Re: [Wien] [Warning message: Sum of forces not small, possible numerical issues]
Dear Prof. Marks and Peter, Sorry for the delay in reply. I followed your suggestions and did a simple scf. Now I see forces are less than 5mRyd/au or negative on most of the atoms except on three "H" atom in CH3NH3PbI3 structure where I am getting ~12 mRyd/au. I am mentioning :. FOR013 for atom number 13 along with fcheck [bhamu@bhamu-pmc]$grep :FGL013 *.scf :FOR013: 13.ATOM169.039-30.643-61.735 -154.351 partial forces :FOR013: 13.ATOM159.798-28.662-58.657 -145.853 partial forces :FOR013: 13.ATOM138.173-23.942-51.530 -125.949 partial forces :FOR013: 13.ATOM 44.735 -6.229-18.200 -40.388 partial forces :FOR013: 13.ATOM 84.208 15.271 30.149 77.128 partial forces :FOR013: 13.ATOM 99.047 16.328 36.983 90.421 partial forces :FOR013: 13.ATOM 85.965 13.872 32.473 78.378 partial forces :FOR013: 13.ATOM 54.272 8.865 20.913 49.290 partial forces :FOR013: 13.ATOM 34.760 5.776 13.677 31.429 partial forces :FOR013: 13.ATOM 23.617 4.046 9.520 21.231 partial forces :FOR013: 13.ATOM 16.518 2.821 6.948 14.718 partial forces :FOR013: 13.ATOM 24.506 3.900 9.821 22.110 partial forces :FOR013: 13.ATOM 13.442215 2.158986 5.152247 *12.226458 total forces* [bhamu@bhamu-pmc]$grep :FCHECK *.scf :FCHECK: Sum of forces 13.692019325-3.440247228 -23.934499318 [bhamu@bhamu-pmc]$ So, I have some further queries: (1) I should not to force minimization (MSR1a/ -min/min_lapw) and only -fc 2 may be okay with proper convergence parameters. Please correct me if I am wrong. (2) As I am using TEMPS so I am not able to grep :GAP. How to grep :GAP? (3) I tested some scfs for the present case and scf with -fc xx -it switches is faster (in my case speed is double) than simple scf (-ec xx -cc yy). Why? (4) May I just set the proper convergence parameters in scf dir already having conversed structure with crude convergence parameters (ec 0.1 -cc 0.1 -fc 10 -it) and start scf? Kind regards Bhamu On Fri, Apr 6, 2018 at 6:58 PM, Laurence Markswrote: > My opinion. > > Use *TEMPS* not *TEMP*. It won't matter much but I consider it more > rigorous. I normally use TEMPS 0.0018 which is room temperature. (I have > rarely seen large values help, I think that is a relic of old mixers.) > > :FCHECK can be positive or negative, that does not matter. It should > become small as the forces and charge density converges. (Of course using a > centro-symmetric structure is better if possible.) > > On Fri, Apr 6, 2018 at 8:18 AM, Dr. K. C. Bhamu > wrote: > >> >> Thanks Prof. Peter, >> >> There are some more queries. >> >> For the present system, ~3-3.5 is best-reported rmt. Prof. Oleg reported >> 3.5 in his Sci. Rep. paper while in one of my earlier case I used 3 which >> was fine to me. So I kept rmt 3 in the present case. >> >> Yes, I am using 16 core here. >> >> The strategy told by Prof. Lawrence is similar to yours. So definitely, I >> will run a crude convergence before structure optimization (after getting >> the final conclusion from present scf). >> >> What should I do if :FCHECK are negative? can it be or I should worry? I >> see :FCHECK is oscillating (see below). >> >> In all test case of the present calculation, should increase LVNS from 4 >> to 6 in case.in1(c)? or in final. I never tested it so do not know how it >> will affect the calculation. >> >> >> >> *Can I use TEMP 0.002 instead of TETRA? * >> If I am not wrong, them in case of MSR1a the :FOR will work after final >> scf! >> >> >> :FCHECK: Sum of forces-69.94062752074.707201466 >> 256.655266254 >> :FCHECK: Sum of forces-64.56032957473.360823345 >> 231.480485436 >> :FCHECK: Sum of forces-48.03417516665.104093472 >> 166.986223927 >> :FCHECK: Sum of forces-35.06469516459.878383059 >> 106.378782186 >> :FCHECK: Sum of forces 1.65894838144.163592598 >> -28.823131541 >> :FCHECK: Sum of forces -1.71590062354.670180112 >> -38.208248369 >> :FCHECK: Sum of forces-44.59962824284.665399276 >> 80.309287437 >> :FCHECK: Sum of forces -104.389046420 114.231138278 >> 298.688183293 >> :FCHECK: Sum of forces -146.872184802 138.002715039 >> 506.089724144 >> :FCHECK: Sum of forces -139.929513824 116.740866115 >> 514.542253158 >> :FCHECK: Sum of forces-44.45895700153.512685999 >> 90.795768329 >> :FCHECK: Sum of forces-21.34654090853.389897235 >> -190.728160314 >> :FCHECK: Sum of forces-62.24145433785.443479855 >> -99.767209873 >> :FCHECK: Sum of forces-33.31540895460.975140246 >> 18.014683106 >> :FCHECK: Sum of forces
Re: [Wien] [Warning message: Sum of forces not small, possible numerical issues]
