diff --git a/README.md b/README.md index dffeee3..6ac95ec 100644 --- a/README.md +++ b/README.md @@ -353,7 +353,7 @@ are allowed for, and the spanning cluster's dimension 0.8% low (details in Here the scale-free range has to be tuned: the temperature to T_c and the magnetic field to zero. Turbulent cascades, and gravitational clustering grown from a nearly scale-invariant start, are scale-free over a range that -nobody tunes: the flow's Reynolds number, or the time gravity has had to act, sets it (sources in +nobody tunes: the flow's own numbers, or the time gravity has had to act, set it (sources in [critical/README.md](critical/README.md#tuned-scale-invariance-against-self-organised-ranges); the clustering turns smooth on the largest scales, as the [opening section](#where-the-intuition-holds-and-where-it-breaks) describes). Real fluids have the same mechanism with three-dimensional exponents: ξ grows 4.27-fold for each tenfold approach diff --git a/critical/README.md b/critical/README.md index 4ed3c28..f710931 100644 --- a/critical/README.md +++ b/critical/README.md @@ -6,14 +6,15 @@ for each of nine seeds and as their mean. No script prints three kinds of number of the checks with 120 other sets of seeds, described under [What it reproduces](#what-it-reproduces); γ/ν's finite-size bias of +0.0012 and the Binder cumulant's offset of 0.0007 at L = 16, worked out from published results; and the 11% and 25% uncertainties of an error bar. `npm test` (`tests/critical.mjs`) checks the numbers -compared with an exact or high-precision value (Newman and Ziff's measured threshold, whose error is negligible -here): the table under "What it reproduces", the ξ table, the measurement-trap cumulant with a fixed count and the -same numbers in the short answer. It checks the cumulant measured the other way only against that one, and the -point where the spanning probability at L = 32 is one half against Ziff and Newman's measurement there. Without -Monte Carlo, it also checks the open-edge shift against their measured crossings for L = 64 to 256. The rest are -printed but not checked: the slopes over other ranges of L, the cluster shares, the size ratios, the point where -the spanning probability at L = 256 is one half, the other comparisons with Ziff and Newman's measurements and each -percolation seed's results. Exact values are quoted from the sources named beside them. Started 18 October 2026. +compared with an exact or high-precision value (Newman and Ziff's measured threshold and Salas and Sokal's V4, +whose errors are negligible here): the table under "What it reproduces", the ξ table, the measurement-trap cumulant +with a fixed count and the same numbers in the short answer. It checks the cumulant measured the other way only +against that one, and the point where the spanning probability at L = 32 is one half against Ziff and Newman's +measurement there. Without Monte Carlo, it also checks the open-edge shift against their measured crossings for +L = 64 to 256. The rest are printed but not checked: the slopes over other ranges of L, the cluster shares, the +size ratios, the point where the spanning probability at L = 256 is one half, the other comparisons with Ziff and +Newman's measurements and each percolation seed's results. Exact and high-precision values are quoted from the +sources named beside them. Started 18 October 2026. The other studies find self-similarity over a range of scales and ask where it stops. This one asks why it appears at all, in the one case where the exact answers are known and both ends of the range can be moved at will. @@ -25,7 +26,7 @@ at all, in the one case where the exact answers are known and both ends of the r zero field. There, patches of aligned spins come in every size from the lattice spacing to the size of the sample, nested inside one another. Away from T_c the nesting stops at the correlation length ξ, which in the two-dimensional model grows as 1/|T − T_c| (exponent ν = 1). -- **A small model lands on or near the exact numbers.** A 2D Ising lattice with seeded random numbers places T_c at +- **A small model lands on or near the known numbers.