[CPMD-list] difficulties with EPR calculations

Silviu Zilberman silviu at Princeton.EDU
Thu May 11 11:11:53 CEST 2006


Hi Reinout,

Thank you for your reply. I did try to remove all keywords from the 
&RESP part, leaving only the EPR keyword. Unfortunately I still 
experience exactly the same problem. Is there anything else wrong with 
my input?

Thanks, Silviu.

Reinout Declerck wrote:
> Hi Silviu,
>
> The 'HYP' keyword invokes the calculation of EPR hyperfine parameters 
> (EPR_HYP.F). This experimental piece of code is functional, but needs 
> some extra input, otherwise it will crash in the way you encountered.
>
> As stated in the manual, you should check the header of EPR_HYP.F how 
> to get it working:
>
> C     == Computes hyperfine parameters (VERY rough method!!!)         ==
> C     == with the method of C.G. Van de Walle and P.E. Blochl         ==
> C     ==                    Phys. Rev. B 47, 4244 (1993)              ==
> C     == how to use this routine                                      ==
> C     == 1) use TM PSP's                                              ==
> C     == 2) force CPMD to read in the atomic pseudo wavefunctions     ==
> C     ==    by putting the following in your input file:              ==
> C     ==         &BASIS                                               ==
> C     ==             PSEUDO AO [NUM]                                  ==
> C     ==             0 1 2 ... [NUM-1]                                ==
> C     ==             ... ! and this for all elements                  ==
> C     ==         &END                                                 ==
> C     == 3) for every element, place a [elementsname].ae in the       ==
> C     ==    same directory where your input file is located           ==
> C     ==    These .ae files must be in the same format as the         ==
> C     ==    USPP pseudopotential generation program of D. Vanderbilt  ==
> C     == 4) run the EPR routine with the option HYP                   ==
> C     ==--------------------------------------------------------------==
>
>
> Please note that the hyperfine routine in its current state will 
> return valid results only for first and second row chemical elements.
>
> Also, please read PRB 73, 115113 (2006) for more details about the 
> accuracy one can expect for g tensor calculations.
>
> Regards,
>
> Reinout
>
> ----- Original Message ----- From: "Silviu Zilberman" 
> <silviu at Princeton.EDU>
> To: <cpmd-list at cpmd.org>
> Sent: Thursday, May 11, 2006 9:28 AM
> Subject: [CPMD-list] difficulties with EPR calculations
>
>
>> Dear All,
>>
>> I am trying to use the new EPR functionality of CPMD (v. 3.11.1). I 
>> prepared a structure for which I optimized the wavefunction (the 
>> geometry is fine). Then, every time I run an EPR calculation the code 
>> crashes with an error message:
>>
>> PROGRAM STOPS IN SUBROUTINE MEMORY_CHECK|  CORRUPTED MEMORY [PROC=  14]
>> p14_15615:  p4_error: : 999
>>
>> (every processor is spitting a similar message, but some also list 
>> the corrupted variables, like
>>
>> ...
>> EIGR      183012405264        OK183017608472        OK      650400
>> psi       183017611280        OK183024425576        OK      851786
>> rhoe      183024427024        OK183027834176        OK      425893
>> v[n0]     183027834896        OK183031242048 CORRUPTED      425893
>> eirop     183031242768 CORRUPTED183032630296 CORRUPTED      173440
>> eivps     183032631312 CORRUPTED183034018840        OK      173440
>> C1t       183034019856        OK183045813784        OK     1474240
>> diffv0a   183045816336        OK183054465248        OK     1081113
>> ...
>>
>> Some processors also complain that the first address of C1t is 
>> corrupted.
>> It doesn't seem to be a memory limit problem:
>>
>> ------------------------------------------------------------------
>> [PEAK NUMBER  115]       PEAK MEMORY      49632012 =  397.1 MBytes
>> [ALL. NUMBER  108]      TOTAL MEMORY      43091907 =  344.7 MBytes
>> ==================================================================
>>
>> and our (dual Opteron) nodes have 2GB of memory each. In addition, 
>> the fact that the same variables are corrupted on all nodes indicates 
>> (I think) that it's a code related issue rather than hardware.
>>
>> Below is the input file. I would greatly appreciate your comments.
>>
>> Thanks, Silviu.
>>
>> --------------------------------------------------------------------------------------------- 
>>
>>
>> INPUT FILE:
>>
>>
>> &CPMD
>> LINEAR RESPONSE
>> RESTART COORDINATES WAVEFUNCTION LATEST
>> CONJUGATE GRADIENT ELECTRONS
>> CONVERGENCE ORBITALS
>>  1.0d-5
>>  MAXSTEP
>>   50000
>>  REAL SPACE WF KEEP
>>  TRAJECTORY XYZ
>>  FILEPATH
>>  /pvfs/silviu/EPR_PDT_active
>> &END
>>
>> &SYSTEM
>>  ANGSTROM
>>  SYMMETRY
>>  1
>>  CELL
>>  13.229425  1.0  1.0  0  0  0
>>  CUTOFF
>>  120.
>>  CHARGE
>>  -2
>> &END
>>
>> &RESP
>>  EPR
>>  OVERLAP
>>  0.1
>>  HYP
>> &END
>>
>> &ATOMS
>>
>> *Fe-q16
>>    LMAX=D
>>    2
>>        4.59378 6.01722 6.64393
>>        7.25407 5.60785 6.76516
>> *S-q6
>>   LMAX=P
>>   3
>>        6.02759 6.9518 8.24564
>>        6.1248 6.84295 5.07668
>>        9.17715 7.06325 6.72235
>> *O-q6
>>   LMAX=P
>>   4
>>        8.4227 4.13464 8.97638
>>        2.99429 4.85191 8.78731
>>        3.14056 4.55461 4.58191
>>        5.64473 3.18001 6.76378
>> *N-q5
>>   LMAX=P
>>   2
>>        8.58601 3.92803 4.56286
>>        2.46491 8.27062 6.4422
>> *C-q4
>>   LMAX=P
>>   10
>>        5.85625 9.29547 6.58413
>>        6.44051 8.63744 5.3385
>>        6.3696 8.7282 7.90236
>>        8.0854 4.58616 5.40449
>>        7.98264 4.74203 8.07122
>>        3.66034 5.30188 7.93443
>>        3.31794 7.46029 6.51862
>>        10.6484 5.98372 6.66049
>>        5.92112 4.34117 6.73045
>>        3.75089 5.14112 5.39089
>>
>> *H-q1
>>   LMAX=S
>>   9
>>        7.4678 8.82514 7.97528
>>        5.9162 9.27653 8.74375
>>        6.12808 10.3697 6.55365
>>        7.54167 8.73005 5.32463
>>        6.0404 9.1328 4.43948
>>        11.5255 6.6054 6.42739
>>        10.8232 5.48561 7.62698
>>        10.5277 5.22004 5.87937
>>        4.76028 9.23949 6.55651
>> &END
>>
>> &DFT
>>  NEWCODE
>>  FUNCTIONAL PBE
>> &END
>>               _______________________________________________
>> CPMD-list mailing list
>> CPMD-list at cpmd.org
>> http://cpmd.org/mailman/listinfo/cpmd-list
>>
>




More information about the CPMD-list mailing list