Problem with epsilon-input- file

Hi Steven,
I'm simulating a real PCF by using SEM images of the devices, then I 
extract the profile from the image,
I create the H5 file and use (set! epsilon-input-file) to pass the 
profile to the solver. Nevertheless, I get  a message of "Segmentation 
fault" in some cases,
even if the H5  dataset is not so large  (about 350 x 256) and my  PC 
has 2 GB of RAM...
This is not a systematic error, if I change  some parameters 
(resolution, size...)  MPB  then accepts the input  file. I highlight
that I sometimes get  "Segmentation fault" even if  the grid size 
matches with the dimensions of the H5 file dataset.
Do you have any idea of the origin of this problem?

Thank you very much,

Andrea

--

-- 
--------------------------------------------------------------------
Andrea Locatelli, Research Associate,
Dipartimento di Elettronica per l'Automazione, Università di Brescia
via Branze, 38 25123 Brescia (Italy)
Phone: +39-030-3715447
Fax: +39-030-380014
E-mail: andrea.locatelli <at> ing.unibs.it
Web: http://nora.ing.unibs.it/loca.htm
--------------------------------------------------------------------

Re: Problem with epsilon-input- file

I haven't seen this before; can you send me an HDF5 dataset and a minimal 
.ctl file that causes it to crash on your machine?

Steven

On Thu, 1 Sep 2005, Andrea Locatelli wrote:
> I'm simulating a real PCF by using SEM images of the devices, then I 
> extract the profile from the image, I create the H5 file and use (set! 
> epsilon-input-file) to pass the profile to the solver. Nevertheless, I 
> get a message of "Segmentation fault" in some cases, even if the H5 
> dataset is not so large (about 350 x 256) and my PC has 2 GB of RAM... 
> This is not a systematic error, if I change some parameters (resolution, 
> size...)  MPB then accepts the input file. I highlight that I sometimes 
> get "Segmentation fault" even if the grid size matches with the 
> dimensions of the H5 file dataset. Do you have any idea of the origin of 
> this problem?

2D tri-hole MPB

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

2-D PCs slab

Dear all,

I am confused when I was looking at the slab-hole.ctl file. For me it
seems that the structure is :

substrate: air, centered at (0,0,1) with thickness of 2

Silicon slab: centered at (0,0,0) with thickness of 0.5

air holes: centered at (0,0,0) periodic distributed, with height of 4

I thought the structure shall be

(define-param h 0.5) ; the thickness of the slab
(define-param eps 12.0) ; the dielectric constant of the slab
(define-param loweps 1.0) ; the dielectric constant of the substrate
(define-param r 0.3) ; the radius of the holes
(define-param supercell-h 4) ; height of the supercell

(set! geometry
      (list (make block (material (make dielectric (epsilon loweps)))
                  (center 0 0 (* 0.25 supercell-h))
                  (size infinity infinity (* 0.5 supercell-h)))
            (make block (material (make dielectric (epsilon eps)))
                  (center 0) (size infinity infinity h))
            (make block (material (make dielectric (epsilon loweps)))
                  (center 0 0 (* -0.25 supercell-h))
                  (size infinity infinity (* 0.5 supercell-h)))
            (make cylinder (material air)
                  (center 0) (radius r) (height supercell-h))
))

I know I must have misunderstood the program and hopefully someone could
help me out.

Also I am wondering if it would be possible to draw the epsilon file and
the field distribution (cross sections). It seems that the comment such as
"h5topng -S 1 epsilon.h5" can not be used.

Thanks,
Mindy

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

Re: 2-D PCs slab

On Fri, 2 Sep 2005, Mindy Lee wrote:
> I am confused when I was looking at the slab-hole.ctl file. For me it
> seems that the structure is :
[...]
> I thought the structure shall be
>
> (set! geometry
>      (list (make block (material (make dielectric (epsilon loweps)))
>                  (center 0 0 (* 0.25 supercell-h))
>                  (size infinity infinity (* 0.5 supercell-h)))
>            (make block (material (make dielectric (epsilon eps)))
>                  (center 0) (size infinity infinity h))
>            (make block (material (make dielectric (epsilon loweps)))
>                  (center 0 0 (* -0.25 supercell-h))
>                  (size infinity infinity (* 0.5 supercell-h)))
>            (make cylinder (material air)
>                  (center 0) (radius r) (height supercell-h))
> ))

Why do you think that you need to add the substrate twice?  It only 
requires a single block below the slab, as in the hole-slab.ctl example 
file included with MPB.

> Also I am wondering if it would be possible to draw the epsilon file and
> the field distribution (cross sections). It seems that the comment such as
> "h5topng -S 1 epsilon.h5" can not be used.

h5topng can certainly be use to draw the dielectric function and the field 
distribution cross-sections.  Why do you think they can't?

Note that for a 3d structure you will need to specify which slice of the 
3d dataset you want h5topng to output.  e.g. to output the z=0 slice, do:

 	mpb-data -r -n 32 epsilon.h5
 	h5topng -0z0 epsilon.h5:data-new

Note that the mpb-data command is required if you want an undistorted 
image, since the epsilon.h5 file is stored for a triangular lattice as 
output by MPB.  See the data analysis tutorial in the manual.

