The trillion-fold trick: why gold refuses to tarnish, and what materia – Good Machine

TL;DR

Tulane University chemical engineers Santu Biswas and Matthew Montemore, publishing in Physical Review Letters, show that gold's famous refusal to tarnish is not because gold dislikes oxygen. It's because gold's surface atoms physically reshuffle themselves, in seconds, into a denser hexagonal "herringbone" pattern — and that pattern makes it a billion to a trillion times harder for oxygen to react.
The paper was published on 21 May 2026. It went mainstream this weekend via a Tulane press release amplified by ScienceDaily, Scientific American, Science News, and the Times of India — a familiar science-virality cycle where the discovery is old, the audience is new.
The finding is one node in a much bigger week for materials science: a Max Delbrück Center team in Berlin unveiled a metamaterial MRI antenna that produces sharper images of the brain and eye; a Finnish team at Jyväskylä and Aalto finally built a 2D topological crystalline insulator that had been predicted for over a decade; and an Osaka University team used a neural network to describe the hidden structure of supercooled water.
The connecting thread across all four: the arrangement of atoms — not the chemistry of the atoms — is doing the work.
For a general reader, the practical thing to know is that "gold doesn't rust" now has a mechanism, and that mechanism suggests a route to better catalysts for vinyl-acetate manufacture, propylene oxide, and vehicle-exhaust cleanup. For a curious reader, the deeper thing is that this is what materials science looks like when it grows up: less periodic table, more geometry.

One trillion-fold

Take a chunk of gold. Cut a fresh face. In that first moment — before anything else happens — the newly exposed surface is a simple square lattice of atoms, and it is chemically vulnerable. If nothing intervened, oxygen molecules in the air would split apart on it and start to bind. Gold would tarnish. Wedding rings would go dull. Byzantine coins would look like Byzantine copper.

Nothing intervenes for very long. Within seconds, the atoms on that face lift, jostle, and re-lock themselves into a denser hexagonal weave — a pattern that, viewed under a scanning tunneling microscope, zigzags in ridges called herringbone reconstruction. This is not a chemical reaction. It is a geometric one. The atoms are the same. Only the arrangement changes.

That arrangement makes gold roughly one billion to one trillion times slower at splitting molecular oxygen. Not somewhat more resistant. Not an order of magnitude. Twelve orders of magnitude. That is the number Biswas and Montemore extracted from quantum-mechanical calculations of the two most common gold surface types.

Which is why gold in the Tutankhamun burial mask looks, at three-and-a-half thousand years, the way it did the day it was hammered.

What happened

Santu Biswas, a postdoctoral fellow, and Matthew Montemore, associate professor in Tulane's chemical engineering department, ran density-functional-theory simulations across the two most common gold crystal faces — Au(100) and Au(111) surfaces that, in their un-reconstructed forms, present atoms in tidy square arrays. They calculated the energy required to split O₂ on those surfaces, both before and after the atoms had reshuffled into the hexagonal reconstruction.

The result surprised even them. The square arrangement was reasonably good at splitting oxygen. The hexagonal arrangement, once formed, was not — because for O₂ to break apart on it, the surface would first have to distort back toward the original square, a thermodynamic hurdle that gold, at ambient conditions, simply doesn't clear. The reconstruction pulls extra gold atoms up out of the bulk and packs them into the surface, closing the geometric doors that oxygen would otherwise walk through.

The paper is titled — with the deadpan precision physics journals prefer — Role of Reconstruction in the Inertness of Gold toward Oxygen. It ran in Physical Review Letters on 21 May. The virality wave arrived this weekend.

What it actually means

For a century, the answer to why is gold noble has been "because gold holds its electrons tightly." That's true, and insufficient. Gold's electrons are held closely enough to make it slow to oxidise, but not slow enough to explain what we actually observe, which is that a gold ring worn against skin for fifty years still shines. The gap between what the electrons predict and what your grandmother's ring demonstrates is the gap Biswas and Montemore just closed.

The mechanism they closed it with is not chemical. It is architectural. Gold's inertness is a structural property of how its surface atoms arrange themselves, not a bulk property of what gold is.

