headers
Liu, Hao | 11 Feb 2002 21:59
Picon

How to deal with a 2D lattice of dielectric rod with a finite hei ght

Hi,  for a 2D lattice of dielectric rod with a finite height in the
z-direction, like a the 2D example in p.16 of the MPB manual, seems we can
simply set the HEIGHT to a finite number in the (set! geometry list...).
While in the (set! grid-size (vector3 32 32 1)), we should replace 1 with a
number denoting the segment size in the z-direction. What else parameters
should be modified accordingly, to adapt the new problem? 
In addition, what if for a problem when the height is very small compared
with the wavelength? For example, when a poforated thin film is involved
where a single segment is quite fit to model the variation of the field in
the z-direction.  Also, thanks for your response for my previous question. 

Regards,

Hao Liu
Posdoc Visiting Fellow
Technologies spatiales/Space Technologies
3A-345
Agence spatiale canadienne/Canadian Space Agency
6767, route de l'Aéroport, St-Hubert, QC J3Y 8Y9
Tél.: (450) 926-6757
hao.liu <at> space.gc.ca

_______________________________________________
mpb-discuss mailing list  -  mpb-discuss <at> ab-initio.mit.edu
http://ab-initio.mit.edu/mailman/listinfo/mpb-discuss
Permalink | Reply | 1 comment |
headers
Steven G. Johnson | 11 Feb 2002 23:55
Picon

Re: How to deal with a 2D lattice of dielectric rod with a finite height

On Mon, 11 Feb 2002, Liu, Hao wrote:
> Hi, for a 2D lattice of dielectric rod with a finite height in the
> z-direction, like a the 2D example in p.16 of the MPB manual, seems we
> can simply set the HEIGHT to a finite number in the (set! geometry
> list...). While in the (set! grid-size (vector3 32 32 1)), we should
> replace 1 with a number denoting the segment size in the z-direction.
> What else parameters should be modified accordingly, to adapt the new
> problem?

You'll also need to change the vertical cell size to some appropriate
supercell.  See, for example, the hole-slab.ctl example file, which models
a finite-height 2d lattice of air holes in a dielectric slab.

> In addition, what if for a problem when the height is very small compared
> with the wavelength?

In that case, what you will find is that the slab modes are only very
weakly guided; there may be only one guided band just beneath the light
line.  (Weak guiding also implies that a larger vertical supercell is
required.)

For more information on the issues of vertical guiding in such structures,
the light line, and the optimal thickness, see for example my paper in PRB
60, p. 5751 (1999)
(http://ojps.aip.org/journal_cgi/getabs?KEY=PRBMDO&cvips=PRBMDO000060000008005751000001&gifs=Yes)

Cordially,
Steven G. Johnson

_______________________________________________
(Continue reading)

Permalink | Reply | 0 comments |
headers
Bertrand P. Boussert | 12 Feb 2002 18:47
Picon
Favicon

enable parallel computing

Hello,

We install mpb on a bi-processor linux cluster. We ran the 
compilation with the options enable-thread and enable-mpi. However we 
are only working on a bi-processor environnement and do not use the 
cluster. Could you confirm that the two options need to be used 
simultaneously. In the documentation, the option enable-mpi is 
specified with a type. What is the value of this type?

Many thanks,

Bertrand Boussert
--

-- 
Bertrand P. Boussert, PhD
Georgia Tech Lorraine	Phone: +33 38 720 3947
Technopole Metz 2000	Secretary: +33 38 720 3939
2/3 rue Marconi		Fax: +33 38 720 3940
57070 Metz, France	Email: bertrand.boussert <at> ece.gatech.edu

Adjunct Assistant Professor of School of Electrical and Computer Engineering
Georgia Institute of Technology
Research Engineer
Georgia Tech Lorraine/GTL-CNRS Telecom Laboratory

_______________________________________________
mpb-discuss mailing list  -  mpb-discuss <at> ab-initio.mit.edu
http://ab-initio.mit.edu/mailman/listinfo/mpb-discuss
Permalink | Reply | 1 comment |
headers
Michael Zedler | 25 Feb 2002 09:47
Picon
Picon

perfect conductors

Hello,

I'd like to know, if perfect conductors will be supported in the (not so
distant) future. If this were true, would this include surface currents?

Thanks in advance,
Michael Zedler

_______________________________________________
mpb-discuss mailing list  -  mpb-discuss <at> ab-initio.mit.edu
http://ab-initio.mit.edu/mailman/listinfo/mpb-discuss
Permalink | Reply | 1 comment |
headers
Steven G. Johnson | 25 Feb 2002 18:01
Picon

Re: perfect conductors

On Mon, 25 Feb 2002, Michael Zedler wrote:
> I'd like to know, if perfect conductors will be supported in the (not so
> distant) future. If this were true, would this include surface currents?

At this point, I'm not sure how to support perfect conductors in a
frequency-domain code like MPB, although I've played with different
approaches.  It is not simply a matter of setting the fields to zero
inside the conductor, because in order to get good convergence it seems
that you need to do this in such a way that (a) it commutes with the
eigen-operator and (b) the eigen-problem remains positive-definite.

(These restrictions are lifted if you are not using iterative methods, and
people in the literature have taken advantage of this.  However, iterative
methods are required to practically treat large and/or 3d problems.  
There are also finite-element methods for treating conductors, but I'm not
planning on implementing them at this point, and they have their own
difficulties.  Another possibility is improved iterative methods, and I'm 
looking into this.)

