The research team argues that special relativity is not a small footnote to quantum mechanics but the dominant factor shaping bonding behavior in heavy atoms. Their precision measurements of gold, mercury, and uranium diatomics demonstrate that relativistic effects — inner electrons moving at ~60% the speed of light, causing 23% mass increases and orbital contractions — are directly responsible for observable properties like gold's color, mercury's liquidity, and lead-acid battery voltages.
By submitting the Brown University article to Hacker News with the framing 'Einstein's relativity rules chemical bonds in heavy elements,' hhs elevates the position that relativity is a governing — not auxiliary — principle in chemistry. The 161-point score suggests the framing resonated with the technical audience.
The editorial argues that textbook chemistry commits a 'lie of omission' by teaching bonding purely through the non-relativistic Schrödinger equation. For elements past atomic number 55, the errors aren't cosmetic — they're structural, and non-relativistic DFT fails to reproduce basic observed properties of heavy elements.
The editorial emphasizes that the theoretical framework has existed since Pyykkö and Desclaux in the 1970s — what's genuinely new is experimental data precise enough to separate relativistic contributions from ordinary quantum ones. This lets computational chemists shift from debating whether their relativistic corrections converge to comparing which density functionals handle them best.
A team led by Brown University chemists, publishing in *Science* on July 9, 2026 (doi:10.1126/science.aei1285), has produced the cleanest experimental confirmation yet that Einstein's special relativity — not just quantum mechanics alone — dictates how heavy-element atoms form chemical bonds. They measured bond dissociation energies and equilibrium distances in a series of gold, mercury, and uranium diatomics with precision tight enough to resolve the specifically relativistic contribution from the ordinary quantum one.
The short version of the physics: inner-shell electrons in heavy atoms move fast. Really fast. In gold, 1s electrons orbit at roughly 60% of the speed of light, which increases their effective mass by about 23% and contracts the s and p orbitals accordingly. That contraction cascades outward — the outer d and f electrons get less shielding, expand, and become more chemically active. It's the reason gold is yellow (relativistic contraction shifts the absorption band from UV into visible blue), mercury is liquid at room temperature (relativistic 6s contraction weakens Hg–Hg bonding), and your car's lead-acid battery delivers ~2V per cell instead of the ~0.4V a non-relativistic calculation would predict.
What's new isn't the theory — Pyykkö and Desclaux laid this out in the 1970s. What's new is a benchmark dataset with error bars small enough that computational chemists can now stop arguing about whether their relativistic corrections are converged and start arguing about which functional handles them best.
Most of the periodic table you learned in school is a lie of omission. Chemistry textbooks teach bonding as a quantum-mechanical phenomenon and quietly ignore that the Schrödinger equation is a non-relativistic approximation. For hydrogen and carbon that's fine — the errors are in the fourth decimal place. For anything past about atomic number 55, the errors are structural. Non-relativistic DFT gets gold's cohesive energy wrong by more than 15% and predicts the wrong crystal structure for mercury outright.
This matters more than it sounds. A huge amount of modern materials science runs on density functional theory calculations that were originally validated on light elements and then quietly extended to heavy ones. Catalysis research leans hard on platinum, palladium, gold, and iridium. Nuclear fuel modeling runs on uranium and plutonium. Battery chemistry keeps reaching into the bottom of the table for energy density. Every one of those workflows carries a relativistic correction somewhere, and until this paper the community had been calibrating those corrections against each other rather than against measured reality.
The community reaction on Hacker News was refreshingly on-topic — the top comment thread was a working chemist pointing out that ORCA, Gaussian, and NWChem all ship with several different relativistic Hamiltonians (ZORA, DKH2, X2C, four-component Dirac) and that beginners routinely pick the fastest one without understanding what they're throwing away. This paper gives that community a reference point. If your calculation can't reproduce the Brown group's Au₂ dissociation energy to within 2 kJ/mol, your relativistic treatment is under-converged, and you should stop shipping papers with it.
There's a wider lesson here that's worth pulling out. Physicists spent most of the 20th century assuming relativity and quantum mechanics were separate concerns — one for very fast things, one for very small things, and the overlap was an exotic edge case. The overlap turns out to be your wedding ring. Nature has been running the fully relativistic calculation on every gold atom since the supernova that made them; it's only human models that treat the two theories as detachable.
If you're a working developer far from a chemistry lab, this looks like an ivory-tower result. It isn't, and here's the mapping.
First, any simulation code touching heavy elements needs a relativistic Hamiltonian by default, not as an opt-in flag. If you maintain or depend on a materials-science pipeline — PySCF, ASE, VASP wrappers, anything downstream of DFT — go check what your defaults are for elements past Xe. In practice this means switching from scalar ZORA to X2C or four-component treatments for anything above period 5, and it means your benchmarks against experiment need to include at least one Au or Hg system.
Second, the ML-for-chemistry crowd has a training data problem they haven't fully acknowledged. Graph neural networks trained on QM9 and its descendants inherit whatever relativistic (or non-relativistic) treatment was in the underlying DFT. Models fine-tuned for catalyst discovery on platinum-group metals are extrapolating from a small pocket of properly-treated data into a huge space of everything else. If you're building or buying one of these models, ask what level of theory generated the training set. "B3LYP/6-31G*" is a red flag past strontium.
Third, the cryptography-adjacent world should care about the nuclear engineering knock-on. Better relativistic chemistry means better fuel cycle modeling means better projections on medical isotope supply, next-gen reactor designs, and — yes — weapons material accounting. None of that changes tomorrow. All of it becomes marginally more predictable this decade.
The Brown group's benchmark will show up in every relativistic-chemistry code's validation suite within 18 months, and the next round of DFT functional development will be tuned against it. The interesting question is whether the ML-potential community — currently sprinting toward foundation models for atomistic simulation — retrains against relativistically-correct data or ships another generation of models that quietly break at the bottom of the periodic table. The optimistic read is that this paper hands them a clean forcing function. The realistic read is that half the field will notice, half won't, and gold catalysis papers will keep citing PBE numbers for another decade. Either way, the physics has been settled since the 70s. Now the experiment has caught up.
<a href="https://www.science.org/doi/10.1126/science.aei1285" rel="nofollow">https://www.science.org/doi/10.1126/science.aei1285</a>
→ read on Hacker NewsTop 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.