That distinction matters far beyond jewellery. It matters because the last five years of materials science have been telling the same story in different accents. Twisted bilayer graphene conducts, insulates, or superconducts depending not on what atoms compose it — it's just carbon — but on the angle between the two layers. Metamaterials bend electromagnetic waves in ways their constituent materials cannot, because their sub-wavelength geometry rewrites the effective physics. Topological insulators carry current on their edges but not through their bulk, a difference that lives in the geometry of their electron wavefunctions, not in their chemistry.

Gold-not-rusting is now a member of that family. Which is why this week's cluster of announcements matters more together than separately.

Why this week is louder than one paper

Materials-science stories go viral in isolation. This one shouldn't. On the calendar it landed on, three other announcements deserve to be read alongside it, and each amplifies the same underlying point.

Berlin, 10 July. A team at the Max Delbrück Center led by doctoral researcher Nandita Saha, under Professor Thoralf Niendorf, unveiled a new MRI antenna built on metamaterial principles that guides radiofrequency fields more efficiently around the deep structures of the brain and the delicate tissues of the eye. Existing scanners; same magnetic field; sharper images. The advance is not a new metal or a new molecule. It is a new geometry for the antenna, engineered at the sub-wavelength scale so that RF propagates the way the imaging needs it to. Metamaterial physics quietly making a hospital scanner see better.

Jyväskylä and Aalto, 11 July. A Finnish group has just experimentally realised a two-dimensional topological crystalline insulator — a class of quantum material predicted in theory more than ten years ago and, until now, absent from any laboratory. They grew a tin telluride film on a substrate whose lattice mismatch compresses the film into strain, and that strain — the geometric distortion of the crystal — is what stabilises the topological edge states. Adjust the strain, adjust the electronic behaviour. This is a candidate platform for room-temperature quantum electronics, which is a claim to hold loosely, but the physics is clean: geometry, again, is doing the work chemistry cannot.

Osaka, 12 July. A University of Osaka team publishing in Communications Chemistry built a neural-network framework to test different descriptors of the structure of supercooled water. Water is chemically boring and structurally strange; its anomalies — density maximum at 4°C, expansion on freezing, unusual heat capacity — are all consequences of how the molecules are arranged, not what they are made of. The AI here is a tool for adjudicating between competing structural theories, not a headline. The headline is that water's weirdness is, once again, a geometry problem.

Four different labs. Four different problems. One quiet consensus: the interesting variable is arrangement.

The quieter story: what happens if you stop the reconstruction

The Tulane paper is not really a paper about jewellery. It is a paper about how to design better catalysts.

Gold is already a workhorse in several industrial oxidation reactions — gold-palladium catalysts for producing vinyl acetate (the precursor to countless plastics and adhesives), gold nanoparticles for oxidising carbon monoxide out of vehicle exhaust, gold-based systems being developed for propylene oxide manufacture. In every case, the goal is the opposite of what protects your wedding ring: you want the gold to split oxygen and hand it off to something else.

Biswas and Montemore's result flips the design question. If the reconstruction is what makes gold inert, then preventing or reversing that reconstruction — chemically, mechanically, or by choosing a support that stabilises the square lattice — could turn ordinary gold into a much better oxidation catalyst. Biswas notes explicitly that adsorbates layered on top of the surface can suppress reconstruction, at which point "the gold can easily oxidize." From an industrial-chemistry perspective, that is a design principle, not a curiosity.

Which is where the wonder becomes the tool.

Where the numbers stop agreeing with the timeline

One honest thing to flag. The paper is dated 21 May 2026. It is being reported globally as new on 11–12 July 2026 — two months later — because Tulane released a press summary this weekend that ScienceDaily, Scientific American (which had actually covered the paper at publication), Science News, the Times of India, and a fanning-out cloud of aggregators picked up in sequence.

This is normal. It is also a small case study in how science virality works: the underlying discovery is fixed, but the audience is variable, and PR cycles decide when a general reader hears about a physics paper. If you are reading this and wondering "did I miss something?" — you didn't. You are catching the second wave. The first wave was quiet and technical. The second wave is loud and general. Both are the same paper.

