### Re: [SIESTA-L] Lattice parameters shorter than crystal after optimization

I can not find any mistake in your calculations. Maybe you can use a different base. But, how do you know that the correct angle is 96^o and not 99^o ? Is this difference very important for you? I mean several theoretical approaches may have some differences with respect to the experiment, but they sill can be used to derive useful results. > Dear Siesta users, > I'm trying to perform the optimization of an unit cell composed by four > molecules of the a neurotransmitter (for reference, the type of system is > similar to: J. Appl. Phys. 129, 234702 (2021); > https://urldefense.com/v3/__https://doi.org/10.1063/5.0054383__;!!D9dNQwwGXtA!VFwPgy7nicqJ4IHnnPgVwKcU8-H_FqWGOqP4Sbhw1XjlP5vvlaNC3lgewdXOsPI0ygo0IQ2_Io4V3TxayhNCrw$ > ). It's a monoclinic crystal with equal alpha and gamma angles. > When I run the geometry optimization, the result is always the shortening > of the a, b, c, alpha and gamma, and the increase of beta (from 96° to > 98-99°). However, it is expected that the a, b, and c parameters to > increase. I thought it could be something related with the > pseudopotencial, and downloaded the PBE Pseudopotencial from different > sources, but nothing change. > Could someone help me to understand what's wrong? > I'm including the parameters I used to control the simulation > #MD.TypeOfRun CG MD.Steps > 2000 MD.MaxDispl 0.001 Bohr MD.MaxForceTol > 0.01 eV/Ang MD.VariableCell T > MD.MaxStressTol 0.02 GPa MD.TargetPressure 0.0 GPa > > # WriteCoorStep TWriteForces TWriteMDHistory > TWriteCoorInitial T > # DM.UseSaveDM TMD.UseSaveXV TMD.UseSaveCG > T > Thank you for your help > -- > SIESTA is supported by the Spanish Research Agency (AEI) and by the > European H2020 MaX Centre of Excellence > (https://urldefense.com/v3/__http://www.max-centre.eu/__;!!D9dNQwwGXtA!UqJJRS4uVg7t-Dbd_eAjfVnwWZ1t2C6FcN2aIQ3zMnpwP4Ll9vMlgtpcq4qP80YzOBfKsToR3FKvor_wWQ0$ > ) > * Dr Zacharias G. Fthenakis ** -- SIESTA is supported by the Spanish Research Agency (AEI) and by the European H2020 MaX Centre of Excellence (http://www.max-centre.eu/)

### Re: [SIESTA-L] Lattice parameters shorter than crystal after optimization

Dear Dr. Fthenakis, Thank you again for answering my questions. My main doubt isn't only the beta angle, but the whole changes in the lattice parameters. There are several crystal structures obtained for this neurotransmitter, and all of them have closer lattice parameters. Besides, it's common to find in the literature that lattice parameters are larger than experimental ones after geometry optimization performed with GGA-PBE functional. However, my results are the opposite. "a", "b", and "c" are ~10% shorter. I'm new to Siesta, so I don't know if there are some forces acting to rearrange the internal parameters... like something trying to push the unit cell. I don't know.. Could be easier to observe if there are any mistakes if I send the input file? By the way, I got the pseudopotentials from Simune website and the NNINC. I mentioned it because a friend had a similar problem when working with Quantum Espresso, and solved it by changing the source of the pseudopotential... I don't know if it could be the same here... -- SIESTA is supported by the Spanish Research Agency (AEI) and by the European H2020 MaX Centre of Excellence (http://www.max-centre.eu/)

### Re: [SIESTA-L] Lattice parameters shorter than crystal after optimization

Maybe the selection of an improper pseudopotential can cause problems. You may have a look here https://departments.icmab.es/leem/siesta/Pseudopotentials/index.html for pseudopotentials for SIESTA. An option is to use the "atoms" code to produce your own pseudopotential, but this is something that has to be done with caution. Maybe trying a different pseudopotential, among those which have already been tested, will solve your problem. > Dear Dr. Fthenakis, > > Thank you again for answering my questions. > > My main doubt isn't only the beta angle, but the whole changes in the > lattice parameters. There are several crystal structures obtained for this > neurotransmitter, and all of them have closer lattice parameters. Besides, > it's common to find in the literature that lattice parameters are larger > than experimental ones after geometry optimization performed with GGA-PBE > functional. However, my results are the opposite. "a", "b", and "c" are > ~10% shorter. > > I'm new to Siesta, so I don't know if there are some forces acting to > rearrange the internal parameters... like something trying to push the > unit cell. I don't know.. > > Could be easier to observe if there are any mistakes if I send the input > file? > > By the way, I got the pseudopotentials from Simune website and the NNINC. > I mentioned it because a friend had a similar problem when working with > Quantum Espresso, and solved it by changing the source of the > pseudopotential... I don't know if it could be the same here... > * Dr Zacharias G. Fthenakis ** -- SIESTA is supported by the Spanish Research Agency (AEI) and by the European H2020 MaX Centre of Excellence (http://www.max-centre.eu/)

### Re: [SIESTA-L] Lattice parameters shorter than crystal after optimization

Dear, Making some tests, I observed 2 things: 1- The system takes ~1800 steps to converge the optimization. During ~700 steps the lattice parameters (a, b, c) increase (beta almost isn't changing). It takes some more steps to the beta angle starts to change, while a, b, c decrease. It repeats in all tests I run. 2- Using non-equal k points (2x4x2 and 4x6x4) the lattice parameter values are closer to the experimental than 4x4x4. However, I continue getting values shorter than experimental. -- SIESTA is supported by the Spanish Research Agency (AEI) and by the European H2020 MaX Centre of Excellence (http://www.max-centre.eu/)

### Re: [SIESTA-L] Lattice parameters shorter than crystal after optimization

It is not an issue that you find lattice parameters smaller than the experimental ones. This may depend on several factors, including the selection of the functional you use. The k-points should be inverse proportional to the lattice parameters. They don't have to be equal. The energy for increasing k-points should converge. According to what you wrote, maybe you need more k-points. On the other hand the number of steps for convergence is too high. If the system is metallic, you may introduce a small electronic temperature, or introduce some mixing. I hope this helps. Zacharias Fthenakis > Dear, > > > Making some tests, I observed 2 things: > > > 1- The system takes ~1800 steps to converge the optimization. During ~700 > steps the lattice parameters (a, b, c) increase (beta almost isn't > changing). It takes some more steps to the beta angle starts to change, > while a, b, c decrease. It repeats in all tests I run. > > > 2- Using non-equal k points (2x4x2 and 4x6x4) the lattice parameter values > are closer to the experimental than 4x4x4. > > > However, I continue getting values shorter than experimental. > > > > > -- > SIESTA is supported by the Spanish Research Agency (AEI) and by the > European H2020 MaX Centre of Excellence > (https://urldefense.com/v3/__http://www.max-centre.eu/__;!!D9dNQwwGXtA!TuCr_mMwS9nHTS06MagdjGH7Bi3JjX9pzEVRBcsRwSkdmZv2gCrAZkXyZzYPBlb9FtOIxjaXpFmaAhtaRQU$ > ) > * Dr Zacharias G. Fthenakis ** -- SIESTA is supported by the Spanish Research Agency (AEI) and by the European H2020 MaX Centre of Excellence (http://www.max-centre.eu/)