Thank you very much, that worked (but not restart)

Dear Eleonora and Aleks,

Thank you so much for the detailed explanation and for sharing the corrected input/output file. That really clarified the issue. I tried your suggested settings with FMIXING and SPINLOCK adjustments, and the geometry optimization is running well with no errors. Have a nice day!

Best,
Aparajita

Hi,

I think that the best way to proceed in this case would be to restart the calculation one or more times. To do this, you just have to make sure that the scratch folder is not deleted when the calculation stops. Basically, you would run a first job with:

FREQCALC END

Then, after the first calculation stops (maybe because you reached a wall clock time limit) you would run a second job restarting from the first as:

FREQCALC RESTART END

In order to make this restart work, you need a few files from the previous job to be placed in the scratch folder of the new job: FREQINFO.DAT, fort.13 and fort.9 (to be renamed fort.20 in the new folder).

If needed, you can repeat this restart process multiple times until completion of the frequency calculation.

Hope this helps,

thank you

very grateful, I will run it again but not sure what to do here if it aborts again

GiacomoAmbrogio

Dear Ambrogio,

Thank you so much! the problem has been perfectly solved.

All the best,

wang

Hey,

Thank you. It works now. This Forum is a great idea!

Dear CRYSTAL community,

We’re excited to share our recent work on accelerating linear algebra operations in the CRYSTAL code using GPUs. Our implementation boosts the performance of self-consistent field (SCF) calculations by offloading key matrix operations like multiplication, diagonalization, inversion, and Cholesky decomposition to GPUs.

In the manuscript, we first analyze the performance and limitations of the standard parallel version of the code (Pcrystal) and then we evaluate the scalability of the new GPU-accelerated approach with 1 to 8 GPUs, observing remarkable scaling. To highlight these improvements, we present benchmark results on different systems, such as the example below.

post_forum_1.png

We expected significant speedups for large systems due to the limited number of k points, each requiring substantial computational effort. To ensure a fair comparison, we ran calculations using the massively parallel version of CRYSTAL (MPPcrystal) on a large MOF structure with over 30000 basis functions. Surprisingly, a single GPU on one node performed comparably to 512–1024 CPU cores running across 4–8 nodes.

To find out more, read the full paper here.

We aim to make this GPU-accelerated version of CRYSTAL available in the upcoming release, allowing all users to benefit from its enhanced performance for large-scale simulations. We look forward to reading your thoughts and discussing potential applications or further improvements.

A big thanks to Lorenzo Donà, Chiara Ribaldone, and Filippo Spiga for their contributions to the development of this code!

Dear Jonas,

Thanks for reaching out and being one of the most active users of these early days of the forum. Your question gives us the chance to clarify some aspects of the output file that might not be obvious to non expert users. Below, I will refer to your output file.

Harmonic Frequencies and IR intensities

To compute harmonic frequencies and IR intensities (with the default approach of the Berry phase) the input looks like:

FREQCALC INTENS ENDFREQ

In the output file, the following table is printed:

HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH EIGENVALUES (EIGV) OF THE MASS WEIGHTED HESSIAN MATRIX AND HARMONIC TRANSVERSE OPTICAL (TO) FREQUENCIES. IRREP LABELS REFER TO SYMMETRY REPRESENTATION ANALYSIS; A AND I INDICATE WHETHER THE MODE IS ACTIVE OR INACTIVE, RESPECTIVELY, FOR IR AND RAMAN; INTEGRATED IR INTENSITIES IN BRACKETS. CONVERSION FACTORS FOR FREQUENCIES: 1 CM**(-1) = 0.4556335E-05 HARTREE 1 THZ = 0.3335641E+02 CM**(-1) HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH MODES EIGV FREQUENCIES IRREP IR INTENS RAMAN (HARTREE**2) (CM**-1) (THZ) (KM/MOL) 1- 1 0.3488E-07 40.9894 1.2288 (A ) A ( 0.40) A 2- 2 0.6077E-07 54.1037 1.6220 (A ) A ( 0.96) A 3- 3 0.6801E-07 57.2344 1.7158 (A ) A ( 2.45) A 4- 4 0.2238E-06 103.8371 3.1130 (A ) A ( 11.47) A [...]

For each mode (or set of degenerate modes) its eigenvalue (in Ha${}^2$), harmonic frequency (in cm${}^{-1}$ and THz) and irreducible representation get printed. In addition, labels specifying whether the mode is IR/Raman active are also displayed (A and I indicate whether the mode is active or inactive, respectively).

