Restated6% 60 of 998 principal statementsChecked28% 17 of 60 restatedScreened29% 20,612 of 70,421 on the dependency pathCompiled47% of 175,187 files, as of 2026-11-20 15:20 EST

Load-bearing statements of the Lean proof

60 statements that the final theorem of the Lean project rests on, written in ordinary notation. Each carries the Lean declaration it restates and how far it has been verified against the source. The dependency graph shows how they depend on one another. The catalogue is incomplete and grows as statements are read.

statement checked against Lean: 17statement read: 41from a worker report, not rechecked: 2

The result

Analytic number theory

Riemann hypothesis for Dirichlet L-functions

statement checked against Lean

Let $N\ge1$ be an integer, let $\chi$ be a Dirichlet character modulo $N$, and let $s\in\mathbb C$. Write $L(s,\chi)$ for its Dirichlet $L$-function. In the half-plane $\Re s>1/2$, $$L(s,\chi)\ne0,$$ except at $s=1$ when $\chi$ is trivial.

Consequently, $$L(s,\chi)=0,\quad0<\Re s<1\quad\Longrightarrow\quad\Re s=1/2.$$ Every zero of the Riemann zeta function $\zeta(s)$ other than its trivial zeros and with $s\ne1$ also has real part $1/2$.

This is the final conclusion for zeta and Dirichlet $L$-functions. It follows from primitive-character zero-freeness, Euler factors, and the functional equation.

Lean
riemannHypothesis, LFunction_ne_zero_of_half_lt_re, LFunction_eq_zero_re
File
Main.lean, Statements.lean

Analytic number theory

Zero-freeness for primitive characters

statement checked against Lean

Let $\psi$ be a primitive Dirichlet character, and let $L(s,\psi)$ be its Dirichlet $L$-function for $s\in\mathbb C$. Then $$L(s,\psi)\ne0\qquad(\Re s>1/2).$$

The weighted prime-sum bound and the Landau theorem give this zero-free half-plane. Euler factors are nonvanishing for $\Re s>0$, so they extend the conclusion to imprimitive characters; the functional equation then gives the critical-line conclusion.

Lean
UpperTheory.FinalTheorems.everyCharacter, Formalism.HC.ZeroFreeAll
File
Development/AutomorphicArithmetic/CharacterBoxes/MainConclusion.lean:10

Analytic reduction

Analytic number theory

Landau theorem on intervals

statement checked against Lean

Let $\Lambda$ be the von Mangoldt function and $\gamma$ Euler's constant. For a nonprincipal Dirichlet character $\chi$ let $g_\chi(x)=\sum_{n\le x}\Lambda(n)\chi(n)\big(\tfrac xn-1\big)+x\,\tfrac{L'}{L}(1,\chi)$, and let $g_1(x)=\sum_{n\le x}\Lambda(n)\big(\tfrac xn-1\big)-x\log x+(1+\gamma)x$.

Let $\chi_1,\dots,\chi_r$ be Dirichlet characters, any principal one having modulus $1$, let $a_1,\dots,a_r>0$ and $G(x)=\sum_i a_i\,\Re g_{\chi_i}(x)$. Fix real numbers $0<\theta<1$, $\lambda>1$, and $C$. If for every $Y_0$ there is $Y\ge Y_0$ such that $$G(x)\le C\,x^{\theta}\qquad\text{for all }x\in[Y,Y^{\lambda}],$$ then $$L(s,\chi_i)\ne0\qquad(\Re s>\theta,\ 1\le i\le r).$$

An upper bound on arbitrarily distant long intervals suffices for zero-freeness. High derivatives of the Mellin transform concentrate on those intervals, contradicting a pole to the right of $\theta$.

