3 Jul 2002 03:37

### Re: Using mpb for quasiperiodic lattices

```
Hi, I was wondering about using mpb to compute the band structures for
quasiperiodic lattices, in addition to periodic ones.  It looks like mpb
usually takes as its input up to 3 lattice vectors, and then generates a
(infinite?) lattice for the calculation.  The exception seems to be
when, say, you're trying to find a state in a point or line defect in
to create a supercell of finite size around the lattice.  Is
this generally true?
By the way, is there any limit to the size of a
supercell you could create around, say, a square lattice of dialectric
rods in 2-D (I think the example on the web has 5 x 5 rods--could you make
this, say, 50 x 50 rods if you wanted, or would it be even worth it, and
how could you tell?)
A quasiperiodic crystal is specified by more than 3 integer linearly
independent basis vectors in 3-D, and more than 2 integer linearly
independent basis vectors in 2-D.  I've got a program that generates these
crystals (it takes as its inputs the kind of lattice geometry you
want, specified by the basis vectors and lattice size, and outputs the
lattice vertices), and I was wondering: would it be possible to get mpb to
read in all the vertices and place objects at them (say spheres, or
cylinders), and then compute the band gaps?  Would there be some
"non-tedious" way for these vertices to be entered in to a list, or would
I have to say, "(make sphere (center x y z)" for each vertex?
The other thing I was thinking about was modifying mpb so that it could
handle more basis vectors and generate quasiperiodic structures, but I'm
getting the feeling that this wouldn't be too easy, is that right?

I had another question: I saw in the data tutorial on the
web the example of making a diamond lattice of spheres, and finding the
```

3 Jul 2002 04:40

### Re: Re: Using mpb for quasiperiodic lattices

```On Tue, 2 Jul 2002, Orion Crisafulli wrote:
> Hi, I was wondering about using mpb to compute the band structures for
> quasiperiodic lattices, in addition to periodic ones.  It looks like mpb
> usually takes as its input up to 3 lattice vectors, and then generates a
> (infinite?) lattice for the calculation.

More accurately, the boundary conditions are (Bloch) periodic, so the
lattice isn't (directly) generated.n

> By the way, is there any limit to the size of a supercell you could
> create around, say, a square lattice of dialectric rods in 2-D (I
> think the example on the web has 5 x 5 rods--could you make this, say,
> 50 x 50 rods if you wanted, or would it be even worth it, and how
> could you tell?)

It is possible to do 50x50 rods, although it would be a big computation
(e.g. you'd have to compute 250 bands before you got up to the first gap,
although the targeted solver ameliorates this).

> A quasiperiodic crystal is specified by more than 3 integer linearly
> independent basis vectors in 3-D, and more than 2 integer linearly
> independent basis vectors in 2-D.  I've got a program that generates
> these crystals (it takes as its inputs the kind of lattice geometry
> you want, specified by the basis vectors and lattice size, and outputs
> the lattice vertices), and I was wondering: would it be possible to
> get mpb to read in all the vertices and place objects at them (say
> spheres, or cylinders), and then compute the band gaps?

Sure, although you'd need a big supercell, and thus a very big
calculation, to represent a large section of the quasi-crystal.
```

4 Jul 2002 17:48

### installation problem with libctl

```Dear Collegues,

after several tries to install MPB on a DEC alpha
running under LINUX this list is my last hope.
Since I do not have su-privileges I did install
all required packages in subdirs of my homedir.
Although guile is recognized and the libctl library
is in place I get during the config-run of mpb
the following error:

checking for guile... yes
checking for guile-config... yes
checking if linking to guile works... yes
checking for libctl dir... /home/fw/LIBCTL/share/libctl
checking for gen-ctl-io... yes
checking for ctl_get_vector3 in -lctl... no
configure: error: Couldn't find the required libctl library.

The configuration routine was started with following switches:
./configure --prefix=/home/fw/MPB --with-hdf5=/home/fw/HDF
--with-libctl=/home/fw/LIBCTL/share/libctl

LDFLAGS=-L/home/fw/FFTW/lib -L/home/fw/LIBCTL
-L/home/fw/LIBCTL/share/libctl

The library directory looks like:
[xp1] ~/LIBCTL/share/libctl-> ls
base  lib  utils
```

