SourceForge.net Logo Opium - pseudopotential generation project

Home

Installation

User's guide

Tutorial

FAQ

Scientific references

Contact


external links:

Download

Mailing list

------------

 

Bennett & Rappe Pseudopotential Library

LDA

GGA

Latest release

Version 3.8 - April 4, 2014

View the 3.7 -> 3.8 Changelog

View the FULL Changelog

Get the Emacs highlighting mode here


Using Opium


Non-interactive execution
Opium is executed in the following way:

opium paramfile logfile command_1 [command_2] [command_3] . . .

Where

  • paramfile is the opium input file
  • logfile is the opium output file
  • command_1 is the first command
  • command_2 is the next command
  • ... etc.

Here is a typical command line:

opium c.param c.log ae ps nl plot vi tc rpt

As of the 1.0.1 release, .param is implied.

Therefore,

opium c c.log ae ps nl plot vi tc rpt

is identical to the previous command line.

Back to top


Interactive execution

Note: Interactive execution is not tested as thoroughly as non-interactive use and should be avoided

Opium can also be used in an interactive mode by not specifying any commands. For example:

opium c c.log

yields an Opium prompt:

opium[param=c|log=log|verb=0]>>

At this prompt one can run one or more commands without having to repeatedly specify the param and log file:

opium[param=c|log=log|verb=0]>> ae ps nl
opium[param=c|log=log|verb=0]>> plot vi
opium[param=c|log=log|verb=0]>> tc
opium[param=c|log=log|verb=0]>> rpt

This gives the same results as the non-interactive command line listed above.

Back to top


Commands

Command line help can be obtained from the opium code by: "opium -c"

Command Description
atomic solve and pseudopotential construction
ae all electron solve of the atom
ps generates pseudopotential
nl pseudopotential solve of the atom for reference state
tc solves the all-elec and pseudo atom for test configurations
all abbreviation for "ae ps nl tc"
pseudoptential output style
pwf generate *.pwf output (for BH)
fhi generate *.fhi output (for ABINIT)
ncppgenerate *.ncpp output (for PWSCF)
recpotgenerate *.recpot output (for CASTEP)
miscellaneous options
rpt generate report file
plot [plot_type]make a plot of type [plot_type]
vtoggle verbosity flag (interactive mode only)

Back to top


Keyblocks

The Opium executable takes one parameter file with a relatively flexible format. Opium relies on FlexiLib for scanning through the supplied parameter file. For a more general discussion about this library take a look at the FlexiLib development page.

For running Opium it is sufficient to know that the parameter file consists of a sequence of so-called "key-blocks". Each of these key-blocks contains a set of related parameters. Some key-blocks are mandatory while others are optional. The order of the individual key-blocks in the parameter file is arbitrary. A key-block is introduced by a "key" set between '[' and ']' characters. The sequence of the following key-block parameters is fixed and needs to match the hard coded order. If an optional key-block is not supplied in the parameter file Opium will use hard coded default values. There are many parameter file examples in the Opium distribution.

Key block help can be obtained from the opium code by: "opium -k"

Back to top


[Atom]

formatexample
atom symbol1 or 2 characters C
number of reference orbitalsinteger 3
nlm, occupation, eigenvalue guess for orb 1 integer,float, - (or float) 100 2.0 -
...... 200 2.0 -
nlm, occupation, eigenvalue guess for orb n ...210 2.0 -0.3
Comments:
An eigenvalue guess of "-" directs code to generate a guess automatically.