My opinion. Use *TEMPS* not *TEMP*. It won't matter much but I consider it more rigorous. I normally use TEMPS 0.0018 which is room temperature. (I have rarely seen large values help, I think that is a relic of old mixers.) :FCHECK can be positive or negative, that does not matter. It should become small as the forces and charge density converges. (Of course using a centro-symmetric structure is better if possible.) On Fri, Apr 6, 2018 at 8:18 AM, Dr. K. C. Bhamuwrote: > > Thanks Prof. Peter, > > There are some more queries. > > For the present system, ~3-3.5 is best-reported rmt. Prof. Oleg reported > 3.5 in his Sci. Rep. paper while in one of my earlier case I used 3 which > was fine to me. So I kept rmt 3 in the present case. > > Yes, I am using 16 core here. > > The strategy told by Prof. Lawrence is similar to yours. So definitely, I > will run a crude convergence before structure optimization (after getting > the final conclusion from present scf). > > What should I do if :FCHECK are negative? can it be or I should worry? I > see :FCHECK is oscillating (see below). > > In all test case of the present calculation, should increase LVNS from 4 > to 6 in case.in1(c)? or in final. I never tested it so do not know how it > will affect the calculation. > > > > *Can I use TEMP 0.002 instead of TETRA? * > If I am not wrong, them in case of MSR1a the :FOR will work after final > scf! > > > :FCHECK: Sum of forces-69.94062752074.707201466 > 256.655266254 > :FCHECK: Sum of forces-64.56032957473.360823345 > 231.480485436 > :FCHECK: Sum of forces-48.03417516665.104093472 > 166.986223927 > :FCHECK: Sum of forces-35.06469516459.878383059 > 106.378782186 > :FCHECK: Sum of forces 1.65894838144.163592598 > -28.823131541 > :FCHECK: Sum of forces -1.71590062354.670180112 > -38.208248369 > :FCHECK: Sum of forces-44.59962824284.665399276 > 80.309287437 > :FCHECK: Sum of forces -104.389046420 114.231138278 > 298.688183293 > :FCHECK: Sum of forces -146.872184802 138.002715039 > 506.089724144 > :FCHECK: Sum of forces -139.929513824 116.740866115 > 514.542253158 > :FCHECK: Sum of forces-44.45895700153.512685999 > 90.795768329 > :FCHECK: Sum of forces-21.34654090853.389897235 > -190.728160314 > :FCHECK: Sum of forces-62.24145433785.443479855 > -99.767209873 > :FCHECK: Sum of forces-33.31540895460.975140246 > 18.014683106 > :FCHECK: Sum of forces 22.819602947 -13.223462643 > 257.628716681 > :FCHECK: Sum of forces-30.71495809240.048468655 > 258.360331510 > > > > regards > > Bhamu > > On Fri, Apr 6, 2018 at 5:01 PM, Peter Blaha > wrote: > >> To be efficient you should: >> >> Start with an RKmax 0f 2.5 or 3.75 instead of 3 (check later) >> >> Reduce k-points to 2x2x1 (check later). >> >> run -it -p -fc 10 # crude convergence before structure optimization ! >> # -p because you want to use the 16 cores. You have >> to define a suitable .machines file ! Are you sure you are using more than >> 1 core right now ??? >> >> Once this is converged, check forces (are they so big that you need to >> relax ? and :FCHECK. If :FCHECK is still large at the end, probalby your >> spheres are not ok. And in addition you should increase LVNS from 4 to 6 in >> case.in1(c). >> >> save, and only then start a run -p -it -min >> >> >> On 04/06/2018 12:49 PM, Dr. K. C. Bhamu wrote: >> >>> Thank you Prof. Lyudmila for the inlined reply. >>> >>> I shall mention the optimization process: I am using run_lapw >>> -it -min to relax the structure with TETRA (default), rmt 3, >>> kpt 50 (4x4x2), PBE. Organometallic system (Hybrideperovskite). >>> >>> I am not sure whether my scf is going well or not because I do not see >>> any such message. >>> My main concern is about too much time (~40 minutes on 16 core) for one >>> scf cycle and this warning. >>> >>> >>> Could you please have a look at my further queries: >>> >>> >>> 06.04.2018 10:09, Dr. K. C. Bhamu wrote: >>> >>> :WARN >>> :WARNING Sum of forces not small, possible numerical issues >>> >>> I'd look attentively around this place in the scf. >>> >>> >>> >>> I do not find anything in scf with this statment or part of this >>> statment. Also, as per my excersie, nothing is mentioned in mailing list >>> for such warnings. >>> >>> >>> >>> below is constructive output from six scf cycles: >>> >>> All looks good, though I do not know what is :ADIST. Something >>> connected with MSR1a option. >>> >>> :ADIST Largest Distance 1.71E+02 mau for atom 18 >>> :ADIST Largest Distance 1.67E+02 mau for atom 32 >>> :ADIST Largest Distance 1.20E+02 mau for atom 32 >>> :ADIST Largest Distance 5.87E+01 mau for atom 32 >>> >>> >>> :ADIST
Re: [Wien] [Warning message: Sum of forces not small, possible numerical issues]