Cordially,
Steven G. Johnson

ERROR: unbound variable: include [brand new Debian installation without Guile]

Hello there,

I have the same problem mentioned before (see previous message):

> Dear MPB users,
> > When I try to run the MPB-1.4.2, the following error is coming.
> >
> > ERROR: In procedure scm_lookup:
> > ERROR: unbound variable: include
>
> This error message about scm_lookup has appeared before on the mailing
> list, and invariable comes because a person has multiple, incompatible
> versions of Guile installed.
>
> Probably, you are using a Linux system that had Guile already
> installed, and you then installed *another* version from source.
>
> What you need to do is to uninstall the versions you installed from
> source ('make uninstall' if you still have the source directory lying
> around), and then install the 'guile-devel' RPM for your Linux
> distribution (as described in the MPB installation manual, both guile
> and guile-devel RPMs are required, but only the former is typically
> installed).
>
> Cordially,
> Steven G. Johnson
>

However, I have a new Debian system (downloaded CDROMs last week),
which did notinstall any other Guile version.  

The version of Guile I installed from sources is 1.6.7 , 
which works ok by itself.

I can execute all the C examples in the subdirectory 
mpb-1.4.2/mpb-ctl/examples

The problem seems to be between guile-1.6.7 and mpb-1.4.2, 
After setting guile variable GUILE_WARN_DEPRECATED to "detailed" 
it displays:

'scm_symbol_value()' deprecated. Use 'scm_lookup' instead.
ERROR: In procedure scm_lookup:
ERROR: unbound variable: include

Is there any way around this glitch to keep using guile-1.6.7 ?
Are the RPM packages still needed for Debian ?
Can you recommend a Guile version that works with Debian for sure ?

Thanks for your assistance.
Ian.

Re: ERROR: unbound variable: include [brand new Debian installation without Guile]

On Sat, 3 Sep 2005, Ian Wright wrote:
> However, I have a new Debian system (downloaded CDROMs last week),
> which did notinstall any other Guile version.

I find that hard to believe, since I thought the Guile libraries were
installed by default as a part of the core system in Debian.

What the default installation is missing is the header files, which are 
installed if you do:
 	apt-get install guile-dev

> The version of Guile I installed from sources is 1.6.7 ,
> which works ok by itself.

Aaaagh.

The entire point of using Debian is that you shouldn't have to install 
practically anything from source.  In fact, MPB is included with the 
latest Debian.  If you just type:

 	apt-get install mpb

it will install MPB and all its dependencies.

Given that you have installed goodness-knows-what from source, however you 
may have screwed up your system beyond your ability to recover from; it 
might be easier for you to wipe your drive and re-install the whole 
system.

> Is there any way around this glitch to keep using guile-1.6.7 ?
> Are the RPM packages still needed for Debian ?

Debian has its own package format and does not normally use RPM.  You 
should hardly ever have to download packages manually, much less install 
from source.  Please read the Debian documentation.  You might want to 
purchase a book such as "Debian GNU/Linux Bible" (from Wiley).

> Can you recommend a Guile version that works with Debian for sure ?

Yes, the version of Guile that comes with the system (like every other 
Linux distribution).

Steven

How to realize perfect metal in MPB?

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

Re: How to realize perfect metal in MPB?

On Sun, 4 Sep 2005, Xianyu Ao wrote:
> Can one calculate the photonic bands of photonic crystals consisting of 
> perfect metal rods by using MPB?

No.

Re: 2-D PCs slab

If my structure is asymmetric, then do I write the structure like:

(set! geometry
      (list (make block (material (make dielectric (epsilon 1.2)))
                  (center 0 0 (* 0.25 supercell-h))
                  (size infinity infinity (* 0.5 supercell-h)))
            (make block (material (make dielectric (epsilon eps)))
                  (center 0) (size infinity infinity h))
            (make block (material (make dielectric (epsilon 1)))
                  (center 0 0 (* -0.25 supercell-h))
                  (size infinity infinity (* 0.5 supercell-h)))
            (make cylinder (material air)
                  (center 0) (radius r) (height supercell-h))
 ))

?

Thanks,
Mindy

> On Fri, 2 Sep 2005, Mindy Lee wrote:
>> I am confused when I was looking at the slab-hole.ctl file. For me it
>> seems that the structure is :
> [...]
>> I thought the structure shall be
>>
>> (set! geometry
>>      (list (make block (material (make dielectric (epsilon loweps)))
>>                  (center 0 0 (* 0.25 supercell-h))
>>                  (size infinity infinity (* 0.5 supercell-h)))
>>            (make block (material (make dielectric (epsilon eps)))
>>                  (center 0) (size infinity infinity h))
>>            (make block (material (make dielectric (epsilon loweps)))
>>                  (center 0 0 (* -0.25 supercell-h))
>>                  (size infinity infinity (* 0.5 supercell-h)))
>>            (make cylinder (material air)
>>                  (center 0) (radius r) (height supercell-h))
>> ))
>
> Why do you think that you need to add the substrate twice?  It only
> requires a single block below the slab, as in the hole-slab.ctl example
> file included with MPB.
>
>> Also I am wondering if it would be possible to draw the epsilon file and
>> the field distribution (cross sections). It seems that the comment such
>> as
>> "h5topng -S 1 epsilon.h5" can not be used.
>
> h5topng can certainly be use to draw the dielectric function and the field
> distribution cross-sections.  Why do you think they can't?
>
> Note that for a 3d structure you will need to specify which slice of the
> 3d dataset you want h5topng to output.  e.g. to output the z=0 slice, do:
>
>  	mpb-data -r -n 32 epsilon.h5
>  	h5topng -0z0 epsilon.h5:data-new
>
> Note that the mpb-data command is required if you want an undistorted
> image, since the epsilon.h5 file is stored for a triangular lattice as
> output by MPB.  See the data analysis tutorial in the manual.
>
> Cordially,
> Steven G. Johnson
>