An unbound valence state can be indicated by making the occupation value negative. This invokes the Hamman generalized state method [10] and (an the occupation is set to 0). You should also specify an eigenvalue guess (can be positive or negative) for the energy of this state. If a "-" is in the eigenvalue guess, the energy of this state is set to 0.0

This is a mandatory keyblock.
Back to top Back to keys


[Pseudo]

formatexample
number of orbitals in pseudopotetntialinteger3
cut-off radius for pseudo orbital 1 float1.5
......1.6
"rc" - cut-off radius for pseudo orbital n ...1.6
(o)ptimized, (k)erker, or (t)m -- pseudo. methodcharacteroptimized
Comments:
This is a mandatory keyblock.
Back to top Back to keys


[Optinfo]formatexample
cut-off wavevector, # bessel fxns for pseudo orb 1float, integer6.00 4
......7.07 10
"qc" - cut-off wavevector, # bessel fxns for pseudo orb n...4.0 5
Comments:
This is a mandatory keyblock if optimized method is chosen in [Pseudo]
Back to top Back to keys


[XC]values
exchange-correlation functional pzlda pwlda or gga
Comments:
pzlda - Perdew-Zunger LDA [7]
pwlda - Perdew-Wang LDA [8]
gga - Perdew-Burke-Ernzerhof GGA [9]
Default is pzlda and lda is an abbreviation for pzlda.
Back to top Back to keys


[Pcc]formatexample
Partial core radiusfloat 0.5
Partial core methodcharacter lfc
Comments:
A partial core radius of 0.0 is treated as meaning no partial core.
There are two pcc methods: "lfc" - Louie, Froyen, and Cohen[4] or "fuchs" - Fuchs and Scheffler[11].
The default core radius is 0.0 (meaning NO partial-core) and the default method (if a radius but no method is specified) is "lfc"
Back to top Back to keys


[Relativity]values
Switch for scalar-relativistic solve nrl or srl
Comments:
nrl - non-relativistic solve
srl - scalar-relativistic solve
Default is nrl
Back to top Back to keys


[Grid]formatexample
Radial grid parameters np,a,b integer,float,float1201 0.0001 0.013
Comments:
The radial grid is defined as: non-rel grid for grid2.gif whichever comes first. This grid is used for all parts of the program
The default parameters are: 1201, 0.0001, 0.013
Back to top Back to keys


[Tol]formatexample
Change in eig. and pot. for AE and NL calcs.float, float 1e-8 1e-6
Comments:
Default is 1e-6 and 1e-8.
Back to top Back to keys


[Configs]formatexample
Number of test configurations integer 3
nlm, occupation, eigenvalue guess for orb 1 integer,float, - (or float) 400 1.5 -
...... 410 6.0 -
nlm, occupation, eigenvalue guess for orb n ...320 9.5 -
Comments:
Note: The valence orbitals must be in same order as the reference state
Back to top Back to keys


[KBdesign]formatexample
Ang. mom. of the local pot. for KB constructioninteger or character s
Number of funcs. for designed non-local procedureinteger3
Units,left edge,right edge,depth for 1st boxstring,float,float,float au 1.0 1.5 -0.3
......gp 500 750 -0.4
Units,left edge,right edge,depth for nth boxstring,float,float,float au 0.0 0.5 -1.3
Comments:
au means left and right edge will be entered in units of bohr, gp means these are specified by grid point.
Default values are: s for the local potential and no augmentation functions.
If the design non-local approach is not used, only the local potential needs to be specified.
Back to top Back to keys


[Loginfo]formatexample
Configuration number used to plot log. derivatives.integer 1
radius, min energy, max energyfloat, float, float 2.2 -2.0 2.0
Comments:
Configuration 0 means the reference state.
Back to top Back to keys


Plotting

Plotting help can also be obtained from the opium code by: "opium -p"

The plotting features of opium are executed in the following way:

opium paramfile logfile plot [plot_type]

where [plot_type] refers to one of the following strings:

Plot typestring
all-electron wavefunctionswa
pseudo & all-electron wavefunctionswp
core, valence and partial core densitypcc
core, valence and partial core densityden
screened pseudopotentialsvs
ionic (descreened) pseudopotentialsvi
q-space pseudo- wavefunctions and potentialsqp
log. deriv. for state listed in [Loginfo] keyblocklogd

