Hello Axel,

I’m glad you answer my question, it’s vert kind of you.

In manual, *compute temp* command say it calculate the temperture by tensor. Tensor should be consider as matrix(in fact ,it is not strict, but it could be in this problem).

Then the trace of matrix is invariant, so we often use the trace to denfine physics.

Therefore, the mean of trace of temperture tensor should be temperture vaule, so I output c_TTTT[1],[2],[3], and their mean is about 0.92

This problem is like Rg tensor, I also calculate the Rg tensor, and output c_Rg[1],c_Rg[2],c_Rg[3], and the value is what i need

So, I’m confused! May be I’m stupid to make a mistake, but I try my best.

Thanks,

Yongjin Ruan

Hello Axel,

I'm glad you answer my question, it's vert kind of you.

In manual, compute temp command say it calculate the temperture by

tensor.

no, it doesn't say so and it doesn't do it. you get the *temperature*

for the compute group by accessing the *scalar* c_TTTT. that is what

the docs say and that is what gives you the temperature. the tensor

c_TTTT[] is a *kinetic energy* tensor. you can easily confirm this

from reading the source code. it looks like "almost 1.0" only because

you are using reduced units (conversion from m*v^2 to energy is 1.0

and k_B is 1.0) and also all computed _extensive_ properties (i.e.

that depend on the number of atoms like kinetic energy) get normalized

(i.e. divided by the total(!) number of atoms). temperature (the

scalar) is an _intensive_ property (does not depend on the number of

atoms) does not get normalized (the number of degrees of freedom for

the group of atoms is already factored in by the compute).

your group as 12144 out of 13200 atoms with a kinetic energy of ~1

epsilon per atom. => 12144/13200 = 0.92

add:

thermo_modify norm no

and you should see a drastic difference in the thermo output for all

extensive properties.

Tensor should be consider as matrix(in fact ,it is not strict, but

it could be in this problem).

Then the trace of matrix is invariant, so we often use the trace to

denfine physics.

Therefore, the mean of trace of temperture tensor should be temperture

vaule, so I output c_TTTT[1],[2],[3], and their mean is about 0.92

This problem is like Rg tensor, I also calculate the Rg tensor, and

output c_Rg[1],c_Rg[2],c_Rg[3], and the value is what i need

So, I'm confused! May be I'm stupid to make a mistake, but I try my

best.

there is a simple way to avoid confusion: read the documentation *all

the way* and also check out the sections that it refers to for further

information.

also, do not jump to conclusions prematurely.

axel.