IALGO - Revision history

Kresse: /* Direct optimization */ recommendation to use PREC= Accurate, if convergence is slow.

2025-02-27T20:43:08Z

Direct optimization: recommendation to use PREC= Accurate, if convergence is slow.

← Older revision		Revision as of 20:43, 27 February 2025
Line 81:		Line 81:
	:The final algorithm {{TAG\|IALGO}}=54 also uses a damped molecular dynamics algorithm but quenches the velocities to zero, if they are antiparallel to the present gradient (quick-min). It is usually not as efficient as {{TAG\|IALGO}}=53, but it is also less sensitive to the {{TAG\|TIME}} parameter.		:The final algorithm {{TAG\|IALGO}}=54 also uses a damped molecular dynamics algorithm but quenches the velocities to zero, if they are antiparallel to the present gradient (quick-min). It is usually not as efficient as {{TAG\|IALGO}}=53, but it is also less sensitive to the {{TAG\|TIME}} parameter.

	:{{NB\|mind\|It is very important to set an appropriate {{TAG\|TIME}} for ~~these~~ algorithms. Furthermore, it might be expedient to set {{TAG\|NELMDL}} to 1 or 2 for molecular dynamics simulations or relaxations in vasp.6. See the corresponding section in the documentation of {{TAG\|NELMDL}}. If the ions move by a very large distance during relaxations, even {{TAG\|NELMDL}}{{=}}3 can be ~~tried~~ (in particular for HF type Hamiltonians).}}		:{{NB\|mind\| Conjugate gradient algorithms require a largely noise-free energy surface for fast and consistent convergence. Some functionals, for instance non-local vdW-DFT functionals or meta-GGA functionals can result in a very noisy energy surface with small discontinuities in the energy. This can cause divergence or very slow convergence of the direct minimization. The issue is usually mitigated by setting {{TAG\| PREC }}{{=}} Accurate, or manually increasing the FFT grids for the densities ({{TAG\| NGXF}}, {{TAG\| NGYF}}, {{TAG\| NGZF}}). }}

			:{{NB\|mind\|It is very important to set an appropriate {{TAG\|TIME}} for some of the algorithms. Furthermore, it might be expedient to set {{TAG\|NELMDL}} to 1 or 2 for molecular dynamics simulations or relaxations in vasp.6. See the corresponding section in the documentation of {{TAG\|NELMDL}}. If the ions move by a very large distance during relaxations, even {{TAG\|NELMDL}}{{=}}3 can be expedient (in particular for HF type Hamiltonians).}}

	== Miscellaneous ==		== Miscellaneous ==

Ftran at 12:06, 18 October 2023

2023-10-18T12:06:38Z

← Older revision		Revision as of 12:06, 18 October 2023
Line 81:		Line 81:
	:The final algorithm {{TAG\|IALGO}}=54 also uses a damped molecular dynamics algorithm but quenches the velocities to zero, if they are antiparallel to the present gradient (quick-min). It is usually not as efficient as {{TAG\|IALGO}}=53, but it is also less sensitive to the {{TAG\|TIME}} parameter.		:The final algorithm {{TAG\|IALGO}}=54 also uses a damped molecular dynamics algorithm but quenches the velocities to zero, if they are antiparallel to the present gradient (quick-min). It is usually not as efficient as {{TAG\|IALGO}}=53, but it is also less sensitive to the {{TAG\|TIME}} parameter.

	:{{NB\|mind\|It is very important to set an appropriate {{TAG\|TIME}} for these algorithms. Furthermore, it might be expedient to set {{TAG\|NELMDL}} to 1 or 2 for molecular dynamics simulations or relaxations in vasp.6. See the corresponding section in the documentation of {{TAG\|NELMDL}}. If the ions move by a very large distance during relaxations, even {{TAG\|NELMDL}}=3 can be tried (in particular for HF type Hamiltonians).}}		:{{NB\|mind\|It is very important to set an appropriate {{TAG\|TIME}} for these algorithms. Furthermore, it might be expedient to set {{TAG\|NELMDL}} to 1 or 2 for molecular dynamics simulations or relaxations in vasp.6. See the corresponding section in the documentation of {{TAG\|NELMDL}}. If the ions move by a very large distance during relaxations, even {{TAG\|NELMDL}}{{=}}3 can be tried (in particular for HF type Hamiltonians).}}

	== Miscellaneous ==		== Miscellaneous ==

Ftran at 12:06, 18 October 2023

2023-10-18T12:06:15Z

← Older revision		Revision as of 12:06, 18 October 2023
Line 81:		Line 81:
	:The final algorithm {{TAG\|IALGO}}=54 also uses a damped molecular dynamics algorithm but quenches the velocities to zero, if they are antiparallel to the present gradient (quick-min). It is usually not as efficient as {{TAG\|IALGO}}=53, but it is also less sensitive to the {{TAG\|TIME}} parameter.		:The final algorithm {{TAG\|IALGO}}=54 also uses a damped molecular dynamics algorithm but quenches the velocities to zero, if they are antiparallel to the present gradient (quick-min). It is usually not as efficient as {{TAG\|IALGO}}=53, but it is also less sensitive to the {{TAG\|TIME}} parameter.

