Calculation of forces/ error minimisation

[saitovilnur]

Hi, My name is Ilnur Saitov and I am trying to use QMCPACK in order to calculate forces in crystalline hydrogen with the C2c structure utilising the "Force" estimator with mode "cep" (CCZ method). However, I've got a huge force errors. The head of result of application of the command "qmca - a s001scalar*" to the dmc part (series 1) is shown below

`qmca -a s001scalar*

avg series 1 LocalEnergy = -51.6687 +/- 0.0011 Variance = 1.9761 +/- 0.0024 Kinetic = 76.740 +/- 0.026 LocalPotential = -128.409 +/- 0.027 ElecElec = -39.5591 +/- 0.0017 IonIon = -55.87 +/- 0.00 ElecIon = -32.981 +/- 0.026 LocalEnergy_sq = 2671.63 +/- 0.11 force = 0.00 +/- 0.00 force_0_0 = 0.0090 +/- 0.0040 force_0_1 = 0.0021 +/- 0.0031 force_0_2 = -0.0041 +/- 0.0017 force_10_0 = 0.0041 +/- 0.0028 force_10_1 = -0.0037 +/- 0.0042 force_10_2 = -0.0079 +/- 0.0024 force_11_0 = -0.0049 +/- 0.0023 force_11_1 = 0.0001 +/- 0.0012 force_11_2 = -0.0092 +/- 0.0030 force_12_0 = -0.0012 +/- 0.0022 force_12_1 = 0.0039 +/- 0.0026 force_12_2 = 0.0158 +/- 0.0019 force_13_0 = 0.0067 +/- 0.0028 force_13_1 = -0.0001 +/- 0.0049 force_13_2 = 0.0099 +/- 0.0019 force_14_0 = 0.0007 +/- 0.0020 force_14_1 = 0.0047 +/- 0.0019 force_14_2 = -0.0095 +/- 0.0018 force_15_0 = -0.0035 +/- 0.0032 force_15_1 = 0.0021 +/- 0.0022 force_15_2 = -0.0019 +/- 0.0016 force_16_0 = -0.0035 +/- 0.0019 force_16_1 = -0.0032 +/- 0.0034 force_16_2 = -0.0060 +/- 0.0023 .... ` The results are averaged over 36 k-point (36 twists).

So I tried to apply backflow transformation, but it didn't improve the errors. I also tried to increase the number of blocks and walkers and it doesn't affect the errors as well.

Could someone give me some recommendations related to the minimisation of force errors?

The input xml-file for one of the twists is also attached.

dmc.g017.twistnum_17.in.txt

Thanks in advance, Ilnur

[jtkrogel]

@rcclay I think this one is for you.