** A 2D Ising lattice with seeded random numbers places T_c at 2.2704 ± 0.0017 J/k_B, where the Binder cumulants of two sizes cross, against Onsager's 2/ln(1 + √2) = 2.2692. It reproduces Yang's spontaneous magnetisation within 0.21%, and gives β/ν = 0.1245 ± 0.0026, γ/ν = 1.751 ± 0.004 and η = 0.250 ± 0.005, against 1/8, 7/4 and 1/4. Site percolation on a square lattice, averaged over nine @@ -40,8 +41,8 @@ at all, in the one case where the exact answers are known and both ends of the r laws within their statistical errors, 0.9% to 1.6%. - **Tuned against self-organised.** Heat the lattice 6% above T_c and the nesting ends at ten lattice spacings; below T_c it ends sooner, at nine lattice spacings only 3% below. Turbulence and cosmic structure are scale-free - over a range that nobody tunes: the flow or the universe sets it, through the Reynolds number or the time gravity - has had to act. + over a range that nobody tunes: the flow or the universe sets it, through the Reynolds number (in an ordinary + fluid) or the time gravity has had to act. ## Why a critical point has no scale @@ -106,7 +107,7 @@ Werner 2001). The square lattice is expected to share it, by universality. ## What it reproduces -| Check | Exact | Model | +| Check | Exact or high-precision | Model | |---|---|---| | Spontaneous magnetisation ⟨\|m\|⟩ at T = 2.0, 2.1, 2.2, L = 64 (Yang 1952) | 0.91132, 0.86875, 0.78476 | 0.91129 ± 0.00035, 0.86913 ± 0.00060, 0.78314 ± 0.00163 | | Energy per spin, same runs (Onsager 1944) | −1.74556, −1.66208, −1.54649 | −1.74536 ± 0.00066, −1.66253 ± 0.00096, −1.54527 ± 0.00129 | @@ -265,10 +266,14 @@ and the field at zero, or p at p_c. The range then runs from the lattice spacing the lattice 6% above T_c and the range ends at ten lattice spacings, 76% above at one; below T_c it ends sooner, at nine lattice spacings only 3% below. Two other scale-free ranges in this repository's subject need no tuning. -- *Turbulent cascades.* A stirred fluid passes energy from the scale at which it is stirred down to the scale at - which viscosity removes it. In between, only the rate of energy flow matters, and the velocity spectrum falls as - k^(−5/3) (Kolmogorov 1941). The flow sets the range itself: its width grows as the Reynolds number to the power - 3/4 (as reviewed by Alexakis and Biferale 2018). This repository does not yet discuss turbulence elsewhere. +- *Turbulent cascades.* A stirred fluid passes energy from the scale at which it is stirred down to a scale set by + the medium. In air and water that is the scale at which viscosity removes it. Between the stirring and viscous + scales only the rate of energy flow matters, and the velocity spectrum falls as k^(−5/3) (Kolmogorov 1941). The + flow sets that range itself: its width grows as the Reynolds number to the power 3/4 (as reviewed by Alexakis and + Biferale 2018). The [turbulence study](../turbulence/README.md) follows the range in a cascade model and lists + what ends real cascades: viscosity in air and water; in the solar wind the ions' own motion, at the ion scales + (about 100 km), below which a steeper kinetic cascade carries on; in molecular clouds the sonic scale, which ends + only the supersonic range. - *Gravitational clustering.* The density ripples of the early universe were close to scale-invariant: their spectrum is a power law in scale, k^(n_s − 1), with n_s = 0.965 ± 0.004, where n_s = 1 would give ripples of equal amplitude on every scale (Planck Collaboration 2020). Gravity has grown them into clustering that is diff --git a/critical/model.js b/critical/model.js index edaf2ab..bd207eb 100644 --- a/critical/model.js +++ b/critical/model.js @@ -20,7 +20,7 @@ * appendix A; near T_c the amplitude ratio 2 is in Pelissetto & Vicari 2002, sec. 3.5). * Exponents beta = 1/8, nu = 1, eta = 1/4, gamma = 7/4 (Pelissetto & Vicari 2002, sec. 3.5). On a periodic L x L * lattice at T_c the Binder cumulant 1 - /(3^2) tends to 1 - V4/3 = 0.61069, V4 = 1.1679229(47) - * (Salas & Sokal 2000). + * (Salas & Sokal 2000), a high-precision value: exact CFT correlators integrated numerically, not a closed form. * Site percolation: an L x L square lattice with open edges, each site occupied with probability p. Newman & Ziff's * (2000) algorithm: occupy sites one at a time in random order, join each to its occupied