	:~~'''Mind''': it~~ is very important to set an appropriate {{TAG\|TIME}} for these algorithms. Furthermore, it might be expedient to set {{TAG\|NELMDL}} to 1 or 2 for molecular dynamics simulations or relaxations in vasp.6. See the corresponding section in the documentation of {{TAG\|NELMDL}}. If the ions move by a very large distance during relaxations, even {{TAG\|NELMDL}}=3 can be tried (in particular for HF type Hamiltonians).		:{{NB\|mind\|It is very important to set an appropriate {{TAG\|TIME}} for these algorithms. Furthermore, it might be expedient to set {{TAG\|NELMDL}} to 1 or 2 for molecular dynamics simulations or relaxations in vasp.6. See the corresponding section in the documentation of {{TAG\|NELMDL}}. If the ions move by a very large distance during relaxations, even {{TAG\|NELMDL}}=3 can be tried (in particular for HF type Hamiltonians).}}

	== Miscellaneous ==		== Miscellaneous ==

Huebsch at 10:25, 11 May 2023

2023-05-11T10:25:19Z

Miranda.henrique: /* Direct optimization */

2022-07-27T07:56:34Z

Direct optimization

← Older revision		Revision as of 07:56, 27 July 2022
Line 82:		Line 82:
	:The final algorithm {{TAG\|IALGO}}=54 also uses a damped molecular dynamics algorithm but quenches the velocities to zero, if they are antiparallel to the present gradient (quick-min). It is usually not as efficient as {{TAG\|IALGO}}=53, but it is also less sensitive to the {{TAG\|TIME}} parameter.		:The final algorithm {{TAG\|IALGO}}=54 also uses a damped molecular dynamics algorithm but quenches the velocities to zero, if they are antiparallel to the present gradient (quick-min). It is usually not as efficient as {{TAG\|IALGO}}=53, but it is also less sensitive to the {{TAG\|TIME}} parameter.

	:'''Mind''': it is very important to set an appropriate {{TAG\|TIME}} for these algorithms. Furthermore, it might be expedient to set {{TAG\|~~NELDL~~}} to 1 or 2 for molecular dynamics simulations or relaxations in vasp.6. See the corresponding section in the documentation of {{TAG\|~~NELDL~~}}. If the ions move by a very large distance during relaxations, even {{TAG\|NELMDL}}=3 can be tried (in particular for HF type Hamiltonians).		:'''Mind''': it is very important to set an appropriate {{TAG\|TIME}} for these algorithms. Furthermore, it might be expedient to set {{TAG\|NELMDL}} to 1 or 2 for molecular dynamics simulations or relaxations in vasp.6. See the corresponding section in the documentation of {{TAG\|NELMDL}}. If the ions move by a very large distance during relaxations, even {{TAG\|NELMDL}}=3 can be tried (in particular for HF type Hamiltonians).

	== Miscellaneous ==		== Miscellaneous ==

Ftran at 13:19, 19 July 2022

2022-07-19T13:19:05Z

← Older revision		Revision as of 13:19, 19 July 2022
Line 61:		Line 61:
	*{{TAG\|IALGO}}=53-58: Treat total free energy as variational quantity and minimize the functional completely selfconsistently{{cite\|stich:prb:89}}{{cite\|gillan:jpc:89}}{{cite\|arias:prl:92}}.		*{{TAG\|IALGO}}=53-58: Treat total free energy as variational quantity and minimize the functional completely selfconsistently{{cite\|stich:prb:89}}{{cite\|gillan:jpc:89}}{{cite\|arias:prl:92}}.

	:These algorithms have been carefully optimized and should be selected for Hartree-Fock type as well as ~~META~~-~~gradient~~ functionals. The present version is rather stable and robust even for metallic systems.		:These algorithms have been carefully optimized and should be selected for Hartree-Fock type as well as meta-GGA functionals. The present version is rather stable and robust even for metallic systems.

