The formula that looked like trouble
There's a number, χ, at the centre of this. It's a ratio: how much energy the
environment delivers during one settling time, divided by how much it costs to leave.
χ > 1 means the environment is delivering more than enough. χ < 1 means it isn't. The
probability of escaping falls off exponentially in χ — and that shape is exactly the
shape of the Arrhenius equation, chemistry's rate law for how fast a reaction goes as a
function of activation energy over temperature. Line the two up and the delivered energy
sits precisely where temperature sits. So the formula seems to say: throughflow is a kind of
temperature.
That's a claim with a bad history. Physicists have spent decades trying to define an
"effective temperature" for driven systems — a stirred fluid, a shaken pile of sand, a
living cell — and it keeps almost working and then not. Temperature is supposed to be one
number: a mercury thermometer and an infrared one should agree. In driven systems that
breaks. Three independently-motivated definitions of effective temperature — from jiggling,
from speed, from escape frequency — agree over a wide range of conditions and then,
measurably, come apart. A framework that calls χ a temperature has walked into an argument
it doesn't need to fight and can't win.
The way out: it was never a thermal idea
Not one character of the formula changes. The word does.
The exponential doesn't come from thermal physics. It comes from a much older, more general
result about rare events — Freidlin–Wentzell large-deviation theory. The result, roughly: if
something unlikely happens, it almost certainly happened by the least unlikely route.
Picture a hiker blown across a mountain range by random gusts. Whether they cross depends on
two things only — the height of the lowest pass, and the strength of the wind. The
probability of crossing falls off exponentially in the pass height, measured in units of
wind strength. That argument never needed the wind to be thermal, never needed a
temperature, never needed equilibrium with anything. It only needed the wind to be weak
compared to the pass — because if it weren't, none of this would be worth writing down.
On this reading, the delivered energy in χ isn't a temperature. It's a noise scale — the
strength of the wind — and a noise scale carries none of the obligations a temperature
carries. It doesn't have to agree across two different ways of measuring it, because it
never claimed to be a state variable of anything.
There's a further twist that makes this more than a lateral move. An escape law is only
worth writing down when escapes are rare. And "rare" and "the barrier is large
compared to the noise" are the same statement — which is exactly the condition the
derivation needs anyway. The regime where the law is useful and the regime where it's
valid turn out to be the same regime. You get the assumption for free, from the fact
that you needed the formula at all. The temperature reading never gets anything for free —
consistency has to be checked separately, every system, every pair of measurements, every
time.
What that move costs: four specific breaks
The reinterpretation secures the shape of the law. What goes inside the
exponent is a separate question, and that's where four real problems live — all four drawn
from published work, catalogued rather than discovered here.
One — the temperature still isn't one number
Already covered above as a reason to stop using the word — but a residue survives that
isn't merely verbal. If the quantity governing escape differs from the quantity governing
jiggling, a number calibrated by measuring flux may not be the number that belongs in the
exponent. Calibration route and application route have to match, and nothing guarantees
they do by default.
Two — shaking is not tilting
Two different things get a marble out of a dip. Shake the table — noise, symmetric,
more shaking means more escape either direction. Tilt the table — not noise at all,
asymmetric, lowers the barrier one side and raises it the other. No amount of
turning up a shaking dial reproduces a tilt.
The distinction is a matter of timescale. If the push reverses direction many times before
an escape could plausibly happen, it's shaking. If it holds steady for longer than the
escape takes, it's a tilt. Published work shows this precisely: an effective-temperature
treatment holds while the drive decorrelates quickly, and has to be abandoned for a
modified-barrier treatment once it doesn't.
Then it gets stranger — escape is fastest at some intermediate persistence, not at
maximum. A resonance: the same reason a swing goes higher when you push in time with it
rather than harder. This shows up across several different well shapes, at almost the same
optimum. And the sharpest consequence in the whole paper: a formula with one number in
it cannot produce a resonance — a resonance is a relationship between two timescales,
and one number can't hold two of them. The resonance is, by itself, proof that something's
missing from the one-number version. (This is the mechanism the chapter's widget runs on —
go back and hunt for the peak if you haven't found it.)
Three — jumps are not jitter
Deliver a fixed amount of energy two ways: constant small jitter, or occasional huge
kicks. Same total, same variance — different escape rates entirely. Under jitter, escape
needs an improbable run of small pushes lining up. Under kicks, the system doesn't bother
— it waits for one big one. The mathematics shows this as a genuine discontinuity in the
optimal escape path, a finite jump rather than a smooth climb.
The size of the error from ignoring this is not a rounding correction. In one measured
case — a tracer particle in a bacterial suspension — the observed escape rate was about
one in a million. A calculation from the particle's variance alone predicted one in 10³².
Twenty-six orders of magnitude off: the difference between "once in the age of the
universe" and "every few seconds."
Four — two numbers aren't enough to describe a push
Build three physically different drives — one intermittent source, five independent ones,
a smooth random one — tuned to identical variance and identical persistence. They produce
different escape rates, different distributions, different everything. Those two numbers
don't determine the escape, not approximately, and no refinement of either fixes it.
The honest narrowing: χ doesn't get to claim it hands you an escape rate. It can
claim to say whether escape is affordable, and rank systems by that. The first
claim dies here; the second survives. What isn't allowed is using the ranking claim in an
argument and the rate claim in a calculation, in the same breath.
What this actually adds — two small things, on purpose
Both are deliberately modest.
A missing parameter. The settling time inside χ belongs to the system — how
fast it recovers from a nudge. Every regime boundary above is set by the persistence of
the drive instead — how long a push holds before it changes. Those are different
quantities, and no combination of what's already in χ produces the second from the first. So:
a second number, ψ — the drive's persistence divided by the system's settling time.
χ tells you whether escape is affordable. ψ tells you which law applies. Same money,
completely different rules about what it gets you, depending on whether you're standing in a
shop or at an auction.
A disagreement the framework didn't know it had. One statement inside it says χ = 1 is
a clean line — above it, escape; below it, none. The exponential itself says that at exactly
χ = 1, the chance of escaping is 37%. Neither is wrong: they're the noiseless and the noisy
versions of the same threshold, and the sharp line is the limit of the soft one as the noise
shrinks. But they can't both be exact at once, and any result proved with the sharp version
now has to carry that label. It's the kind of thing that only shows up when the formalism
gets written out properly and the two halves get held next to each other — which is the
argument for having written this page at all.
None of the mountain-pass reframing above is new mathematics, and it doesn't try to be —
quasipotential ≡ Waddington landscape has been established for about fifteen years
(Bhattacharya et al.; Wang et al., PNAS; Ao; Ge & Qian). Named here rather than
left for a reviewer to find, the way this framework tries to handle prior art generally: say
so early, credit what's already someone else's, then state what's actually different — which
in this case is the scale it's applied at and the separability into χ and ψ, not the
identification itself.