neighbours with weighted * union-find, and note when a cluster first touches both the top and the bottom row (spans). Averages at a given p diff --git a/critical/table.mjs b/critical/table.mjs index e3fe225..9b450af 100644 --- a/critical/table.mjs +++ b/critical/table.mjs @@ -1,4 +1,4 @@ -// Prints the Monte Carlo numbers quoted in critical/README.md (issue #20) and the exact values beside them. Usage: node +// Prints the Monte Carlo numbers quoted in critical/README.md (issue #20) and the known values beside them. Usage: node // critical/table.mjs (about 65 s). The runs are the same fixed-seed runs that tests/critical.mjs makes, so the numbers // match; not every number printed here is checked (critical/README.md says which are). Percolation runs with nine // seeds, and its checks use their mean: printed here with each seed's values, the standard error of the mean, the @@ -7,7 +7,11 @@ // (node critical/table.mjs scatter, which used to print eight of those seeds on their own, now prints the same.) // No script prints the numbers critical/README.md takes from the one-off repeat with 120 other sets of seeds, or the // finite-size shifts it works out from published formulas other than the open-edge shifts printed below; the README -// names them. +// names them. Not every known value printed is exact, and each that is not is marked where it is printed: Newman & +// Ziff's p_c and Salas & Sokal's V4 are high-precision values whose errors are negligible here (the percolation means +// print p_c as the reference value), the 3D Ising exponents are conformal-bootstrap estimates (Kos, Poland, +// Simmons-Duffin & Vichi 2016), b0 and a1 are the coefficients Ziff & Newman use, the open-edge shifts built from them +// keep only the leading term, and Ziff & Newman's Table 1 values are measurements. import { createRequire } from 'node:module'; const CR = createRequire(import.meta.url)('./model.js'); const { TC, KC, EXP, PERC } = CR; @@ -16,19 +20,21 @@ const pm = (j, d) => `${j.value.toFixed(d)} ± ${j.err.toFixed(d)}`; const col = (s, w) => String(s).padStart(w); console.log('Critical point study (issue #20). J = k_B = 1: temperatures in J/k_B, lengths in lattice spacings.\n'); -console.log('Exact values (infinite lattice)'); +console.log('Known values: exact unless marked, for the infinite lattice unless a size is given'); console.log(` T_c = 2/ln(1 + sqrt 2) = ${TC.toFixed(6)}, K_c = 1/T_c = ${KC.toFixed(7)}; energy per spin at T_c ${CR.energyExact(TC).toFixed(6)} (-sqrt 2)`); console.log(` Ising exponents: beta = ${EXP.beta}, nu = ${EXP.nu}, eta = ${EXP.eta}, gamma = ${EXP.gamma}; ` + `beta/nu = ${EXP.beta / EXP.nu}, gamma/nu = ${EXP.gamma / EXP.nu}`); console.log(` correlation length above T_c 1/(ln coth K - 2K); near T_c 1/(4(K_c - K)) = ${(1 / (4 * KC)).toFixed(4)}/t, t = (K_c - K)/K_c; ` + `below T_c 1/(4(K - K*)), K* the dual coupling`); -console.log(` Binder cumulant at T_c on a periodic square, L -> infinity: 1 - 1.1679229/3 = ${(1 - 1.1679229 / 3).toFixed(5)}`); -console.log(` percolation: p_c = ${PERC.pc}, D = 91/48 = ${PERC.D.toFixed(4)} (2 - 5/48); open edges put the crossing of ` + - `R_L1(p) and R_L2(p) above p_c by (b0/a1)(1/L1 - 1/L2)/(L2^(3/4) - L1^(3/4)), b0 = ${PERC.b0}, a1 = ${PERC.a1}: ` + +console.log(` Binder cumulant at T_c on a periodic square, L -> infinity: 1 - 1.1679229/3 = ${(1 - 1.1679229 / 3).toFixed(5)} ` + + "(Salas & Sokal's high-precision V4)"); +console.log(` percolation: p_c = ${PERC.pc} (Newman & Ziff's measurement), D = 91/48 = ${PERC.D.toFixed(4)} (2 - 5/48); ` + + `to leading order, open edges put the crossing of R_L1(p) and R_L2(p) above p_c by (b0/a1)(1/L1 - 1/L2)/(L2^(3/4) - L1^(3/4)), ` + + `b0 = ${PERC.b0}, a1 = ${PERC.a1} (the coefficients Ziff & Newman use): ` + `+${CR.edgeShift(32, 128).toFixed(5)} for L = 32/128, +${CR.edgeShift(64, 256).toFixed(5)} for 64/256`); const nu3 = 0.629971; -console.log(` 3D Ising (liquid-vapour critical points): nu = ${nu3}, eta = 0.0362978; a tenfold approach to T_c multiplies xi by ` + - `10^nu = ${Math.pow(10, nu3).toFixed(2)} in 3D, ${Math.pow(10, EXP.nu).toFixed(0)} in 2D`); +console.log(` 3D Ising (liquid-vapour critical points): nu = ${nu3}, eta = 0.0362978 (conformal-bootstrap estimates, Kos