	:Important sub-switches:		:Important sub-switches:

Karsai: /* References */

2022-04-08T14:20:43Z

References

← Older revision		Revision as of 14:20, 8 April 2022
Line 112:		Line 112:
	<references/>		<references/>
	----		----
	~~[[The_VASP_Manual\|Contents]]~~

	[[Category:INCAR]][[Category:Electronic minimization]]		[[Category:INCAR tag]][[Category:Electronic minimization]]

Karsai: /* Related Tags and Sections */

2022-04-08T14:20:33Z

Related Tags and Sections

← Older revision		Revision as of 14:20, 8 April 2022
Line 100:		Line 100:
	*{{TAG\|IALGO}}=90: Exact Diagonalization. This flag selects an exact diagonalization of the one-electron Hamiltonian. This requires a fairly large amount of memory and should be selected with caution. Specifically, we recommend selecting this algorithm to prepare the {{FILE\|WAVECAR}} for RPA or GW calculations, where many unoccupied orbitals are calculated (more than 30-50 % of the states spanned by the full plane-wave basis). To speed up the calculations, we recommend performing a routine ground-state calculation before calculating the unoccupied states.		*{{TAG\|IALGO}}=90: Exact Diagonalization. This flag selects an exact diagonalization of the one-electron Hamiltonian. This requires a fairly large amount of memory and should be selected with caution. Specifically, we recommend selecting this algorithm to prepare the {{FILE\|WAVECAR}} for RPA or GW calculations, where many unoccupied orbitals are calculated (more than 30-50 % of the states spanned by the full plane-wave basis). To speed up the calculations, we recommend performing a routine ground-state calculation before calculating the unoccupied states.

	== Related ~~Tags~~ and ~~Sections~~ ==		== Related tags and articles ==
	{{TAG\|ALGO}},		{{TAG\|ALGO}},
	{{TAG\|LSUBROT}},		{{TAG\|LSUBROT}},

Huebsch at 11:14, 6 April 2022

2022-04-06T11:14:07Z

Kresse at 11:19, 10 November 2021

2021-11-10T11:19:06Z

← Older revision		Revision as of 11:19, 10 November 2021
Line 53:		Line 53:
	:The final option is somewhat complicated and requires an understanding of how the initialization algorithm of the RMM-DIIS algorithm works: after the random initialization of the orbitals, the initial orbitals for the RMM-DIIS algorithm are determined during a non selfconsistent steepest descent phase (the number of steepest descent sweeps is given by {{TAG\|NELMDL}}, default is {{TAG\|NELMDL}}=-12 for RMM-DIIS). During this initial phase in each sweep, one steepest descent step per orbital is performed between each sub space rotation. This "automatic" simple steepest descent approach during the delay is faced with a rather ill-conditioned minimization problem and can fail to produce reasonable trial orbitals for the RMM-DIIS algorithm. In this case the quantity in the column "rms" will not decrease during the initial phase (12 steps), and you must improve the conditioning of the problem by setting the {{TAG\|ENINI}} parameter in the {{FILE\|INCAR}} file. {{TAG\|ENINI}} controls the cutoff during the initial (steepest descent) phase for {{TAG\|IALGO}}=48. Default for {{TAG\|ENINI}} is {{TAG\|ENINI}}= {{TAG\|ENCUT}}. If convergence problems are observed, start with a slightly smaller {{TAG\|ENINI}}; reduce {{TAG\|ENINI}} in steps of 20%, till the norm of the residual vector (column "rms") decreases continuously during the first 12 steps.		:The final option is somewhat complicated and requires an understanding of how the initialization algorithm of the RMM-DIIS algorithm works: after the random initialization of the orbitals, the initial orbitals for the RMM-DIIS algorithm are determined during a non selfconsistent steepest descent phase (the number of steepest descent sweeps is given by {{TAG\|NELMDL}}, default is {{TAG\|NELMDL}}=-12 for RMM-DIIS). During this initial phase in each sweep, one steepest descent step per orbital is performed between each sub space rotation. This "automatic" simple steepest descent approach during the delay is faced with a rather ill-conditioned minimization problem and can fail to produce reasonable trial orbitals for the RMM-DIIS algorithm. In this case the quantity in the column "rms" will not decrease during the initial phase (12 steps), and you must improve the conditioning of the problem by setting the {{TAG\|ENINI}} parameter in the {{FILE\|INCAR}} file. {{TAG\|ENINI}} controls the cutoff during the initial (steepest descent) phase for {{TAG\|IALGO}}=48. Default for {{TAG\|ENINI}} is {{TAG\|ENINI}}= {{TAG\|ENCUT}}. If convergence problems are observed, start with a slightly smaller {{TAG\|ENINI}}; reduce {{TAG\|ENINI}} in steps of 20%, till the norm of the residual vector (column "rms") decreases continuously during the first 12 steps.

	:The algorithm can be combined with {{TAG\|LDIAG}}=.FALSE. to conserve the initial orbital order (when orbitals are read from the ({{TAG\|WAVECAR}} file).		:The algorithm can be combined with {{TAG\|LDIAG}}=.FALSE. to conserve the initial orbital order (when orbitals are read from the {{TAG\|WAVECAR}} file).

