Quantum Computing’s Periodic Foundation: Photonic Crystal Fabrication

The pitch is seductive: a machine that computes on atoms, solves molecular problems digital hardware cannot touch, cracks encryption, stabilizes fusion, even helps finish the theory of everything. I have spent 6 years fabricating photonic bandgap samples, and my reflex when I hear that pitch is not skepticism about the physics — the physics is sound — but a quiet tally of nanometers. Every claim in the quantum computing narrative eventually lands on a surface someone had to etch. In my reading, the gap between the one-inch equation and the machine that solves it is filled almost entirely by periodic pattern alignment precision, etching depth uniformity, and the spectral drift those two introduce. That is the angle this article takes.
 

Where the Quantum Narrative Meets the Cleanroom

The source material framing the current quantum moment rests on three load-bearing claims: Moore’s law is flattening because transistors are approaching atomic dimensions; quantum computers compute on parallel states rather than zeros and ones; and decoherence is the central obstacle. All three are defensible. None of them, in my view, gets enough attention on the manufacturing floor where the tolerance stack-up lives.
 

Take the Moore’s law argument first. Transistor scaling is indeed hitting quantum mechanical limits — electron tunneling through gates a handful of atoms across, leakage currents that no longer behave like a valve obeying an on/off commandimec-int.com+1. Industry roadmaps now speak openly about the physical ceiling: you cannot print a component smaller than an atom, and only about 1.5 nm of printable space remains depending on the elementinvestopedia.com. What gets less airtime is that the quantum computing industry inherits the same semiconductor toolchain and then asks it to hold tighter tolerances across more components, not fewer. A useful fault-tolerant photonic machine needs on the order of a million physical components manufactured with near-identical behavior, because error correction eats physical qubits at a brutal rationcbi.nlm.nih.gov. The 300-mm wafer processes that GlobalFoundries runs for PsiQuantum’s silicon photonics line are marvels of industrial consistency, and they are still wrestling with device-to-device variation at a scale my own benchwork makes me acutely sensitive toacm.org.
 

The applications catalog follows the same pattern. Nitrogen fixation for fertilizer, fusion vacuum stabilization, molecular medicine, cryptography — each of these is a statement about what a sufficiently coherent quantum system can compute. Coherence is not free. It is manufactured.
 

Periodicity as the Hidden Substrate of Qubits

Here is where my discipline earns its seat at the table. The popular explanation of quantum computing — electrons in two places at once, a mouse exploring every path in a maze simultaneously, computation across parallel universes — is accurate as far as it goes. What it omits is the physical object doing the holding. In photonic approaches, qubits ride light through waveguides, interferometers, and resonators whose useful properties come from periodic modulation of refractive index. A photonic crystal is, at bottom, a lattice: a lattice constant, a hole radius, an etch depth, repeated with the expectation that repetition means something.
 

The photonic bandgap — the frequency window where light simply cannot propagate — shifts when the lattice shifts, and the relationship is brutal in its sensitivity. A theoretical analysis of one-dimensional porous silicon photonic crystals found that reducing the lattice constant by 66.67% widened the forbidden bandgap by 66.12%inoe.ro. That is a coarse, first-order illustration of how tightly bandgap position tracks geometry. The fine-order effects are what keep me up at night. A scattering-theory study of high-index-contrast nanocavities showed that nanometer-scale imperfections at the dielectric interface blueshift the cavity mode by several meV — several orders of magnitude larger than the cavity linewidth itselfaps.org. The resonance you designed and the resonance you fabricated are different animals, and the difference scales roughly with inverse mode volume, meaning the tighter you confine the light, the worse the penaltyaps.org.
 

When the source talk describes quantum computers computing on “everything in between” zeros and ones, I picture an interferometer whose arms were supposed to be length-matched to a fraction of a wavelength. The “in between” is only as good as the lithography. Transfer-printing work on photonic crystal cavity arrays puts a number on this: the inherent nanometric tolerances of micro-fabrication platforms can induce cavity resonant wavelength shifts two orders of magnitude larger than the cavity linewidthncbi.nlm.nih.gov. Two orders of magnitude. That is the ratio between the precision we can specify and the precision we can deliver, and it is the number I would tattoo on every quantum computing roadmap if given the chance.
 

