Newton-X CS 2.6/COLUMBUS 7.3: Unexpected Number of MR-SDCI Roots and Missing Nonadiabatic Coupling Vectors

[Akshaya Kumar Rauta]

Hello,

I am facing a problem with the Newton-X CS 2.6 interface with COLUMBUS 7.3 while setting up a two-state nonadiabatic dynamics calculation.

An important point is that the example calculations supplied with Newton-X run successfully on the same system and produce the expected results, including the required quantities. Therefore, the Newton-X/COLUMBUS installation and interface appear to be working in general. The problem occurs specifically with my own molecular system/input.

For my calculation, Newton-X is explicitly configured for two states:

&input nat = 8 nstat = 2 nstatdyn = 2 istep = 0 dt = 0.1 t = 0 tmax = 100 prog = 1.0 thres = 100 killstat = 1 timekill = 0 ndamp = 0 lvprt = 1 kt = 1 mem = 200 Etot_jump = 0.25 Etot_drift = 0.25 nxrestart = 0 zpecorr = 0 /

Thus:

nstat = 2 nstatdyn = 2

My COLUMBUS ciudgin also specifies two roots:

NROOT = 2 RTOLCI = 1e-4,1e-4

When COLUMBUS is run directly, it correctly calculates two MR-SDCI roots:

mr-sdci # 1 1 -227.8282441869 0.0000E+00 3.5271E-26 1.3119E-13 1.0000E-04 

mr-sdci # 1 2 -227.6438335287 -1.7053E-13 0.0000E+00 1.5341E-13 1.0000E-04

However, when the calculation is called through Newton-X, MR-SDCI gives four roots:

mr-sdci # 1 1 -227.7682178922 2.0442E+02 6.4304E-02 2.9924E-01 1.0000E-03 

mr-sdci # 1 2 -227.5869324305 2.0424E+02 0.0000E+00 2.7076E-01 1.0000E-04 

mr-sdci # 1 3 -227.3441990706 2.0400E+02 0.0000E+00 3.4593E-01 1.0000E-04

 mr-sdci # 1 4 -227.1881340427 2.0384E+02 0.0000E+00 4.5003E-01 1.0000E-04

In addition, the nonadiabatic coupling (NAC) vectors are not calculated for my system.

My columbus.par is:

memcol = 1600 cver = 7 cirestart = 0 quad_conv = 600 mc_conv = 0 ci_conv = 0 ivmode = 8 mocoef = 1 prt_mo = 0 ci_iter = 30 citol = 0.0001 reduce_tol = 1 all_grads = 1 grad_lvl = MCSCF mull_pop = 0

The main point that I find difficult to understand is:

The supplied Newton-X/COLUMBUS examples run correctly on the same installation, but my system does not.

For my system:

• Direct COLUMBUS → 2 MR-SDCI roots, as requested

• Newton-X + COLUMBUS → 4 MR-SDCI roots

• control.dyn → nstat = 2, nstatdyn = 2

• ciudgin → NROOT = 2

• NAC vectors → not calculated

• Supplied Newton-X examples → run correctly

Could you please suggest which input files or parameters I should compare with the working example?

In particular:

1. Why might Newton-X/COLUMBUS use four MR-SDCI roots for my system despite nstat = 2 and NROOT = 2?

2. Does Newton-X modify ciudgin or generate another CI input during the electronic-structure call?

3. Could the number of roots be related to the DRT/reference-space setup rather than NROOT alone?

4. What input controls whether the NAC vectors are calculated by COLUMBUS?

5. Could the unexpected four-root calculation and missing NAC vectors originate from the same input/setup problem?

Software versions:
Newton-X CS: 2.6
COLUMBUS: 7.3

I would be happy to provide the complete COLUMBUS input directory and Newton-X input/output files if needed.

Thank you.