	:A final note concerns the mixing: {{TAG\|IALGO}}=48 dislikes too abrupt mixing. Since the RMM-DIIS algorithm always stays in the space spanned by the initial orbitals, and too strong mixing (large {{TAG\|AMIX}}, small {{TAG\|BMIX}}) might require discontinuous changes of the orbitals, the initial mixing must not be too sizable for {{TAG\|IALGO}}=48. Try to reduce {{TAG\|AMIX}} and increase {{TAG\|BMIX}} if you suspect such a situation. Increasing {{TAG\|NBANDS}} also helps in this situation.		:A final note concerns the mixing: {{TAG\|IALGO}}=48 dislikes too abrupt mixing. Since the RMM-DIIS algorithm always stays in the space spanned by the initial orbitals, and too strong mixing (large {{TAG\|AMIX}}, small {{TAG\|BMIX}}) might require discontinuous changes of the orbitals, the initial mixing must not be too sizable for {{TAG\|IALGO}}=48. Try to reduce {{TAG\|AMIX}} and increase {{TAG\|BMIX}} if you suspect such a situation. Increasing {{TAG\|NBANDS}} also helps in this situation.
Line 78:		Line 78:

	:For the damped MD algorithm ({{TAG\|IALGO}}=53, {{TAG\|ALGO}}=Damped), a sensible {{TAG\|TIME}} step is even more important. In this case {{TAG\|TIME}} is not automatically adjusted, and the user is entirely responsible to chose an appropriate value. Too small time-steps slow the convergence significantly, whereas too large values will always lead to divergence. It is sensible to optimize this value, in particular, if many different configurations are considered for a particular system. It is recommended to start with a small step size {{TAG\|TIME}}, and to increase {{TAG\|TIME}} by a factor 1.2 until the calculations diverge. The largest stable step {{TAG\|TIME}} should then be used for all calculations.		:For the damped MD algorithm ({{TAG\|IALGO}}=53, {{TAG\|ALGO}}=Damped), a sensible {{TAG\|TIME}} step is even more important. In this case {{TAG\|TIME}} is not automatically adjusted, and the user is entirely responsible to chose an appropriate value. Too small time-steps slow the convergence significantly, whereas too large values will always lead to divergence. It is sensible to optimize this value, in particular, if many different configurations are considered for a particular system. It is recommended to start with a small step size {{TAG\|TIME}}, and to increase {{TAG\|TIME}} by a factor 1.2 until the calculations diverge. The largest stable step {{TAG\|TIME}} should then be used for all calculations.
	The damped MD algorithm can be combined with {{TAG\|LDIAG}}=.FALSE. to conserve the initial orbital order (when orbitals are read from the ({{TAG\|WAVECAR}} file).		The damped MD algorithm can be combined with {{TAG\|LDIAG}}=.FALSE. to conserve the initial orbital order (when orbitals are read from the {{TAG\|WAVECAR}} file).

	:The final algorithm {{TAG\|IALGO}}=54 also uses a damped molecular dynamics algorithm but quenches the velocities to zero, if they are antiparallel to the present gradient (quick-min). It is usually not as efficient as {{TAG\|IALGO}}=53, but it is also less sensitive to the {{TAG\|TIME}} parameter.		:The final algorithm {{TAG\|IALGO}}=54 also uses a damped molecular dynamics algorithm but quenches the velocities to zero, if they are antiparallel to the present gradient (quick-min). It is usually not as efficient as {{TAG\|IALGO}}=53, but it is also less sensitive to the {{TAG\|TIME}} parameter.

← Older revision		Revision as of 11:19, 10 November 2021
Line 53:		Line 53:
	:The final option is somewhat complicated and requires an understanding of how the initialization algorithm of the RMM-DIIS algorithm works: after the random initialization of the orbitals, the initial orbitals for the RMM-DIIS algorithm are determined during a non selfconsistent steepest descent phase (the number of steepest descent sweeps is given by {{TAG\|NELMDL}}, default is {{TAG\|NELMDL}}=-12 for RMM-DIIS). During this initial phase in each sweep, one steepest descent step per orbital is performed between each sub space rotation. This "automatic" simple steepest descent approach during the delay is faced with a rather ill-conditioned minimization problem and can fail to produce reasonable trial orbitals for the RMM-DIIS algorithm. In this case the quantity in the column "rms" will not decrease during the initial phase (12 steps), and you must improve the conditioning of the problem by setting the {{TAG\|ENINI}} parameter in the {{FILE\|INCAR}} file. {{TAG\|ENINI}} controls the cutoff during the initial (steepest descent) phase for {{TAG\|IALGO}}=48. Default for {{TAG\|ENINI}} is {{TAG\|ENINI}}= {{TAG\|ENCUT}}. If convergence problems are observed, start with a slightly smaller {{TAG\|ENINI}}; reduce {{TAG\|ENINI}} in steps of 20%, till the norm of the residual vector (column "rms") decreases continuously during the first 12 steps.		:The final option is somewhat complicated and requires an understanding of how the initialization algorithm of the RMM-DIIS algorithm works: after the random initialization of the orbitals, the initial orbitals for the RMM-DIIS algorithm are determined during a non selfconsistent steepest descent phase (the number of steepest descent sweeps is given by {{TAG\|NELMDL}}, default is {{TAG\|NELMDL}}=-12 for RMM-DIIS). During this initial phase in each sweep, one steepest descent step per orbital is performed between each sub space rotation. This "automatic" simple steepest descent approach during the delay is faced with a rather ill-conditioned minimization problem and can fail to produce reasonable trial orbitals for the RMM-DIIS algorithm. In this case the quantity in the column "rms" will not decrease during the initial phase (12 steps), and you must improve the conditioning of the problem by setting the {{TAG\|ENINI}} parameter in the {{FILE\|INCAR}} file. {{TAG\|ENINI}} controls the cutoff during the initial (steepest descent) phase for {{TAG\|IALGO}}=48. Default for {{TAG\|ENINI}} is {{TAG\|ENINI}}= {{TAG\|ENCUT}}. If convergence problems are observed, start with a slightly smaller {{TAG\|ENINI}}; reduce {{TAG\|ENINI}} in steps of 20%, till the norm of the residual vector (column "rms") decreases continuously during the first 12 steps.

