full orbital relaxation in the core-ionized RASSCF, RASSI crashes with:

  RASSI CANNOT CONTINUE. THE PROBLEM AT HAND
  IS PROBABLY NOT SOLUBLE. THE TWO ORBITAL
  SPACES ARE TOO DISSIMILAR.
  LU PARTITIONING IS TROUBLESOME. DIAGONAL
  ELEMENT NR.                    30  IS TOO SMALL:
  IN MATRIX CXA IT IS   1.00000000000000
  IN MATRIX CYB IT IS  3.788449464205840E-005
  EVEN AFTER OPTIMAL PIVOT-TRANSFORMATION.

Element 30 is the first RAS2 valence orbital, not the core. Inspecting the converged orbitals confirms that orbital 30 has rotated and swapped with the inactive space during the core-hole optimization, which is causing the error in RASSI. Interestingly, this rotation reproduces across cc-pVDZ and 6-31+G**, but the optimization works without any error when using aug-cc-pVDZ or 6-31G**.

I attempted to use the CRPR keyword (Manual sec. 4.2.47.1) with CRPR = 29 33.0 to address this, but it did not resolve the issue. Unfortunately, there is very little documentation or literature available on best practices for CRPR.

I was eventually able to get both RASSCF and RASSI to complete successfully by using SUPSYM to lock the active space and prevent the sigma-pi mixing:
SupSym = 2; 1 29; 6 30 31 32 33 34 35

While this bypasses the error, it introduces a severe energy penalty. Compared to an unconstrained run, the SUPSYM absolute energy is approximately 1.5 to 2.0 eV higher, and the relative state spacings differ by up to 1.8 eV.

This leads me to a few questions for the community:

1. Is there a recommended way to keep neutral and core-ionized active-orbital indices aligned to calculate valid Dyson pole strengths without artificially freezing the orbital rotations (which clearly suppresses critical core-hole relaxation)?

2. Are there any recommended settings, test sets, or sample inputs demonstrating the proper use of CRPR for this type of problem?

3. I am planning a production run across 10,000 time-resolved geometries. What is the best practical strategy here? Is the SUPSYM active-space constraint an accepted workaround in the literature (i.e., do the relative quantities and splittings remain physically meaningful for TR-XPS despite the absolute energy shift), or does this level of constraint introduce an unacceptable geometry-dependent bias across a trajectory?

I really appreciate your time and help.

Best regards,
Arshad

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,204

Re: Dyson RASSI biorthonormalization failure for core-ionization

I can't say much about "best practices" or expected errors, but I have a couple of thoughts:

1. If you have 3 orbitals in RAS3 and you allow up to 6 electrons... what's the difference from including those orbitals in RAS2?

2. To stabilize the RAS2 space you could try increasing the number of roots in the state-average (with the intent of including states that excite from the problem orbital), or increasing the active space size to include the "new" inactive orbital that wants to come in.

3. If you get your desired active space with some basis set, you could try projecting that result to other basis sets (using e.g. Tools/mort).

arshadmehmood118

Member

Registered: 2026-05-28

Posts: 2

Re: Dyson RASSI biorthonormalization failure for core-ionization

Hi @Ignacio,
Thank you so much for your quick help.
I tested your suggestions, and both restricting the electrons in RAS3 and increasing the number of roots in the neutral RASSCF step solved the issue.

I really appreciate your time. You are a lifesaver.

Best regards.