Coherent Light
Copyright© 2026 by Stories2tell
Chapter 12: Reproduction and Control
The first attempt to reproduce the effect failed on a Wednesday evening three days after the campus lab session, and the failure was the specific kind that was most difficult to manage: not a clean negative result that ruled out a hypothesis, but an ambiguous partial result that suggested the approach was correct and the implementation was insufficient.
I had spent the three days between the lab session and the first garage attempt doing what I always did with a problem that exceeded my current framework — reading everything relevant I could find, building the most accurate model the available literature supported, identifying the gaps between the model and what I had observed. The reading had been extensive and had produced, at the end of seventy hours spread across three days, the following useful conclusions: nothing in the published photonics literature described an effect consistent with what I had observed. Nothing in the quantum optics literature described it. Nothing in the nonlinear optics literature, the quantum information literature, or the adjacent territories of condensed matter physics that touched on quantum dot behavior described anything that resembled a self-bounded dark hemisphere with optical depth cues consistent with a distant star field.
This was either because the effect was genuinely novel, which had implications I was not yet ready to examine, or because the effect was a known phenomenon described in a framework I hadn’t yet reached, which was the more conservative hypothesis and the one I was treating as primary until the evidence required otherwise.
What I had from the lab session was a complete record of the electromagnetic environment that had produced the effect — the welding generator’s specifications, the variable frequency drive’s switching frequency, the interaction spectrum from the analyzer — and a complete record of my system’s operating parameters at the moment of the transition. What I did not have was either of the two pieces of equipment from the renovation team, which had finished their work in the adjacent wing and departed.
The garage setup was not the campus lab setup. The quantum dot array I had been running in the lab was a more developed version of the architecture I had in the garage — better characterized dots, more precise phase control, a detection system with higher dynamic range. The garage version was the prototype that the lab version had been developed from, which meant it shared the essential architecture but lacked some of the refinements that might or might not be relevant to reproducing the effect.
The electromagnetic environment was the larger problem.
What the renovation equipment had produced was a specific combination of broadband noise and structured interference between two interacting systems — the generator and the variable frequency drive — that I could characterize but not easily replicate. The interaction had a specific spectral signature that was the product of the particular operating parameters of both devices under the particular load conditions of that afternoon. Reproducing it required either the same equipment under the same conditions, which was not available, or a substitute that produced the same spectral signature through different means.
I spent the first evening building the substitute.
The welding generator’s contribution to the noise spectrum was primarily a set of harmonics from its alternator, combined with the conducted emissions from the welding arc itself — a complex, high-amplitude broadband source with energy concentrated in specific frequency bands. I could approximate this with a combination of a signal generator driving a power amplifier and a spark gap device I built from components in the parts bins, which produced the arc-related broadband component. The variable frequency drive’s contribution was more structured — a switching waveform at a specific frequency with its harmonics — which I could reproduce more accurately with a second signal generator set to the correct frequency and waveform shape.
The interaction between the two, which was the component I was least confident about, depended on the specific impedance environment of the building’s electrical infrastructure — the ground loops, the power line coupling, the distributed capacitance of the wiring. The garage had its own electrical infrastructure, different from the campus lab’s, which meant the interaction would be different even if the individual sources were correctly specified.
This was the gap I could not close from first principles. I would have to close it empirically.
The first attempt used the signal sources I had built and the garage array at the operating parameters recorded from the lab session. I ran it for two hours and produced nothing that resembled the lab effect — the array’s output was disturbed by the artificial electromagnetic environment in ways that were consistent with simple interference rather than the regime transition I was looking for. The noise was degrading the coherence rather than enhancing it, which was the opposite of what I needed.
I recorded the result and went to bed.
The second attempt was the following evening. I had spent the day reconsidering the coupling mechanism — the way the electromagnetic environment interacted with the feedback architecture specifically, rather than with the array as a whole. The feedback system was the component most sensitive to external electromagnetic influence, because it was a control loop with gain, and control loops with gain amplified disturbances as well as signals. The question was whether the noise from the renovation equipment had been coupling primarily into the feedback loop or into the array elements directly.