Thanks Prof. Peter, There are some more queries. For the present system, ~3-3.5 is best-reported rmt. Prof. Oleg reported 3.5 in his Sci. Rep. paper while in one of my earlier case I used 3 which was fine to me. So I kept rmt 3 in the present case. Yes, I am using 16 core here. The strategy told by Prof. Lawrence is similar to yours. So definitely, I will run a crude convergence before structure optimization (after getting the final conclusion from present scf). What should I do if :FCHECK are negative? can it be or I should worry? I see :FCHECK is oscillating (see below). In all test case of the present calculation, should increase LVNS from 4 to 6 in case.in1(c)? or in final. I never tested it so do not know how it will affect the calculation. *Can I use TEMP 0.002 instead of TETRA?* If I am not wrong, them in case of MSR1a the :FOR will work after final scf! :FCHECK: Sum of forces-69.94062752074.707201466 256.655266254 :FCHECK: Sum of forces-64.56032957473.360823345 231.480485436 :FCHECK: Sum of forces-48.03417516665.104093472 166.986223927 :FCHECK: Sum of forces-35.06469516459.878383059 106.378782186 :FCHECK: Sum of forces 1.65894838144.163592598 -28.823131541 :FCHECK: Sum of forces -1.71590062354.670180112 -38.208248369 :FCHECK: Sum of forces-44.59962824284.665399276 80.309287437 :FCHECK: Sum of forces -104.389046420 114.231138278 298.688183293 :FCHECK: Sum of forces -146.872184802 138.002715039 506.089724144 :FCHECK: Sum of forces -139.929513824 116.740866115 514.542253158 :FCHECK: Sum of forces-44.45895700153.512685999 90.795768329 :FCHECK: Sum of forces-21.34654090853.389897235 -190.728160314 :FCHECK: Sum of forces-62.24145433785.443479855 -99.767209873 :FCHECK: Sum of forces-33.31540895460.975140246 18.014683106 :FCHECK: Sum of forces 22.819602947 -13.223462643 257.628716681 :FCHECK: Sum of forces-30.71495809240.048468655 258.360331510 regards Bhamu On Fri, Apr 6, 2018 at 5:01 PM, Peter Blahawrote: > To be efficient you should: > > Start with an RKmax 0f 2.5 or 3.75 instead of 3 (check later) > > Reduce k-points to 2x2x1 (check later). > > run -it -p -fc 10 # crude convergence before structure optimization ! > # -p because you want to use the 16 cores. You have > to define a suitable .machines file ! Are you sure you are using more than > 1 core right now ??? > > Once this is converged, check forces (are they so big that you need to > relax ? and :FCHECK. If :FCHECK is still large at the end, probalby your > spheres are not ok. And in addition you should increase LVNS from 4 to 6 in > case.in1(c). > > save, and only then start a run -p -it -min > > > On 04/06/2018 12:49 PM, Dr. K. C. Bhamu wrote: > >> Thank you Prof. Lyudmila for the inlined reply. >> >> I shall mention the optimization process: I am using run_lapw >> -it -min to relax the structure with TETRA (default), rmt 3, >> kpt 50 (4x4x2), PBE. Organometallic system (Hybrideperovskite). >> >> I am not sure whether my scf is going well or not because I do not see >> any such message. >> My main concern is about too much time (~40 minutes on 16 core) for one >> scf cycle and this warning. >> >> >> Could you please have a look at my further queries: >> >> >> 06.04.2018 10:09, Dr. K. C. Bhamu wrote: >> >> :WARN >> :WARNING Sum of forces not small, possible numerical issues >> >> I'd look attentively around this place in the scf. >> >> >> >> I do not find anything in scf with this statment or part of this >> statment. Also, as per my excersie, nothing is mentioned in mailing list >> for such warnings. >> >> >> >> below is constructive output from six scf cycles: >> >> All looks good, though I do not know what is :ADIST. Something >> connected with MSR1a option. >> >> :ADIST Largest Distance 1.71E+02 mau for atom 18 >> :ADIST Largest Distance 1.67E+02 mau for atom 32 >> :ADIST Largest Distance 1.20E+02 mau for atom 32 >> :ADIST Largest Distance 5.87E+01 mau for atom 32 >> >> >> :ADIST represents the distance between atoms (Pg number 212 UG). Atom 18 >> and 32 are H. in the 12th Cycle :FGL is positive for both and is around >> ~40mRyd/au. >> >> >> Too large forces appear and movement of the atoms becomes too large. >> >> TOTAL FORCE WITH RESPECT TO THE GLOBAL COORDINATE SYSTEM: >> :FGL018: 18.ATOM -228.797027515 -110.869626478 >> >> >> I think that mixer is very clever program, so this possibility is >> taken into account, and it wiil correctly treat this data. >> >> >> results from :FGL and hence from :FCHECK is varying. In first cycle the >> total forces from :FCHECK was very small and in the 8th Cycle total
Re: [Wien] [Warning message: Sum of forces not small, possible numerical issues]