5 Jul 2002 09:48

### Re: installation problem with libctl

```Hi there,

I have the same problem before. I recompiled everything with the
same compilers and it works for me. Hope that helps.

Ze

On Thu, 4 Jul 2002 weik <at> mbi-berlin.de wrote:

> Dear Collegues,
>
> after several tries to install MPB on a DEC alpha
> running under LINUX this list is my last hope.
> Since I do not have su-privileges I did install
> all required packages in subdirs of my homedir.
> Although guile is recognized and the libctl library
> is in place I get during the config-run of mpb
> the following error:
>
> checking for guile... yes
> checking for guile-config... yes
> checking if linking to guile works... yes
> checking for libctl dir... /home/fw/LIBCTL/share/libctl
> checking for gen-ctl-io... yes
> checking for ctl_get_vector3 in -lctl... no
> configure: error: Couldn't find the required libctl library.
>
> The configuration routine was started with following switches:
> ./configure --prefix=/home/fw/MPB --with-hdf5=/home/fw/HDF
```

5 Jul 2002 17:52

### Re: installation problem with libctl

```On Thu, 4 Jul 2002 weik <at> mbi-berlin.de wrote:
> checking for ctl_get_vector3 in -lctl... no
> configure: error: Couldn't find the required libctl library.
>
> The configuration routine was started with following switches:
> ./configure --prefix=/home/fw/MPB --with-hdf5=/home/fw/HDF
>               --with-libctl=/home/fw/LIBCTL/share/libctl

First, you're making life hard for yourself by installing everything in
its own directory.  This is Unix, and it is not oriented towards 1
directory per program.  Just install everything under /home/fw/install or
something like that, and you'll only need to add a single directory to

> The linker environment was set:
> LDFLAGS=-L/home/fw/FFTW/lib -L/home/fw/LIBCTL
>                  -L/home/fw/LIBCTL/share/libctl

bzzt.

LDFLAGS="-L/home/fw/FFTW/lib -L/home/fw/LIBCTL/lib -L/home/fw/HDF/lib ..."

LDFLAGS should only be the lib directories, where the libraries
reside.  (Notice how easier it would have been to have just
-L/home/fw/install/lib)

Note that you'll also need to add the various include directories to
installation section of the MPB manual.

```

5 Jul 2002 19:24

### cygwin

```
Hello,

I'd like to know, if somebody was able to compile mpb on cygwin. Up to now I
could make guile, libctl and fftw work, but I experience problems with hdf5,
lapack, blas and atlas.
Anybody out there who has a statically linked mpb.exe? ;)

Michael Zedler

_______________________________________________
mpb-discuss mailing list  -  mpb-discuss <at> ab-initio.mit.edu
http://ab-initio.mit.edu/mailman/listinfo/mpb-discuss

```
6 Jul 2002 03:53

### Re: cygwin

```On Fri, 5 Jul 2002, Michael Zedler wrote:
> I'd like to know, if somebody was able to compile mpb on cygwin. Up to now I
> could make guile, libctl and fftw work, but I experience problems with hdf5,
> lapack, blas and atlas.

Yes, a couple of people have reported success with CygWin.

> Anybody out there who has a statically linked mpb.exe? ;)

If someone wants to put one together, I'd be happy to post it on the MPB
web page.  However, it would require more than just an mpb.exe, because
Guile and MPB need to load several Scheme files upon startup.

Steven

_______________________________________________
mpb-discuss mailing list  -  mpb-discuss <at> ab-initio.mit.edu
http://ab-initio.mit.edu/mailman/listinfo/mpb-discuss

