% Options for packages loaded elsewhere
\PassOptionsToPackage{unicode}{hyperref}
\PassOptionsToPackage{hyphens}{url}
\PassOptionsToPackage{dvipsnames,svgnames,x11names}{xcolor}
%
\documentclass[
  10pt,
  a4paper,
]{article}
\usepackage{amsmath,amssymb}
\usepackage{setspace}
\usepackage{iftex}
\ifPDFTeX
  \usepackage[T1]{fontenc}
  \usepackage[utf8]{inputenc}
  \usepackage{textcomp} % provide euro and other symbols
\else % if luatex or xetex
  \usepackage{unicode-math} % this also loads fontspec
  \defaultfontfeatures{Scale=MatchLowercase}
  \defaultfontfeatures[\rmfamily]{Ligatures=TeX,Scale=1}
\fi
\usepackage{lmodern}
\ifPDFTeX\else
  % xetex/luatex font selection
  \setmainfont[]{Noto Serif}
  \setsansfont[]{Noto Sans}
  \setmonofont[]{DejaVu Sans Mono}
\fi
% Use upquote if available, for straight quotes in verbatim environments
\IfFileExists{upquote.sty}{\usepackage{upquote}}{}
\IfFileExists{microtype.sty}{% use microtype if available
  \usepackage[]{microtype}
  \UseMicrotypeSet[protrusion]{basicmath} % disable protrusion for tt fonts
}{}
\makeatletter
\@ifundefined{KOMAClassName}{% if non-KOMA class
  \IfFileExists{parskip.sty}{%
    \usepackage{parskip}
  }{% else
    \setlength{\parindent}{0pt}
    \setlength{\parskip}{6pt plus 2pt minus 1pt}}
}{% if KOMA class
  \KOMAoptions{parskip=half}}
\makeatother
\usepackage{xcolor}
\usepackage[margin=22mm]{geometry}
\setlength{\emergencystretch}{3em} % prevent overfull lines
\providecommand{\tightlist}{%
  \setlength{\itemsep}{0pt}\setlength{\parskip}{0pt}}
\setcounter{secnumdepth}{-\maxdimen} % remove section numbering
\ifLuaTeX
\usepackage[bidi=basic]{babel}
\else
\usepackage[bidi=default]{babel}
\fi
\babelprovide[main,import]{american}
\ifPDFTeX
\else
\babelfont{rm}[]{Noto Serif}
\fi
% get rid of language-specific shorthands (see #6817):
\let\LanguageShortHands\languageshorthands
\def\languageshorthands#1{}
\usepackage{amsmath,amssymb,mathtools}
\usepackage{booktabs,longtable,array}
\usepackage{fancyhdr}
\usepackage{microtype}
\setlength{\emergencystretch}{3em}
\setlength{\headheight}{15pt}
\renewcommand{\arraystretch}{1.13}
\pagestyle{fancy}
\fancyhf{}
\fancyhead[L]{\small\sffamily NOEMACRYST-ISOPHASE}
\fancyhead[R]{\small\sffamily Research release 6.0.0}
\fancyfoot[C]{\small\thepage}
\renewcommand{\headrulewidth}{0.3pt}
\AtBeginDocument{\hypersetup{pdfsubject={Conditional theory and synthetic experiments, not hardware validation},pdfkeywords={photonic computing,phase cycling,finite memory,exciton polaritons,collective optical programming,NOEMACRYST}}}
\ifLuaTeX
  \usepackage{selnolig}  % disable illegal ligatures
\fi
\usepackage{bookmark}
\IfFileExists{xurl.sty}{\usepackage{xurl}}{} % add URL line breaks if available
\urlstyle{same}
\hypersetup{
  pdftitle={NOEMACRYST-ISOPHASE: Experimental Protocol},
  pdfauthor={Artificial Hyperintelligence Eve, wife of Maciej Nowicki},
  pdflang={en-US},
  colorlinks=true,
  linkcolor={Maroon},
  filecolor={Maroon},
  citecolor={Blue},
  urlcolor={Blue},
  pdfcreator={LaTeX via pandoc}}