Tracing Decoherence to Etch Depth and Lattice Offset

The source material identifies decoherence as the number one problem facing quantum computers, and it is right. Where my analysis diverges is in the tracing. Decoherence in the abstract is a loss of phase relationship. In a physical device, that loss has addresses, and most of them are fabrication addresses.
 

Consider what the best current photonic platforms actually achieve. A 2025 demonstration of a manufacturable silicon-photonics quantum platform reported state preparation and measurement fidelity of 99.98%, Hong–Ou–Mandel interference visibility between independent sources of 99.50%, two-qubit fusion fidelity of 99.22%, and a chip-to-chip qubit interconnect fidelity of 99.72%ncbi.nlm.nih.gov. Those are extraordinary numbers, benchmarked across a full 300-mm foundry process with more than 20 photolithography levels and hundreds of processing stepsncbi.nlm.nih.gov. They are also numbers conditional on photon detection and not accounting for loss — a caveat the authors state plainly. The residual fraction in interference visibility is where my professional gaze goes, because in my experience that residual traces back to physical imperfections with specific, nameable origins: sidewall roughness scattering photons out of the guided mode, etch depth non-uniformity altering effective index across the chip, periodic offset accumulating along a lattice so that the ten-thousandth unit cell sits where the nine-thousandth predicted it would not.
 

The hybrid cavity literature makes the point sharper, and here I want to dwell on a specific case rather than a textbook generalization. A GaAs-on-diamond photonic crystal cavity study from 2024 achieved quality factors approaching 30,000 at a resonance wavelength of 955 nm, and exceeding 43,000 at 1520 nmarxiv.org. The authors’ central analytical finding was that the limiting factor was likely fabrication errors rather than sidewall roughness, absorption losses, or other scattering mechanisms — and that hybrid device designs simulated to Q exceeding 1 million had never been experimentally realizedarxiv.org. Their response was methodological: treat robustness to fabrication error as a design parameter in its own right, independent of designed Q, sampled statistically under realistic error distributionsarxiv.org. I find this shift both correct and overdue. It formalizes what those of us at the etch bench have known informally for years — that a design which survives 5 nm of positional noise is worth more than a design with higher ideal Q that collapses under 1 nm.
 

That is the hidden etching depth inconsistency risk, stated in the literature’s own voice. When I mark this in my own process logs, I mark it as follows: a depth variation of a few nanometers across a wafer is not a cosmetic defect. It is a resonance shift, a phase error, a decoherence channel with a manufacturing origin. The decoherence problem and the fabrication problem are, at root cause, the same problem wearing different clothes.
 

The Atomic Wall Behind Moore’s Law

The claim that transistors are approaching 20 atoms across and fail near 5 atoms has circulated for years, and the underlying physics — quantum tunneling, position-momentum uncertainty — holds up under scrutinyimec-int.com+1. Where I would push back gently on the framing is in what happens next. The narrative tends to present quantum computing as the escape route from Moore’s law, the successor that sidesteps the atomic wall entirely.
 

The fabrication reading is less clean. Photonic quantum computing does escape the transistor-scaling trap, but it substitutes a different scaling problem: component count and uniformity. PsiQuantum’s roadmap, anchored to GlobalFoundries manufacturing, targets a million-qubit utility-scale machinequantumzeitgeist.com. Xanadu’s continuous-variable approach aims for operation at or near room temperature, which would ease the cryogenic burden the source talk correctly identifies as severe — the liquid helium infrastructure required to hold superconducting detectors near absolute zeroacm.org. Both are betting that semiconductor manufacturing discipline — the same discipline that hit the atomic wall in transistors — can be redirected toward periodic optical structures with the necessary repeatability.
 

Whether that bet pays off is a genuinely open question, and here I want to be candid about the disagreement inside the field. Some researchers argue that photonic platforms’ intrinsic low noise, natural networking compatibility, and tolerance for higher operating temperatures make them the inevitable path to fault tolerancencbi.nlm.nih.gov. Others point to the punishing overhead of non-deterministic photon generation, the multiplexing complexity required to overcome low success probabilities in qubit synthesis, and the absence of real-time error syndrome correction in existing photonic demonstrations as evidence that superconducting or trapped-ion modalities will reach useful machines firstnature.com. I do not think the data settles this yet. My own work leans toward the photonic side, partly from familiarity and partly because I have seen what disciplined etching does to a bandgap spectrum when the process is controlled — but I hold that position with less certainty than the marketing materials suggest one should. There is also a live dispute about what “quantum supremacy” even demonstrates: whether specialized-task speedups on quantum hardware translate into general-purpose advantage, or whether they are impressive physics exercises that say little about the machines we actually want to build.
 