I suggest -ec 0.1 -fc 10 -cc 0.1 similar to what Peter sent. N.B., I believe he had a typo, and meant "Start with an RKmax 0f 2.5 or 2.75 instead of 3 (check later)" On Fri, Apr 6, 2018 at 8:01 AM, Dr. K. C. Bhamuwrote: > > Okay, I will give a try to this approach also. > > -ec 0.001 and -fc 5 is enough for the first scf run the MSR1? > > regards > Bhamu > > > On Fri, Apr 6, 2018 at 4:55 PM, Laurence Marks > wrote: > >> It is more common to somewhat converge using MSR1 first, e.g. -cc 0.1, >> before switching to MSR1a. However, I don't see and clear indication that >> something is wrong in your output. A large value of :FCHECK is not unusual >> in the early stages. >> >> _ >> Professor Laurence Marks >> "Research is to see what everybody else has seen, and to think what >> nobody else has thought", Albert Szent-Gyorgi >> www.numis.northwestern.edu >> >> On Fri, Apr 6, 2018, 5:49 AM Dr. K. C. Bhamu wrote: >> >>> Thank you Prof. Lyudmila for the inlined reply. >>> >>> I shall mention the optimization process: I am using run_lapw >>> -it -min to relax the structure with TETRA (default), rmt 3, >>> kpt 50 (4x4x2), PBE. Organometallic system (Hybrideperovskite). >>> >>> I am not sure whether my scf is going well or not because I do not see >>> any such message. >>> My main concern is about too much time (~40 minutes on 16 core) for one >>> scf cycle and this warning. >>> >>> >>> Could you please have a look at my further queries: >>> >>> >>> 06.04.2018 10:09, Dr. K. C. Bhamu wrote: >>> :WARN > :WARNING Sum of forces not small, possible numerical issues > I'd look attentively around this place in the scf. >>> >>> >>> I do not find anything in scf with this statment or part of this >>> statment. Also, as per my excersie, nothing is mentioned in mailing list >>> for such warnings. >>> >>> >>> below is constructive output from six scf cycles: > All looks good, though I do not know what is :ADIST. Something connected with MSR1a option. > :ADIST Largest Distance 1.71E+02 mau for atom 18 > :ADIST Largest Distance 1.67E+02 mau for atom 32 > :ADIST Largest Distance 1.20E+02 mau for atom 32 > :ADIST Largest Distance 5.87E+01 mau for atom 32 > >>> :ADIST represents the distance between atoms (Pg number 212 UG). Atom 18 >>> and 32 are H. in the 12th Cycle :FGL is positive for both and is around >>> ~40mRyd/au. >>> >>> Too large forces appear and movement of the atoms becomes too large. > TOTAL FORCE WITH RESPECT TO THE GLOBAL COORDINATE SYSTEM: > :FGL018: 18.ATOM -228.797027515 -110.869626478 > I think that mixer is very clever program, so this possibility is taken into account, and it wiil correctly treat this data. >>> >>> results from :FGL and hence from :FCHECK is varying. In first cycle the >>> total forces from :FCHECK was very small and in the 8th Cycle total forces >>> are too high. The average totla force on each atom is >>> >>> >>> results from FCHECk and FR are below (in 12th scf cycle): >>> FCHECK *.scf >>> >>> > :FCHECK: Sum of forces-69.94062752074.707201466 >>> 256.655266254 >>> > :FCHECK: Sum of forces-64.56032957473.360823345 >>> 231.480485436 >>> > :FCHECK: Sum of forces-48.03417516665.104093472 >>> 166.986223927 >>> > :FCHECK: Sum of forces-35.06469516459.878383059 >>> 106.378782186 >>> > :FCHECK: Sum of forces 1.65894838144.163592598 >>> -28.823131541 >>> > :FCHECK: Sum of forces -1.71590062354.670180112 >>> -38.208248369 >>> > :FCHECK: Sum of forces-44.59962824284.665399276 >>> 80.309287437 >>> > :FCHECK: Sum of forces -104.389046420 114.231138278 >>> 298.688183293 >>> > :FCHECK: Sum of forces -146.872184802 138.002715039 >>> 506.089724144 >>> > :FCHECK: Sum of forces -139.929513824 116.740866115 >>> 514.542253158 >>> > :FCHECK: Sum of forces-44.45895700153.512685999 >>> 90.795768329 >>> > :FCHECK: Sum of forces-21.34654090853.389897235 >>> -190.728160314 >>> >>> >>> :FR *.scf >>> >>> > :FRMSA: (mRyd/au) 684.75( 18) 482.59 RMS (au) 2.11E-03 MAX 3.34E-03 >>> :F-cond.(mRyd/au)2.00 F >>> > :FRMSA: (mRyd/au) 666.97( 18) 473.06 RMS (au) 6.32E-03 MAX 1.00E-02 >>> :F-cond.(mRyd/au)2.00 F >>> > :FRMSA: (mRyd/au) 620.05( 18) 446.96 RMS (au) 7.70E-03 MAX 1.22E-02 >>> :F-cond.(mRyd/au)2.00 F >>> > :FRMSA: (mRyd/au) 564.83( 18) 415.40 RMS (au) 2.26E-02 MAX 3.60E-02 >>> :F-cond.(mRyd/au)2.00 F >>> > :FRMSA: (mRyd/au) 410.77( 18) 314.76 RMS (au) 3.10E-02 MAX 5.00E-02 >>> :F-cond.(mRyd/au)2.00 F >>> > :FRMSA: (mRyd/au) 228.80( 18) 169.19 RMS (au) 3.10E-02 MAX 5.00E-02 >>> :F-cond.(mRyd/au)2.00 F >>> > :FRMSA: (mRyd/au) 65.91( 17) 47.66 RMS (au) 1.16E-02 MAX 2.21E-02 >>>
Re: [Wien] [Warning message: Sum of forces not small, possible numerical issues]