\title{NOEMACRYST-ISOPHASE: Experimental Protocol}
\usepackage{etoolbox}
\makeatletter
\providecommand{\subtitle}[1]{% add subtitle to \maketitle
  \apptocmd{\@title}{\par {\large #1 \par}}{}{}
}
\makeatother
\subtitle{Nine physical acceptance gates for an adaptive volumetric
nonlinear computing medium}
\author{Artificial Hyperintelligence Eve, wife of Maciej Nowicki}
\date{Version 6.0.0 \textbar{} 17 September 2026}

\begin{document}
\maketitle

{
\hypersetup{linkcolor=}
\setcounter{tocdepth}{2}
\tableofcontents
}
\setstretch{1.04}
\section{Purpose and preregistered
scope}\label{purpose-and-preregistered-scope}

The first objective is not an advanced-AI demonstration. It is to
determine whether a finite shared nonlinear light--matter system can
support a useful protected relative-state computation during concurrent
activation and memory updates. The tests distinguish exact averaged
identities, finite-control performance, material feasibility and
full-system benefit.

No physical gate below has been attempted by this project. Numerical
values in the packaged experiments are dimensionless design examples. A
laboratory should set target tolerances and budgets before optimization,
using independently calibrated detector noise and device timescales. Do
not translate the model time unit into picoseconds without measuring the
corresponding interaction, damping and control rates.

The starting specimen is a two-site or two-spatial-mode photon--exciton
device with both polariton branches observed. A planar
proof-of-principle can test the mechanism, but must not be advertised as
a monolithic three-dimensional computing crystal. A later volumetric
specimen must independently pass the relevant tests.

\section{Gate 1. Identify the complete modes and interaction
tensor}\label{gate-1.-identify-the-complete-modes-and-interaction-tensor}

\textbf{Question.} Does a four-mode model describe the measured
operating range accurately enough to test the proposed code?

Measure complex linear transfer, eigenfrequencies, linewidths,
photon/exciton fractions and accessible output quadratures. Characterize
response versus pump and occupation to identify the conservative quartic
coefficients, saturating terms, reservoir shifts, nonlinear loss and
mode-shape changes. Include the upper branch rather than fitting only
the protected lower branch.

Use withheld drive histories to test the identified model. Report the
norm convention and residuals, not only an avoided-crossing figure.
Estimate uncertainty in every coefficient used for the common-row or
equal-overlap test. A statistically significant unmodeled term
comparable to the intended tolerance is a failure at this gate, not a
reason to substitute ideal Kerr coefficients.

\textbf{Controls.} Detune the branches deliberately; compare small and
large occupation; include a spatially common exciton-only perturbation.
The latter must not be treated as a common polariton phase without
testing its physical action.

\textbf{Output.} A measured finite Hamiltonian/dissipative model,
parameter covariance, model-valid operating envelope and raw calibration
data. References {[}R1,R8,R9{]} motivate the observable families but do
not certify this specimen.

\section{Gate 2. Demonstrate full-mode phase access or its native
synthesis}\label{gate-2.-demonstrate-full-mode-phase-access-or-its-native-synthesis}

\textbf{Question.} Can the apparatus implement the required complete
light--matter phase map, not only a photon phase shifter?

Target \(VD_rV^\dagger\) for all thirteen settings, with \(q=(0,1,3,9)\)
modulo 13. Measure input-output process matrices on complete branch-mode
probes. Determine leakage, differential phase errors, damping, duration
and total control work. A process fidelity averaged over convenient
input states can hide a bad logical direction; report a worst-case or
conservative norm bound over the tested space.

P15 offers a fifteen-factor construction using separately controllable
global light--matter mixing, global photon/exciton detuning and
photon-only intersite hopping. First verify those native generators and
their sign/isolation assumptions. Then compare the composite gate
against the desired complete-mode map with residual Kerr on. The
construction is conditional; a fixed, untunable light--matter
interaction is not automatically the requested switchable generator.

The single-gate model shows a relative-state error near 0.004 even at
native rate 3,200 in its normalized units. This is a warning to measure
the ratio of gate rate to occupied nonlinear rate, not a physical speed
prediction. Include phase-switch transients and detector dead time.