	:The algorithm can be combined with {{TAG\|LDIAG}}=.FALSE. to conserve the initial orbital order (when orbitals are read from the ({{TAG\|WAVECAR}} file).		:The algorithm can be combined with {{TAG\|LDIAG}}=.FALSE. to conserve the initial orbital order (when orbitals are read from the {{TAG\|WAVECAR}} file).

	:A final note concerns the mixing: {{TAG\|IALGO}}=48 dislikes too abrupt mixing. Since the RMM-DIIS algorithm always stays in the space spanned by the initial orbitals, and too strong mixing (large {{TAG\|AMIX}}, small {{TAG\|BMIX}}) might require discontinuous changes of the orbitals, the initial mixing must not be too sizable for {{TAG\|IALGO}}=48. Try to reduce {{TAG\|AMIX}} and increase {{TAG\|BMIX}} if you suspect such a situation. Increasing {{TAG\|NBANDS}} also helps in this situation.		:A final note concerns the mixing: {{TAG\|IALGO}}=48 dislikes too abrupt mixing. Since the RMM-DIIS algorithm always stays in the space spanned by the initial orbitals, and too strong mixing (large {{TAG\|AMIX}}, small {{TAG\|BMIX}}) might require discontinuous changes of the orbitals, the initial mixing must not be too sizable for {{TAG\|IALGO}}=48. Try to reduce {{TAG\|AMIX}} and increase {{TAG\|BMIX}} if you suspect such a situation. Increasing {{TAG\|NBANDS}} also helps in this situation.
Line 78:		Line 78:

	:For the damped MD algorithm ({{TAG\|IALGO}}=53, {{TAG\|ALGO}}=Damped), a sensible {{TAG\|TIME}} step is even more important. In this case {{TAG\|TIME}} is not automatically adjusted, and the user is entirely responsible to chose an appropriate value. Too small time-steps slow the convergence significantly, whereas too large values will always lead to divergence. It is sensible to optimize this value, in particular, if many different configurations are considered for a particular system. It is recommended to start with a small step size {{TAG\|TIME}}, and to increase {{TAG\|TIME}} by a factor 1.2 until the calculations diverge. The largest stable step {{TAG\|TIME}} should then be used for all calculations.		:For the damped MD algorithm ({{TAG\|IALGO}}=53, {{TAG\|ALGO}}=Damped), a sensible {{TAG\|TIME}} step is even more important. In this case {{TAG\|TIME}} is not automatically adjusted, and the user is entirely responsible to chose an appropriate value. Too small time-steps slow the convergence significantly, whereas too large values will always lead to divergence. It is sensible to optimize this value, in particular, if many different configurations are considered for a particular system. It is recommended to start with a small step size {{TAG\|TIME}}, and to increase {{TAG\|TIME}} by a factor 1.2 until the calculations diverge. The largest stable step {{TAG\|TIME}} should then be used for all calculations.
	The damped MD algorithm can be combined with {{TAG\|LDIAG}}=.FALSE. to conserve the initial orbital order (when orbitals are read from the ({{TAG\|WAVECAR}} file).		The damped MD algorithm can be combined with {{TAG\|LDIAG}}=.FALSE. to conserve the initial orbital order (when orbitals are read from the {{TAG\|WAVECAR}} file).