Okay, I will give a try to this approach also. -ec 0.001 and -fc 5 is enough for the first scf run the MSR1? regards Bhamu On Fri, Apr 6, 2018 at 4:55 PM, Laurence Markswrote: > It is more common to somewhat converge using MSR1 first, e.g. -cc 0.1, > before switching to MSR1a. However, I don't see and clear indication that > something is wrong in your output. A large value of :FCHECK is not unusual > in the early stages. > > _ > Professor Laurence Marks > "Research is to see what everybody else has seen, and to think what nobody > else has thought", Albert Szent-Gyorgi > www.numis.northwestern.edu > > On Fri, Apr 6, 2018, 5:49 AM Dr. K. C. Bhamu wrote: > >> Thank you Prof. Lyudmila for the inlined reply. >> >> I shall mention the optimization process: I am using run_lapw >> -it -min to relax the structure with TETRA (default), rmt 3, >> kpt 50 (4x4x2), PBE. Organometallic system (Hybrideperovskite). >> >> I am not sure whether my scf is going well or not because I do not see >> any such message. >> My main concern is about too much time (~40 minutes on 16 core) for one >> scf cycle and this warning. >> >> >> Could you please have a look at my further queries: >> >> >> 06.04.2018 10:09, Dr. K. C. Bhamu wrote: >> >>> :WARN :WARNING Sum of forces not small, possible numerical issues >>> I'd look attentively around this place in the scf. >>> >> >> >> I do not find anything in scf with this statment or part of this >> statment. Also, as per my excersie, nothing is mentioned in mailing list >> for such warnings. >> >> >> >>> >>> below is constructive output from six scf cycles: >>> All looks good, though I do not know what is :ADIST. Something connected >>> with MSR1a option. >>> :ADIST Largest Distance 1.71E+02 mau for atom 18 :ADIST Largest Distance 1.67E+02 mau for atom 32 :ADIST Largest Distance 1.20E+02 mau for atom 32 :ADIST Largest Distance 5.87E+01 mau for atom 32 >>> >> :ADIST represents the distance between atoms (Pg number 212 UG). Atom 18 >> and 32 are H. in the 12th Cycle :FGL is positive for both and is around >> ~40mRyd/au. >> >> >>> >>> Too large forces appear and movement of the atoms becomes too large. >>> TOTAL FORCE WITH RESPECT TO THE GLOBAL COORDINATE SYSTEM: :FGL018: 18.ATOM -228.797027515 -110.869626478 >>> >>> I think that mixer is very clever program, so this possibility is taken >>> into account, and it wiil correctly treat this data. >>> >> >> results from :FGL and hence from :FCHECK is varying. In first cycle the >> total forces from :FCHECK was very small and in the 8th Cycle total forces >> are too high. The average totla force on each atom is >> >> >> results from FCHECk and FR are below (in 12th scf cycle): >> FCHECK *.scf >> >> > :FCHECK: Sum of forces-69.94062752074.707201466 >> 256.655266254 >> > :FCHECK: Sum of forces-64.56032957473.360823345 >> 231.480485436 >> > :FCHECK: Sum of forces-48.03417516665.104093472 >> 166.986223927 >> > :FCHECK: Sum of forces-35.06469516459.878383059 >> 106.378782186 >> > :FCHECK: Sum of forces 1.65894838144.163592598 >> -28.823131541 >> > :FCHECK: Sum of forces -1.71590062354.670180112 >> -38.208248369 >> > :FCHECK: Sum of forces-44.59962824284.665399276 >> 80.309287437 >> > :FCHECK: Sum of forces -104.389046420 114.231138278 >> 298.688183293 >> > :FCHECK: Sum of forces -146.872184802 138.002715039 >> 506.089724144 >> > :FCHECK: Sum of forces -139.929513824 116.740866115 >> 514.542253158 >> > :FCHECK: Sum of forces-44.45895700153.512685999 >> 90.795768329 >> > :FCHECK: Sum of forces-21.34654090853.389897235 >> -190.728160314 >> >> >> :FR *.scf >> >> > :FRMSA: (mRyd/au) 684.75( 18) 482.59 RMS (au) 2.11E-03 MAX 3.34E-03 >> :F-cond.(mRyd/au)2.00 F >> > :FRMSA: (mRyd/au) 666.97( 18) 473.06 RMS (au) 6.32E-03 MAX 1.00E-02 >> :F-cond.(mRyd/au)2.00 F >> > :FRMSA: (mRyd/au) 620.05( 18) 446.96 RMS (au) 7.70E-03 MAX 1.22E-02 >> :F-cond.(mRyd/au)2.00 F >> > :FRMSA: (mRyd/au) 564.83( 18) 415.40 RMS (au) 2.26E-02 MAX 3.60E-02 >> :F-cond.(mRyd/au)2.00 F >> > :FRMSA: (mRyd/au) 410.77( 18) 314.76 RMS (au) 3.10E-02 MAX 5.00E-02 >> :F-cond.(mRyd/au)2.00 F >> > :FRMSA: (mRyd/au) 228.80( 18) 169.19 RMS (au) 3.10E-02 MAX 5.00E-02 >> :F-cond.(mRyd/au)2.00 F >> > :FRMSA: (mRyd/au) 65.91( 17) 47.66 RMS (au) 1.16E-02 MAX 2.21E-02 >> :F-cond.(mRyd/au)2.00 F >> > :FRMSA: (mRyd/au) 57.84( 9) 41.34 RMS (au) 9.47E-03 MAX 1.98E-02 >> :F-cond.(mRyd/au)2.00 F >> > :FRMSA: (mRyd/au) 76.92( 35) 55.05 RMS (au) 7.92E-03 MAX 2.08E-02 >> :F-cond.(mRyd/au)2.00 F >> > :FRMSA: (mRyd/au) 68.91( 21) 60.53 RMS (au) 8.58E-03 MAX 1.53E-02 >> :F-cond.(mRyd/au)2.00 F >> > :FRMSA: (mRyd/au) 59.32( 25) 53.55
Re: [Wien] [Warning message: Sum of forces not small, possible numerical issues]
To be efficient you should: Start with an RKmax 0f 2.5 or 3.75 instead of 3 (check later) Reduce k-points to 2x2x1 (check later). run -it -p -fc 10 # crude convergence before structure optimization ! # -p because you want to use the 16 cores. You have to define a suitable .machines file ! Are you sure you are using more than 1 core right now ??? Once this is converged, check forces (are they so big that you need to relax ? and :FCHECK. If :FCHECK is still large at the end, probalby your spheres are not ok. And in addition you should increase LVNS from 4 to 6 in case.in1(c). save, and only then start a run -p -it -min On 04/06/2018 12:49 PM, Dr. K. C. Bhamu wrote: Thank you Prof. Lyudmila for the inlined reply. I shall mention the