If the coupling was primarily into the feedback loop, then the relevant parameter was not the full spectral signature of the noise but the component of the noise that fell within the feedback loop’s bandwidth. The feedback loop had a specific bandwidth determined by its design — the frequency range over which it actively responded to input. Noise outside that bandwidth would be filtered. Noise inside that bandwidth would be amplified.
I recalculated the overlap between the renovation equipment’s noise spectrum and the feedback loop’s bandwidth. The overlap was concentrated in a specific frequency range — lower than I had initially focused on, in the range where the variable frequency drive’s fundamental switching frequency and its second harmonic fell. This was a narrower target than the full broadband noise, which meant I could produce it more accurately with the signal generators I had.
The second attempt used a modified noise source focused on this frequency range, coupled specifically into the feedback loop’s input rather than radiated into the general electromagnetic environment of the garage. The array’s response was different immediately — the coherence metrics on the monitoring system showed a different pattern of disturbance, less degradation and more of the structured response I had been looking for.
The hemisphere did not appear. But the array’s behavior in the presence of the modified noise was qualitatively different from the first attempt, and the difference was in the direction of the lab effect rather than away from it. I recorded everything and noted in the notebook: the coupling pathway hypothesis is supported. The regime transition requires additional conditions not yet identified.
The third through seventh attempts occupied the following two weeks and produced a sequence of partial results that were, collectively, more informative than any single success would have been. Each attempt modified one variable based on the previous result, and each result constrained the parameter space further. The hemisphere did not appear in any of these attempts, but the array’s behavior moved progressively closer to the transition point — the coherence metrics showing the specific pattern I had recorded in the lab in the seconds before the transition, then collapsing back rather than crossing into the new regime.
The collapse was the problem. The system was approaching the threshold and then failing to cross it, which meant something was missing — a condition necessary for the transition that I had not yet identified.
I went back to the lab records.
The transition in the campus lab had been abrupt — the threshold crossing I had described to Patrick months ago, the system shifting from one regime to another at a specific point rather than gradually. But the approach to the threshold had not been abrupt. The array’s behavior had been changing for approximately thirty seconds before the transition, shifting in the direction of the new regime in a way that I had noted in the lab but had not yet fully analyzed in the garage attempts.
I pulled the lab data and looked at the thirty seconds before the transition in detail.
The feedback gain had been at a specific setting — not the setting I typically used, but a setting I had moved to approximately two minutes before the transition while trying to optimize the array’s coherence in the presence of the electromagnetic disturbance. I had recorded the gain value but had not flagged it as significant because at the time I had been focused on the visual output rather than the system parameters.
The gain value was higher than my standard operating range. Significantly higher — at the upper edge of the feedback system’s stable operating region, the point where the loop gain was approaching but not yet reaching the instability threshold. I had been running the garage attempts at my standard gain setting, which was chosen for stability under normal operating conditions. The lab transition had occurred at a gain setting that was chosen for performance under abnormal conditions and that I would not normally have used.
The interaction between the elevated gain and the structured electromagnetic noise was the missing condition. The noise was driving the feedback system, and the elevated gain was amplifying that drive to the point where the system crossed into the new regime. At standard gain the drive was insufficient. At elevated gain it was sufficient, but only in combination with the correctly structured noise.
The eighth attempt used the elevated gain setting.
The hemisphere appeared at eleven forty-seven on a Thursday evening, seventeen days after the campus lab session, in the northeast corner of the garage above the optical bench.
It was smaller than the campus version — approximately twenty-five centimeters in diameter — and the boundary was less sharp. But the interior was dark in the characteristic way, and the points of light were there, and the depth cues were consistent with what I had seen in the lab. I sat very still for approximately ten seconds and looked at it.
Then I began working.
To read the complete story you need to be logged in:
Log In or
Register for a Free account
(Why register?)
* Allows you 3 stories to read in 24 hours.