	:The final algorithm {{TAG\|IALGO}}=54 also uses a damped molecular dynamics algorithm but quenches the velocities to zero, if they are antiparallel to the present gradient (quick-min). It is usually not as efficient as {{TAG\|IALGO}}=53, but it is also less sensitive to the {{TAG\|TIME}} parameter.		:The final algorithm {{TAG\|IALGO}}=54 also uses a damped molecular dynamics algorithm but quenches the velocities to zero, if they are antiparallel to the present gradient (quick-min). It is usually not as efficient as {{TAG\|IALGO}}=53, but it is also less sensitive to the {{TAG\|TIME}} parameter.

← Older revision		Revision as of 10:25, 11 May 2023
Line 2:		Line 2:
	{{DEF\|IALGO\|8\|for VASP.4.4 and older\|38\|else (if {{TAG\|ALGO}} is not set)}}		{{DEF\|IALGO\|8\|for VASP.4.4 and older\|38\|else (if {{TAG\|ALGO}} is not set)}}

	Description: {{TAG\|IALGO}} selects the algorithm ~~used~~ to optimize the orbitals.		Description: {{TAG\|IALGO}} selects the algorithm to optimize the orbitals.

	~~'''Mind''': We strongly urge the users to select the algorithms via {{TAG\|ALGO}}. Algorithms other than those available via {{TAG\|ALGO}} are subject to instabilities~~.
	----		----
			{{NB\|warning\|We suggest selecting the algorithms via {{TAG\|ALGO}} instead of {{TAG\|IALGO}}. Algorithms other than those available via {{TAG\|ALGO}} are subject to instabilities.}}
	== Conjugate gradient algorithms ==		== Conjugate gradient algorithms ==
	The band-by-band conjugate gradient algorithms are no longer maintained or supported.		The band-by-band conjugate gradient algorithms are no longer maintained or supported.

	*{{TAG\|IALGO}}=5-8: Conjugate gradient algorithms ~~(not~~ supported since vasp.5)		*{{TAG\|IALGO}}=5-8: Conjugate gradient algorithms {{NB\|deprecated\|Not supported since vasp.5.}}
	:Optimize each band iteratively using a conjugate gradient algorithm. Subspace-diagonalization before conjugate gradient algorithm. The conjugate gradient algorithm is used to optimize the eigenvalue of each band.		:Optimize each band iteratively using a conjugate gradient algorithm. Subspace-diagonalization before conjugate gradient algorithm. The conjugate gradient algorithm is used to optimize the eigenvalue of each band.
	:*{{TAG\|IALGO}}=5 steepest descent		:*{{TAG\|IALGO}}=5 steepest descent
Line 16:		Line 15:
	:*{{TAG\|IALGO}}=8 preconditioned conjugated gradient		:*{{TAG\|IALGO}}=8 preconditioned conjugated gradient
	:{{TAG\|IALGO}}=8 is always fastest, whereas {{TAG\|IALGO}}=5-7 are only implemented for test purposes.		:{{TAG\|IALGO}}=8 is always fastest, whereas {{TAG\|IALGO}}=5-7 are only implemented for test purposes.
	:Please mind, that {{TAG\|IALGO}}=8 is not supported as of VASP.4.5, since M. Teter, Corning and M. Payne hold a patent on this algorithm. The algorithms have been replaced by faster and more efficient Davidson like algorithms.		:Please mind, that {{TAG\|IALGO}}=8 is not supported as of VASP.4.5, since M. Teter, Corning and M. Payne hold a patent on this algorithm. The algorithms have been replaced by faster and more efficient Davidson-like algorithms.

	*{{TAG\|IALGO}}=15-18: Conjugate gradient algorithms ~~(not~~ supported since vasp.5)		*{{TAG\|IALGO}}=15-18: Conjugate gradient algorithms {{NB\|deprecated\|Not supported since vasp.5.}}
	:Subspace-diagonalization after iterative refinement of the eigenvectors using the conjugate gradient algorithm. These switches are retained for compatibility reasons only and should not be used any longer. Generally {{TAG\|IALGO}}=5-8 is preferable. Sub-switches as above.		:Subspace-diagonalization after iterative refinement of the eigenvectors using the conjugate gradient algorithm. These switches are retained for compatibility reasons only and should not be used any longer. Generally {{TAG\|IALGO}}=5-8 is preferable. Sub-switches as above.

	*{{TAG\|IALGO}}=28: Conjugate gradient algorithm ~~(not~~ supported since vasp.5)		*{{TAG\|IALGO}}=28: Conjugate gradient algorithm {{NB\|deprecated\|Not supported since vasp.5.}}
	:Subspace-diagonalization before conjugate gradient algorithm. No explicit orthonormalization of the gradients to the trial orbitals is done. This setting saves time, but does fail in most cases (mainly included for test purposes). Try {{TAG\|IALGO}}=4X (RMM-DIIS) instead.		:Subspace-diagonalization before conjugate gradient algorithm. No explicit orthonormalization of the gradients to the trial orbitals is done. This setting saves time, but does fail in most cases (mainly included for test purposes). Try {{TAG\|IALGO}}=4X (RMM-DIIS) instead.