```
12 Jul 2002 09:17

### scampi

```Hi !

With the help of our PC-Gurus I was able to compile mpb-mpi for scampi
First our mpicc had a bigger bug, but then it works.
But mpicc for scampi has still a problem

The PC-Gurus told me that

-DVERSION_STRING="MIT Photonic Bands 1.3, Copyright (C) 1999-2001 by MIT."
in the Makefile for main program has unusual quotings. We've changed it
and all works fine

VERSION_STRING = "MIT Photonic Bands 1.3, Copyright (C) 1999-2001 by
MIT."
-DVERSION_STRING='"'\$(VERSION_STRING)'"'

bye
Andreas

>
> greulich <at> psc-master:/homes2/psc-master/phokri/mpb-1.3/mpb-ctl>make
> ...
> mpicc  -L/homes2/psc-master/phokri/lib -g -O2 -Wall -W -Wbad-function-cast
> -Wcast-qual -Wpointer-arith -Wcast-align -pedantic main.o ctl-io.o mpb.o
> epsilon_file.o geom.o      ../src/maxwell/maxwell.a ../src/matrixio/matrixio.a
> ../src/matrices/matrices.a ../src/util/util.a -lctl
> -L/homes2/psc-master/phokri/GUILE/lib -lguile -lm -lhdf5 -lz -lfftw_mpi
> /homes2/psc-master/phokri/LAPACK/lapack_LINUX.a
> /homes2/psc-master/phokri/LAPACK/blas_LINUX.a -lfftw -lm
> -L/homes2/psc-master/phokri/lib -L/usr/lib/gcc-lib/i686-pc-linux-gnu/2.95.3
```

15 Jul 2002 16:22

### Re: cygwin

```Dear Michael,
Do You still need an assistance? I just compiled all You need for cygwin and I
Best Wishes,
Pavel S. Ivanov

Michael Zedler wrote:

> Hello,
>
> I'd like to know, if somebody was able to compile mpb on cygwin. Up to now I
> could make guile, libctl and fftw work, but I experience problems with hdf5,
> lapack, blas and atlas.
> Anybody out there who has a statically linked mpb.exe? ;)
>
> Michael Zedler
>
> _______________________________________________
> mpb-discuss mailing list  -  mpb-discuss <at> ab-initio.mit.edu
> http://ab-initio.mit.edu/mailman/listinfo/mpb-discuss

_______________________________________________
mpb-discuss mailing list  -  mpb-discuss <at> ab-initio.mit.edu
http://ab-initio.mit.edu/mailman/listinfo/mpb-discuss

```
17 Jul 2002 16:44

### h5utils under Cygwin

```Hello  Steven, Hello all !

I installed mpb 1.3 and another packages mentioned in the section
'Installation' of MPB 1.3 manual under latest Cygwin.

I have a problem with h5utils. I tried versions 1.7.1 and 1.7.2. I
receive an error, then I try to convert the file 'epsilon.h5' to file in
vis5d, text of png formats.
I am attaching 'config.log' file. During the compilation I receive the
following massage:

h5tov5d.c: In function `output_v5d':
h5tov5d.c:234: warning: passing arg 7 of `v5dCreate' from incompatible
pointer type

And after running 'h5tov5d.exe epsilon.h5' I received the message:

Segmentation fault (core dumped)

The file 'StackDump' is also attached. Is it possible to solve this
problem ? Or, may be, anybody can suggest me another way for
converting/viewing of files in h5d format?
Yours,
Pavel S. Ivanov
```
Attachment (config.log): application/x-unknown-content-type-logfiletype, 11 KiB
```Exception: STATUS_ACCESS_VIOLATION at eip=61080570
eax=3D357EC6 ebx=85B03790 ecx=00000000 edx=00000000 esi=0003A5F9 edi=0080D9E0
ebp=0080D9A8 esp=0080D980 program=C:\CYGWIN\USR\LOCAL\BIN\H5TOV5D.EXE
```