Lilienfeld 1925 Characterize Stability Under Sustained Operation

# Characterize Stability Under Sustained Operation: A Verdict Measured Once Is Not a Device That Works

Step 24 asked whether the Step 23 verdict generalizes across specimens. Step 25 asks a different question that no prior step can answer: does a single specimen's verdict survive the passage of time, continued electrical operation, and ordinary storage? Every measurement in Steps 15 through 24 was taken within one bounded session on a specimen handled under tightly controlled conditions. None of them establishes that a specimen passing Step 23 today will still pass tomorrow, after a week on a shelf, or after the hours of continuous operation a usable amplifier stage requires.

This question matters specifically for this device because two mechanisms identified earlier in the arc are time-dependent by nature rather than momentary. Step 19 named copper sulfide as a mixed ionic–electronic conductor in which sustained bias can rearrange the film's composition. Step 18 named humidity and surface-conduction pathways as drift sources that build over time rather than appearing instantly. A verification program that stops at Step 24 has shown the device can pass once; Step 25 is what shows whether it keeps passing.

## 1. Separate three distinct timescales that drift can live on

Operational drift during active use, storage drift with no bias applied, and cycling drift from repeated power-on and power-off transitions are physically distinct and must not be collapsed into one "stability" number. A specimen that holds steady under continuous bias but degrades badly on a shelf is a storage problem; one that is stable in both of those conditions but shifts every time it is powered up is a cycling problem, and each points to a different physical cause and a different mitigation.

Define three separate protocols before testing begins:

  • operational drift — the Step 19 baseline and Step 23 gain measurement repeated at fixed intervals while the specimen remains under continuous bias and signal exercise between measurements;
  • storage drift — the identical measurements repeated after the specimen sits unbiased and unconnected for a declared interval, under declared temperature and humidity;
  • cycling drift — the identical measurements repeated after each of a declared number of full power-on, operate, power-off, rest sequences.
Three distinct drift timescales for a Lilienfeld specimen under long-term study Three parallel timelines show the same specimen exercised under continuous operation, left in unbiased storage, and cycled through repeated power-on and power-off sequences, with the identical baseline and gain checkpoint measurement repeated at declared intervals on each timeline. THREE DRIFT MECHANISMS, THREE SEPARATE TIMELINES Operation, storage, and power cycling stress the device differently and must be tracked apart THREE PARALLEL PROTOCOLS, SAME CHECKPOINT MEASUREMENT OPERATIONAL continuous bias and signal exercise between checkpoints STORAGE unbiased, unconnected, declared temperature and humidity between checkpoints CYCLING a declared number of full power-on, operate, power-off, rest sequences between checkpoints t0t1t2t3tN each checkpoint repeats the identical Step 19 baseline and Step 23 gain measurement WHY COLLAPSING THEM LOSES INFORMATION OPERATIONAL ONLYstable under bias but degrades on the shelf points to a storage-chemistry problemCYCLING ONLYstable in both above but shifts every power-up points to a contact or thermal-shock problemONE NUMBERa single combined drift figure cannot distinguish these causes or guide a fix

## 2. Reuse the existing measurement protocols as checkpoints, not a new instrument

Every checkpoint on every timeline repeats the unmodified Step 19 baseline sweep and the unmodified Step 23 power-gain measurement, using the same instrument settings, bias points, and signal amplitude established there. Introducing a different, faster "monitoring" measurement for convenience would make checkpoints incomparable to the original Step 19 and Step 23 results and to each other. The only new element is the repetition schedule; the measurement itself must be the one already validated.

At each checkpoint, also repeat the Step 20 leakage check and the Step 18 isolation check. A specimen whose terminal characteristic looks unchanged but whose control-electrode leakage has grown, or whose isolation has degraded, is not actually stable — it may simply be approaching a failure that the baseline and gain measurements alone would not yet reveal.

## 3. Fit a trend, not just a start-versus-end comparison

A two-point comparison between the first and last checkpoint cannot distinguish a monotonic trend from a one-time shift followed by stability, and it discards the information in every intermediate point. Fit the checkpoint sequence for each quantity of interest — zero-bias conductance $g(0)$, extracted $R_{out}$, and $G_{avail}$ at the Step 23 reference frequency and bias — to a declared functional form, for example

$$ X(t)=X_{0}+\Delta X\left(1-e^{-t/\tau}\right), $$

reporting the asymptotic shift $\Delta X$ and the time constant $\tau$, or an explicit statement that the data does not support this form if the trend is not monotonic or saturating. Compare $\Delta X$ against the Step 19 resolvable-change threshold $\delta I_{min}$ and the Step 23 gain uncertainty, so that a "drift" smaller than the measurement's own repeatability is correctly reported as not resolved rather than as a confirmed trend.

