LTMP2 - Tutorial: Difference between revisions

Latest revision as of 10:12, 19 July 2022

Overview > MP2 > LTMP2 > stochastic LTMP2 > High energy contributions using stochastic LTMP2 > List of tutorials

On this page, we explain how to perform a calculation using the LTMP2^[1] algorithm. Make sure you successfully completed the preparation steps Hartree-Fock ground state and Hartree-Fock virtuals.

NOTE: If you use this algorithm, please cite reference ^[1] in your publication in addition to the standard VASP reference.

The INCAR file

The LTMP2 calculation can simply be performed using the following INCAR file

ALGO = ACFDTRK 
LMP2LT = .TRUE.
NOMEGA = # number of tau points (see below)
NBANDS = # same value as in the Hartree-Fock unoccupied step ( = number of plane-waves)
ENCUT = # same value as in the Hartree-Fock step
LORBITALREAL = .TRUE.
PRECFOCK = Fast
KPAR = # parallelization (see below)

Make sure that VASP reads the WAVECAR file from the Hartree-Fock virtuals step. The setting for PRECFOCK is strongly recommended, since the code heavily relies on real space grid FFTs.

NOMEGA flag

The number of $\tau$ -points is controlled by the NOMEGA flag. This is necessary to calculate the Laplace transformed energy denominator (see Ref ^[1] for details),

${\displaystyle \frac{1}{\varepsilon_i + \varepsilon_j - \varepsilon_a -\varepsilon_b} = - \int_0^\infty \textrm e^{-(\varepsilon_i + \varepsilon_j - \varepsilon_a -\varepsilon_b)\tau} \; \textrm d \tau \;. }$

Usually it is sufficient to set NOMEGA=6. For materials with a small bandgap it is worth checking if the MP2 energy changes with increasing NOMEGA (e.g. 8 or 10). Note, that the MP2 energy diverges with 1/bandgap, independent of NOMEGA.

Parallelization

The LTMP2 algorithm is a high-performance code and can easily be used on many CPUs. Both OpenMP and MPI is supported. We recommend to use MPI for parallelization since the code possesses an almost ideal parallelization efficiency. OpenMP should only be used to increase the shared memory, if necessary.

In order to activate the efficient MPI parallelization use the KPAR flag in the following way (note that the usual meaning of the KPAR flag becomes obsolete in the LTMP2 algorithm). KPAR specifies the number of plane-waves treated in parallel. Ideally, set KPAR to half of the used MPI ranks. If this results in memory issues, further decrease KPAR (such that KPAR is alway a divisor of the used MPI ranks) or increase the number of OpenMP threads.

Example for 512 CPUs

MPI ranks: 512
OpenMP threads per rank: 1

KPAR = 256

To decrease the memory requirement you can alternatively set KPAR to 128 or 64 and so on. Or also try
MPI ranks: 256
OpenMP threads per rank: 2

KPAR = 128

References

↑ ^a ^b ^c T. Schäfer, B. Ramberger, and G. Kresse, J. Chem. Phys. 146, 104101 (2017).

[schaefer2017-1] T. Schäfer, B. Ramberger, and G. Kresse, J. Chem. Phys. 146, 104101 (2017).

[1]

