The Folding Threshold

The engine does not pass from a finished figure into an undifferentiated haze at one universal line. It stretches, folds, returns, and confines at every step. What changes across parameter space is the balance among those actions.

The corpus now gives us three different things that an early sketch could easily collapse into one:

1. an exact geometric fold, where the Jacobian becomes singular;
2. an exact independent-phase law for mean logarithmic area change;
3. an empirical onset band where the maximal Lyapunov exponent crosses zero.

They speak to one another. They are not interchangeable.

The product $F_{\mathrm{fold}}=\mu|\alpha|$ organizes the passage, but it does not erase the two parameters. The first threshold concerns directional separation; the second concerns total area.

The operation: stretch, return, confine

A fold does two apparently opposite things at once. Along one direction it magnifies a small difference. Across another it presses a region thin. It then returns the elongated image over itself, putting states that were far apart into close geometric contact.

That is the engine’s native Solve et Coagula. Solve is not a territory to the right of a line, and Coagula is not a territory to the left. They are simultaneous directional actions of the same bounded map.

The Jacobian is the spine

At $\varepsilon=0$, the determinant of the engine’s Jacobian is exact:

$$\det J = -\alpha^2 \left[ \mu^2\sin(\alpha x)\sin(\alpha y) + \cos(\alpha x)\cos(\alpha y) \right].$$

The same expression can be read at four levels:

reading	object	what it tells us
geometry	$\det J=0$	where the map loses local invertibility and folds
image geometry	the forward image of $\det J=0$	the caustic traced by the fold
local analysis	$\log\lvert\det J\rvert$	the logarithmic area response of one step
long-run dynamics	$\langle\log\lvert\det J\rvert\rangle=\lambda_1+\lambda_2$	whether a typical orbit contracts or expands area on average

The first two are geometric. The last one is dynamical and depends on the measure sampled by the orbit. A critical curve is therefore not an attractor-density plot, and an exact caustic is not by itself a certificate of chaos.

The exact trace law

Average over independent uniform torus phases. The determinant closes:

$$\mathbb{E}_{\mathrm{torus}}\!\left[\log|\det J|\right] = \log\!\left(\frac{\alpha^2(\mu^2+1)}{4}\right).$$

For $\mu>0$ and $\alpha\neq0$, define

$$F_{\mathrm{fold}}=\mu|\alpha|.$$

Then the same law becomes

$$\mathbb{E}_{\mathrm{torus}}\!\left[\log|\det J|\right] = 2\log F_{\mathrm{fold}} -2\log2 +\log(1+\mu^{-2}).$$

Three pieces matter:

• $2\log F_{\mathrm{fold}}$ is the leading folding coordinate;
• $-2\log2$ is an exact analytic intercept under the uniform-torus phase law;
• $\log(1+\mu^{-2})$ is the exact finite-$\mu$ correction—not merely its leading approximation $1/\mu^2$.

The intercept is not an isolated curiosity. It belongs to the same analytic family as the engine’s cumulant ladder: at $\mu=1$, the normalized mean is $-\ln2$ and higher cumulants open into Dirichlet $\eta$ and zeta values. The fuller ladder is developed in The Eta Ladder.

There is one essential boundary around that exact result. A real orbit averages against its physical invariant measure, not an automatically uniform torus. Equality between the formula above and an observed Lyapunov sum is therefore a mixing statement to be checked in the parameter region and basin being studied.

Two thresholds, not one

The trace and the maximal exponent answer different questions:

event	criterion	folding coordinate	meaning
chaos onset	$\lambda_1=0$	$F_{\mathrm{fold}}\approx1.30$	nearby trajectories begin to separate exponentially
area neutrality	$\lambda_1+\lambda_2=0$	$F_{\mathrm{fold}}=\dfrac{2\mu}{\sqrt{\mu^2+1}}\to2$	mean logarithmic area contraction is exhausted in the uniform-phase law

The first is empirical and soft. At $\varepsilon=0$, the measured onset occupies approximately

$$F_{\mathrm{fold}}\in[1.275,\,1.325],$$

and shifts to about $1.29$ at $\varepsilon=\Delta$. Local periodic windows and basin changes keep this from becoming a knife-edge theorem.

The second follows from the exact trace by setting its mean to zero:

$$\alpha^2(\mu^2+1)=4.$$

Expressed through the folding coordinate, its exact finite-$\mu$ locus tends to the memorable large-$\mu$ value $F_{\mathrm{fold}}=2$. That value is the area-neutral asymptote, not the onset of chaos.

Between the positive-$\lambda_1$ onset and area neutrality lies the characteristic dissipative strange-attractor regime:

$$\lambda_1>0, \qquad \lambda_1+\lambda_2<0, \qquad \lambda_2<-\lambda_1.$$

One direction solves—nearby trajectories separate. The other coagulates more strongly—the total area still contracts. Chaos and confinement are not rivals there; each requires the other.

An organizer, not an oracle

$F_{\mathrm{fold}}$ is a powerful coordinate because it captures the leading scale of the determinant trace and organizes the measured onset. It is not a universal function that fixes $\lambda_1$ independently of how its product was made.

At fixed $\mu|\alpha|$, the maximal exponent still spreads. Frequency $\alpha$ leads the observed change, while $\mu$ reshapes the fold, and the orbit’s invariant measure decides which local stretches are actually visited. In the measured comparison, frequency carried roughly a $1.8{:}1$ lead. That is an empirical directional result, not a replacement exponent to publish as a law.

Why can the product organize total area better than it predicts the largest exponent? Because the determinant knows only the product of the two local singular values. It constrains $\lambda_1+\lambda_2$. The split between $\lambda_1$ and $\lambda_2$ also depends on direction, rotation, recurrence, and the measure sampled by the orbit.

The shape of the fold is not the strength of chaos

The geometry makes the distinction visible. At $\mu=0$, the determinant-zero set is a horizontal-and-vertical grid, and its forward image is an exact square of side $2$. At $\mu=1$, the determinant-zero set becomes diagonal lines, and their forward image is a point together with an exact circle of radius $2$.

Those square and circle forms are caustics—images of the singular set—not the singular sets themselves. Their full-phase-plane form is independent of nonzero $\alpha$. The circle can therefore remain exact while $\alpha$ is raised into a strongly chaotic regime. Fold shape and chaos strength are different readings of the same Jacobian spine.

This geometric lane is developed in Caustica.

Where the claim stops

The mature folding statement has four clean edges:

1. The determinant trace is exact for the independent uniform-torus phase law; agreement with an orbit average is parameter- and basin-dependent.
2. The chaos onset near $1.30$ is a measured band, not a universal sharp constant.
3. $F_{\mathrm{fold}}$ organizes the passage but does not determine $\lambda_1$ by itself.
4. The determinant-zero set, its caustic, and the orbit-density envelope are three different objects.

Within those edges, the result is stronger than a single threshold story. One coordinate links geometry to dynamics. Two thresholds separate directional instability from area neutrality. And the interval between them gives Solve et Coagula its precise dynamical meaning: stretch without escape, dissolution held inside form.

:: APPARATUS ::

Sources for this page

Full bibliography →

1. §

   Cosmopoiesis · living edition

   The public kernel definition, parameter meanings, iteration protocol, and dynamical vocabulary used throughout the scholia.

   Read the live canon ↗
2. §

   Mathematica · living edition

   The canonical constants, algebraic identities, and computed closures used by the engine and its derived findings.

   Read the live canon ↗