Relate Growth Conditions to Film Properties

Which measurements during growth are worth comparing with the properties measured afterward? Use a notebook to join the evidence by sample, inspect relationships, and decide what to measure next. This example follows a thin-film campaign with process-temperature logs, early RHEED measurements, and post-growth sheet resistance. We first examine a relationship across all films, then account for temperature to see whether the RHEED measurement adds useful information.

​

Start with one row per physical sample

The campaign contains Samples A–L, with four films at each of three temperature settings. Each sample has a process log and an early RHEED recording. Sheet resistance was measured afterward using the same method. Open Analysis, select New Notebook, and enter:

Investigate which growth measurements track sheet resistance in this campaign. Join process logs, early RHEED measurements, and post-growth resistance by physical sample. Show data coverage and units. Start with growth temperature, then test whether a consistently measured RHEED feature adds information after accounting for temperature. Keep correlations separate from causal claims.

Click Create Notebook. Check that the joined table associates each measurement with the correct physical sample. See Creating a Notebook for the full instructions. This is an excerpt of the 12-sample example. Keep missing values visible and report the sample count used in each comparison. Several recordings of one film are measurements of the same sample, not independent films.

​

Define comparable measurements before plotting

For the RHEED feature, the example takes the median streak width from seconds 20–30 after deposition begins, using the same measurement view. Confirm that this interval belongs to the same process phase in every film. Pixel widths are comparable only when the recording geometry, image scale, and feature selection support that comparison. A change in camera setup could otherwise look like a change in the material. For resistance, verify units and measurement conditions. Sheet resistance and two-point resistance are different quantities; keep the recorded measurement type explicit.

Show the recording and interval used for each sample, the number of usable frames, and the resistance measurement method. Exclude samples from a comparison when the required measurements are not comparable, and explain each exclusion.

​

Inspect the relationship across the campaign

The first panel below shows lower sheet resistance at higher temperature settings in this fictional campaign. The logarithmic resistance axis makes the spread at each setting visible. Inspect individual points and group sizes before fitting a model. Four films per setting provide a small exploratory comparison, especially if one measurement could move the apparent trend.

​

Account for a competing explanation

Temperature could affect both the RHEED feature and resistance. A relationship across all films might therefore reflect temperature rather than additional information in the RHEED measurement. Ask a narrower question:

Within each temperature setting, does early streak width still track resistance? Show separate groups. Compare each film’s resistance with the median for its temperature group, and check whether a single point or group drives the result.

The second panel divides each film’s resistance by its temperature group’s median. A value above 1 means resistance is above that group’s median. In the fictional data, larger streak widths accompany larger relative resistance within each group. This grouping helps examine one competing explanation. It does not control every relevant growth condition or show that streak width causes resistance. With only a few films per setting, the group medians are also uncertain.

​

Turn the finding into a useful next experiment

A provisional handoff could say:

Temperature separates the resistance levels in this example. Within each temperature group, early streak width also tracks resistance relative to the group median. Repeat comparable films at a fixed temperature and preserve the same RHEED geometry to test whether this relationship holds on new samples.

Keep those new films separate when evaluating a predictive model. Choosing features and reporting prediction accuracy on the same small set would not show how well the model works on new growths.

Summarize the samples included, the relationships observed, the competing explanations checked, and a follow-up measurement plan. Identify which conclusions are descriptive and what would require new data.

Save the notebook so colleagues can inspect the joins and exclusions. Record the proposed experiment and its context on the project.

​

Adapt the investigation

For example, compare plasma conditions with post-growth roughness while accounting for recipe family. The reusable approach is to join the right measurements, make them comparable, inspect alternatives, and test the relationship on new data. If you first need a suitable comparison run, see Choose a Reference Growth.