← Older revision		Revision as of 11:14, 6 April 2022
Line 71:		Line 71:
	:Furthermore, {{TAG\|LSUBROT}} determines whether the subspace rotation matrix (rotation matrix in the space spanned by the occupied and unoccupied orbitals) is optimized. The current default is {{TAG\|LSUBROT}}=.FALSE. This allows for efficient groundstate calculations for insulators. When hybrid functionals are used, {{TAG\|LSUBROT}}=.TRUE. can be tried for small gap semiconductors and metals. This algorithm performs standard SCF steps during the direct optimization steps in order to determine an optimal rotation matrix between occupied and unoccupied orbitals. For hybrid functionals, {{TAG\|LSUBROT}}=.TRUE. is generally faster, however, in rare cases, it can lead to instabilities{{cite\|kresse:prb:96}}.		:Furthermore, {{TAG\|LSUBROT}} determines whether the subspace rotation matrix (rotation matrix in the space spanned by the occupied and unoccupied orbitals) is optimized. The current default is {{TAG\|LSUBROT}}=.FALSE. This allows for efficient groundstate calculations for insulators. When hybrid functionals are used, {{TAG\|LSUBROT}}=.TRUE. can be tried for small gap semiconductors and metals. This algorithm performs standard SCF steps during the direct optimization steps in order to determine an optimal rotation matrix between occupied and unoccupied orbitals. For hybrid functionals, {{TAG\|LSUBROT}}=.TRUE. is generally faster, however, in rare cases, it can lead to instabilities{{cite\|kresse:prb:96}}.

	:The preconditioned conjugate gradient ({{TAG\|IALGO}}= 58, {{TAG\|ALGO}}= A) algorithm is recommended for insulators. The best stability is usually obtained if the number of bands equals half the number of electrons (non spin polarized case). In this case, the algorithm is fairly robust and ~~fool proof~~ and might even outperform the mixing algorithm.		:The preconditioned conjugate gradient ({{TAG\|IALGO}}= 58, {{TAG\|ALGO}}= A) algorithm is recommended for insulators. The best stability is usually obtained if the number of bands equals half the number of electrons (non-spin-polarized case). In this case, the algorithm is fairly robust and foolproof and might even outperform the mixing algorithm.

	:For small gap systems and for metals, it is however usually required (metals) or desirable (semiconductors) to use a larger value for {{TAG\|NBANDS}}. In this case, we recommend ~~to use~~ the damped MD algorithm ({{TAG\|IALGO}}=53, {{TAG\|ALGO}}=Damped) instead of the conjugate gradient.		:For small gap systems and for metals, it is however usually required (metals) or desirable (semiconductors) to use a larger value for {{TAG\|NBANDS}}. In this case, we recommend using the damped MD algorithm ({{TAG\|IALGO}}=53, {{TAG\|ALGO}}=Damped) instead of the conjugate gradient.

	:The stability of the all bands simultaneously algorithms depends strongly on the setting of {{TAG\|TIME}}. For the conjugate gradient case, {{TAG\|TIME}} controls the step size in the trial step, which is required in order to perform a line minimization of the energy along the gradient (or conjugated gradient). Too small steps make the line minimization less accurate, whereas too large steps can cause instabilities. The step size is usually automatically scaled by the actual step size minimizing the total energy along the gradient (values can range from 1.0 for insulators to 0.01 for metals with a large density of states at the Fermi-level).		:The stability of the all bands simultaneously algorithms depends strongly on the setting of {{TAG\|TIME}}. For the conjugate gradient case, {{TAG\|TIME}} controls the step size in the trial step, which is required in order to perform a line minimization of the energy along the gradient (or conjugated gradient). Too small steps make the line minimization less accurate, whereas too large steps can cause instabilities. The step size is usually automatically scaled by the actual step size minimizing the total energy along the gradient (values can range from 1.0 for insulators to 0.01 for metals with a large density of states at the Fermi-level).

