Errors in compiling meam package in Linux

Hi everyone,
When i compile meam package in Linux , the same errors always existed.
Up to now, i have no idea about the reason and resolutions for those errors.
I attached the detials of my operation, hope you can give some advises.
(Lammps: 26May13 version), I used fortran and openmpi.

I compile it by following steps:
Step: 1) In the directory …/lib/meam— First delete the files of .a, .f, .mod; Then make -f Makefile.gfortran.
get following information:

gfortran O -fpic - fno-second-undersocre c meam_data F。

gfortran O -fpic

  • fno-second-undersocre cmeam_setup_done F。

gfortran O -fpic

  • fno-second-undersocre cmeam_setup_global F。

gfortran O -fpic

  • fno-second-undersocre cmeam_setup_param F。

gfortran O -fpic

  • fno-second-undersocre cmeam_dens_init F。

gfortran O -fpic

  • fno-second-undersocre cmeam_dens_final F。

gfortran O
-fpic- fno-second-undersocre cmeam_force F。

gfortran O
-fpic- fno-second-undersocre cmeam_cleanup F。

AR RV libmeam。一meam_data
.Omeam_setup_done.O meam_setup_global.O meam_setup_param.O meam_dens_init.O meam_dens_final.O meam_force.O meam_cleanup.o。

AR:rv libmeam.a。

meam_data.o。

meam_setup_done.o。

meam_setup_global.o。

meam_setup_para.o

截图3.png

截图4.png

Your errors are occurring at link time, probably b/c you are not including all
the correct libs needed to link a C++ application (LAMMPS) with a Fortran (MEAM)
lib. Which libs are needed is a function of your compilers. Section_start.html
of the manual explains this.

In src/MAKE/Makefile.openmpi you should not need to change anything except

in this section:

compiler/linker settings

specify flags and libraries needed for your compiler

CC = mpic++
CCFLAGS = -O2
-funroll-loops -fstrict-aliasing -Wall -W -Wno-uninitialized
SHFLAGS = -fPIC
DEPFLAGS = -M

LINK = mpic++
LINKFLAGS = -O
LIB = -lstdc++
SIZE = size

ARCHIVE = ar
ARFLAGS = -rcsv
SHLIBFLAGS = -shared

and you should not add any variables.

Any Fortran-dependent libs should be added to lib/meam/Makefile.lammps.

Again this is explained in Section_start.html. There are several example
Makefile.lammps.* files i that dir that are useful for different Fortran compilers.

Steve

截图3.png

截图4.png

Hi Steve,
Thank you for your replies.
According to your advises, i know some settings should be modified to the makefile.lammps in …/lib/meam file.

The following three settings should be correctly modified.
Followings are my modifications:
meam_SYSINC =
meam_SYSLIB = -lgfortran
meam_SYSPATH = -L/public/software/intel/Compiler/11.1/059/lib/intel64
For meam_SYSPATH , i have no ideas. So, i refers to the following information:
截图1.png
However, there were same errors: eg. undefined reference to " meam_force"

Best regardes.

截图3.png

截图4.png

-lgfortran is not an Intel Fortran compiler lib, so those settings

are not consistent. If you compiled with the GNU fortran compiler
than use -lgfortran. If you compiled with the Intel Fortran compiler,
then you need to list Intel Fortran libs for SYSLIB.
See the example Makefile.lammps files in lib/meam. Cases for
both those compilers are there.

Steve

截图4.png

截图3.png

截图1.png

Hi Steve,
Following your advice, I modified the compiling of meam package.
I could make it, lmp_openmpi can correctly operate the in.meam file. In the in.meam file, there are 10000 atoms of TiN.
However, i found the calculating speed of linux 48 cores (parallel) is much solwer than that of windows 4 cores (parallel).
eg. for 5000 timesteps, linux needs 2 hrs, while windows only needs less than 1 hrs.
Are this speed dofference related to the settings of compiling files, like makefile.ifort and makefile.openmpi?
Followings are my settings of those two files.
1)make -f Makefile.ifort
2)make yes-meam
3)make openmpi.

截图1.png

截图3.png

截图4.png

If both versions ran, I doubt that the compilation
details are the issue. What do you get
if you run the Linux version on 4 cores, same
as the Windows? Just for a couple 100 timesteps.
Can you post both log files. Are you sure that
both versions are actually running in parallel?

Steve

截图4.png

截图3.png

截图1.png

Hi everyone,
There maybe a stupid question to someone. However, it confuses me.

compute 1 precipitate voronoi/atom surface all

compute 2 precipitate reduce sum c_1[3]
Using those two commands, we can know the surface area of the “precipitate” group (the
hull of the precipitate) .
So my confusion is about units of the surface area.
According to the “units” command.
i think the units of the surface area could be like this.
Metal: A2
LJ: sigma
2
real: A*2
yes or not?

