User Tools

Site Tools


exercises:2015_pitt:mp2

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Next revision
Previous revision
exercises:2015_pitt:mp2 [2015/03/05 11:31] – created vondeleexercises:2015_pitt:mp2 [2020/08/21 10:15] (current) – external edit 127.0.0.1
Line 1: Line 1:
-====== MP2 and RPA ======+====== MP2 and RPA =====
  
-TODO+In order to go beyond GGA and hybrid DFT, one option is to include wavefunction correlation terms. Recently, MP2 and RPA correlation have been added to CP2K [[doi>10.1016/j.cpc.2014.10.021]], [[doi>10.1021/ct4002202]], [[doi>10.1021/ct300531w]]. The implementation is aimed at condensed phase calculations (see e.g. [[doi>10.1021/jz401931f]] and [[doi>10.1021/jz501985w]]), and in the case of MP2 provides energies, forces and stress at O(N^5) cost, while RPA provides energies at O(N^4) cost. 
 + 
 +However, significant computer resources need to be available for condensed phase calculations. We will start with gas phase calculations first, even though RI-GPW is not particularly efficient in this case. 
 + 
 +===== 1. Task: Benzene dimer MP2 binding energy ===== 
 + 
 +Employ the provided input file to compute the benzene dimer binding energy. The provided dimer geometry is optimized already. To obtain the energy of the monomer, geometry optimization is necessary (is it?). 
 + 
 +Topics: 
 +   * RI approach (''RI_MP2_GPW'' and ''RI_AUX_BASIS_SET''
 +   * Wavelet solver (''PSOLVER  WAVELET'', ''CENTER_COORDINATES''
 +   * gas phase HFX calculation  
 +   * Parallel efficiency (''NUMBER_PROC'', ''MEMORY'', ''BLOCK_SIZE''
 + 
 +===== 2. Task: Benzene monomer RPA energy: frequency integration ===== 
 + 
 +RPA is HFX+RPA correlation. It can be performed with HFX orbitals and eigenvalues, but also based on e.g. GGA or hybrid orbitals. Two advantages over MP2 are : 1) scaling is O(N^4) (for RI-dRPA with frequency integration). 2) it is applicable for systems with a small gap. 
 + 
 +Here, we look at the convergence of the RPA energy as a function of the number of integration points (''RPA_NUM_QUAD_POINTS''). Change this parameter in the range 6-20. For large systems the time needed for the calculations is proportional to the number of integration points. 
 + 
 +Use the following section to change to RPA: 
 + 
 +<code> 
 +      ! with WF correlation 
 +      &WF_CORRELATION 
 +        ! use the RI-GPW approach 
 +        METHOD  RI_RPA_GPW 
 +        &WFC_GPW 
 +        &END 
 +        &RI_RPA 
 +          ! number of quadrature points, essential for accurate energies. 
 +          ! small gap systems need more points 
 +          RPA_NUM_QUAD_POINTS 
 +          ! essentially always use minimax 
 +          MINIMAX 
 +          &HF 
 +            FRACTION 1.0000000 
 +            &SCREENING 
 +              EPS_SCHWARZ 1.0E-8 
 +              SCREEN_ON_INITIAL_P FALSE 
 +            &END SCREENING 
 +          &END HF 
 +        &END RI_RPA 
 +        MEMORY    1800 
 +        NUMBER_PROC 
 +      &END 
 +</code> 
 + 
 +It is strongly recommended to use the ''MINIMAX'' integration scheme.  
 + 
 +===== 3. task : condensed phase systems ===== 
 + 
 +Condensed phase systems are computationally are demanding. However, example input files can be found as part of [[src>tests/QS/benchmark_mp2_rpa/64-H2O/ | the CP2K benchmarks]]. 
 + 
 +While the corresponding runtimes are available online [[performance:magnus-h2o-64-ri-mp2]] 
 + 
 +For large runs, using a mixed MPI+OMP setup will be beneficial, and these calculations can also exploit GPUs. 
 + 
 +====== Required files ====== 
 + 
 +input file for an RI-MP2 calculation on a benzene dimer 
 + 
 +<code - benzene_dimer.inp> 
 +&GLOBAL 
 +  PROJECT      benzene_dimer 
 +  RUN_TYPE     ENERGY 
 +&END GLOBAL 
 + 
 +&FORCE_EVAL 
 +  METHOD         Quickstep 
 +  &DFT 
 +    ! specification of basis and potential files (cp2k/data) 
 +    BASIS_SET_FILE_NAME    ./HFX_BASIS 
 +    POTENTIAL_FILE_NAME    ./HF_POTENTIALS 
 +    &MGRID 
 +      CUTOFF    400 
 +    &END MGRID 
 + 
 +    &QS 
 +      METHOD GPW 
 +      EPS_DEFAULT  1.0E-10 
 +      EPS_PGF_ORB  1.0E-8 
 +    &END QS 
 + 
 +    ! standard OT 
 +    &SCF 
 +      SCF_GUESS ATOMIC 
 +      EPS_SCF 1.0E-6 