Checkpoint trend fitting and stability disposition across three drift protocols A plot of a tracked quantity across repeated checkpoints on all three timelines shows a saturating trend fit against the baseline repeatability band, beside a disposition table scoring each timeline as stable, drifting but resolved, or failed based on whether the fitted shift exceeds the measurement uncertainty and whether a hard failure occurred. A TREND NEEDS A FIT, NOT JUST A FIRST AND LAST POINT Compare the fitted shift against the measurement's own repeatability before calling it drift CHECKPOINT TREND ACROSS THE THREE PROTOCOLS checkpoint time, t0 through tN tracked quantity, e.g. g(0) repeatability band operational, within band cycling, resolved shift storage, abrupt change STABILITY DISPOSITION PER TIMELINE STABLEfitted shift stays within the baseline and gain repeatability bounds through all checkpointsDRIFTING, RESOLVEDshift exceeds repeatability but saturates and leakage and isolation remain within limitsFAILEDabrupt change, non-saturating trend, or a Step 18 or Step 20 limit exceeded at any checkpointREPORTdisposition and fitted parameters for each of the three timelines separately, never merged into one number

## 4. Treat a hard failure as data, not as an aborted run

If a specimen fails Step 18's isolation check or Step 20's leakage limit at any checkpoint, or stops responding to the control electrode altogether, record the checkpoint at which this happened, the full measurement taken at that checkpoint, and every prior checkpoint's trend leading up to it. This is itself a reportable stability result — a mean or median time-to-failure across the specimens carried into this step — not a discarded run. Do not replace a failed specimen mid-protocol with a fresh one and report only the survivor's data; that silently selects for the most stable specimens and overstates the population's actual stability.

## 5. Report the operating envelope the stability result is conditioned on

Every drift and failure result is conditioned on the temperature, humidity, bias level, and signal amplitude used during the protocol. State this envelope explicitly, and do not extrapolate a storage-drift result measured at one humidity to a claim about behavior at a different humidity, or an operational-drift result measured at the Step 21 bias points to a claim about behavior at a different bias. If resources permit, repeat a subset of the protocol at a second point in the envelope — elevated humidity for storage, elevated bias for operation — specifically to test whether the stability result itself depends on the operating condition, since a specimen that is stable at room humidity but fails quickly at high humidity has a qualitatively different limitation than one that fails regardless of humidity.

## 6. Carry the population structure from Step 24 forward

Run this protocol across the same pre-declared, independence-documented population Step 24 established, not a new ad hoc set of specimens chosen because they happened to be on hand. Report the stability disposition per specimen and the resulting rate — the fraction of specimens remaining stable, drifting-but-resolved, or failed, at each checkpoint — using the same binomial interval discipline Step 24 applied to the pass rate. A stability claim made from one or two convenient specimens is exactly the single-device generalization problem Step 24 was written to correct, reintroduced at the time axis instead of the specimen axis.

## 7. Minimum Step 25 record

Retain:

  • the three separate protocol definitions — operational, storage, cycling — with their checkpoint schedule, declared temperature and humidity, and bias and signal conditions, fixed before testing began;
  • confirmation that every checkpoint measurement reused the unmodified Step 18, 19, 20, and 23 protocols without substitution;
  • the complete per-checkpoint, per-specimen data for $g(0)$, $R_{out}$, $G_{avail}$, leakage, and isolation, across the full population carried forward from Step 24;
  • the fitted trend parameters $\Delta X$ and $\tau$ for each tracked quantity on each timeline, or the explicit statement that no saturating fit was supported;
  • the comparison of each $\Delta X$ against the Step 19 and Step 23 repeatability bounds;
  • every hard failure, its checkpoint, and the full trend leading up to it, with no failed specimen silently replaced;
  • the operating envelope each result is conditioned on, and any secondary-envelope data collected to test envelope dependence;
  • the per-specimen stability disposition and the resulting population rate with its confidence interval, reported separately for each of the three timelines.

The historical geometry and claimed effect are grounded in [J. E. Lilienfeld, US Patent 1,745,175](https://patents.google.com/patent/US1745175A/en). The specific drift mechanisms this step tests for — ionic rearrangement under sustained bias and humidity-sensitive surface conduction — were identified during the low-level measurement work of Steps 18 through 20 and are tracked here using the same guarded, blank-referenced measurement discipline described in the [Keithley Low Level Measurements Handbook](https://www.tek.com/en/documents/product-article/keithley-low-level-measurements-handbook---8th-edition).

## Characterize Stability Under Sustained Operation’s Place in the Process Lineage

Step 24 showed whether the Step 23 verdict holds across a population of specimens measured once. Step 25 now asks whether that verdict holds for any one specimen across time, under continuous operation, in storage, and across power cycling, using the identical checkpoint measurements and the same population-level statistical discipline rather than a new instrument or a hand-picked survivor. Only a population that remains stable, or drifts in a bounded and characterized way, across all three timelines supports treating the Step 23 and Step 24 results as a property of a usable device rather than of a single favorable measurement session.

Take lilienfeld 1925 characterize stability under sustained operation further

Ask the copilot about this term, or have our engineers assess it against your process.

Ask CFSGPT about this → Request a technical assessment → Open the CFS app