Complex Wave Functions in Large Simulation Cells
kevin rasch
function produces nan and inf results in VMC.
STEP 1:
At Portland APS, Lucas tells me to use the "PRINT_WF_VALS" keyword in
my method and try different values of MAGNIFY (specifically that I may
need smaller instead of larger MAGNIFY). So I try VMC for 5 steps, 10
blocks, 1 walker using Qwalk-wagner executing on my MacBookPro (make
PLATFORM=Darwin).
MAGNIFY = 0.1
-----------------------
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
MAGNIFY = 1.0
------------------------
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
MAGNIFY = 10.0
-------------------------
node 0 amp -321.5451824
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
node 0 amp -inf
MAGNIFY = 20.0
-------------------------
node 0 amp -209.2553391
node 0 amp -339.5200086
node 0 amp -309.7604736
node 0 amp -299.6622649
node 0 amp -331.7791386
node 0 amp -311.6537747
node 0 amp -339.9160482
node 0 amp -306.8876066
node 0 amp -330.9910569
node 0 amp -348.236941
This has seemingly gotten better, however, using the TEST method on
the end configuration gives 0 in every position, e.g.
#############checking derivatives
#######################
####electron 0 #####
#######################
0 0 OK
0 0 OK
0 0 OK
laplacian
0 0 OK
#######################
####electron 1 #####
#######################
0 0 OK
0 0 OK
0 0 OK
laplacian
0 0 OK
#######################
####electron 2 #####
#######################
0 0 OK
0 0 OK
0 0 OK
laplacian
0 0 OK
etc.
furthermore, the energy computed is not reasonable (kinetic energy is
0). Further increasing the MAGNIFY doesn't change this and
eventually, e.g. MAGNIFY = 200, results in the error:
"Error Did over 1000 shifts and we're still out of the simulation
cell. There's probably something wrong. Position : nan"
STEP 2:
I added to the PRINT_WF_VALS output in Vmc_method.cpp at about line
227 additional cout statements to print the other values stored in
wfval. Again running VMC for 5 steps, 10 blocks, 1 walker using Qwalk-
wagner executing on my MacBookPro (make PLATFORM=Darwin):
MAGNIFY = 20.0
-------------------------
node 0 amp(0,0) -209.2553391
node 0 amp(0,1) 4.485050304e-308
node 0 amp(0,2) 0
node 0 phase(0,0) -0
node 0 phase(0,1) 646.1938284
node 0 cvals(0,0) (-209.2553391,0)
node 0 cvals(0,1) (0,0)
node 0 cvals(0,2) (-0,646.1938284)
node 0 amp(0,0) -339.5200086
node 0 amp(0,1) 4.1342844e-309
node 0 amp(0,2) 2.419126577e-44
node 0 phase(0,0) -3.141592654
node 0 phase(0,1) 894.8800174
node 0 cvals(0,0) (-339.5200086,0)
node 0 cvals(0,1) (0,0)
node 0 cvals(0,2) (-3.141592654,894.8800174)
node 0 amp(0,0) -309.7604736
node 0 amp(0,1) 4.485050304e-308
node 0 amp(0,2) 1.688550142e-50
node 0 phase(0,0) -0
node 0 phase(0,1) 893.2316627
node 0 cvals(0,0) (-309.7604736,0)
node 0 cvals(0,1) (0,0)
node 0 cvals(0,2) (-2.000007153,6.389360643)
node 0 amp(0,0) -299.6622649
node 0 amp(0,1) 4.485050304e-308
node 0 amp(0,2) 5.781236649e-41
node 0 phase(0,0) -0
node 0 phase(0,1) 955.1740631
node 0 cvals(0,0) (-299.6622649,0)
node 0 cvals(0,1) (0,0)
node 0 cvals(0,2) (-2.000007153,8.494266894)
node 0 amp(0,0) -331.7791386
node 0 amp(0,1) 4.485050304e-308
node 0 amp(0,2) 2.07730639e-49
node 0 phase(0,0) -3.141592654
node 0 phase(0,1) 7.084846579
node 0 cvals(0,0) (-331.7791386,0)
node 0 cvals(0,1) (0,0)
node 0 cvals(0,2) (-2.000007153,7.889754405)
node 0 amp(0,0) -311.6537747
node 0 amp(0,1) 4.485050304e-308
node 0 amp(0,2) 3.480537753e-50
node 0 phase(0,0) -0
node 0 phase(0,1) 1.124657769e-312
node 0 cvals(0,0) (-311.6537747,0)
node 0 cvals(0,1) (0,0)
node 0 cvals(0,2) (-2.000007153,5.885732892)
node 0 amp(0,0) -339.9160482
node 0 amp(0,1) 4.485050304e-308
node 0 amp(0,2) 9.334397574e-48
node 0 phase(0,0) -0
node 0 phase(0,1) 4.456191164e-313
node 0 cvals(0,0) (-339.9160482,0)
node 0 cvals(0,1) (0,0)
node 0 cvals(0,2) (-2.000007153,7.752554148)
node 0 amp(0,0) -306.8876066
node 0 amp(0,1) 4.485050304e-308
node 0 amp(0,2) 2.523667453e-45
node 0 phase(0,0) -3.141592654
node 0 phase(0,1) 9.336781483e-313
node 0 cvals(0,0) (-306.8876066,0)
node 0 cvals(0,1) (0,0)
node 0 cvals(0,2) (-2.000007153,2.300600327)
node 0 amp(0,0) -330.9910569
node 0 amp(0,1) 4.485050304e-308
node 0 amp(0,2) 7.643665371e-51
node 0 phase(0,0) -3.141592654
node 0 phase(0,1) 9.548981062e-313
node 0 cvals(0,0) (-330.9910569,0)
node 0 cvals(0,1) (0,0)
node 0 cvals(0,2) (-2.000007153,2.57311471)
node 0 amp(0,0) -348.236941
node 0 amp(0,1) 4.485050304e-308
node 0 amp(0,2) 2.47661477e-41
node 0 phase(0,0) -3.141592654
node 0 phase(0,1) 4.880590322e-313
node 0 cvals(0,0) (-348.236941,0)
node 0 cvals(0,1) (0,0)
node 0 cvals(0,2) (-2.000007153,18.4046744)
This is strange to say the least because my MacBookPro reports that
DBL_MIN is 2.22507e-308 and so I don't understand how one gets the
result above of 4.880590322e-313.
The first step is to corral the origin of this problem. Once I/we
know where the calculation first goes bad, then there is a chance to
correct it. Any suggestions from the more veteran coders are welcome/
requested/begged for.
Lucas Wagner
Please post your input files. It's a bit strange that the amplitude
is *decreasing* as you go forward. It looks like MAGNIFY is still too
low; try 25 or so.
Lucas
> --
> To unsubscribe, reply using "remove me" as the subject.
>