Hai-Long Yao

State Key Laboratory for Mechanical Behavior of Materials
School of Materials Science and Engineering, Xi’an Jiaotong University
Xianning West Road 28, Xi’an, Shaanxi, 710049 China
Tel: +86 29 82665299
Fax: +86 29 83237910
E-mail:[email protected]…127…

截图1.png

截图3.png

截图4.png

Hi daniel,
There maybe a stupid question to someone. However, it confuses me.

compute 1 precipitate voronoi/atom surface all

compute 2 precipitate reduce sum c_1[3]
Using those two commands, we can know the surface area of the “precipitate” group (the
hull of the precipitate) .
So my confusion is about units of the surface area.
According to the “units” command.
i think the units of the surface area could be like this.
Metal: A2
LJ: sigma
2
real: A*2
yes or not?

Another confusion :
If the precipitate group is crushed into several parts, it means the surface area of the precipitate group increases.
In this case, can the “compute voronoi/atom” command correctly calculate the surface area of the precipitate group (including several parts)?

Regardes

Hai-Long Yao

State Key Laboratory for Mechanical Behavior of Materials
School of Materials Science and Engineering, Xi’an Jiaotong University
Xianning West Road 28, Xi’an, Shaanxi, 710049 China
Tel: +86 29 82665299
Fax: +86 29 83237910
E-mail:[email protected]…127…

截图1.png

截图3.png

截图4.png

compute 1 precipitate voronoi/atom surface all

compute 2 precipitate reduce sum c_1[3]
Using those two commands, we can know the surface area of the “precipitate” group (the
hull of the precipitate) .

No! These commands do not compute the hull! If “surface all” is specified the internal voronoi faces are counted as well. You need to create a “matrix” group, which consists of all-precipitate and specify “surface matrix”.

Dear steve and all,
From most published papers,a modified charge transfer-embedded atom method
potential can be used for simulations of metal-metal oxide systems( eg. Al-O).
However, i cannot find a responding pair-style in lammps. Is this potential not implemeted in lammps?
If yes, i wonder how to make it.
Best regards.

Hai-Long Yao

State Key Laboratory for Mechanical Behavior of Materials
School of Materials Science and Engineering, Xi’an Jiaotong University
Xianning West Road 28, Xi’an, Shaanxi, 710049 China
Tel: +86 29 82665299
Fax: +86 29 83237910
E-mail:[email protected]…127…

截图1.png

截图3.png

截图4.png

I believe you mean the Streitz-Mintmire potential? It is not yet released.

Ray

截图3.png

截图1.png

截图4.png

Dear Shan,
Thank you for your reply.
I intend to simulate the Al-O system. Althrough “MODIFIED EMBEDDED ATOM METHOD CALCULATIONS OF INTERFACES” gave some
parameters for Al-O interaction in MEAM potential, i think there are not enough. So, i wonder to know and use the Streitz-Mintmire potential.
According to 'MODIFIED EMBEDDED ATOM METHOD CALCULATIONS OF INTERFACES‘,the Al-O.meam file is following:
lattce(1,2)=‘b1’
Ec(1,2) = 4.000
alpha(1,2) = 4.5
re(1,2) = 1.97
re(1,1) = 2.86
re(2,2) = 1.21
rc = 4.0
delr = 0.1
bkgd_dyn=1
Maybe lack of parameters of Cmin(i,j,k) and Cmax(i,j,k) induce the useless of the Al-O.meam file
Could you give some advice about the Cmin(i,j,k) and Cmax(i,j,k)?
Best regards

截图1.png

截图3.png

截图4.png

You can’t use Streitz Mintmire yet because it is not yet available.

Ray

截图4.png

截图1.png

截图3.png

It’s a pity for such a famous potential. I meam the potential developped by Baskes MI.
Some papers (eg.Attachment ) used the meam potential for Al_Al2O3 interface and the their references are “Baskes MI. Modified embedded atom method calculations of interfaces. Sandia 1996;96:8484.” and "M.I. Baskes, Modified embedded atom method calculations of interfaces, in: S.

Nishijima, H. Onodera (Eds.), Sandia National Laboratories Livermore, USA,1996.".

However, they didn’t show the complete parameters for the Al-O.meam file.
Can anyone give hints excepting follows ?

lattce(1,2)=‘b1’
Ec(1,2) = 4.000
alpha(1,2) = 4.5
re(1,2) = 1.97
re(1,1) = 2.86
re(2,2) = 1.21
rc = 4.0
delr = 0.1
bkgd_dyn=1

截图1.png

截图3.png

截图4.png

Al2O3_Al__Effect of porosity on the interface behavior of an Al2O3–aluminum composite A molecular dynamics study.pdf (546 KB)

Baskes - .pdf (5.08 MB)

I don’t understand your comment of “a pity for such a famous potential”…

Anyway, Streitz-Mintmire and MEAM are not related and using one does not necessarily require another.

Your subject title refers to S-M, but you are really asking a question about MEAM. I suggest you open a new thread with a proper title to get maximum attention.

Ray

截图4.png

截图1.png

截图3.png