 +      MAX_SCF 40 
 +      &OT 
 +        MINIMIZER       CG 
 +        PRECONDITIONER  FULL_SINGLE_INVERSE 
 +      &END 
 +      &OUTER_SCF 
 +        EPS_SCF 1.0E-6 
 +        MAX_SCF 10 
 +      &END 
 +    &END SCF 
 +    
 +    ! Non periodic calculation needs Poisson solver: use wavelet solver 
 +    &POISSON             
 +      PERIODIC NONE 
 +      PSOLVER  WAVELET 
 +    &END POISSON 
 + 
 +    &XC 
 +      ! no XC functional 
 +      &XC_FUNCTIONAL NONE                        
 +      &END XC_FUNCTIONAL 
 +      ! and 100%HFX 
 +      &HF          
 +        FRACTION    1.0  
 +        &SCREENING       
 +          EPS_SCHWARZ          1.0E-9 
 +        &END SCREENING 
 +        &MEMORY 
 +          MAX_MEMORY  1800   
 +        &END 
 +      &END HF 
 +      ! with WF correlation 
 +      &WF_CORRELATION            
 +        ! use the RI-GPW approach 
 +        METHOD  RI_MP2_GPW 
 +        &WFC_GPW 
 +        &END 
 +        ! block sizes and memory affect parallel efficiency 
 +        &RI_MP2                      
 +          BLOCK_SIZE              
 +        &END 
 +        MEMORY    1800 
 +        NUMBER_PROC 
 +      &END 
 +    &END XC 
 +  &END DFT 
 +   
 +  &SUBSYS 
 +    ! sufficiently large non-periodic unit cell 
 +    &CELL 
 +      ABC [angstrom]  15 15 15 
 +      PERIODIC NONE           ! Non periodic calculation. 
 +    &END CELL 
 +     
 +! specification of an external file with coordinates  
 +    &COORD 
 +  H         8.5709951714        6.1617188657        9.1769626364 
 +  C         8.0387778483        5.9757981379        8.2528397815 
 +  C         6.6427474511        5.9561428224        8.2430496957 
 +  H         6.0937543445        6.1249650636        9.1610309894 
 +  C         5.9546039990        5.7314555716        7.0475649921 
 +  H         4.8717398728        5.7196290593        7.0404902271 
 +  C         6.6635660065        5.5279574532        5.8611612413 
 +  H         6.1300650069        5.3536860275        4.9350551393 
 +  C         8.0615723757        5.5446390494        5.8709776417 
 +  H         8.6121087088        5.3836499080        4.9523278088 
 +  C         8.7479484762        5.7683467907        7.0666774749 
 +  H         9.8306982448        5.7893185194        7.0736980974 
 +  H         6.4284687494        8.8405192314        5.8228620343 
 +  H         8.9057373702        8.8752743844        5.8387264727 
 +  C         6.9607954533        9.0255100555        6.7471184900 
 +  C         8.3568440363        9.0439429763        6.7567676586 
 +  C         6.2520232187        9.2329626290        7.9335080936 
 +  C         9.0454365692        9.2673522387        7.9522880615 
 +  H         5.1692478273        9.2131682112        7.9263504668 
 +  H        10.1283042427        9.2780238064        7.9594286415 
 +  C         6.9387312641        9.4553058890        9.1292884396 
 +  C         8.3367433685        9.4704785560        9.1388664843 
 +  H         6.3883177411        9.6163935534       10.0479844468 
 +  H         8.8703567259        9.6435467092       10.0651664724 
 +    &END 
 + 
 +    ! keep atoms away from box borders,  
 +    ! a requirement for the wavelet Poisson solver 
 +    &TOPOLOGY 
 +      &CENTER_COORDINATES      
 +      &END 
 +    &END TOPOLOGY 
 +  
 +    ! MP2 needs correlation consistent basis set 
 +    ! RI-MP2 needs an auxiliary basis set 
 +    ! We employ HF pseudo potentials 
 +    &KIND H 
 +      BASIS_SET         cc-TZV2P-GTH 
 +      RI_AUX_BASIS_SET  TZ_fiodena_opt 
 +      POTENTIAL         GTH-HF-q1       
 +    &END KIND 
 +    &KIND C 
 +      BASIS_SET         cc-TZV2P-GTH 
 +      RI_AUX_BASIS_SET  TZ_fiodena_opt 
 +      POTENTIAL         GTH-HF-q4 
 +    &END KIND 
 +  &END SUBSYS 
 +&END FORCE_EVAL 
 + 
 +! how to propagate the system, selection via RUN_TYPE in the &GLOBAL section 
 +&MOTION 
 + &GEO_OPT 
 +   OPTIMIZER BFGS ! Good choice for 'small' systems (use LBFGS for large systems) 
 +   MAX_ITER  100 
 +   MAX_DR    [bohr] 0.003 ! adjust target as needed 
 +   &BFGS 
 +   &END 
 + &END 
 +&END 
 +</code>
  
exercises/2015_pitt/mp2.1425555069.txt.gz · Last modified: 2020/08/21 10:15 (external edit)