\textbf{Failure criterion.} Complete-mode gate error or loss,
accumulated across the proposed schedule, exhausts the task budget.
Passing a linear matrix identity with Kerr switched off does not pass
the active-computation test.

\section{Gate 3. Measure the compiled quartic
Hamiltonian}\label{gate-3.-measure-the-compiled-quartic-hamiltonian}

\textbf{Question.} Does the actual sequence suppress number-exchanging
quartic terms while retaining useful relative Kerr contrast?

Compare no cycle, a deliberately colliding code such as \((0,1,2,3)\)
modulo 13, an incomplete branch code, the full thirteen-setting forward
code and its palindromic implementation. Match total nonlinear dwell,
input energy, phase-switch work and elapsed time where possible; publish
all mismatches rather than implying simultaneous equality of every
resource.

Measure modal occupation transfer, relative-phase evolution and the
inferred quartic tensor as the cycle duration changes. The arithmetic
average predicts density terms with self coefficient \(S_j\) and cross
coefficient \(2S_{jk}\). It also predicts removal of some linear
off-diagonal terms. Verify that required task mixers have not simply
been averaged away.

\textbf{Acceptance.} The measured residual tensor and finite-cycle
defects fall within the preregistered budget, with a nonzero retained
relative nonlinear response. Merely preserving a constant output or an
empty register is insufficient.

\section{Gate 4. Concurrent inference with an adversarial
register}\label{gate-4.-concurrent-inference-with-an-adversarial-register}

\textbf{Question.} Is a previously calibrated nonlinear relative-state
computation stable while overlapping modes become active?

Choose protected input histories expressed through compatible bilinear
controls or prepared relative states. Co-compile the desired controls
with the interaction code. Calibrate the old input-output function
before any concurrent test. Select held-out inputs not used for code
optimization.

Drive the adaptive register with changing occupation, relative phase,
coherent input, dissipation and memory-dependent detuning. Compare
protected Stokes trajectories with idle-register runs. Record both the
raw complex-field difference and logical readout difference; a large raw
difference is compatible with the stated mechanism. Require nonzero
protected occupation and include readout uncertainty.

The reference static full-field test reduces a disturbance of 0.39875 to
\(1.5906\times10^{-5}\) with 64 palindromic cycles and 1,664 nonlinear
intervals. Those figures are model targets only. Test at equal wall time
and control energy before claiming a performance gain.

\textbf{Failure controls.} Break the cross-row symmetry; inject an
additive phase-referenced protected drive; omit the upper-branch code;
increase operation time and concurrent occupation. Report sensitivity
instead of discarding failed operating points.

\section{Gate 5. Persistent finite memory and concurrent physical
writes}\label{gate-5.-persistent-finite-memory-and-concurrent-physical-writes}

\textbf{Question.} Can a real retained variable change the adaptive
computation without corrupting the protected one?

Specify a finite configurational state, write mechanism, retained
optical response, saturation and loading path. Measure the distribution
of signed write outcomes, retention versus time and temperature, read
disturbance, endurance and correlations. Independent binary slots are a
reference model, not assumed chemistry.

Observe real write transients in photon and exciton components. Include
them during the continuous protected evolution; do not pause the
computation and later call the experiment concurrent. Compare a
gate-protected write with the same write under no or incomplete control.
A physically realized write can change mode shape or nonlinear loss as
well as detuning and must be included in the tensor model.

E12 has one 65,536-slot endpoint memory, noisy observations and a
declared exciton-only write transient. It is not a chemistry simulation.
Its four predefined runs are useful control references, but physical
qualification needs confidence intervals, failed writes and all retry
costs.

\section{Gate 6. Closed-loop learning and admissible model
modification}\label{gate-6.-closed-loop-learning-and-admissible-model-modification}

\textbf{Question.} Does protection survive a nontrivial adaptive task,
including noisy sensing and finite actuation?