et al. 2016); ` + + `a tenfold approach to T_c multiplies xi by 10^nu = ${Math.pow(10, nu3).toFixed(2)} in 3D, ${Math.pow(10, EXP.nu).toFixed(0)} in 2D`); console.log('\nBelow T_c (Wolff clusters, L = 64, 3000 measurements each)'); console.log(col('T', 6) + col('<|m|>', 22) + col('Yang', 10) + col('energy', 22) + col('Onsager', 11) + col('xi (exact)', 12)); @@ -93,10 +99,11 @@ const cols = [['p_c, L = 32/128', r => r.wide[0].p, pool.wide[0].p, PERC.pc, 5, console.log(col('seed', 6) + cols.map(c => col(c[0], 22)).join('')); pool.each.forEach((r, i) => console.log(col(seeds[i], 6) + cols.map(c => col(pm(c[1](r), c[4]), 22)).join(''))); console.log(col('mean', 6) + cols.map(c => col(pm(c[2], c[4]), 22)).join('') + '\n'); -for (const [name, f, m, exact, d, shift] of cols) { +for (const [name, f, m, ref, d, shift] of cols) { + const label = ref === PERC.pc ? 'reference' : 'exact'; const jk = pool.each.reduce((a, r) => a + f(r).err, 0) / pool.each.length; - console.log(` ${name}: mean ${pm(m, d)}, exact ${exact.toFixed(d)}, mean - exact ${(m.value - exact).toFixed(d)}` + - (shift ? ` (open-edge shift +${shift.toFixed(d)})` : '') + `; scatter between seeds ${m.sd.toFixed(d)}, mean jackknife error ${jk.toFixed(d)}`); + console.log(` ${name}: mean ${pm(m, d)}, ${label} ${ref.toFixed(d)}, mean - ${label} ${(m.value - ref).toFixed(d)}` + + (shift ? ` (leading-order open-edge shift +${shift.toFixed(d)})` : '') + `; scatter between seeds ${m.sd.toFixed(d)}, mean jackknife error ${jk.toFixed(d)}`); } // How the slope of ln(mass) against ln L changes with L: slopes between successive sizes, and how much leaving out // L = 32 raises the fitted slope, seed by seed (mean ± standard error over the nine seeds). diff --git a/index.html b/index.html index 7bf1ef9..589baeb 100644 --- a/index.html +++ b/index.html @@ -235,7 +235,7 @@

One question, asked at every scale.

  • Gravitational atom beside hydrogen: the same wave equation at two scales, drawn at matched size.
  • Quantum paths: where a recorded path's fractal dimension 2 starts and stops.
  • The critical point: a magnet whose spins point only up or down repeats itself across scales only when tuned to its critical point; away from it the repetition stops at the correlation length.
  • -
  • Turbulence: between the size where a flow is stirred and the size where viscosity takes over, the same energy flux passes through every scale; the bursts do not scale exactly, and where the cascade ends depends on the medium.
  • +
  • Turbulence: from the size where a flow is stirred down to a smallest size, the same energy flux passes through every scale; the bursts do not scale exactly, and what sets that smallest size depends on the medium: viscosity in air and water, the ions' own motion in the solar wind.
  • Cosmic homogeneity: the scale where galaxy clustering stops looking fractal. Model and checks written; issue 17 tracks publication.
  • Open questions: halos within halos. The issue list is the programme.
  • diff --git a/tests/critical.mjs b/tests/critical.mjs index 8a835a3..b4b1e2a 100644 --- a/tests/critical.mjs +++ b/tests/critical.mjs @@ -1,12 +1,12 @@ // Checks the critical-point study (critical/model.js, issue #20) against exact or high-precision results: Onsager's -// T_c and energy, Yang's magnetisation, the exact correlation length, the 2D Ising exponents, the Binder cumulant at -// T_c, the measurement trap in cluster runs, Newman and Ziff's measured percolation threshold, 0.59274621(13), whose -// error is negligible here, and the 91/48 dimension; and against Ziff and Newman's measurements of the same -// percolation system at finite sizes: their crossings of two sizes test the open-edge shift, and one point tests the -// runs. Run: npm test (or node tests/critical.mjs). -// Each check states the source, the value compared with (exact, Newman and Ziff's threshold, or in two checks Ziff and -// Newman's measurements), the model's value and the tolerance. Monte Carlo runs use fixed seeds and integer-only -// random numbers, so every run, on every Node version, prints the same numbers. Ising error bars are jackknife +// T_c and energy, Yang's magnetisation, the exact correlation length, the 2D Ising exponents, Salas and