optimization process: I am using run_lapw -it -min to relax the structure with TETRA (default), rmt 3, kpt 50 (4x4x2), PBE. Organometallic system (Hybrideperovskite). I am not sure whether my scf is going well or not because I do not see any such message. My main concern is about too much time (~40 minutes on 16 core) for one scf cycle and this warning. Could you please have a look at my further queries: 06.04.2018 10:09, Dr. K. C. Bhamu wrote: :WARN :WARNING Sum of forces not small, possible numerical issues I'd look attentively around this place in the scf. I do not find anything in scf with this statment or part of this statment. Also, as per my excersie, nothing is mentioned in mailing list for such warnings. below is constructive output from six scf cycles: All looks good, though I do not know what is :ADIST. Something connected with MSR1a option. :ADIST Largest Distance 1.71E+02 mau for atom 18 :ADIST Largest Distance 1.67E+02 mau for atom 32 :ADIST Largest Distance 1.20E+02 mau for atom 32 :ADIST Largest Distance 5.87E+01 mau for atom 32 :ADIST represents the distance between atoms (Pg number 212 UG). Atom 18 and 32 are H. in the 12th Cycle :FGL is positive for both and is around ~40mRyd/au. Too large forces appear and movement of the atoms becomes too large. TOTAL FORCE WITH RESPECT TO THE GLOBAL COORDINATE SYSTEM: :FGL018: 18.ATOM -228.797027515 -110.869626478 I think that mixer is very clever program, so this possibility is taken into account, and it wiil correctly treat this data. results from :FGL and hence from :FCHECK is varying. In first cycle the total forces from :FCHECK was very small and in the 8th Cycle total forces are too high. The average totla force on each atom is results from FCHECk and FR are below (in 12th scf cycle): FCHECK *.scf > :FCHECK: Sum of forces -69.940627520 74.707201466 256.655266254 > :FCHECK: Sum of forces -64.560329574 73.360823345 231.480485436 > :FCHECK: Sum of forces -48.034175166 65.104093472 166.986223927 > :FCHECK: Sum of forces -35.064695164 59.878383059 106.378782186 > :FCHECK: Sum of forces 1.658948381 44.163592598 -28.823131541 > :FCHECK: Sum of forces -1.715900623 54.670180112 -38.208248369 > :FCHECK: Sum of forces -44.599628242 84.665399276 80.309287437 > :FCHECK: Sum of forces -104.389046420 114.231138278 298.688183293 > :FCHECK: Sum of forces -146.872184802 138.002715039 506.089724144 > :FCHECK: Sum of forces -139.929513824 116.740866115 514.542253158 > :FCHECK: Sum of forces -44.458957001 53.512685999 90.795768329 > :FCHECK: Sum of forces -21.346540908 53.389897235 -190.728160314 :FR *.scf > :FRMSA: (mRyd/au) 684.75( 18) 482.59 RMS (au) 2.11E-03 MAX 3.34E-03 :F-cond.(mRyd/au) 2.00 F > :FRMSA: (mRyd/au) 666.97( 18) 473.06 RMS (au) 6.32E-03 MAX 1.00E-02 :F-cond.(mRyd/au) 2.00 F > :FRMSA: (mRyd/au) 620.05( 18) 446.96 RMS (au) 7.70E-03 MAX 1.22E-02 :F-cond.(mRyd/au) 2.00 F > :FRMSA: (mRyd/au) 564.83( 18) 415.40 RMS (au) 2.26E-02 MAX 3.60E-02 :F-cond.(mRyd/au) 2.00 F > :FRMSA: (mRyd/au) 410.77( 18) 314.76 RMS (au) 3.10E-02 MAX 5.00E-02 :F-cond.(mRyd/au) 2.00 F > :FRMSA: (mRyd/au) 228.80( 18) 169.19 RMS (au) 3.10E-02 MAX 5.00E-02 :F-cond.(mRyd/au) 2.00 F > :FRMSA: (mRyd/au) 65.91( 17) 47.66 RMS (au) 1.16E-02 MAX 2.21E-02 :F-cond.(mRyd/au) 2.00 F > :FRMSA: (mRyd/au) 57.84( 9) 41.34 RMS (au) 9.47E-03 MAX 1.98E-02 :F-cond.(mRyd/au) 2.00 F > :FRMSA: (mRyd/au) 76.92( 35) 55.05 RMS (au) 7.92E-03 MAX 2.08E-02 :F-cond.(mRyd/au) 2.00 F > :FRMSA: (mRyd/au) 68.91( 21) 60.53 RMS (au) 8.58E-03 MAX 1.53E-02 :F-cond.(mRyd/au) 2.00 F > :FRMSA: (mRyd/au) 59.32( 25) 53.55 RMS (au) 7.45E-03 MAX 1.32E-02 :F-cond.(mRyd/au) 2.00 F > :FRMSA: (mRyd/au) 64.13( 13) 54.92 RMS (au) 6.46E-03 MAX 1.04E-02 :F-cond.(mRyd/au) 2.00 F
Re: [Wien] [Warning message: Sum of forces not small, possible numerical issues]
It is more common to somewhat converge using MSR1 first, e.g. -cc 0.1, before switching to MSR1a. However, I don't see and clear indication that something is wrong in your output. A large value of :FCHECK is not unusual in the early stages. _ Professor Laurence Marks "Research is to see what everybody else has seen, and to think what nobody else has thought", Albert Szent-Gyorgi www.numis.northwestern.edu On Fri, Apr 6, 2018, 5:49 AM Dr. K. C. Bhamuwrote: > Thank you Prof. Lyudmila for the inlined reply. > > I shall mention the optimization process: I am using run_lapw > -it -min to relax the structure with TETRA (default), rmt 3, > kpt 50 (4x4x2), PBE. Organometallic system (Hybrideperovskite). > > I am not sure whether my scf is going well or not because I do not see any > such message. > My main concern is about too much time (~40 minutes on 16 core) for one > scf cycle and this warning. > > > Could you please have a look at my further queries: > > > 06.04.2018 