To get the code, go to release page on Github to get the stable version (v 2.1.0).

A nuclear motion quantum mechanics solver library written in C++.

This project is a set of toy codes written by Yu Zhai.

In the past, I wrote yDVR as a set of functions. There are many reasons, but the core one is that computational chemists prefer Procedural programming (?).

The drawback of procedural programming is the interface can be very complex and impossible to remember, even hard to check. E.g., in yDVR 1.3, the following function is used to perform a PODVR computation

int sincPODVR1d(yScalar m, int N, int NPO, yScalar a, yScalar b,
    const DoubleFunction1d& potential, yVector& x, yVector& E,
    yMatrix& wf, yMatrix& H_PODVR);

Even myself is driven crazy when using it in my new projects.

Now in yDVR 2.x, you can use the following code

Oscillator co(m, pot);
SincDVR sincdvr(co,1.33,4.3,100 );
PODVR podvr(sincdvr, 10);

The readability has been improved and if you notice, now PODVR calculation can based on DVRs other than sinc-DVR because PODVR is now a decorator. (Yes!) Although in version 2.0 there is only sinc-DVR available, other representations can be easily added in the future.

So you will not able to generate a list of potential point and then run yDVR directly. By default, now yDVR is a shared library libydvr.so. You have to write a piece of C++ code and call yDVR library to perform the calculation.

Is that more difficult to use? I think the readability of the input file (If we stick to the concept) is much improved thank to the expressibility of C++ language. You can read the src/podvr_test.cc as an example.

Some suggestions are given for those who are not familiar with modern C++ (Because users are more likely to be physical chemists rather than professional C++ programmers).

• Potential function is used as argument of Oscillator class constructor. It can be any callable objects, including functions, lambda expressions, and class with operator() overloaded. I suggest you to use lambda expression to convert the units in-line.
• Reference is what I like, and when design the interface, I avoid using pointers. The good part of reference is that the user can release themselves from * & -> . thing, and for most of time you can directly use stack variables instead of new-ing something from heap. Therefore, you do not need to do the memory management.

Now we have OrthogonalMDDVR class. It can handle multi-dimension DVR problem.

In this project cmake is used as the build system. For those who not familiar with cmake, the following instruction is made.

cd /path/to/yDVR
mkdir build
cd build
ccmake .. # and chage the parameters as you wish
make
make install

This project is experimental with no warranty.

Here I thank the Eigen project. In 1.x version I used TinyXML-2 project to deal with I/O.

Also thank Prof. Hui Li. He makes me aware that this field is very interesting.

• Do not work well with PGI compiler pgc++ 18.4-0 64-bit target on x86-64 Linux -tp haswell because the project is based on Eigen. As a computational chemist I prefer leave this problem to Eigen development team ;-)