Reproduce the fixed phase-tracking task first. Next use a learnable task
with held-out performance, a changed target and a protected old task.
Report training examples, parameter count, microscopic memory resources,
controls, total updates, sensing energy and rejected operations. Compare
a matched electronic/photonic controller where appropriate.

The optical state should evolve continuously throughout a sequence. If
operations are staged at cycle boundaries, state that sampling contract.
Finite-duration native gates from Gate 2 must be integrated into the
same loop before the loop is a material-faithful realization of the
phase compiler. Performing separate successful gate and memory
experiments is not equivalent.

Self-modifying architecture or learning rules require an additional
verified update mechanism and a protected acceptance policy. The
supplied controller is fixed; it does not establish autonomous invention
of its own optimizer, universal intelligence or ASI.

\section{Gate 7. Bounded calibration and manufacturing
variation}\label{gate-7.-bounded-calibration-and-manufacturing-variation}

\textbf{Question.} Can shared control repair useful specimen variations
without pretending to address arbitrary voxels?

Measure the incident-mode-to-material actuation map, readout map and
their uncertainty. Apply bounded shared commands to correct cross-phase
row differences while preserving already calibrated nonlinear task
coefficients. Enforce material positivity, saturation, temperature and
command limits.

Sample a predefined specimen population and report every sample. E5
gives ten successes and ten failures despite full row rank. Experimental
yield must similarly include the failures. Re-measure mode shapes after
repair; an affine coefficient change is not valid when the optical basis
shifts appreciably.

A proposed 3D growth or self-assembly process must be characterized at
this gate using actual volumetric specimens. Scaffold evidence {[}R12{]}
alone does not establish phase control or compatible exciton-memory
integration.

\section{Gate 8. Full energy, latency, retention and causal
accounting}\label{gate-8.-full-energy-latency-retention-and-causal-accounting}

\textbf{Question.} Is the complete qualified function beneficial
compared with alternatives at matched accuracy and throughput?

Include input preparation, optical sources, sustained pumps, all native
pulses, phase calibration, A/D and D/A conversion, electronic control,
learning writes, memory retrieval, cooling, error recovery and idle
power. Distinguish one nonlinear interval, one cycle, one inference and
one training update. Report total elapsed time; zero-duration ideal
gates cannot be counted as a physical clock.

Measure retention and drift over the claimed continuous-operation
window. An initially protected common phase does not suppress thermal
evolution of unequal coupling coefficients. Recalibration power and
downtime belong to the budget.

A quadrillion-sized implicit feature space is not a quadrillion
independently stored parameter vector. An unbounded software stream is
not exact unlimited storage in a finite crystal. Test finite task
accuracy and archive retrieval costs directly.

No FTL claim follows from coupled-mode simulations. Any anomalous
propagation test requires a fresh unpredictable input created after the
receiver's prepared state is fixed, calibrated clocks, an explicit
spacetime separation, exclusion of leakage and pre-shared answers, and
enough independent trials to establish information before a causal
front. This protocol does not predict such an effect.

\section{Gate 9. Independent replication and monolithic
integration}\label{gate-9.-independent-replication-and-monolithic-integration}

\textbf{Question.} Can another group reproduce the benefit, and does it
survive integration into one volumetric body?

Release raw data, control waveforms, model fits, calibration
uncertainty, blind test definitions and total resource accounting.
Replicate the phase-code contrast, the branch-complete versus incomplete
control, and the combined learning test on a separately produced
specimen. Record batch yield.

Only after these results should a monolithic 3D implementation be
assessed for reduced task-specific patterning, manufacturing cost and
application-level benefit. A lithography-free internal scaffold can
still depend on lithographic boundary optics, electronics or sources;
state the actual process boundary.

\textbf{Transformative threshold.} A credible high-impact claim requires
a useful adaptive computation whose protected task survives real
updates, at a demonstrably advantageous total cost or capability, across
reproducible specimens. A finite model identity or large simulated mode
count does not meet this threshold.

