*Handouts: printout of deltashell_boundstates.cpp,
LAPACK fortran routine documentation
*

In this session, we'll consider the bound-state
Schrodinger equation in momentum
space, which takes the form of an integral equation.

*
Your goals for this session:
*

- Finish any interesting (to you!) parts of Session 15.
- Fill out the SET course evaluation.
- Look at an example of solving an integral equation, which is a momentum-space bound-state problem, using fortran library routines (from LAPACK) and the C++ "new" and "delete" functions

Thank you in advance for your comments, which are invaluable in planning future versions of this computational physics course. Some things I'd appreciate comments on:

- Please comment on the pace: too fast, too slow, not enough depth?
- Do the background notes work? (Do you need them further in advance?)
- Any topics you would omit? Any ones you would spend more time on? Any topics not covered you think should be covered?
- Do you think I should choose a text to go along with the course, even if we don't follow it closely? (Suggestions?)
- Would it be better to introduce C++ classes from the beginning?
- If you had a choice between computational physics courses taught entirely with Mathematica, entirely with C++, or entirely with Matlab, which would you choose? Or do you think a mix is more useful?
- As always, suggestions for improvements to any aspect of the course would be appreciated.

As discussed in the background notes, the bound-state Schrodinger
equation in momentum space can be solved as a matrix eigenvalue problem.
But because the matrix to be diagonalized is *NOT* symmetric, we
can't use the (limited) gsl routines. So we'll use LAPACK, a fortran
library, as an exercise in calling fortran routines from C++.
We'll also dynamically allocate the one- and two-dimensional arrays
using "new" and "delete".

The delta-shell potential is a convenient choice for a test potential because we can find the eigenvalues analytically and it takes a simple form in momentum space. It has only one bound state; if we only print out the negative-energy (E<0) eigenvalues then we should find just one. See the background notes for additional details.

- What are the units of lambda in the delta-shell potential?
[Note: hbar=1 here.]

- For mu=0.5, b=2.0, and lambda=-1.0, use Mathematica
to find the correct eigenvalue
for the (single) bound state in the delta-shell potential by
using the appropriate equation in the background notes. [Hint: FindRoot]
What is the answer?

- Look at deltashell_boundstates.cpp. Compare the prototype
declaration for the fortran routine dgeev to the documentation for
dgeev.f (both are given on printouts) to see how they match up.
Read the notes in the
documentation at the beginning of the deltashell_boundstates.cpp code.
Questions?

- Note how all but one of the arrays (the exception is Eigvec_left)
are dynamically allocated and removed using "new" and "delete".
Matrices are allocated using double pointers; note the construction.
However, to pass them to the fortran routine, we've "flattened" them
into single arrays (see how Hmat_passed is formed from Hmat).
What does flattening mean?

- Compile, link, and run deltashell_boundstates.cpp. Look at the
makefile to see what libraries are linked. The "-lg2c" is essential
when using g++
but easy to overlook. Run with lambda=-1.0 but specify on your own
the number of gauss points to use and the upper bound for the
integrations of momentum. The latter may be much larger than you
would guess because the delta-shell potential is actually rather
poorly behaved in momentum space!
Based on how the potential falls off with k, what do you think is a
reasonable momentum cutoff? (What should you do to discover how the
potential falls off?)

- Run the code
with different numbers of gauss points and different momentum
cutoffs to try to improve the predicted eigenvalue.
Estimate the precision of your best value for the bound-state
energy.
How could you improve the numerical accuracy of the calculation?
(For example, would it help to change "job" when calling "gauss"? [Hint,
hint.])

- What does the momentum-space wave-function look like (it is
printed out to a file)?
- Does it fall off at large k? Does it oscillate? Is it well behaved
at the origin? Can you account for its characteristics by some simple
arguments?

- The analytic wave function in coordinate space is given in the background notes. Try to calculate the momentum space wave function from this (as discussed in the notes) using Mathematica.
- Is the momentum wave function normalized? What is the
normalization condition?

- Does it fall off at large k? Does it oscillate? Is it well behaved
at the origin? Can you account for its characteristics by some simple
arguments?
- Bonus: Verify that, regardless of the potential's strength, there is only a single bound state.
- Bonus: Modify the code to calculate the l=1 bound state.

furnstahl.1@osu.edu