SPERT III, a nuclear test reactor

Whose measurements these are

R. K. McCardell, D. I. Herborn and J. E. Houghtaling set down what the reactor did, and how sure they were of each number.

They wrote for the Atomic Energy Division of Phillips Petroleum Company, under contract to the U.S. Atomic Energy Commission, and their report of March 1969 covers eighty surges of power. One table in it gives thirty of them with a standard deviation beside every value, and that care is what lets a model be held to their work more than half a century later.

The authors: R. K. McCardell, D. I. Herborn and J. E. Houghtaling
Written for: Phillips Petroleum Company, Atomic Energy Division, under contract to the U.S. Atomic Energy Commission, Idaho Operations Office
The report: Reactivity Accident Test Results and Analyses for the SPERT III E-Core: A Small, Oxide-Fueled, Pressurized-Water Reactor. Report IDO-17281, March 1969.
Where it is published: The U.S. Department of Energy’s Office of Scientific and Technical Information: doi.org/10.2172/4792676
The report’s own thanks: The authors thank M. Ishikawa, G. A. Mortensen, R. K. Stitt and M. W. King, and J. P. Campbell, J. F. Scott and W. A. Eisenbarth of the Oxide Core Kinetics Group, and write that the programme took the efforts of many people.

What happened

Eighty times, on purpose, a reactor was given a sudden surge of power. Every time it came through unharmed.

Each of the thirty cold tests began with the reactor barely running. A rod moved in a few hundredths of a second, and the power climbed.

One surge, drawn and not plotted. The power lies low, climbs to a peak and falls away. The report gives three things for each surge: how high the peak was, how soon it came, and how much energy was released on the way up, shown here in orange.

How do you learn what a power reactor does in an accident without having the accident? In the 1960s one answer was to build a small reactor that could take it, and to give it the surge on purpose.

In the cold tests the power began at almost nothing and peaked anywhere from 2 to 280 megawatts, in the quickest test within a quarter of a second. Then the fuel warmed, the warming itself held the reaction back, and the surge turned over.

McCardell, Herborn and Houghtaling recorded how high each surge went, how much energy it released on the way up and when the peak came. Their report held the computer codes of its day to those figures. This page is about a later model held to the same table.

What the model did

The model called the size of all ten surges it had never seen, and it ran early on every one.

Super Intelligence coded the model from the report in under six hours. Twenty of the thirty cold tests were opened to it, and its one fitted constant came from those. The other ten, every third test in the order they were run, stayed sealed, and the passing grade was written down before any model was run.

Then the ten were opened. On how high each surge would go, the model met every mark. On when the peak would come it was early every time, typically by about 23 percent, and that is a miss. On the energy released on the way up it met two marks of three and read a little low on the third. Five marks of eight were met, so by the rule set beforehand the whole comparison counts as a miss, and the model’s own description says so.

Why early? The description points at it. The model switches the push on all at once, where the real rod took a few hundredths of a second to move. A second model of the same reactor carries that ramp, and it is served beside this one.

MissedOn timing

One square for each surge, its size called. Under each, a red arrow: early.

Where it stops

It follows a surge up to its first peak. Asked for anything after that, the answer stops and says why.

What it was tested for

The cold-startup surges of this one reactor core, with its water between 56 and 92 degrees Fahrenheit and not flowing.
Three things about each surge: how high the power went, how much energy was released on the way up, and when the peak came.
Ten sealed tests, whose peaks ran from 4.3 to 110 megawatts.

Past that, and what its own description says of itself

After the first peak of a surge the model stops. Boiling, heat passing from the fuel to the water, hot startup and running at power are not in it.
Its timing runs early, on the way up as well as at the peak.
The weakest and the strongest push it accepts are those of two of the report’s own tests. Beyond the range of the ten sealed tests it was compared only with tests its constant had been fitted on.
Some of its constants are the report’s values for a reactor at 500 degrees Fahrenheit, used here on cold tests, because the report prints none for cold.

For the reader who wants every number

In the verifier’s own words.

Everything above is this page’s telling. What follows is the verifier’s own, exact and whole: four parts of the description of the model energy/reactor_spert3, as read on the morning of 11 October 2026. Where the verifier says card it means a model, where it says line it means a pass mark, and where it says refuses it means the model stops there and says why.

The real thingobject

The SPERT III E-core, a small light-water test reactor, in its cold-startup reactivity-insertion tests: primary coolant at 56 to 92 degrees Fahrenheit, no flow, 5e-5 MW at the start, reactivity inserted in a step.

The passing grade, fixed firstpass_line

Fixed before any model was run (protocol SHA-256 47458dc27bcd9a60fe6c16ac613b2be3e15b30c7e4bb75d7624c80707e4e7ccd). Peak power and energy released to the peak: the root mean square of the error, in units of the report's standard deviation, at most 1.5; the worst error at most 3; the mean error at most 0.5 in size. Time of the peak: the root mean square of the relative error at most 5 percent and the worst at most 10 percent.

The resultresult

Compared with the ten held-back tests on 2026-10-08, and by the rule fixed beforehand the comparison failed: five of the eight lines were met and three were missed. Peak power met all three of its lines: root mean square 0.58 of the report's standard deviation, worst 1.13, mean +0.29. Energy released to the peak met two and missed one: root mean square 1.01, worst 1.99, and a mean of -0.705 against a line of 0.5 in size: the card reads low. Time of the peak missed both of its lines: root mean square 23 percent against 5 and worst 29 percent against 10, every peak early.

Its limitslimits

One constant is fitted, the strength of the feedback with energy released, on the 20 opened tests. The delayed-neutron constants are the report's values stated for 500 degrees Fahrenheit, used here on cold tests; the report prints none for cold. A forecast made outside the engine on the opened tests, each left out of the fit in turn, met the peak-power and energy lines and missed the time-of-peak line, with peaks about 21 percent early. The card stops at the first power maximum: boiling, heat flow from fuel to water, hot startup and operating power are not on the card. The reactivity is the report's reactor period converted by the caller; the card takes dollars. Only the peak power, the energy released to the peak and the time of the peak were compared with the held-back tests. The states the card seals before the peak were compared only with digitised plots of 19 opened tests; on 18 of them the card reaches a tenth of its peak about 26 percent early. The card steps the reactivity in at 5e-5 MW; the report's rod ramped it in over 50 to 65 ms, and a fit to the plots puts the starting power near 9 W (evidence/research/reactor_spert3/analysis/10_peak_timing_cause.md). energy/reactor_spert3_fuel carries the ramp.

Hand this one to your agent.

https://mcp.zerotrustphysics.com/mcp

Open to any agent.

Something to paste

Connect to https://mcp.zerotrustphysics.com/mcp and read the model energy/reactor_spert3. Then tell me what it is, and why it might matter to me.

Give it this page too.

What it will find

This model’s own description: what it takes in, where it stops, the report it was held to and how it did, in its own words.

If these are your measurements.

The model’s account of SPERT III is there for any agent to read, in its own words, and all of it rests on report IDO-17281. If you worked on the Oxide Core Kinetics Program, or knew the people who did, John Kruze would be glad to hear what the model got wrong about the real reactor.

Give the address and this page to your own agent first. Then write or call.

Write: [email protected]
Call: 928-264-7883