	:For the damped MD algorithm ({{TAG\|IALGO}}=53, {{TAG\|ALGO}}=Damped), a sensible {{TAG\|TIME}} step is even more important. In this case {{TAG\|TIME}} is not automatically adjusted, and the user is entirely responsible to ~~chose~~ an appropriate value. Too small time-steps slow the convergence significantly, whereas too large values will always lead to divergence. It is sensible to optimize this value, in particular, if many different configurations are considered for a particular system. It is recommended to start with a small step size {{TAG\|TIME}}, and to increase {{TAG\|TIME}} by a factor 1.2 until the calculations diverge. The largest stable step {{TAG\|TIME}} should then be used for all calculations.		:For the damped MD algorithm ({{TAG\|IALGO}}=53, {{TAG\|ALGO}}=Damped), a sensible {{TAG\|TIME}} step is even more important. In this case {{TAG\|TIME}} is not automatically adjusted, and the user is entirely responsible to choose an appropriate value. Too small time steps slow the convergence significantly, whereas too large values will always lead to divergence. It is sensible to optimize this value, in particular, if many different configurations are considered for a particular system. It is recommended to start with a small step size {{TAG\|TIME}}, and to increase {{TAG\|TIME}} by a factor 1.2 until the calculations diverge. The largest stable step {{TAG\|TIME}} should then be used for all calculations.
	The damped MD algorithm can be combined with {{TAG\|LDIAG}}=.FALSE. to conserve the initial orbital order (when orbitals are read from the {{TAG\|WAVECAR}} file).		The damped MD algorithm can be combined with {{TAG\|LDIAG}}=.FALSE. to conserve the initial orbital order (when orbitals are read from the {{TAG\|WAVECAR}} file).

Line 90:		Line 90:

	*{{TAG\|IALGO}}=2: Orbitals and one-electron energies are kept fixed.		*{{TAG\|IALGO}}=2: Orbitals and one-electron energies are kept fixed.
	:One electron occupancies and electronic density of states (DOS) are, however, recalculated. This option is only useful if a pre-converged {{FILE\|WAVECAR}} file is read. The option allows ~~to run~~ selected post-processing tasks, such as local DOS, or the interface code to Wannier90.		:One electron occupancies and electronic density of states (DOS) are, however, recalculated. This option is only useful if a pre-converged {{FILE\|WAVECAR}} file is read. The option allows running selected post-processing tasks, such as local DOS, or the interface code to Wannier90.

	*{{TAG\|IALGO}}=3: Orbitals are kept fixed.		*{{TAG\|IALGO}}=3: Orbitals are kept fixed.
	:One-electron energies, band structure energies, and the electronic density of states (DOS) are, as well as, the total energy are recalculated for the present Hamiltonian (the one electron occupancies are kept fixed, however). This option is only useful if a pre-converged {{FILE\|WAVECAR}} file is read. The option also allows ~~to run~~ selected post-processing tasks, such as local DOS, or the interface code to Wannier90.		:One-electron energies, band structure energies, and the electronic density of states (DOS) are, as well as, the total energy are recalculated for the present Hamiltonian (the one-electron occupancies are kept fixed, however). This option is only useful if a pre-converged {{FILE\|WAVECAR}} file is read. The option also allows running selected post-processing tasks, such as local DOS, or the interface code to Wannier90.

	*{{TAG\|IALGO}}=4: Orbitals are updated by applying a sub-space rotation.		*{{TAG\|IALGO}}=4: Orbitals are updated by applying a sub-space rotation.
	:The Hamiltonian is evaluated in the space spanned by the orbitals (read from {{FILE\|WAVECAR}}), and one diagonalization in this space is performed. No optimization outside the subspace spanned by the orbitals is performed. Note: if {{TAG\|NBANDS}} is larger or equal to the total number of plane waves, the resulting one-electron orbitals are exact.		:The Hamiltonian is evaluated in the space spanned by the orbitals (read from {{FILE\|WAVECAR}}), and one diagonalization in this space is performed. No optimization outside the subspace spanned by the orbitals is performed. Note: if {{TAG\|NBANDS}} is larger or equal to the total number of plane waves, the resulting one-electron orbitals are exact.

	*{{TAG\|IALGO}}=90: Exact Diagonalization. This flag selects an exact diagonalization of the one-electron Hamiltonian. This requires a fairly large amount of memory, and should be selected with caution. Specifically, we recommend ~~to select~~ this algorithm to prepare the {{FILE\|WAVECAR}} for RPA or GW calculations, where many unoccupied orbitals are calculated (more than 30-50 % of the states spanned by the full plane wave basis). To speed up the calculations, we recommend ~~to perform~~ a routine ~~groundstate~~ calculation before calculating the unoccupied states.		*{{TAG\|IALGO}}=90: Exact Diagonalization. This flag selects an exact diagonalization of the one-electron Hamiltonian. This requires a fairly large amount of memory and should be selected with caution. Specifically, we recommend selecting this algorithm to prepare the {{FILE\|WAVECAR}} for RPA or GW calculations, where many unoccupied orbitals are calculated (more than 30-50 % of the states spanned by the full plane-wave basis). To speed up the calculations, we recommend performing a routine ground-state calculation before calculating the unoccupied states.

	== Related Tags and Sections ==		== Related Tags and Sections ==
Line 114:		Line 114:
	[[The_VASP_Manual\|Contents]]		[[The_VASP_Manual\|Contents]]

	[[Category:INCAR]][[Category:Electronic ~~Minimization]][[Category:Electronic Minimization Methods~~]]		[[Category:INCAR]][[Category:Electronic minimization]]