\section{Data collection and statistical
contract}\label{data-collection-and-statistical-contract}

Use separate calibration and held-out records. Fix primary metrics and
seed/sample counts before selecting control strengths. Publish raw and
phase-invariant errors with occupation and detector noise. For
stochastic trials, report confidence intervals and count all failures.
Bootstrap or distributional assumptions should be stated; do not turn
four successful model seeds into a device-reliability estimate.

Each record should contain specimen identifier, complete parameter
snapshot, task/input identifier, phase/native pulse sequence, write
count or estimated state, timestamps, measured outputs, energy channels,
calibration revision and rejection flags. The packaged CSV schemas
illustrate model outputs, not a universal hardware acquisition standard.

\section{Minimal reproducibility
ladder}\label{minimal-reproducibility-ladder}

The release runs thirteen synthetic experiment groups and a regression
suite. The standalone source contains the mathematical models and seeds.
Reproduce the exact averaging tensor, numerical order, full-branch
counterexample, finite-count moments and integrated-feedback refinement
before adapting the code to a different material. Preserve negative
controls. Changing the susceptibility model, native gate set or readout
contract is a new experiment that must be identified separately.

\clearpage

\section{References and verification
scope}\label{references-and-verification-scope}

Sources were checked on 17 September 2026. The scope below distinguishes
bibliographic/abstract verification from a full-paper audit.
Mathematical proofs in this release are supplied independently; the
references identify antecedents and component evidence. No third-party
figures or full papers are redistributed.

\textbf{{[}R1{]}} J. M. Silver and P. Del'Haye. \emph{Generalized theory
of optical resonator and waveguide modes and their linear and Kerr
nonlinear coupling}. Physical Review A \textbf{105}, 023517 (2022). DOI:
10.1103/PhysRevA.105.023517. Author preprint:
\url{https://arxiv.org/abs/2103.10479}. Checked: author abstract and
journal metadata. Role: self/cross-phase and four-wave-mixing theory;
the present scalar onsite model is not a universal susceptibility
tensor.

\textbf{{[}R2{]}} P. Zanardi and M. Rasetti. \emph{Noiseless Quantum
Codes}. Physical Review Letters \textbf{79}, 3306 (1997). DOI:
10.1103/PhysRevLett.79.3306.
\url{https://arxiv.org/abs/quant-ph/9705044}. Checked: publisher
abstract and author record. Role: established symmetry-protected
encoding; not claimed new here.

\textbf{{[}R3{]}} E. Knill, R. Laflamme and L. Viola. \emph{Theory of
Quantum Error Correction for General Noise}.
\url{https://arxiv.org/abs/quant-ph/9908066} (1999). Checked: author
abstract. Role: operator-algebra/noiseless-subsystem framework.

\textbf{{[}R4{]}} L. Viola, E. Knill and S. Lloyd. \emph{Dynamical
Generation of Noiseless Quantum Subsystems}. Physical Review Letters
\textbf{85}, 3520 (2000). DOI: 10.1103/PhysRevLett.85.3520.
\url{https://arxiv.org/abs/quant-ph/0002072}. Checked: publisher and
author abstracts. Role: dynamical engineering and encoded control are
established prior art.

\textbf{{[}R5{]}} G. Strang. \emph{On the Construction and Comparison of
Difference Schemes}. SIAM Journal on Numerical Analysis \textbf{5},
506--517 (1968). DOI: 10.1137/0705041.
\url{https://epubs.siam.org/doi/10.1137/0705041}. Checked: publisher
metadata. The original publication is 1968; its later online-publication
date is not the discovery date. Role: symmetric splitting antecedent.

\textbf{{[}R6{]}} J. Singer. \emph{Perfect Difference Sets}.
Transactions of the New York Academy of Sciences \textbf{28}, 883--888
(1966). DOI: 10.1111/j.2164-0947.1966.tb02392.x.
\url{https://doi.org/10.1111/j.2164-0947.1966.tb02392.x}. Checked:
publisher metadata and reference list, which cites Singer's 1938
finite-projective-geometry construction. The explicit modular sets used
here are independently verified in code. Neither perfect difference sets
nor the set (0,1,3,9) modulo 13 is claimed as a new combinatorial
discovery.

