[Up][Tutorial main page]

Mini spectrum calculation

So far you have performed spectroscopy on a tiny configuration which is close to free field. To enable you to perform a slightly more realistic calculation we have generated 20 configurations in the quenched approximation at &beta=5.8 (which corresponds to a lattice spacing of approximately 0.145 fm) with lattice size 12³x24.

You should keep the quark mass heavy so that chroma takes only a few minutes or less to run. This should be the case if you choose kappa=0.109 for the Wilson action, which corresponds to roughly the charm quark mass. For the clover action a similar quark mass is obtained using kappa=0.103 (where the nonpertubative value for the clover coefficient is roughly 2.14).

The aim is to generate meson and/or baryon correlators on the 20 configurations, strip the XML data files using strip_hadspec and then to calculate the effective masses. If you don't have time to generate propagators for all the configurations, you can use some that we have already computed. In this case, your input XML only has to have the inline measurements for reading a propagator in and then performing the sink smearing before calculating the spectrum.

They are saved in SZINQIO format and so you will need to use this as the key for the <cfg_type> tag in the input file.

The smearing parameters that were used are the following:

• Gauge invariant Gaussian smearing: wvf_param=2.8284, wvfIntPar=20.
• APE link smearing: link_smear_fact=2.5, link_smear_num=10.

Running on a single configuration

This should be straightforward. Now that the inversion takes the order of 20 seconds or more it is easier to see the differences in timings if you use

• a different inverter - BiCGStab instead of CG,
• preconditioning - or not,
• a different action.

Try it out if you have time. You could also decrease the quark mass (increase kappa) to make the differences even more noticable. For the Wilson case, kappa_critical (roughly) = 0.1615 (corresponding to when the quark mass is zero) and for clover, kappa_critical (roughly) = 0.1317. You should not use a kappa value close to or higher than kappa_critical.

Running on 20 configurations

For 20 configurations you need 20 input files. It is tedious to create these by hand and normally people develop scripts to do it automatically. Since you can also use a script to run chroma (20 times), you can also generate the input file in the same script.

You can have a look at the bash scripts which were used to generate some of the propagators (and correlators) here:

• propSEED.ini.xml - this is the generic input file.
• create_xml.sh - this creates all the input files by using the sed command to replace the string CONFIGNUMBER, which appears in propSEED.ini.xml, with a configuration number. For the configurations we generated the configuration number runs from 1 to 20.
• invert_all.sh - this runs chroma for each of the 20 configurations.

If you download these files, remember to make the script files (those ending in .sh) executable with the chmod command,

before running them. You will also have to modify the scripts and input XML, for example, to have the correct location of the configurations and also your chroma executable if you want to use them to generate results.

Stripping the data files

To strip the 20 data files containing the correlators into simple text format use the strip_hadspec command:

where "hadspec.dat.xml1 hadspec.dat.xml2 ... hadspec.dat.xml20" is the list of data files (replace the ... with the missing 17 file names).

Calculating the effective masses

You can now obtain a crude estimate of the masses of the particles you have simulated by calculating effective masses. Denoting the correlator as C[t], the effective mass is given by m[t+0.5]=-log(C[t+1]/C[t]).

Download the effective mass progam: effmass.cc. It can be compiled using the command:

To run the program use

where pion.D8544.DG2p8284_1.DG2p8284_1.SS is the correlator file. You will have generated 2 files with the names:

• avg_pion.D8544.DG2p8284_1.DG2p8284_1.SS - contains the correlator, averaged over the configurations.
• effmass_pion.D8544.DG2p8284_1.DG2p8284_1.SS - contains the effective mass.

The files have the format:

0.5 2.97109 0.0192236
1.5 2.29875 0.0199504
2.5 2.2993 0.0189994
3.5 2.25501 0.0181033
4.5 2.23543 0.0194734
.
.
.

the time, the average and the error.

You can plot the file, for example, with gnuplot. Start gnuplot with the command

and then plot the file with

Before plotting the average of the correlator you can set the y axis to log scale using the command

For more information on gnuplot, look at the gnuplot website.

You can compare how the effective masses look for the correlators that have different smearings, for example, a smeared sink or a point sink.

[Up][Tutorial main page]