Back to top


The log file

The log file is the main output file for Opium. It contains all comments, warnings, and errors. It is important to read the log file carefully for such things. The output for the following commands will be described below:


Welcome message:

The log file begins with a welcome message and writes out most of the parameters of the caluclation:


 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
           OPIUM  Version:      1.0.2                             
           =====================================            
 See http://opium.sourceforge.net for help and information                 
 Copyright 2004 : The OPIUM project                         
 time of execution: Thu Aug 19 16:49:42 2004

 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 Reading parameter file:  <cu.param> ...
 File prefix             : cu 
 Element                 : Copper     (Cu) 
 Z                       : 29 
 
 Number of all-electron orbitals   : 8 
 Number of pseudo       orbitals   : 3 
 
 Pseudopotential is non-relativistic 
 Exchage-correlation functional is : Perdew-Zunger LDA 
 The s potential is used for the KB construction 
 Optimized (RRKJ) pseudopotential method will be used 

 nl    cutoff radii    q-max      # bessel functions
 4s       1.700        5.500              10 
 4p       1.900        5.500              10 
 3d       2.000        7.100              10 

 Reference Configuration: core  <-/-> valence 
 1s2  2s2  2p6  3s2  3p6  <-/-> 4s0.75  4p0.25  3d9 

 Grid definitions: 
 a_grid=1.00e-04    b_grid=1.30e-02    # points=1131 
 r(1)=3.25e-05      r(np)=7.91e+01    

 dEmax tolerance:  1.00e-08      dVmax tolerance:  1.00e-06     

 Starting calculation...

When running non-interactively, this section is at the beginning of every log file. In interactive mode this section is written after every carriage return.

Back to top Back to log


All electron calculation (ae):

The first step to creating the pseudopotential is to perform the all-electron or 'AE' calculation. Here is the all-electron output for a copper atom:

 ======================================================================== 
 Begin AE calculation
 ======================================================================== 
 Performing non-relativistic AE calculation...  
 
 iter       Etot             Ebs             Ehxc       de_max   dv_max
   1    -3606.1898110   -2096.5570427   -1509.6327684 0.63E+00 0.47E+01
   2    -3291.9228178   -1947.2651070   -1344.6577108 0.65E+00 0.10E+01
   3    -3282.8597030   -1930.6798294   -1352.1798735 0.13E+00 0.18E+00
   .
   .
   .
   .
  32    -3274.5405170   -1919.4887891   -1355.0517279 0.67E-07 0.66E-07
  33    -3274.5405211   -1919.4887909   -1355.0517302 0.28E-07 0.13E-07
  34    -3274.5405219   -1919.4887907   -1355.0517312 0.44E-08 0.26E-07
 
 After    34 iterations...
 Energy:   -3274.54052195  Ebs:   -1919.48879071  Ehxc:   -1355.05173124
 
   Orbital    Filling           Eigenvalues        Norm(rc->oo)
    |100>      2.000         -642.7023469980
    |200>      2.000          -77.4339827449
    |210>      6.000          -68.1106530315
    |300>      2.000           -9.2360669866
    |310>      6.000           -6.3320340017
    |400>      0.750           -1.0243790023       0.8162396303
    |410>      0.250           -0.5972120528       0.8703851995
    |320>      9.000           -1.4633653098       0.0485860817

 ======================================================================== 
 End AE calculation
 ======================================================================== 

The first part of the AE section shows the convergence of the Total energy (Etot), the orbital energy (Ebs) and the sum of the Hartree and exchange/correlation energy (Ehxc). Also, the largest change in the eigenvalues and the potential is printed.

The next part is the summary of the eigenvalues and partial norm of each wavefunction (integral from the cutoff radius to infinity). The partial norm, is only calculated for valence states.