@@ Line 1: / Line 1: @@
-(UNDER CONSTRUCTION)
+{{Template:MP2 - Tutorial}}
+On this page, we explain how to perform a calculation using the '''LTMP2'''<ref name="schaefer2017"/> algorithm. Make sure you successfully completed the preparation steps [[MP2 ground state calculation#Preparation:_the_Hartree-Fock_ground_state|Hartree-Fock ground state]] and [[MP2 ground state calculation#Calculating_the_unoccupied_Hartree-Fock_orbitals|Hartree-Fock virtuals]].
-On this page, we explain how to perform a calculation using the '''LTMP2'''<ref name="schaefer2017"/> algorithm. Make sure you successfully completed the preparation steps [[MP2 ground state calculation#Preparation:_the_Hartree-Fock_ground_state|Hartree-Fock ground state]] and [[MP2 ground state calculation#Calculating_the_unoccupied_Hartree-Fock_orbitals|Hartree-Fock virtuals]].
+'''NOTE:''' ''If you use this algorithm, please cite reference <ref name="schaefer2017"/> in your publication in addition to the standard VASP reference.''
 == The INCAR file ==
 The LTMP2 calculation can simply be performed using the following {{TAG|INCAR}} file
-  ALGO = ACFDTRK
+  {{TAGBL|ALGO}} = ACFDTRK
-  LMP2LT = .TRUE.
+  {{TAGBL|LMP2LT}} = .TRUE.
-  NOMEGA = # number of tau points (see below)
+  {{TAGBL|NOMEGA}} = # number of tau points (see below)
-  NBANDS = # same valule as in the Hartree-Fock unoccupieds step ( = number of plane-waves)
+  {{TAGBL|NBANDS}} = # same value as in the Hartree-Fock unoccupied step ( = number of plane-waves)
-  ENCUT = # same value as in the Hartree-Fock step
+  {{TAGBL|ENCUT}} = # same value as in the Hartree-Fock step
-  LORBITALREAL = .TRUE.
+  {{TAGBL|LORBITALREAL}} = .TRUE.
-  PRECFOCK = Fast
+  {{TAGBL|PRECFOCK}} = Fast
-Make sure that VASP reads the WAVECAR file from the [[MP2 ground state calculation#Calculating_the_unoccupied_Hartree-Fock_orbitals|Hartree-Fock virtuals]] step.
+ {{TAGBL|KPAR}} = # parallelization (see below)
+Make sure that VASP reads the WAVECAR file from the [[MP2 ground state calculation#Calculating_the_unoccupied_Hartree-Fock_orbitals|Hartree-Fock virtuals]] step. The setting for [[PRECFOCK|PRECFOCK]] is strongly recommended, since the code heavily relies on real space grid FFTs.
 ==== NOMEGA flag ====
-The number of <math>\tau</math>-points is set with the NOMEGA flag. This is necessary to calculate the Laplace transformed energy denominator (see Ref <ref name="schaefer2017"/>  for details),
+The number of <math>\tau</math>-points is controlled by the NOMEGA flag. This is necessary to calculate the Laplace transformed energy denominator (see Ref <ref name="schaefer2017"/>  for details),
 <math>
 \frac{1}{\varepsilon_i + \varepsilon_j - \varepsilon_a -\varepsilon_b} = - \int_0^\infty \textrm e^{-(\varepsilon_i + \varepsilon_j - \varepsilon_a -\varepsilon_b)\tau} \; \textrm d \tau \;.
-</math>.
+</math>
+Usually it is sufficient to set {{TAG|NOMEGA}}=6. For materials with a small bandgap it is worth checking if the MP2 energy changes with increasing {{TAG|NOMEGA}} (e.g. 8 or 10). Note, that the MP2 energy diverges with 1/bandgap, independent of {{TAG|NOMEGA}}.
 == Parallelization ==
-The LTMP2 algorithm is a high-performance code and can easily be used on many CPUs. Both OpenMP and MPI is supported.
+The LTMP2 algorithm is a high-performance code and can easily be used on many CPUs. Both OpenMP and MPI is supported. We recommend to use MPI for parallelization since the code possesses an almost ideal parallelization efficiency. OpenMP should only be used to increase the shared memory, if necessary.
+In order to activate the efficient MPI parallelization use the KPAR flag in the following way (note that the usual meaning of the KPAR flag becomes obsolete in the LTMP2 algorithm). KPAR specifies the number of plane-waves treated in parallel. Ideally, set KPAR to half of the used MPI ranks. If this results in memory issues, further decrease KPAR (such that KPAR is alway a divisor of the used MPI ranks) or increase the number of OpenMP threads.
+==== Example for 512 CPUs ====
+MPI ranks: 512 <br />
+OpenMP threads per rank: 1 <br />
+ KPAR = 256
+To decrease the memory requirement you can alternatively set KPAR to 128 or 64 and so on. Or also try <br />
+MPI ranks: 256 <br />
+OpenMP threads per rank: 2 <br />
+ KPAR = 128
 == References ==
@@ Line 31: / Line 46: @@
 </references>
+----
+[[Category:Many-body perturbation theory]][[Category:MP2]][[Category:Tutorials]]