AIMD simulations with Umbrella Sampling

[yang_shao-horn1]

We are currently working trying to set up MD simulations using umbrella sampling to extract thermodynamic properties of the saddle point using VASP. Specifically, We are interested in constraining ion 17 with a harmonic potential to prevent its jump away from the saddle point.
We have managed to define the ICONST file using the X, Y, Z flags, and status=8 (see ICONST below). However, we are encountering errors where VASP is looking for SPRING_K, SPRING_R0, and SPRING_V0 flags (see INCAR flags below), for which we are unable to find the necessary documentation and we are finding it difficult to understand what input values to provide.

Does anybody know how the challenges can be overcome?

[alexey.tal]

Dear yang_shao-horn1,

Could you please provide all the relevant input and output files for your calculation according to the forum guidelines.

[yang_shao-horn1]

Thanks for approving the post. I have included the inputs below.
Here are specific challenges we are facing:
We could not find documentation on the units of SPRING_K
We do not understand/find documentation for what flags/format should be inserted in SPRING_R0

ICONST
X 17 8
Y 17 8
Z 17 8

INCAR
#ALGO = Fast
IBRION = 0
LPLANE = False
LREAR = Auto
NSIM = 1
POTIM = 1
NSW = 10000
TEBEG = 1273
TEEND = 1273
MDALGO=21
SMASS = 0
HILLS_BIN = NSW
SPRING_K = 0.1 0.1 0.1
SPRING_R0 = -0.000056 7.615347 -0.000006
SPRING_V0 = 0

EDIFF = 1.E-5
ENCUT = 616.6635
ISMEAR = 0
LREAL = .FALSE.
ISPIN = 2
LCHARG = False
LWAVE = False
NELM = 2000
SIGMA = 0.05
NELECT = 312.0

POSCAR
La8 Al8 O23
1.0
7.640318 0.000062 0.015830
-0.000056 7.615347 -0.000006
0.015978 -0.000148 7.672853
La Al O
8 8 23
direct
0.007645 0.994207 0.993535 La
0.004765 0.997129 0.511398 La
0.002977 0.502700 0.996131 La
0.002062 0.500872 0.515296 La
0.495255 0.994359 0.994965 La
0.497107 0.998889 0.510168 La
0.498171 0.505052 0.996635 La
0.495296 0.507436 0.516775 La
0.228366 0.243290 0.252677 Al
0.255557 0.250624 0.755618 Al
0.250808 0.750341 0.255181 Al
0.249368 0.749003 0.753988 Al
0.764230 0.262895 0.256112 Al
0.752973 0.252512 0.755228 Al
0.745124 0.728400 0.257691 Al
0.751116 0.755162 0.754019 Al
0.600939 0.102779 0.248887 O
0.272072 0.235280 0.012011 O
0.993068 0.224118 0.241422 O
0.252036 0.000078 0.273394 O
0.258313 0.266920 0.496022 O
0.007675 0.258102 0.763972 O
0.241564 0.000133 0.743250 O
0.244153 0.763251 0.004117 O
0.991319 0.759868 0.257475 O
0.250618 0.491164 0.237184 O
0.251137 0.731445 0.505059 O
0.999859 0.744157 0.750712 O
0.253719 0.499618 0.769879 O
0.748084 0.286086 0.004166 O
0.770659 0.233857 0.506973 O
0.502624 0.249573 0.746266 O
0.758351 0.002132 0.777232 O
0.754436 0.745062 0.012483 O
0.501016 0.743574 0.250169 O
0.737424 0.493160 0.282819 O
0.747775 0.784651 0.496952 O
0.500601 0.747390 0.755779 O
0.749527 0.507214 0.732664 O

[yang_shao-horn1]

inputs_output.zip

[alexey.tal]

It looks like the problem is in SPRING_V0, not in SPRING_R0. I see that you only provided one coordinate for SPRING_V0. If all three coordinates are provided, the code runs through.

The comment in the code indicates that the units of SPRING_K is a.u..

It is true that the documentation of this part is lacking. We will work on improving that.

[apolet]

I was also looking for documentation or ideally an example such as this one on how to set up a harmonic-potential constraint properly. So I am just seconding the question in this thread. It would be very helpful to me if you could add some of these to the wiki. Thank you!

[alexey.tal]

I contacted a person who is an expert on umbrella sampling in VASP, so I got some clarification on your question.

The units of SPRING_K are eV/(units of coordinate)^2. For instance, if it's applied to a distance, SPRING_K has units of eV/A^2, if it is applied to an angle/torsion, it's eV/(rad.)^2.

SPRING_R0 is the position of the minimum. The units are the same as those of the coordinate the potential acts upon.

SPRING_V0 is the velocity at which R0 is shifted during MD and it is used for some special simulation protocols (e.g., a version of the slow growth), not for a typical umbrella sampling. In principle, SPRING_V0 is only optional (unlike SPRING_K and SPRING_R0) and the default value is 0. However, in VASP5.4.4 which you are using, an error is shown if SPRING_V0 is not provided. VASP6 just shows a warning if SPRING_V0 is not provided and the code should run normally.

If you have access to VASP 6, I would recommend that you use it instead.

Upon you requests, we will accelerate our work on the tutorial and the documentation for umbrella sampling, so please bare with us.

[apolet]

This answer is very helpful. Thank you very much!