Back to top Back to log


Pseudopotential generation (ps):

The next section is the generation of the pseudopotential or 'PS' calculation. In this example, the optimized pseudopotential method is used:


 ======================================================================== 
 Begin PS construction
 ======================================================================== 
 Optimized Pseudopotential Generation
 
 ==================
 Pseudizing state : |400>
 eigenvalue       : -1.024379
 qc               :  5.500000
 # bessel fxns    : 10
 point nearest rc :  837
 rc               :  1.700000
 actual rc        :  1.707837
 
 Psi(rc)          :     0.3826553303
 Slope at rc      :    -0.0534146711
 Curvature at rc  :    -0.5396774115
 Del^2 Psi at rc  :    -0.6022298326
 Log Deriv at rc  :    -0.1395895126
 
 Starting KE minimization...
  step #     theta      slope       curv       KEresid     coeffsum
     1    0.001697   -0.076480   22.534734    0.000114    0.063402
     2   -0.000648    0.034498   26.600879    0.000103    0.059346
     3   -0.001585    0.023718    7.481439    0.000084    0.062049
     4   -0.002179    0.022481    5.158509    0.000059    0.074217
    11   -0.000046    0.002442   26.290474    0.000044    0.073522
 
 # steps:    15
 Resid KE (Ry)     :        0.0000430933
 Norm error        :     0.139E-15
 Continuity error  :     0.000E+00
 Curvature error   :    -0.899E-14
 
 Bessel wavevectors and final coefficients
   1        0.4835608470        0.3552464562
   2        2.6620880178       -0.2981574809
   3        4.5414761759        0.0137704608
   4        6.3975547999        0.0025853601
   5        8.2461851245        0.0002364149
   6       10.0914767940       -0.0004722742
   7       11.9349823306        0.0009239812
   8       13.7774199990       -0.0010797913
   9       15.6191681976        0.0013383903
  10       17.4604453288       -0.0011563335
 
 Convergence term (Ry/e) :     0.0000430933
 Occupation              :     0.7500000000
 Convergence error (mRy) :     0.0323199834
 Convergence error (meV) :     0.4397392295
 

The first section displays the state being pseudized, its eigenvalue, selected wavevector cut-off (qc), and the number of basis functions used in the bessel expansion. Next, the grid point nearest to the desired cut-off (pseudization) radius, the desired cut-off radius, and then actual cut-off radius used are displayed.

The next block shows the parameters which define the soon to be constructed pseduo-wavefunction. These are: the AE wavefunction, it's first two derivatives, the Laplacian and the logarithmic derivative at the cut-off radius.

The next section shows the convergence of the residual kinetic energy minimization. The step #, step size (theta), slope and curvature at theta=0, the residual KE and the sum of the Bessel coefficients. After the minimization converges, the final # of steps, residual KE and the errors in partial norm, value at rc, and curvature at rc are all reported.

The last section shows the set of bessel wavevectors, their coefficients and the final convergence information. The convergence term is just the residual kinetic energy. The convergence error is the convergence term multiplied by the occupation number.

IMPORTANT : The convergence error should be checked carefully. This value signifies the level of error that this state contributes to the total energy at the minimal cut-off energy (Ecut) is qc^2 Ry. Even if you obtain excellent transferability for your pseudoptential at, say, qc = 3.0 (Ecut = 9.0 Ry), you still must have a small convergence error at this qc for your results to be correct.

Once the first state is pseudized, the rest are done in turn.


 ------------------------------
 Descreening potential
 valence charge          :  10.000000
 core    charge          :   0.000000
 
 ----Solving the Schrodinger equation for all states----
 State: 4s  AE eigenvalue =  -1.024379  PS eigenvalue =  -1.024379
 State: 4p  AE eigenvalue =  -0.597212  PS eigenvalue =  -0.597212
 State: 3d  AE eigenvalue =  -1.463365  PS eigenvalue =  -1.463365

The next part of the 'PS' output shows total valence and core charge. The core charge will always be zero unless a partial core correction is used in which case the partial core charge is printed.