Suggestions are appreciated, of course.

Cordially,
Steven G. Johnson

_______________________________________________
mpb-discuss mailing list  -  mpb-discuss <at> ab-initio.mit.edu
http://ab-initio.mit.edu/mailman/listinfo/mpb-discuss
Permalink | Reply | 0 comments |
headers
Michael Zedler | 26 Feb 2002 10:55
Picon
Picon

p4 performance, intel mkl, intel compilers

Hello,

I was just wondering if anybody has already tried compiling mpb using the
intel math kernel library and intel f90/c++ compilers? Can I expect a
significant speed gain?

Thanks in advance
Michael Zedler

_______________________________________________
mpb-discuss mailing list  -  mpb-discuss <at> ab-initio.mit.edu
http://ab-initio.mit.edu/mailman/listinfo/mpb-discuss
Permalink | Reply | 1 comment |
headers
Michael Zedler | 26 Feb 2002 19:05
Picon
Picon

Re: p4 performance, intel mkl, intel compilers


>On Tue, 26 Feb 2002, Michael Zedler wrote:
>> I was just wondering if anybody has already tried compiling mpb using the
>> intel math kernel library and intel f90/c++ compilers? Can I expect a
>> significant speed gain?
>
>Almost all of the performance of MPB is determined by the FFTW and BLAS
>libraries, so it does not matter much what compiler you use for MPB per
>se.  Last I checked, the Intel compiler did not improve FFTW much (if at
>all), but I don't know how it would affect a fast BLAS like ATLAS.

Is BLAS not implemented in the math kernel library
(http://www.intel.com/software/products/mkl/mklfeatures_new.htm)? Can one
expect this implementation to be superior to atlas?

regards,
Michael

_______________________________________________
mpb-discuss mailing list  -  mpb-discuss <at> ab-initio.mit.edu
http://ab-initio.mit.edu/mailman/listinfo/mpb-discuss
Permalink | Reply | 1 comment |
headers
Liu, Hao | 1 Mar 2002 18:04
Picon

honeycomb lattice air columns in dielectric

Hi, Have you ever tried honeycomb lattice air columns in dielectric?  Any
information is appreciated. 

Best Regards,

Hao Liu
Posdoc Visiting Fellow
Technologies spatiales/Space Technologies
3A-345
Agence spatiale canadienne/Canadian Space Agency
6767, route de l'Aéroport, St-Hubert, QC J3Y 8Y9
Tél.: (450) 926-6757
hao.liu <at> space.gc.ca

_______________________________________________
mpb-discuss mailing list  -  mpb-discuss <at> ab-initio.mit.edu
http://ab-initio.mit.edu/mailman/listinfo/mpb-discuss
Permalink | Reply | 1 comment |
headers
Allan, Douglas C Dr | 1 Mar 2002 14:27

testing of new installations and upgrades

> Dear Steven,
> A recent email exchange among Steve Lantz and you and me has prompted me to make this suggestion.  You
probably have some suite of tests that you run when you make changes to mpb, or especially, when someone
else contributes changes, to check for any obvious brokenness.  If you don't, my first suggestion is that
you should.  If you do, then my second suggestion is, would you take one or a small number of those tests and
make their ctl files and output files public?  I.e. on your web site?  I have in mind that you would always run
at least 1 or a small number of short jobs with every new build, and confirm that the answers are the same as
the last build, or differ for reasons that you understand.  Then you could keep the ctl files and output
files of the latest release on your web site for the benefit of users.  This is exactly the sort of test Steve
Lantz was trying to run.
> 
> For example, we have an extensive set of tests we run in the ABINIT project.  In that case the source
contributions come from all over the world, and we have very little control over the quality of what is
contributed, EXCEPT that we can insist that every new contribution gets the same results for all the
standard tests.  The testing is automated of course, as are the diffs against older results.  The same tests
can be run by someone installing a new build or an upgrade to see if something is broken.  This testing has
been extremely useful for us.
> 
> If you already do all this and I overlooked it on your web site, please accept my apology.
> 
> Regards,
> Doug Allan
> 

_______________________________________________
mpb-discuss mailing list  -  mpb-discuss <at> ab-initio.mit.edu
http://ab-initio.mit.edu/mailman/listinfo/mpb-discuss
Permalink | Reply | 1 comment |
headers
Steven G. Johnson | 1 Mar 2002 18:58
Picon

Re: honeycomb lattice air columns in dielectric

On Fri, 1 Mar 2002, Liu, Hao wrote:
> Hi, Have you ever tried honeycomb lattice air columns in dielectric?  Any
> information is appreciated. 

I have not tried this myself, but there is a honey-rods.ctl (honeycomb
lattice of rods in air) example file included with MPB that you could
easily adapt to try this out.

Steven

PS. I would venture a guess that the honeycomb lattice of air columns
would not have a complete gap (i.e. for both polarizations).  In the limit
where the holes are touching, it resembles a triangular lattice of rods,
which does not have a complete gap.  In contrast, the honeycomb lattice of
rods resembles a triangular lattice of air holes, which can have a
complete gap.

_______________________________________________
mpb-discuss mailing list  -  mpb-discuss <at> ab-initio.mit.edu
http://ab-initio.mit.edu/mailman/listinfo/mpb-discuss
Permalink | Reply | 0 comments |

Gmane