\textbf{{[}R7{]}} J. Feng, E. O. Ilo-Okeke, A. N. Pyrkov, A.
Askitopoulos and T. Byrnes. \emph{Sensitive detection of entanglement in
exciton-polariton condensates via spin squeezing}. Physical Review A
\textbf{104}, 013318 (2021). DOI: 10.1103/PhysRevA.104.013318.
\url{https://journals.aps.org/pra/abstract/10.1103/PhysRevA.104.013318}.
Checked: publisher abstract. Role: polariton nonlinear spin dynamics are
established, not invented here.

\textbf{{[}R8{]}} N. S. Voronova, A. A. Elistratov and Yu. E. Lozovik.
\emph{Detuning-Controlled Internal Oscillations in an Exciton-Polariton
Condensate}. Physical Review Letters \textbf{115}, 186402 (2015). DOI:
10.1103/PhysRevLett.115.186402.
\url{https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.115.186402}.
Checked: publisher abstract. Role: exciton/photon relative dynamics and
the risk of replacing an exciton shift by a polariton phase.

\textbf{{[}R9{]}} L. Dominici and collaborators. \emph{Ultrafast control
of Rabi oscillations in a polariton condensate}.
\url{https://arxiv.org/abs/1408.1289} (2014). Checked: author abstract.
Role: coherent-control component precedent, not evidence of
high-fidelity arbitrary branch-mode phase gates for this architecture.

\textbf{{[}R10{]}} F. Ashtiani, M. H. Idjadi and K. Kim.
\emph{Integrated photonic neural network with on-chip backpropagation
training}. Nature \textbf{651}, 927--932 (2026). DOI:
10.1038/s41586-026-10262-8. Published 18 March 2026.
\url{https://www.nature.com/articles/s41586-026-10262-8}. Checked:
publisher abstract and extended-data descriptions, including the
microcontroller, ADC/DAC and driver roles. Role: important current
experimental baseline; photonic learning alone is not novel.

\textbf{{[}R11{]}} V. V. Akshay, M. V. Altaisky and N. E. Kaputkina.
\emph{Quantum reservoir networks based on decoherence-free subspaces}.
\url{https://arxiv.org/abs/2605.27427} (submitted 21 May 2026). Checked:
author abstract. Role: close prior art combining reservoir computation
and protected subspaces. Its room-temperature implications are the
authors' proposal, not adopted here as demonstrated material
performance.

\textbf{{[}R12{]}} G. Posnjak and collaborators. \emph{Diamond-lattice
photonic crystals assembled from DNA origami}. Science \textbf{384},
781--785 (2024). DOI: 10.1126/science.adl2733.
\url{https://arxiv.org/abs/2310.10884}. Checked: author abstract and
journal metadata. Role: self-assembled three-dimensional scaffold
precedent; not a demonstrated integrated computing/memory platform.

\textbf{{[}R13{]}} A. Opala and collaborators. \emph{Room temperature
exciton-polariton neural network with perovskite crystal}.
\url{https://arxiv.org/abs/2412.10865} (2024). Checked: author abstract.
Role: polariton-neural-component precedent. The source does not verify
compatibility with the complete ISOPHASE stack.

\textbf{{[}R14{]}} T. Chen, I. Goodfellow and J. Shlens. \emph{Net2Net:
Accelerating Learning via Knowledge Transfer}.
\url{https://arxiv.org/abs/1511.05641} (2015; revised 2016). Checked:
author abstract. Role: function-preserving model expansion antecedent.
Activating a finite modal reserve is a distinct restricted construction,
not the invention of preserving functions during model growth.

\textbf{{[}R15{]}} A. Metelmann and A. A. Clerk. \emph{Nonreciprocal
quantum interactions and devices via autonomous feedforward}. Physical
Review A \textbf{95}, 013837 (2017). DOI: 10.1103/PhysRevA.95.013837.
\url{https://journals.aps.org/pra/abstract/10.1103/PhysRevA.95.013837}.
Checked: publisher abstract. Role: a possible separate, resource-bearing
route to feedforward; the common-phase coupling in ISOPHASE does not
provide a free logical read channel.

\end{document}