Next, the Schrodinger equation is solved for all valence states to ensure that the pseudopotential does indeed yield the correct AE eigenvalue for the reference state.


 ---Semilocal ghost testing---
 Local state: 4s
 
 Test  state: 4p
 KB energy :  12.467713  KB strength:   0.388156  KB cosine:   0.031133
 el0       :  -0.618203  el1        :  -0.177050  eig      :  -0.597212
 No ghosts!  Ekb>0  and el0 < eig < el1
 
 Test  state: 3d
 KB energy : -18.167960  KB strength:  14.284050  KB cosine:  -0.786222
 el0       :  -0.205692  el1        :  -0.085889  eig      :  -1.463365
 No ghosts!  Ekb<0  and eig < el0
 
 No ghosts for local potential: 4s
 ------------------------------
 
 Local state: 4p
 
 Test  state: 4s
 KB energy : -18.202096  KB strength:   0.590062  KB cosine:  -0.032417
 el0       :  -1.070932  el1        :  -0.288800  eig      :  -1.024379
     !GHOST! : 4s  -1.024379  Should be lower than   -1.070932
 
 Test  state: 3d
 KB energy : -14.172017  KB strength:  11.600853  KB cosine:  -0.818575
 el0       :  -0.202399  el1        :  -0.085093  eig      :  -1.463365
 No ghosts!  Ekb<0  and eig < el0
 
  !WARNING! Ghosts for local potential: 4p
 ------------------------------
 
 Local state: 3d
 .
 .
 . 
 ------------------------------

 ======================================================================== 
 End PS construction
 ======================================================================== 


The final section of the PS calculation is a loop over all valence states to check for the existence of ghost states. This ghost testing shows you the possible angular momentum channels that could be used as the local potential. For instance, above we see that the 'p' potential is not a good choice for the local potential since it will result in ghosts.

Please see [5] for more information.

Back to top Back to log


Non-local calculation (nl):

After the semi-local pseudopotential is constructed, the Kleinman-Bylander non-local form is tested. This step is referred to as the NL step. The eigenvalues and partial norms for the NL section should agree with AE section since the pseudopotential was constructed to do this.

To do the NL step, a local potential must be defined, usually just one of the valence potentials. Opium also has the ability to use the sum of one or a series of step functions and a valence potential.


 ======================================================================== 
 Begin NL calculation
 ======================================================================== 
 Using the s potential as the local potential
 
 iter       Etot             Ebs             Ehxc       de_max   dv_max
   1      -83.2901858     -14.0878751     -69.2023107 0.00E+00 0.17E-07
 
 Converged in 1 iteration (probably reference state)
 Energy:     -83.29018577  Ebs:     -14.08787505  Ehxc:     -69.20231072
 
   Orbital    Filling           Eigenvalues        Norm(rc->oo)
    |100>      0.750           -1.0243790023       0.8162396540
    |210>      0.250           -0.5972120528       0.8703851995
    |320>      9.000           -1.4633653098       0.0485860792
 

Again, notice how the NL calculation reproduces the AE results. Of course, the total energies are different since there are no core electrons.

  ---Non-local ghost testing---
 Local state: 1s
 
 Test  state: 2p
 KB energy :  12.467713  KB strength:   0.388156  KB cosine:   0.031133
 el0       :  -0.618203  el1        :  -0.177050  eig      :  -0.597212
 No ghosts!  Ekb>0  and el0 < eig < el1
 
 Test  state: 3d
 KB energy : -18.167960  KB strength:  14.284050  KB cosine:  -0.786222
 el0       :  -0.205692  el1        :  -0.085889  eig      :  -1.463365
 No ghosts!  Ekb<0  and eig < el0
 ------------------------------
 
 No ghosts present for local potential

 

 ======================================================================== 
 End NL calculation
 ======================================================================== 


The last part of the NL section is another round of ghost testing. These results should be the same as the PS ghost testing if the local potential is just chosen from the valence. If the designed non-local method is used (some function(s) added to a valence potential) this can change the ghost behavior so this should be checked.