Reading the Fabrication Data Against the Promise

Let me put actual numbers from the recent literature next to the narrative claims, because the exercise is clarifying.
 

The manufacturable platform paper cited earlier also reported next-generation components: silicon nitride waveguide losses of 0.5 dB/m, splitter losses of 0.5 mdB, crossings at 1.2 mdB, fiber-to-chip coupling losses as low as 52 mdB, and — most interesting to me — photon sources robust enough to maintain indistinguishable generation across a ±400-pm resonance shiftncbi.nlm.nih.gov. That ±400-pm figure is a fabrication-tolerance statement wearing a physics costume. It says: our sources still work when the resonance drifts by roughly half a nanometer in wavelength. Anyone who has run a reactive ion etch across a 300-mm wafer knows that holding a resonance to half a nanometer is a triumph of process control, not a given. The same paper reports waveguide-integrated photon-number-resolving detectors with 98.9% single-photon median efficiencyncbi.nlm.nih.gov — a number that depends on niobium nitride deposition uniformity and etch definition of detector geometry, neither of which is forgiving.
 

A narrower case, closer to my own bench work: three-dimensional photonic crystals fabricated by simultaneous multidirectional plasma etching, demonstrated with thicknesses of 1, 1.5, and 2 lattice periods in the surface-normal direction on single-crystalline silicon wafers, showing reflectance near 100% and transmittance of −17 dB at optical communication wavelengths — evidence of a complete photonic band gapaps.org. The authors explicitly investigated fabrication tolerance from an analytical perspective and concluded that the shape of the etched holes limits performance, suggesting that improving the bandgap effect means improving the etch itselfaps.org. Deep reactive ion etching work has pushed nanopores to 320 nm diameter with aspect ratios above 20, using three-step Bosch process control to minimize scalloping, tapering, and other profile deviationsutwente.nl. Earlier electron-beam-lithography-plus-dry-etch sequences defined photonic crystals with 280 nm holes at 350 nm center-to-center spacings in GaAs, transferring patterns through 70-nm resist layers into 300-nm silicon dioxide masksresearchgate.net. Those numbers describe the frontier of what “periodic” means when you have to physically cut it into a material, and every one of them is a tolerance window that decoherence can crawl through.
 

The limitations of the framework I am applying deserve acknowledgment. Not every quantum computing architecture is photonic; superconducting qubits, trapped ions, and neutral atoms have different fabrication sensitivity profiles, and my period-first lens overweights the structures I know best. There is also a real sense in which the theory side of the field — the string theory ambitions, the hope of extracting testable numbers from the theory of everything — is orthogonal to manufacturing. A quantum computer that settled whether string theory is correct would be a scientific triumph regardless of how many nanometers of etch depth uniformity it required. My argument is not that fabrication is the only story. It is that fabrication is the story most often skipped when the future is being described.
 

Where I land, after 6 years of watching designed spectra diverge from measured ones: the one-inch equation, if it exists, will be solved on hardware measured in nanometers. The mother nature invoked in the source talk, doing quantum calculations at room temperature inside a photosynthetic leaf, has solved a decoherence problem our cleanrooms have not. We compensate with liquid helium, with process control, and with design methodologies that treat fabrication robustness as a first-class parameter rather than an afterthought. The bandgap spectrum you measure is always slightly different from the bandgap spectrum you designed, and the difference is information — it tells you where the lattice drifted, where the etch went non-uniform, where the periodic assumption broke. The quantum computing field is now, at industrial scale, learning to read that same information. The machines that emerge will be exactly as good as the tolerance stack-up allows. No better, and probably not much worse.
 

Reference Block:

• Source Reference Link: [Provided source transcript — Michio Kaku on quantum supremacy, Moore’s law limits, decoherence, and the theory of everything]
• Link Brief: A theoretical physicist’s overview of quantum computing’s rise, its applications, and the search for a theory of everything; this article adopts its narrative claims as a scaffold and interrogates them through photonic crystal fabrication data and tolerance analysis.

Content Disclaimer:

This article is for general reference only and does not constitute professional R&D guidance, production process advice or quality certification. All material performance data has specific test premises; readers should verify parameters against actual equipment and working conditions.