Emily's pages: Questions · Vocabulary · Language · Studies reproduced · Experiment proposal · The essay
Two measurements that would settle two open questions
Companion to question 10 of [Questions we need Emily to answer](/writing/questions-for-emily-testing-a-theory-of-the-interface). Written 10 November 2026. The predictions are from our programs; the numbers behind them are agent-read from papers and have not been checked by a person.
If the organisers run experiments on promising models, these are the cheapest runs we know of that turn an inconclusive cell into a supported or rejected one. Each one is a point at which our surviving candidates disagree by many times the measurement's repeatability and no measurement we hold has been made.
1. SiC: does the removal of interface states stop?
The question. After oxidation, an argon anneal lowers the interface-state density on SiC. The data we hold (the SiC sources in our library) fit two forms equally well: one in which the removal runs to completion and one in which it stops at a level. They agree on every anneal that has been measured.
The run.
| Sample | 4H-SiC as in the sources we hold, oxidized at 1300 °C (the paper’s own face, doping and ambient) |
| Then | argon anneal at 1300 °C for 50 min (3000 s) |
| Measure | interface-state density at 0.3, 0.4 and 0.5 eV below the conduction band, by the method the source paper used |
| Control | the same oxidation with no anneal, and with the 5 min anneal already in the literature |
What the two forms predict (per cm² per eV, at 0.3 / 0.4 / 0.5 eV below the band edge): the form that keeps removing gives 8.7e11 / 3.9e11 / 1.9e11; the form that stops gives 2.1e12 / 1.4e12 / 9.2e11. They differ by about 13 repeatability tolerances at the first level. A 10 h anneal at 1300 °C separates them just as far, if the 50 min run is not possible.
What each outcome means. Low values: removal continues, and the model that stops is rejected. High values: it stops, and any claim about a "rate of removal" is only about the early part of the anneal. Anything between: neither form is right, and the structure search has to be reopened.
2. Silicon: what survives a nitrogen anneal at 700 °C?
The question. Two programs for fixed charge in (111) silicon survive every measurement we hold. They differ in how the charge relaxes during an anneal after oxidation.
The run.
| Sample | (111) silicon, dry oxidation at 900 °C |
| Then | 1 h in nitrogen at 700 °C |
| Measure | fixed charge from capacitance-voltage, with the same hydrogen treatment as the paper we match (to be agreed) |
| Control | the same oxidation with no post-oxidation anneal |
Prediction: 4.4e11 per cm² for the first candidate, 3.4e11 for the second. The gap is 2.1 tolerances, so it needs at least four wafers per condition to read: the repeats at 1000 °C in the paper we depend on scatter by 9 to 38% between identical runs (see the repeatability note in our analysis), which is as large as the gap.
3. If a third run is possible: matched wafers
One oxidation, split across wafers, with the cooling rate and the hydrogen exposure after oxidation varied on purpose and nothing else. Every reproduction we have run is limited by these two numbers being unstated.
What we are not asking for
No new equipment. Each run is a recipe in a standard furnace line and a capacitance-voltage measurement, and for each we will provide the prediction in writing before the measurement is made, with the frozen program that produced it.