Back to top Back to log


Transferability testing: (tc)

After the ae, ps, and nl steps are completed with small convergence errors and no ghosts for the desired local potential, the transferability must now be checked. Transferability measures how well a pseudopotential performs in environments other than the reference configuration (the configuration used in the generation step). The all-electron and pseudo eigenvalues and partial norms are computed for each test configuration (from the [Configs] keyblock) and written to the log.


<<<do_tc>>>

 ===============Configuration 1 AE Calc=============== 
 
 iter       Etot             Ebs             Ehxc       de_max   dv_max
   1    -3656.6376989   -2099.4527600   -1557.1849389 0.63E+00 0.56E+01
   2    -3261.0263906   -1928.1127034   -1332.9136872 0.75E+00 0.14E+01
   .
   .
   .
  31    -3274.9371373   -1906.6704499   -1368.2666874 0.18E-06 0.62E-07
  32    -3274.9371415   -1906.6704488   -1368.2666927 0.74E-08 0.14E-06
 
 After    32 iterations...
 Energy:   -3274.93714150  Ebs:   -1906.67044877  Ehxc:   -1368.26669273
 
   Orbital    Filling           Eigenvalues        Norm(rc->oo)
    |100>      2.000         -642.2116645779
    |200>      2.000          -76.8977081730
    |210>      6.000          -67.5794262456
    |300>      2.000           -8.7475989839
    |310>      6.000           -5.8528500789
    |400>    0.000        -0.8700997488      0.8418328331 
    |410>    0.000        -0.4810316412      0.8970980498 
    |320>   10.000        -1.0362847356      0.0736856317 
.
.
.

 ===============Configuration 1 NL: Calc ===============
 Using the s potential as the local potential
 
 iter       Etot             Ebs             Ehxc       de_max   dv_max
   1      -90.6881561     -14.6336531     -76.0545030 0.00E+00 0.79E+00
   2      -85.5600101     -11.4740303     -74.0859799 0.22E+00 0.18E+00
   .
   .
   .
  23      -83.6927176     -10.3388674     -73.3538502 0.17E-07 0.22E-07
  24      -83.6927175     -10.3388673     -73.3538501 0.63E-08 0.14E-07
 
 After    24 iterations...
 Energy:     -83.69271746  Ebs:     -10.33886734  Ehxc:     -73.35385011
 
   Orbital    Filling           Eigenvalues        Norm(rc->oo)
   |100>     0.000       -0.8597891449      0.8454717157 
   |210>     0.000       -0.4756488215      0.8994808662 
   |320>    10.000       -1.0338867340      0.0703706398 
 .
 .
 .
 . 
   ------------------------------------------------
   ================================================

Back to top Back to log


The report file

You could compute the norm and eigenvalue differences by hand using the log output, but it is much more convenient to generate a 'report' file using the rpt command. The report command summarizes key information concerning the pseudopotential.

The first part of the report file is a dump of the parameter file:

##########################################################
#    Opium Report File                                   #
##########################################################
     Opium version:      1.0.2

### copy of the parameter file #######################

[Atom]
Cu
8             
100  2.00  -  
200  2.00  -    
210  6.00  -    
300  2.00  -    
310  6.00  
.
.
.

Next, the AE output is summarized:

### AE report ########################################

 AE:Orbital    Filling       Eigenvalues[Ry]          Norm
    ----------------------------------------------------------
	100	 2.000	    -642.7023469980
	200	 2.000	     -77.4339827449
	210	 6.000	     -68.1106530315
	300	 2.000	      -9.2360669866
	310	 6.000	      -6.3320340017
	400	 0.750	      -1.0243790023	  0.8162396303
	410	 0.250	      -0.5972120528	  0.8703851995
	320	 9.000	      -1.4633653098	  0.0485860817

      E_tot =    -3274.5405219488 Ry

Next, the convergence error and ghost testing results are printed. The first column is the valence state, the second column is the convergence error per electron. Next, the error per electron is multiplied by the occupation of the state to yield the convergence error in the reference state and is reported in mRy as well as meV. The last column states whether a ghost was found when this state was used as the local potential. We see that only the s potential is a valid choice for the local potential. If the ghost testing was inconclusive, a '?' will be printed.

