| 1 |
+++ |
| 2 |
title = "Reflectometry: Fit Pt layer" |
| 3 |
weight = 50 |
| 4 |
+++ |
| 5 |
|
| 6 |
### Reflectometry: Fit Pt layer |
| 7 |
|
| 8 |
In this example, we want to demonstrate how to fit experimental reflectivity data |
| 9 |
that was obtained by a time-of-flight experiment with unpolarized neutrons. |
| 10 |
Experimental data is available for a sample of a roughly 50 nm thick platinum layer on top of a silicon substrate that is published in [this repository](https://doi.org/10.5281/zenodo.4072376). |
| 11 |
|
| 12 |
The mesaurements were made by Timothy Charlton, Haile Ambaye and Michael Fitzsimmons (ORNL) |
| 13 |
on a sample provided by Eric Fullerton (UCSD). |
| 14 |
|
| 15 |
|
| 16 |
#### Fit model |
| 17 |
|
| 18 |
We describe the above experiment by a three-layer model, where as usual the top layer is |
| 19 |
the vacuum and the substrate layer is the silicon substrate. |
| 20 |
On top of the silicon substrate, we place the platinum layer. |
| 21 |
The materials of both layers are described by their SLD, where we use literature values |
| 22 |
for both silicon as well as platinum and keep them constant throughout the fitting procedure. |
| 23 |
|
| 24 |
The main parameters of the sample are stored in dictionary, where they are defined by a unique `name` |
| 25 |
and the following six parameters are utilized: |
| 26 |
|
| 27 |
* Beam intensity: `intensity` |
| 28 |
|
| 29 |
We explicitly fit the beam intensity, in order to compensate for possible experimental errors |
| 30 |
and to circumvent problems with the rather large variance in the reflectivity data at low $Q$-values. |
| 31 |
|
| 32 |
* Roughness on top of the Pt layer: `r_pt/nm` |
| 33 |
|
| 34 |
* Roughness on top of the Si substrate: `r_si/nm` |
| 35 |
|
| 36 |
* Thickness of the Pt layer: `t_pt/nm` |
| 37 |
|
| 38 |
* The relative $Q$-resolution: `q_res/q` ([c.f.](/ref/instr/scan/tof-resolution)) |
| 39 |
|
| 40 |
* A $Q$-offset: `q_offset` |
| 41 |
|
| 42 |
This global offset is introduced to account for uncertainties in the angle at which the measurement is performed. |
| 43 |
|
| 44 |
Due to saturation of the detector it is possible that the intensity at low $Q$-values (i.e. at high count rates) |
| 45 |
is underestimated. |
| 46 |
Furthermore, there is a rather large variance in the data that also leads to a rather bad fit in this region. |
| 47 |
Therefore, we neglect the data in the low $Q$-region by choosing a cutoff at $Q_{\text{min}} = $ 0.18. |
| 48 |
This value is selected by hand after performing several fits and visually selecting a good result. |
| 49 |
|
| 50 |
##### $Q$-offset |
| 51 |
|
| 52 |
Currently, BornAgain does not have an API support for an offset of the $Q$-axis. |
| 53 |
Therefore, we need to shift the $Q$-axis before performing a simulation |
| 54 |
```python |
| 55 |
q_axis = q_axis + parameters["q_offset"] |
| 56 |
``` |
| 57 |
|
| 58 |
This shift then needs to be counter-transformed when returning the results in the `qr(result)` function |
| 59 |
```python |
| 60 |
q = numpy.array(result.result().axis(ba.Axes.QSPACE)) - q_offset |
| 61 |
``` |
| 62 |
|
| 63 |
##### Initial parameters |
| 64 |
|
| 65 |
In order to successfully fit this example, we chose some sane starting values and |
| 66 |
the example code that is fully given below, can be run with the following command: |
| 67 |
```python |
| 68 |
python3 Pt_layer_fit.py fit |
| 69 |
``` |
| 70 |
This performs a simulation with the initial parameters and yields the following result: |
| 71 |
|
| 72 |
{{< galleryscg >}} |
| 73 |
{{< figscg src="/img/draw/Pt-Layer-initial.png" width="550px" caption="Reflectivity with the initial parameters before fitting">}} |
| 74 |
{{< /galleryscg >}} |
| 75 |
|
| 76 |
Immediately afterwards the fit is performed. |
| 77 |
|
| 78 |
|
| 79 |
#### Fit result |
| 80 |
|
| 81 |
In order to run the fitting procedure, the following command can be issued: |
| 82 |
|
| 83 |
```python |
| 84 |
python3 Pt_layer_fit.py fit |
| 85 |
``` |
| 86 |
|
| 87 |
We need to allow a few seconds computational time and BornAgain should compute the following result |
| 88 |
|
| 89 |
{{< galleryscg >}} |
| 90 |
{{< figscg src="/img/draw/Pt-Layer-final.png" width="550px" caption="Reflectivity with the parameters obtained from our fit">}} |
| 91 |
{{< /galleryscg >}} |
| 92 |
|
| 93 |
|
| 94 |
If the `fit` keyword is omitted from the command line |
| 95 |
```python |
| 96 |
python3 Pt_layer_fit.py |
| 97 |
``` |
| 98 |
a simulation is performed with our fit results and one should obtain the result shown above. |
| 99 |
|
| 100 |
{{< show-ex file="fit/specular/Pt_layer_fit.py" >}} |