Raman intensities

Raman intensities can be computed via a coupled-perturbed approach by inserting the INTRAMAN keyword followed by the INTCPHF block in the input deck:

FREQCALC INTRAMAN INTCPHF END ENDFREQ

Raman intensities are computed for each independent component of the polarizability tensor (xx, xy, xz, yy, yz, zz, labeled as "Single Crystal" in the output file) and are also averaged to mimic polycrystalline powder samples (total, parallel polarisation, perpendicular polarisation averages are printed in the output).

POLYCRYSTALLINE ISOTROPIC INTENSITIES (ARBITRARY UNITS) MODES FREQUENCIES I_tot I_par I_perp ---------------------------------------------------------------- 1- 1 40.9894 (A ) 0.46 0.27 0.19 2- 2 54.1037 (A ) 7.35 4.23 3.12 3- 3 57.2344 (A ) 12.79 8.82 3.96 4- 4 103.8371 (A ) 13.66 7.89 5.77 SINGLE CRYSTAL DIRECTIONAL INTENSITIES (ARBITRARY UNITS) MODES FREQUENCIES I_xx I_xy I_xz I_yy I_yz I_zz ---------------------------------------------------------------------------- 1- 1 40.9894 (A ) 0.00 0.37 0.02 0.63 0.00 0.21 2- 2 54.1037 (A ) 3.17 0.69 0.00 4.35 3.66 10.05 3- 3 57.2344 (A ) 3.82 3.54 0.02 3.50 0.03 27.53 4- 4 103.8371 (A ) 2.57 1.81 0.01 16.25 3.34 19.62

For more details on such polycrystalline averages, please refer to sections 8.4 and 8.7 of the CRYSTAL23 manual.

Raman spectrum

A continuous Raman spectrum can be simulated by use of the RAMSPEC block, as in:

FREQCALC INTRAMAN INTCPHF END RAMSPEC END ENDFREQ

The simulated spectrum is printed in an external file named RAMSPEC.DAT that contains several columns: column 1 with frequencies in cm${}^{-1}$, columns 2-4 with polycrystalline intensities (total, parallel, perpendicular), columns 5-10 with single crystal intensities (xx, xy, xz, yy, yz, zz).

Effect of Temperature and Laser wavelength

The effect of temperature and laser wavelength on computed Raman intensities can be accounted for by use of the RAMANEXP keyword, as in:

FREQCALC INTRAMAN INTCPHF END RAMANEXP 298 532 RAMSPEC END ENDFREQ

Here we set 298 K for the temperature and 532 nm for the laser wavelength. This option modifies the values of all computed Raman intensities (in the output and in the RAMSPEC.DAT file accordingly).

Please, note that other properties (harmonic frequencies and IR intensities) are not affected by this option and thus remain unchanged in the output.

Plots

When CRYSPLOT reads the CRYSTAL output file it only plots the total intensity of the polycrystalline powder model.

When CRYSPLOT reads the RAMSPEC.DAT file it plots all components:

Screenshot 2025-03-13 at 12.47.37.png

Other plotting tools can be used to plot specific columns of the RAMSPEC.DAT file (e.g., CRYSTALClear, gnuplot).

Thank you very much Jacques and Alessandro.
Now it works :).
Best regards
Xavier

Hi,
Thank you very much for your reply!

Best regards,
Masoud

Dear Jacques,

Thank you for your reply which indeed solve the problem.

Best regards
Xavier

Thank you Giacomo Ambrogio.
I am greatful to the CRYSTAL code forum
This will help alot.

Best Wishes

thank you

Dear CRYSTALlers,

The Theoretical Chemistry Group of the University of Torino and the Department of Chemistry and the Thomas Young Centre at Imperial College London are organising the

MSSC2025 Summer School on Ab initio Modelling in Solid State Chemistry

logo_Full.png

to be held in Torino (Italy), 8-12 September 2025. The School is designed for Master and Ph.D. students, as well as for post-docs and researchers who have an interest in getting or strengthening a background in Computational Solid State Chemistry, Physics, Materials Science, Surface- and Nano-Science.

The week-long School consists of morning lectures and afternoon hands-on tutorial sessions, where the formal framework and functionalities of the CRYSTAL electronic structure package will be explored.

Registration opens on March 15 2025. See the School website for further details.

The School takes place in the elegant city of Torino (Italy):

Torino.png

­
While we strongly encourage in-person participation, we also offer the possibility to attend remotely through streaming of morning lectures and afternoon hands-on tutorials.

Participants will have the opportunity to present their research at a poster session.

See you in Torino!

job314 Yes, it should print the geometry of the last optimization step. Please, be aware that the CIFPRT and CIFPRTSYM options are not as general as one would like them to be. For systems where the primitive and crystallographic cells differ, they may result in an incomplete list of atoms. In those cases, I recommend switching symmetry off with the SYMMREMO option in combination with TESTGEOM, just to generate the .cif files correctly.