6.2 KiB
LEM single-histogram example
This worked example repeats the LEM 2025 energy scans using
single-histogram fits, not asymmetry fits. The original
files in /tmp/lemfit were asymmetry fits and were used only
for initial parameter estimates. They were not overwritten.
Data and model
- Data set 1: runs 1014-1027, 3.8 K, blue circles.
- Data set 2: runs 1033-1045, 10 K, red squares.
- Sample: 85 nm Nb/Al2O3; nominal field 100 G; energy approximately 1-18 keV.
- Fit type: Single histogram
(
fittype 0), groups21 25and23 27; nominal fit interval 0-8 microseconds, packing 100.
Both count-histogram groups are fitted together for each physical run. The polarization model is
Pj(t) = A0exp [−(σt)2/2]cos (2πγμBt + ϕj).
The MSR theory is asymmetry, simpleGss,
TFieldCos with fun1 = gamma_mu * par3, where
parameter 3 is Bfield. Thus field is fitted directly in G,
rather than converted from a plotted frequency later. Sgm
is in inverse microseconds and Asy is dimensionless.
Each group has its own phase, normalization and constant background.
Asy, Sgm and Bfield are common to
the two groups within one physical run. No parameters
are shared between physical runs: these are 27
independent fits, not one global fit across the energy scan.
Note
No musrt0 -g correction or explicit t0/data override was
applied. musrfit reads the LEM timing information and reports its
automatic data range fallback (start t0 + 51 bins, end 66601). The
nominal interval is 0-8 microseconds; the fallback excludes
approximately the first 10 ns. The retained command logs record these
warnings.
Inspect the fitted run
The screenshots below show the actual application after importing the
fitted 1014_his.msr. The fits themselves were executed with
the command-line script described below; the screenshots do not imply a
browser fit job was executed. Importing this file also restores its
recognized GUI settings.
musrview 1014_his.msr --png output.
PLOT 0 contains runs 1 2 and
lifetimecorrection; both histogram groups are
shown.Compare the energy scans
- Fit each run range independently and export one parameter table per
scan. The supplied
dataset1.datanddataset2.datcontain the new fits. - Open Parameter plots, load the first table and add
the second as an overlay. Select
dataE(implantation energy in keV) for X. - Use a 3 x 1 layout, with
Bfield,SgmandAsyin separate panels. Overlay both data sets in every panel. - Set data set 1 blue and data set 2 red using the marker style editor's This data set scope. Keep error bars enabled. Resize the plot window to give each panel approximately 4:3 proportions.
All 27 fits produced finite parameters and positive uncertainties for the three plotted quantities. Reduced chi-square ranges from 0.933 to 2.003 (809 degrees of freedom per fit). In particular, the lowest-energy fit is not statistically ideal: successful minimization does not establish that one Gaussian cosine fully describes the data. Inspect residuals and fit range dependence before drawing quantitative physical conclusions.
Reproduce and inspect
The repository directory
docs/examples/lem2025-singlehist-energy-scan contains all
27 fitted MSRs, their starting-file backups and logs, raw and normalized
msr2data tables, metadata, a fit summary and the fit script. Raw
detector data are not included. The summary records their SHA256 hashes.
The documented run used musrfit 1.12.1 with ROOT 6.38.00.
From the repository root, with musrfit, msr2data and dump_header on PATH:
python3 docs/examples/lem2025-singlehist-energy-scan/fit_all.py \
--data-dir /mnt/data/nemu/his/2025 \
--work-dir /tmp/lem-singlehist-new \
--reference-dir /tmp/lemfit
Use your own data directory containing
lem25_his_1014.root etc. Choose an empty work directory;
existing results are not overwritten. The optional reference directory
supplies starting values only. Without it, the script uses documented
generic initial values, which need not reproduce the same minimum in
every case.
The fit commands are musrfit RUN_his.msr -e -t -u 2 -k;
each MSR contains MINIMIZE, HESSE and
SAVE. The export commands use
msr2data "[RUN ...]" _his data new -o datasetN.raw.dat.
Only run suffixes in column headings are removed in
datasetN.dat so both scans have matching parameter names;
numerical values and uncertainties are unchanged.
Download the tables used in the comparison: data set 1 <../examples/lem2025-singlehist-energy-scan/dataset1.dat>
and data set 2 <../examples/lem2025-singlehist-energy-scan/dataset2.dat>.