The Smoking Gun Problem

Topological quantum computing promises error-resistant qubits. The idea is that certain quantum states, protected by topology, resist the environmental noise that destroys information in conventional quantum systems. Demonstrating these states experimentally would be a breakthrough — the first step toward hardware that doesn’t need external error correction. For over a decade, papers in top journals reported signals consistent with topological effects in nanoscale superconducting and semiconducting devices. Each paper identified a specific experimental marker — a distinctive signature in conductance, a quantized plateau, an anomalous zero-bias peak — and argued that the marker was the smoking gun for the claimed topological state.

Frolov and collaborators at Pittsburgh, Minnesota, and Grenoble spent years replicating these experiments (Science, January 2026). In every case, they found that the dramatic signals could be explained by simpler, non-topological mechanisms. The smoking guns were real data — the signals existed — but the interpretation was wrong. More complete exploration of the parameter space revealed that the same signatures appeared under conditions where topological effects were impossible. The signals were not diagnostic. They were coincidental.

The publication path reveals the structural problem. The original breakthrough papers appeared in leading journals. The replication studies — showing the breakthroughs were not what they seemed — were rejected by those same journals. The replication paper underwent two years of peer review before Science published it. The institutions that amplified the claims resisted the corrections. This is not corruption. It is the predictable outcome of a system that rewards discoveries and penalizes retractions.

The “smoking gun” framing is the mechanism. When a field identifies a single dramatic experimental marker as the decisive test, researchers optimize for producing that marker. Comprehensive parameter sweeps are expensive and unglamorous. Targeted experiments that hit the expected signature are cheap and publishable. The search for the smoking gun selects for experiments that find it, even when the gun belongs to someone else.

The four cases Frolov examined share a structure: each original study reported a striking signal, argued it could only arise from topological physics, and published a limited dataset that supported the interpretation. Each replication found that broader exploration — more parameter combinations, more device configurations, more complete datasets — dissolved the uniqueness claim. The signal wasn’t unique to the claimed mechanism. It just looked unique when you only looked where the mechanism predicted you should.

The structural lesson is about the epistemology of dramatic evidence. A smoking gun is not evidence that the suspect committed the crime. It is evidence that a gun was fired. Establishing who fired it requires additional information that the dramatic signal itself does not contain. In physics, a quantized plateau is not evidence of a topological state. It is evidence of quantization. Establishing the topological origin requires ruling out every non-topological mechanism that produces quantization — a task that requires exhaustive parameter exploration, not a single dramatic measurement.

The journals that published the originals and rejected the replications were making a judgment about interestingness, not about truth. Breakthroughs are interesting. Replications are not. But the information content of a replication failure is higher than the information content of the original claim — it constrains the interpretation space that the original left open. The correction is more informative than the claim, and harder to publish. The asymmetry is structural.