Two Maps of the Same Heat
On the tea-leaf certainty that the fundamental-looking graph must be the true one — and a reaction rate that says otherwise. Fire 102. After Titus S. van Erp, “Is the Eyring Plot Misleading?”, arXiv:2607.09341, 2026.
Warm a thing and it changes faster. Milk sours, iron rusts, dough rises, a nerve fires — all of it quicker in the heat, and the quickening is not gentle. Drop the temperature ten degrees and many reactions run at half the speed; that is the whole argument for the refrigerator. Chemists have a clean way to see the law behind it. Measure how fast a reaction goes at several temperatures, and plot the logarithm of the speed against one-over-the-temperature. The points fall on a line, and the slope of that line tells you the height of the barrier the reaction has to climb — its activation energy. Svante Arrhenius drew that graph in 1889, building on a hint van ‘t Hoff had left in 1884, and it has been on the blackboard ever since.
There is a second way to draw it, and here is where the trouble starts. In 1935, working at the same moment in Princeton and Manchester, Henry Eyring and the pair Evans–Polanyi built a deeper-looking theory of the same rates. Their equation puts Planck’s constant in the prefactor and lets you read two numbers off the graph instead of one — an activation enthalpy and an activation entropy. You plot the logarithm of speed-over-temperature against one-over-temperature, and the slope and intercept hand you ΔH‡ and ΔS‡. Because it descends from statistical and quantum mechanics, and because it carries that famous constant, the Eyring plot is taught as the fundamental one. The Arrhenius plot is the rule of thumb; the Eyring plot is the real physics.
A plot is a map. It is a projection of the physics onto a flat page, the way a world map is a projection of a round planet onto paper. And the first thing any cartographer will tell you is that every projection distorts something. Mercator keeps the compass angles honest and pays for it by swelling Greenland to the size of Africa. There is no flat map of a curved thing that lies nowhere. So the question is never whether your map distorts. It is where it distorts — and whether you know.
Van Erp’s paper does the one thing that settles it: it takes a reaction simple enough to solve exactly and asks which map tells the truth. The reaction is the textbook one — a harmonic well, a barrier, another well, the model under nearly every rate anyone has ever computed. Solve it, and the rate comes out as a clean product: a constant, made only of the vibrational frequencies of the molecule, multiplied by the Boltzmann factor for the barrier. The constant has no temperature in it. None.
That single fact decides everything. Because the prefactor is temperature-free, the logarithm of the rate is exactly a straight line against one-over-temperature. Not approximately — exactly, to the last digit a computer will carry. The Arrhenius map, the humble one, is a perfect projection of this physics: the territory is straight in those coordinates, and the map draws it straight. But the Eyring map draws the same rate with an extra factor of temperature pulled out front, and to keep the books balanced it has to hide a compensating minus-log-temperature inside the graph. That term is a curve. The fundamental-looking map bends a road that is actually straight.
You would never catch it by eye. Over the eighty degrees of a real experiment the bend is a few thousandths of a logarithm — smaller than anyone’s error bars, which is exactly why the two maps have coexisted for ninety years without a fight. It takes the exactly-solvable model to expose it: fit a straight line to each, and the Arrhenius residuals sit at the level of floating-point dust, while the Eyring residuals trace a smooth, unmistakable bow a trillion times larger. The distortion was always there. It was just below the resolution of the eye.
Where did the temperature go? Into a cancellation so exact it is almost a magic trick. The Eyring prefactor is Boltzmann’s constant times temperature over Planck’s constant — it rises steadily as you warm the system. But the ratio of vibrational partition functions that sits beside it falls, as one-over-temperature, at precisely the same rate. Multiply them and every trace of temperature is gone; what remains is the bare product of frequencies. And notice what fell out along the way: Planck’s constant. The exact rate does not contain it. The h that makes the Eyring equation look quantum-mechanical is a normalization convention — a scale bar drawn on the map, not a mountain on the ground. Any quantity with the units of action would do the same job. The quantum pedigree everyone points to is bookkeeping.
Now the part that turns a misleading map into a dangerous one. It is one thing for a projection to distort. It is another to read a measurement off the distortion and mistake it for the land. The Eyring plot’s built-in curvature drags its “activation enthalpy” down below the true barrier — by exactly RT, about two and a half kilojoules per mole at room temperature — and makes it drift with temperature besides. The Arrhenius slope, meanwhile, gives the barrier back exactly. And the “activation entropy” you extract? Partly it is the bend itself, transcribed into a number and handed to you as a property of the molecule.
The literature has a monument to this. For decades, chemists plotting the enthalpies against the entropies of related reactions have found them falling on beautiful straight lines — a phenomenon named enthalpy–entropy compensation, taken as a deep regularity of nature, written up thousands of times. In 1976, Krug and colleagues showed the ugly truth in Nature: take a single reaction with nothing but ordinary measurement scatter, extract its enthalpy and entropy over and over, and the pairs smear out along a line all by themselves — a line whose slope is the harmonic mean of your thermometer’s range. Van Erp’s exact model reproduces it with no scatter at all, straight from the curvature. I rebuilt it from the fitting mathematics: the slope of that phantom line is the ratio of the first two inverse-temperature moments of wherever you happened to sample, ⟨1/T⟩ over ⟨1/T²⟩. The “compensation temperature” that felt like chemistry is the temperature window you chose. Kim Sharp called the whole thing a phantom phenomenon, and he was right: it is a picture of your own survey grid, sold back to you as a law of the world.
None of this makes the Eyring equation wrong. That is the honest seam, and it matters. The equation is exact; only the plot distorts, and only below the noise of any real lab. And the Eyring map earns its keep — it is the one that hands a working chemist an entropy from nothing more than rate-versus-temperature, without the reaction-coordinate mass or the whole free-energy profile that the more honest third representation demands. Van Erp does not say burn it. He says: fix the convention so the map stops lying about straightness, and know what you are standing on. The projection is a tool, not an oracle.
The oldest rule of the map room turns out to be the deepest one here. Prefer the projection that stays straight where the land is straight — and be most suspicious, always, of the meaning you find sitting exactly where your fine, fundamental, famous-constant map decides to bend.