Sokal's +// high-precision Binder cumulant at T_c, the measurement trap in cluster runs, Newman and Ziff's measured percolation +// threshold, 0.59274621(13), whose error is negligible here, and the 91/48 dimension; and against Ziff and Newman's +// measurements of the same percolation system at finite sizes: their crossings of two sizes test the open-edge shift, +// and one point tests the runs. Run: npm test (or node tests/critical.mjs). +// Each check states the source, the value compared with (exact or high-precision, or in two checks Ziff and Newman's +// measurements), the model's value and the tolerance. Monte Carlo runs use fixed seeds and integer-only random +// numbers, so every run, on every Node version, prints the same numbers. Ising error bars are jackknife // estimates over 40 blocks. The percolation checks use the mean over nine seeds, the study seed and eight others, with // the standard error of that mean as error bar: the scatter between the nine seeds over 3. Each tolerance is a fixed // number, 3 to 5 standard deviations of the noisiest estimate it covers, plus any finite-size shift: worked out from @@ -50,8 +50,9 @@ function check(name, ok, detail, src) { } // A Monte Carlo estimate {value, err} against a reference value: passes if ref - lo <= model <= ref + hi, where tol is // one number (lo = hi = tol) or [lo, hi], and the error bar is finite (a jackknife replicate can fail and give NaN). -// `label` names the value compared with when it is a published measurement rather than an exact value: 'reference' for -// Newman and Ziff's threshold, whose error (1.3e-7) is negligible here, 'measured' for a finite-size measurement. +// `label` names the value compared with when it is not an exact value: 'reference' for Newman and Ziff's measured +// threshold and Salas and Sokal's V4 (exact CFT correlators integrated numerically), whose errors (1.3e-7, and 1.6e-6 +// in U) are negligible here, 'measured' for a finite-size measurement. function near(name, est, want, tol, src, why, d, label = 'exact') { const [lo, hi] = Array.isArray(tol) ? tol : [tol, tol], dev = est.value - want; const finite = Number.isFinite(est.err), ok = finite && dev >= -lo && dev <= hi; @@ -147,7 +148,8 @@ near(`T_c from the crossing of the Binder cumulants of L = 64 and 128`, cross(64 near('Binder cumulant at T_c, mean of L = 16 to 128', crit.binderMean, 1 - 1.1679229 / 3, 0.005, 'Salas & Sokal 2000: 1 - V4/3, V4 = 1.1679229(47)', 'tolerance 4.5 sigma (sigma = 0.0011), 4.2 sigma of the seed-to-seed scatter ' + '(0.0012); 4.3 and 4.0 after the mean\'s +0.00024 finite-size shift, most of it from L = 16, which sits 0.0007 above its limit ' + - '(Salas & Sokal\'s V4(L); the 120 repeat seeds average +0.0002 +- 0.0001); the sizes agree with each other within their errors'); + '(Salas & Sokal\'s V4(L); the 120 repeat seeds average +0.0002 +- 0.0001); the sizes agree with each other within their errors', + 4, 'reference'); const Ls = crit.Ls.join(', '); near(`beta/nu from <|m|> ~ L^(-beta/nu) at T_c, least squares over L = ${Ls}`, crit.betaNu, EXP.beta / EXP.nu, 0.012, 'Pelissetto & Vicari 2002, sec. 3.5: beta = 1/8, nu = 1', 'tolerance 4.6 sigma (sigma = 0.0026), 3.9 sigma of the seed-to-seed scatter (0.0031)'); @@ -170,7 +172,7 @@ const trap = CR.studyTrap(); near(`Binder cumulant at T_c, L = ${trap.L}, measured after a fixed number of cluster flips (${trap.n} measurements)`, trap.fixed, 1 - 1.1679229 / 3, [0.0035, 0.004], 'Salas & Sokal 2000: 1 - V4/3', 'tolerance: at L = 16 U sits +0.0007 above its limit ' + '(Salas & Sokal\'s measured V4(16) = 1.16586(38)), so the high side is 4.1 sigma (sigma = 0.0008) above that shift and the low ' + - 'side, 4.4 sigma, allows none; 4.4 and 4.7 sigma of the seed-to-seed scatter (0.00075)'); + 'side, 4.4 sigma, allows none; 4.4 and 4.7 sigma of the seed-to-seed scatter (0.00075)', 4, 'reference'); const gap = trap.until.value - trap.fixed.value, gapErr = Math.hypot(trap.until.err, trap.fixed.err); check('Measuring each time the flipped clusters add up to L^2 spins over-weights ordered states: the Binder cumulant comes out higher', gap / gapErr > 10, `U = ${fmt(trap.until.value)} ± ${fmt(trap.until.err)} that way, against ${fmt(trap.fixed.value)} ± ` +