spectroscopy-channel-calibration

Made by Brynjar Morka Mæhlum for the subject "TFY4255 Materialfysikk", as a part of Brynjars "TFY4520 Nanoteknologi, fordypningsprosjekt".

---

Use the Binder link below to open this repository in an online and interactive way:

---

Goal of this notebook:

Use a spectrum with at least two known peaks to calibrate the channel width of a detector, and then use the calibration to plot an unknown spectrum.

To calibrate a spectrum we calculate:

• dispersion [keV / channel], which is the width of each channel
• offset [channels], which is the distance from first data entry to keV=0

When we have calibrated a spectrum, we could also find:

• energy resolution
• quantities of each element by area under the peak
  • bear in mind: this is dependent on signal-to-background, efficiency of the detector, ...
• ... ?

---

What we will do:

1. Read in the data of a known spectrum
2. Fit the data to a gaussian
3. Calibrate the x-axis
4. Plot the calibrated data
5. Use the calibrated x-axis on an unknown spectrum from the same source

---

Organisation of the repository

This is the file tree:

│   .gitignore
│   channel_calibration.ipynb
│   environment.yml
│   LICENSE
│   README.md
│
├───helper_files
│   │   calibration.py
│   │   gaussian_fitting.py
│   │   plotting.py
│   │   read_data.py
│   │   saving_json.py
│   │   spectrum_dict.py
│   │   __init__.py
│
├───Lab3_data
│       SEM_known_Cu.msa
│       SEM_known_NiO_on_Mo.msa
│       SEM_unknown.msa
│       TEM_known_NiO_on_Mo_A.emsa
│       TEM_known_NiO_on_Mo_B.emsa
│       TEM_unknown.msa
│       XRF_known_Cu.mca
│       XRF_known_Pb.mca
│       XRF_no_sample.mca
│       XRF_unknown.mca
│       XRF_unknown_2nd.mca
│
├───Lab3_data_calibrated
│       SEM_known_Cu_calibrated.json  # example of a plot
│
└───plots
        SEM_known_Cu_calibrated.html  # example of a plot

The main notebook is called "channel_calibration.ipynb", which contains a step-by-step calibration of a known spectrum, and then using that on an unknown spectrum.

To make the notebooks shorter and more understandable, I've made the folder "helper_files" with functions that you will use. The helper functions are documented with NumPy docstrings, which you can read by running a cell with the function name followed by a questionmark:

function_name?

>>> prints the docstring of function_name

---

Info on the data files

• SEM_known_Cu.msa

  • Bulk Cu
  • Collected on Hitatchi TM4000 at 15 kV

• SEM_unknown.msa

  • 0.3 mm thick unknown
  • Hitatchi TM4000 at 15 kV

• TEM_known_NiO_on_Mo.emsa

  • NiO calibration specimen [6]
  • Collected at Jeol 2100 at 200 kV using a Oxford Instrument 80mm2 SDD

• TEM_unknown.msa

  • Unknown material crushed and deposited on a 300 mesh Cu TEM grid with a holey 20 nm C-support
  • Data collected at Jeol 2100 at 200 kV using a Oxford Instrument 80mm2 SDD

• XRF_known_Cu.msa

  • Bulk Cu
  • MoK X-ray source, AmpTek® energy dispersive detector

• XRF_unknown.msa

  • unknown material, 0.3 mm thick
  • MoK X-ray source AmpTek® energy dispersive detector

• ---

Challenges I've had

• What is the actual zero / start of the data? (solved)
  • What to do with the zero-peak which some instruments have?
  • Some (or all?) spectra starts measuring before 0. How do I set the 0 right?
• Some datasets (eg. .emsa of GaAs_30keV) starts with negative keV values. (solved)
• Not possible to recognize all peaks, since some are low.
  • Without bg removal one peak is usually fitted in the middle as the bg with a very high std
• For some reason it did not work to use the raw x values from .emsa when fitting, while using channels as int worked.
  • Also fitting is slower / worse with non-normalized values.
    • Might be because it is easier to guess correct amp, mu, and std with normalized values
  • workaround: fit with channels and intensity normalized to 1