### PS report ########################################

    Orbital  Conv. error: [mRy/e]            [mRy]             [meV]     Ghost
    --------------------------------------------------------------------------
	400            0.0430933109      0.0323199832      0.4397392268	    no
	410            0.0172931190      0.0043232797      0.0588216796	   yes
	320            0.1638535915      1.4746823237     20.0642327596	   yes

                  Tot. error =           1.5113255866     20.5627936661

Next, the NL test results are summarized. The ghost testing column in this table shows whether one of the non-local potentials gives a ghost given the choice of local potential.

### NL report ########################################


 NL:Orbital    Filling       Eigenvalues[Ry]          Norm       Ghost
    ------------------------------------------------------------------
	100	 0.750	      -1.0243790023	  0.8162396540	    no
	210	 0.250	      -0.5972120528	  0.8703851995	    no
	320	 9.000	      -1.4633653098	  0.0485860792	    no

      E_tot =      -83.2901858014 Ry

Finally, the transferability tests are summarized and the errors are computed and printed.

### TC report ########################################

 AE:Orbital    Filling       Eigenvalues[Ry]          Norm
    ----------------------------------------------------------
	100	 2.000	    -642.2116645779
	200	 2.000	     -76.8977081730
	210	 6.000	     -67.5794262456
	300	 2.000	      -8.7475989839
	310	 6.000	      -5.8528500789
	400	 0.000	      -0.8700997488	  0.8418328331
	410	 0.000	      -0.4810316412	  0.8970980498
	320	10.000	      -1.0362847356	  0.0736856317

      E_tot =    -3274.9371415016 Ry

 NL:Orbital    Filling       Eigenvalues[Ry]          Norm       Ghost
    ------------------------------------------------------------------
	100	 0.000	      -0.8597891450	  0.8454717157	    no
	210	 0.000	      -0.4756488216	  0.8994808662	    no
	320	10.000	      -1.0338867340	  0.0703706398	    no

      E_tot =      -83.6927174850 Ry

 AE-NL:Orbital Filling       Eigenvalues[mRy]         Norm[1e-3] 
 AE-NL- --------------------------------------------------------------
 AE-NL- 100	 0.000	     -10.3106038212	 -3.6388825155	
 AE-NL- 210	 0.000	      -5.3828196377	 -2.3828163585	
 AE-NL- 320	10.000	      -2.3980016394	  3.3149918622	
 AE-NL-  total error =	      18.0914250983	  9.3366907362

 =====================================================================
 AE:Orbital    Filling       Eigenvalues[Ry]          Norm
    ----------------------------------------------------------
	100	 2.000	    -642.6711429101
.
.
.
.
 =====================================================================

The last section is the comparison of the change in energy between configuration 'i' and 'j' (configuration "0" is the reference) for the AE and NL atoms. This is another quantity that can be used to measure transferability.

  Comparison of total energy differences.           
   DD_ij = (E_i - E_j)_all-electron - (E_i - E_j)_pseudo     

 AE-NL-  i   j          DD[mRy]        DD[meV] 
 AE-NL- ------------------------------------------
 AE-NL-   0   1        -5.912131     -80.438677
 AE-NL-   0   2        -0.006575      -0.089456
 AE-NL-   0   3       -22.830914    -310.630563
 AE-NL-   1   2         5.905556      80.349221
 AE-NL-   1   3       -16.918783    -230.191886
 AE-NL-   2   3       -22.824339    -310.541106

Back to top


Last updated: Jul 21, 2008