CLUPAN v0.7 documentation

8. Optimization of clusters

«  7. Estimation of effective cluster interaction   ::   Contents   ::   9. Structure selection for DFT calculations  »

8. Optimization of clusters

8.1. Theoretical backgrounds

Optimization of clusters using the genetic algorithm

See [ga1] and [ga2].

[ga1]G L W Hart, V Blum, M J Walorski and A Zunger, Nat. Mater. 4, 391 (2005).
[ga2]V Blum, G L W Hart, M J Walorski and A Zunger, Phys. Rev. B 72, 165113 (2005).

8.2. gasa

  • Searching for the set of clusters that minimizes the CV score using the genetic algorithm.
  • An initial population can be generated using the simulated annealing.
  • Multi-thread calculations using OPENMP are available only for the simulated annealing.

Input files

Output files

  • gasa.out

8.3. wgasa

  • Searching for the set of clusters that minimizes the weighted CV score using the genetic algorithm.

Input files

Output files

  • wgasa.out

8.4. GASA.in

Example

The set of clusters is optimized using the genetic algorithm. The cluster set is composed of no.0-5 clusters and 5 clusters selected among no.6-52 clusters.

NLOOP       = 0         # only performing genetic algorithm
BASECLUSTER = 0 1 2 3 4 5
NPOP        = 25
NELITE      = 1
NALLCLUSTER = 53
MININDEX    = 6
MAXINDEX    = 52
NCLUSTER    = 5
MAXGENE     = 300
PMATING     = 0.9
PMUTATION   = 0.03

The initial population is generated by simulated annealing and the set of clusters is optimized using the genetic algorithm. The cluster set is composed of no.0-5 clusters and 5 clusters selected among no.6-52 clusters.

NLOOP       = 100
TEMPINIT    = 1
TEMPFINAL   = 0.001
TEMPMUL     = 0.9
BASECLUSTER = 0 1 2 3 4 5
NPOP        = 25
NELITE      = 1
NALLCLUSTER = 53
MININDEX    = 6
MAXINDEX    = 52
NCLUSTER    = 5
MAXGENE     = 300
PMATING     = 0.9
PMUTATION   = 0.03

8.4.1. NALLCLUSTER tag

Number of all clusters included in CORRELATION file.

  • Default : none
  • Example : NALLCLUSTER = 53

8.4.2. NCLUSTER tag

Number of clusters for selection.

  • Default : none
  • Example : NCLUSTER = 5

8.4.3. MININDEX tag, MAXINDEX tag

Minimum and maximum cluster index for the optimization.

  • Default : MININDEX = 0, MAXINDEX = NALLCLUSTER - 1
  • Example : MININDEX = 5
  • Example : MAXINDEX = 52

8.4.4. BASECLUSTER tag

Cluster indexes included for ECI evaluation.

  • Default : empty
  • Example : BASECLUSTER = 0 1 2 3 4 5

8.4.5. EXCLUSTER tag

Cluster indexes excluded for the optimization.

  • Default : empty
  • Example : EXCLUSTER = 15 23 34

8.4.6. NPOP tag

Number of cluster sets included in the population.

  • Default : 30
  • Example : NPOP = 25

8.4.7. NELITE tag

Number of elites (good cluster sets) kept for the next optimization step.

  • Default : 1
  • Example : NELITE = 5

8.4.8. PMATING tag

Probability of mating.

  • Default : 0.9
  • Example : PMATING = 0.95

8.4.9. PMUTATION tag

Probability of mutation.

  • Default : 0.03
  • Example : PMUTATION = 0.05

8.4.10. NLOOP tag

Number of steps for simulated annealing at one temperature. If you perform only the genetic algorithm, set NLOOP = 0.

  • Default : 0
  • Example : NLOOP = 100

8.4.11. TEMPINIT tag, TEMPFINAL tag

Initial and final temperatures in the simulated annealing.

  • Default : none
  • Example : TEMPINIT = 1
  • Example : TEMPFINAL = 0.001

8.4.12. TEMPMUL tag

Exponential base used in an automatic setting of temperatures in the simulated annealing. Set values ranged from 0 to 1.

  • Default : none
  • Example : TEMPMUL = 0.9

«  7. Estimation of effective cluster interaction   ::   Contents   ::   9. Structure selection for DFT calculations  »