What this means for you

Addressed to a general reader.

If you own gold jewellery, the practical answer is: nothing changes. Your ring was already fine. It is now fine for a reason.
If you follow materials science, chemistry, or condensed-matter physics, the useful reframe is: geometry-first thinking has crossed from exotic (graphene, topological insulators, metamaterials) into elemental (the surface of a plain gold crystal). The next generation of catalyst-design papers will lean on surface-reconstruction control the way the last generation leaned on alloying and nanoparticle support.
If you are a science communicator or teacher, the classroom line has just changed. "Gold is inert because its electrons are held tightly" is no longer sufficient. "Gold is inert because its surface atoms rearrange into a pattern that oxygen cannot break" is the new sentence, and it is more accurate, more visual, and more teachable.
If you are curious about where this leads, watch industrial-catalysis literature over the next twelve to twenty-four months for gold-based oxidation catalysts designed around suppressed reconstruction. That is where the paper stops being about wedding rings and starts being about vinyl-acetate plants and exhaust systems.

For most readers, the honest answer is that this is a knowledge update rather than an action update, and that is fine. Some of the best science reporting is the kind that leaves you with a new sentence in your head, not a new to-do on your list.

Uncertainty ledger

Scope of the mechanism. The paper studies two of gold's most common surface types. A third (already known to be inherently hexagonal) was not the focus. Real gold objects present many crystal faces, grain boundaries, and defects. How much of your ring is protected by this exact mechanism versus adjacent ones is not resolved.
Alloys. The result is for pure gold. Real jewellery is usually alloyed with silver, copper, palladium, or nickel, and those alloys tarnish differently. The Tulane mechanism explains pure-gold behaviour cleanly; it does not automatically explain 18-karat gold.
Catalyst translation. "Reversing reconstruction to make gold a better oxidiser" is a design principle, not a product. Whether specific adsorbates or supports can reverse reconstruction and stabilise a working catalyst at industrial temperatures and pressures is an open experimental question.
The wider cluster. The metamaterial MRI advance, the 2D topological crystalline insulator, and the supercooled-water AI descriptor work are each early-stage. Framing them together, as this piece does, is an editorial call about a pattern — not a claim that any single one of them is a settled technology.

Bottom Line

Gold's refusal to tarnish is not chemistry. It is architecture. The surface atoms reshuffle themselves into a geometry that shuts oxygen out by twelve orders of magnitude, and that answer — landed by a two-author team at Tulane in Physical Review Letters, and going global this weekend — is the plainest evidence yet that materials science has quietly moved its centre of gravity from what atoms are to how atoms are arranged. Read it alongside this week's metamaterial MRI antenna in Berlin, the long-predicted 2D topological insulator finally built in Finland, and the neural-network dissection of supercooled water in Osaka, and the pattern is unmistakable: geometry is winning.

Sources

Tier 1. Biswas, S., & Montemore, M. M. Role of Reconstruction in the Inertness of Gold toward Oxygen. Physical Review Letters, 21 May 2026. DOI: 10.1103/g3bc-t1qv
Tier 1. Scientific American — Scientists Discover Why Gold Doesn't 'Rust' (May 2026 primary coverage)
Tier 1. Science News — Physics Explains Why Gold Stays Pristine (22 May 2026)
Tier 2. ScienceDaily / Tulane University press release — Why gold never tarnishes has finally been explained (11 July 2026)
Tier 2. ScienceDaily — New MRI breakthrough reveals the brain and eye like never before (10 July 2026) — Max Delbrück Center, Berlin; Saha & Niendorf
Tier 2. ScienceDaily — Physicists finally build a quantum material predicted more than a decade ago (11 July 2026) — University of Jyväskylä & Aalto University, Finland
Tier 2. SciTechDaily — AI Just Uncovered a Hidden Secret Inside Water (12 July 2026) — University of Osaka, Communications Chemistry
Tier 3. Times of India — mass-audience amplification of the Tulane finding (12 July 2026)