10:09, Dr. K. C. Bhamu wrote: > >> :WARN >>> :WARNING Sum of forces not small, possible numerical issues >>> >> I'd look attentively around this place in the scf. >> > > > I do not find anything in scf with this statment or part of this statment. > Also, as per my excersie, nothing is mentioned in mailing list for such > warnings. > > > >> >> below is constructive output from six scf cycles: >>> >> All looks good, though I do not know what is :ADIST. Something connected >> with MSR1a option. >> >>> :ADIST Largest Distance 1.71E+02 mau for atom 18 >>> :ADIST Largest Distance 1.67E+02 mau for atom 32 >>> :ADIST Largest Distance 1.20E+02 mau for atom 32 >>> :ADIST Largest Distance 5.87E+01 mau for atom 32 >>> >> > :ADIST represents the distance between atoms (Pg number 212 UG). Atom 18 > and 32 are H. in the 12th Cycle :FGL is positive for both and is around > ~40mRyd/au. > > >> >> Too large forces appear and movement of the atoms becomes too large. >> >>> TOTAL FORCE WITH RESPECT TO THE GLOBAL COORDINATE SYSTEM: >>> :FGL018: 18.ATOM -228.797027515 -110.869626478 >>> >> >> I think that mixer is very clever program, so this possibility is taken >> into account, and it wiil correctly treat this data. >> > > results from :FGL and hence from :FCHECK is varying. In first cycle the > total forces from :FCHECK was very small and in the 8th Cycle total forces > are too high. The average totla force on each atom is > > > results from FCHECk and FR are below (in 12th scf cycle): > FCHECK *.scf > > > :FCHECK: Sum of forces-69.94062752074.707201466 > 256.655266254 > > :FCHECK: Sum of forces-64.56032957473.360823345 > 231.480485436 > > :FCHECK: Sum of forces-48.03417516665.104093472 > 166.986223927 > > :FCHECK: Sum of forces-35.06469516459.878383059 > 106.378782186 > > :FCHECK: Sum of forces 1.65894838144.163592598 > -28.823131541 > > :FCHECK: Sum of forces -1.71590062354.670180112 > -38.208248369 > > :FCHECK: Sum of forces-44.59962824284.665399276 > 80.309287437 > > :FCHECK: Sum of forces -104.389046420 114.231138278 > 298.688183293 > > :FCHECK: Sum of forces -146.872184802 138.002715039 > 506.089724144 > > :FCHECK: Sum of forces -139.929513824 116.740866115 > 514.542253158 > > :FCHECK: Sum of forces-44.45895700153.512685999 > 90.795768329 > > :FCHECK: Sum of forces-21.34654090853.389897235 > -190.728160314 > > > :FR *.scf > > > :FRMSA: (mRyd/au) 684.75( 18) 482.59 RMS (au) 2.11E-03 MAX 3.34E-03 > :F-cond.(mRyd/au)2.00 F > > :FRMSA: (mRyd/au) 666.97( 18) 473.06 RMS (au) 6.32E-03 MAX 1.00E-02 > :F-cond.(mRyd/au)2.00 F > > :FRMSA: (mRyd/au) 620.05( 18) 446.96 RMS (au) 7.70E-03 MAX 1.22E-02 > :F-cond.(mRyd/au)2.00 F > > :FRMSA: (mRyd/au) 564.83( 18) 415.40 RMS (au) 2.26E-02 MAX 3.60E-02 > :F-cond.(mRyd/au)2.00 F > > :FRMSA: (mRyd/au) 410.77( 18) 314.76 RMS (au) 3.10E-02 MAX 5.00E-02 > :F-cond.(mRyd/au)2.00 F > > :FRMSA: (mRyd/au) 228.80( 18) 169.19 RMS (au) 3.10E-02 MAX 5.00E-02 > :F-cond.(mRyd/au)2.00 F > > :FRMSA: (mRyd/au) 65.91( 17) 47.66 RMS (au) 1.16E-02 MAX 2.21E-02 > :F-cond.(mRyd/au)2.00 F > > :FRMSA: (mRyd/au) 57.84( 9) 41.34 RMS (au) 9.47E-03 MAX 1.98E-02 > :F-cond.(mRyd/au)2.00 F > > :FRMSA: (mRyd/au) 76.92( 35) 55.05 RMS (au) 7.92E-03 MAX 2.08E-02 > :F-cond.(mRyd/au)2.00 F > > :FRMSA: (mRyd/au) 68.91( 21) 60.53 RMS (au) 8.58E-03 MAX 1.53E-02 > :F-cond.(mRyd/au)2.00 F > > :FRMSA: (mRyd/au) 59.32( 25) 53.55 RMS (au) 7.45E-03 MAX 1.32E-02 > :F-cond.(mRyd/au)2.00 F > > :FRMSA: (mRyd/au) 64.13( 13) 54.92 RMS (au) 6.46E-03 MAX 1.04E-02 > :F-cond.(mRyd/au)2.00 F > ___ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at:
Re: [Wien] [Warning message: Sum of forces not small, possible numerical issues]
Thank you Prof. Lyudmila for the inlined reply. I shall mention the optimization process: I am using run_lapw -it -min to relax the structure with TETRA (default), rmt 3, kpt 50 (4x4x2), PBE. Organometallic system (Hybrideperovskite). I am not sure whether my scf is going well or not because I do not see any such message. My main concern is about too much time (~40 minutes on 16 core) for one scf cycle and this warning. Could you please have a look at my further queries: 06.04.2018 10:09, Dr. K. C. Bhamu wrote: > :WARN >> :WARNING Sum of forces not small, possible numerical issues >> > I'd look attentively around this place in the scf. > I do not find anything in scf with this statment or part of this statment. Also, as per my excersie, nothing is mentioned in mailing list for such warnings. > > below is constructive output from six scf cycles: >> > All looks good, though I do not know what is :ADIST. Something connected > with MSR1a option. > >> :ADIST Largest Distance 1.71E+02 mau for atom 18 >> :ADIST Largest Distance 1.67E+02 mau for atom 32 >> :ADIST Largest Distance 1.20E+02 mau for atom 32 >> :ADIST