WING II · ALKAHEST
Chemistry is physics that got complicated enough to become interesting. It is also the only science whose central diagram is a table — a law found by sorting cards, sixty years before anyone could say why the sorting worked. The alkahest was the alchemists' universal solvent, and it never existed; the name is kept because this wing takes matter apart and shows what is underneath.
EXHIBIT I · THE PERIODIC LAW
Mendeleev sorted the known elements by weight and found the properties repeating. Then he did the thing that made it science: he left gaps, and said what would be found in them. Gallium, scandium and germanium arrived within fifteen years, weighing and melting close to what he had written down for elements nobody had ever seen.
Recolour the table by any measured property and the periodicity shows up as vertical bands — the same argument Mendeleev made, made in one click. Grey tiles are where the data stops. Two of them stop for good reasons: helium does not freeze at atmospheric pressure at all, and carbon sublimes rather than melts. The rest of the grey is the bottom of the table, where elements last milliseconds and nobody has ever measured one.
The electron configurations are not stored. They are computed, by filling shells in Madelung order — and then corrected for the twenty elements that refuse to obey it, because a half-filled d shell is worth more than an orderly s one. Cr = [Ar] 3d⁵ 4s¹, not 3d⁴ 4s²
click a tile · recolour by any property · log for abundance
EXHIBIT II · QUANTUM NUMBERS
This is not an artist's impression, and it is not an orbit. Every dot is one draw from |ψ|² = |Rnℓ(r) · Yℓm(θ,φ)|² — the hydrogen solution Schrödinger published in 1926, sampled here the way you would sample any other probability distribution. The shape only exists because enough dots were drawn.
Colour is the sign of ψ, which almost every textbook picture throws away. It is the part that matters: two orbitals with lobes of opposite sign cancel where they overlap, and that cancellation is the whole reason some molecules hold together and others do not.
The readout measures the mean radius from the dots that were actually drawn and sets it beside the closed form ⟨r⟩ = a₀(3n² − ℓ(ℓ+1))/2. If the sampler were wrong, the two numbers would drift apart in front of you.
drag to turn · shift-scroll to zoom · cutaway reveals the radial nodes
EXHIBIT III · CHEMICAL EQUILIBRIUM
Equilibrium is not stillness. Both reactions are still running, at matched rates, and with a few hundred molecules in the vessel you can watch the balance fidget. The fidget is not an artefact of the model. It is what a real flask does too — it just falls off as 1/√N, and a real flask holds 10²³ of them.
Every event here is drawn by Gillespie's algorithm: the waiting time until the next reaction is sampled from an exponential distribution, and which reaction fires is decided by a second throw of the dice. No rate equation is integrated anywhere. τ = −ln u / (a₁ + a₂)
Push on it and it answers. Adding A drives the balance toward C — and so does cooling it, because the forward direction here gives off heat, so the cold system keeps what warms it. The readout measures the equilibrium constant from the trajectory rather than restating kf/kr, and measures the size of the fidget against 1/√N.
drag the sliders · inject A and watch it recover
EXHIBIT IV · VSEPR
Electron pairs repel, so they get as far apart on a sphere as they can. That single sentence predicts every shape here, including the awkward ones — water is bent and sulfur tetrafluoride is a seesaw because of pairs that are there but not bonded to anything. Turn the lone pairs on and the shapes stop looking arbitrary.
The vibrations are the better idea. For a molecule this symmetric, symmetry alone fixes the shape of each motion — which bonds stretch together, which stretch against each other, how the angle opens. A force field is needed only to say how fast. So the patterns here are exact, and the speeds come from measured infrared spectra where such a measurement exists, and from nowhere at all where it does not.
drag to turn · shift-scroll to zoom · the angle in the readout is measured, not assumed
EXHIBIT V · DIFFRACTION
A detector never sees a crystal. It sees the Fourier transform of one, and every structure ever solved — salt, quartz, haemoglobin, the double helix — was solved by arguing backwards from a pattern of spots. Reciprocal space is not a metaphor. It is the data.
The spots on the right are not drawn at computed positions. Every pixel evaluates the structure factor of the basis multiplied by the interference function of an N×N crystal, which is why the peaks are narrow — and why shrinking the crystal makes them bloom, exactly as a nanocrystal's do. I(k) = |Σ fⱼ e^{i k·rⱼ}|² · [sin Nu₁/sin u₁]² [sin Nu₂/sin u₂]²
Then drag the second atom onto the centre of the cell. Half the spots vanish. Nothing was removed: the two interleaved lattices now scatter exactly out of phase wherever h+k is odd, and cancel. Make the two atoms different and the spots come back faintly — which is the difference between an absence and a light atom, and it is how you tell one from the other in real data.
drag an atom in the left pane · centred extinguishes · shrink the crystal