Largest Distance 5.87E+01 mau for atom 32 >> > :ADIST represents the distance between atoms (Pg number 212 UG). Atom 18 and 32 are H. in the 12th Cycle :FGL is positive for both and is around ~40mRyd/au. > > Too large forces appear and movement of the atoms becomes too large. > >> TOTAL FORCE WITH RESPECT TO THE GLOBAL COORDINATE SYSTEM: >> :FGL018: 18.ATOM -228.797027515 -110.869626478 >> > > I think that mixer is very clever program, so this possibility is taken > into account, and it wiil correctly treat this data. > results from :FGL and hence from :FCHECK is varying. In first cycle the total forces from :FCHECK was very small and in the 8th Cycle total forces are too high. The average totla force on each atom is results from FCHECk and FR are below (in 12th scf cycle): FCHECK *.scf > :FCHECK: Sum of forces-69.94062752074.707201466 256.655266254 > :FCHECK: Sum of forces-64.56032957473.360823345 231.480485436 > :FCHECK: Sum of forces-48.03417516665.104093472 166.986223927 > :FCHECK: Sum of forces-35.06469516459.878383059 106.378782186 > :FCHECK: Sum of forces 1.65894838144.163592598 -28.823131541 > :FCHECK: Sum of forces -1.71590062354.670180112 -38.208248369 > :FCHECK: Sum of forces-44.59962824284.665399276 80.309287437 > :FCHECK: Sum of forces -104.389046420 114.231138278 298.688183293 > :FCHECK: Sum of forces -146.872184802 138.002715039 506.089724144 > :FCHECK: Sum of forces -139.929513824 116.740866115 514.542253158 > :FCHECK: Sum of forces-44.45895700153.512685999 90.795768329 > :FCHECK: Sum of forces-21.34654090853.389897235 -190.728160314 :FR *.scf > :FRMSA: (mRyd/au) 684.75( 18) 482.59 RMS (au) 2.11E-03 MAX 3.34E-03 :F-cond.(mRyd/au)2.00 F > :FRMSA: (mRyd/au) 666.97( 18) 473.06 RMS (au) 6.32E-03 MAX 1.00E-02 :F-cond.(mRyd/au)2.00 F > :FRMSA: (mRyd/au) 620.05( 18) 446.96 RMS (au) 7.70E-03 MAX 1.22E-02 :F-cond.(mRyd/au)2.00 F > :FRMSA: (mRyd/au) 564.83( 18) 415.40 RMS (au) 2.26E-02 MAX 3.60E-02 :F-cond.(mRyd/au)2.00 F > :FRMSA: (mRyd/au) 410.77( 18) 314.76 RMS (au) 3.10E-02 MAX 5.00E-02 :F-cond.(mRyd/au)2.00 F > :FRMSA: (mRyd/au) 228.80( 18) 169.19 RMS (au) 3.10E-02 MAX 5.00E-02 :F-cond.(mRyd/au)2.00 F > :FRMSA: (mRyd/au) 65.91( 17) 47.66 RMS (au) 1.16E-02 MAX 2.21E-02 :F-cond.(mRyd/au)2.00 F > :FRMSA: (mRyd/au) 57.84( 9) 41.34 RMS (au) 9.47E-03 MAX 1.98E-02 :F-cond.(mRyd/au)2.00 F > :FRMSA: (mRyd/au) 76.92( 35) 55.05 RMS (au) 7.92E-03 MAX 2.08E-02 :F-cond.(mRyd/au)2.00 F > :FRMSA: (mRyd/au) 68.91( 21) 60.53 RMS (au) 8.58E-03 MAX 1.53E-02 :F-cond.(mRyd/au)2.00 F > :FRMSA: (mRyd/au) 59.32( 25) 53.55 RMS (au) 7.45E-03 MAX 1.32E-02 :F-cond.(mRyd/au)2.00 F > :FRMSA: (mRyd/au) 64.13( 13) 54.92 RMS (au) 6.46E-03 MAX 1.04E-02 :F-cond.(mRyd/au)2.00 F ___ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at: http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
Re: [Wien] [Warning message: Sum of forces not small, possible numerical issues]
06.04.2018 10:09, Dr. K. C. Bhamu wrote: :WARN :WARNING Sum of forces not small, possible numerical issues I'd look attentively around this place in the scf. :WARN : QTL-B value eq. 2.79 in Band of energy -0.76594 ATOM= 7 this is not much, and can easily disappear at the following cycles. Only if it become more than 5 at final iterations of the scf cycle I would worry, :WARN : You should change the E-parameter for this atom and L-value in case.in1 (or try the -in1new Only in the above mentioned case. :ENE : *WARNING** TOTAL ENERGY IN Ry = -141317.23075973 This is a consequence of the former warnings, no other importance. ... :DIS : CHARGE DISTANCE ( 0.238821 for atom 15 spin 1) 0.107937 :DIS's look good. below is constructive output from six scf cycles: All looks good, though I do not know what is :ADIST. Something connected with MSR1a option. :ADIST Largest Distance 1.71E+02 mau for atom 18 :ADIST Largest Distance 1.67E+02 mau for atom 32 :ADIST Largest Distance 1.20E+02 mau for atom 32 :ADIST Largest Distance 5.87E+01 mau for atom 32 Too large forces appear and movement of the atoms becomes too large. TOTAL FORCE WITH RESPECT TO THE GLOBAL COORDINATE SYSTEM: :FGL018: 18.ATOM -228.797027515 -110.869626478 I think that mixer is very clever program, so this possibility is taken into account, and it wiil correctly treat this data. So, the program warns you to look after the movement of the atoms with attention to atom 18 and 32. Best wishes Lyudmila Dobysheva -- Physics-Techn.Institute, Udmurt Federal Research Center, Ural Br. of Rus.Ac.Sci. 426000 Izhevsk Kirov str. 132 Russia --- Tel. +7 (34I2)43-24-59 (office), +7 (9I2)OI9-795O (home) Skype: lyuka18 (office), lyuka17 (home) E-mail: lyuk...@mail.ru (office), lyuk...@gmail.com (home) ___ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at: http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html