Lean
Deduction.zeroFree_of_rpow_window
File
Development/AutomorphicArithmetic/DedekindLFunctions/LandauTheorems/ZeroFreeOfRpowWindow.lean:11

Analytic number theory

Weighted prime-sum bound

statement checked against Lean

Use the smoothed sums $g_\chi$ and $g_1$ of Landau theorem on intervals. Let $\psi$ be a primitive Dirichlet character of order $o$. In the following sum, $\chi$ runs over the primitive characters inducing the nontrivial powers of $\psi$: $$G_\psi(x)=g_1(x)+\frac1{3o+1}\sum_\chi\Re g_\chi(x).$$ There are constants $e_D,e_L$ such that, for every sufficiently small $\varepsilon>0$, there are arbitrarily large $Y$ for which $$G_\psi(x)\le x^{1/2+(e_D+e_L+7)\varepsilon}\qquad(Y\le x\le Y^{3/2}).$$

This is the estimate to which the Landau theorem is applied. The Fourier identity reduces it to the prime-power count and the bounds for auxiliary twists.

Lean
UpperTheory.Comparison₂.core_gm, level_gm_le_on_family_scales
File
Development/AutomorphicArithmetic/PrimitiveCosets/CoreGm.lean:12

Analytic number theory

Mellin transform of a smoothed character sum

statement read

Let $\chi$ be a nonprincipal Dirichlet character of positive modulus. With $\Lambda$ the von Mangoldt function, put $$g_\chi(y)=\sum_{1\le n\le y}\Lambda(n)\chi(n)(y/n-1)+y\frac{L'}L(1,\chi).$$ For $\Re s>1$, $$\int_1^\infty g_\chi(y)y^{-s-1}\,dy=-\frac{(L'/L)(s,\chi)-s(L'/L)(1,\chi)}{s(s-1)}.$$

The transform connects the smoothed prime sum to the logarithmic derivative of the $L$-function. The character need not be primitive.

Lean
Deduction.mellin_Gchi_of_ne_one
File
Development/AutomorphicArithmetic/DedekindLFunctions/LandauTheorems/mellinGchiOfNeOne.lean:116

Analytic number theory

Derivatives of the Laplace transform

statement read

Let $g:\mathbb R\to\mathbb R$ be measurable, with $|g(u)|\le C e^{2u}$ for $u\ge0$ and some real $C$. Let $Z\subset\{\Re s>0\}$ be discrete. Suppose $\Phi$ is holomorphic on $\{\Re s>0\}\setminus Z$, agrees with $\int_0^\infty g(u)e^{-su}\,du$ for $\Re s>2$, and at each $\rho\in Z$ has a simple pole with nonzero residue, positive real when $\rho$ is real. Discreteness is relative to the open right half-plane. Then, for integers $k\ge0$ and $\Re s>2$, $$\frac{(-1)^k}{k!}\Phi^{(k)}(s)=\int_0^\infty g(u)\frac{u^k e^{-su}}{k!}\,du.$$

This identity expresses high derivatives as integrals against kernels concentrated in short relative intervals. The pole assumptions are retained for the later zero-free criterion.

Lean
Deduction.landau_coeff
File
Development/AutomorphicArithmetic/DedekindLFunctions/LandauTheorems/LandauCoeff.lean:10

Analytic number theory

One-sided bound for high derivatives

statement read

Use $g,\Phi,Z$ and all assumptions of Derivatives of the Laplace transform. Fix $0<\theta<1$, $\delta>0$. Suppose that for every $\eta>0$ and $U_0\in\mathbb R$ there are $U\ge U_0$ and $R:\mathbb R\to\mathbb R$ such that, on $I_U=[U,(1+\delta)U]$, $R\ge0$, $g\le R$, and $R(u)e^{-\theta u}$ is integrable with $$\int_{I_U}R(u)e^{-\theta u}\,du\le e^{\eta U}.$$ There are $\sigma>2$ and a real $B$ such that, for every $\eta>0$ and real $U_0$, some $U\ge U_0$ and integer $k\ge0$ with $k\ge U$ satisfy, simultaneously for every $\gamma\in\mathbb R$, $$\left|\frac{(-1)^k\Phi^{(k)}(\sigma+i\gamma)}{k!}\right|+\Re\frac{(-1)^k\Phi^{(k)}(\sigma)}{k!}\le\frac{B e^{\eta U}}{(\sigma-\theta)^k}.$$

The interval majorant bounds high derivatives of the Laplace transform. A single derivative order works simultaneously for every imaginary shift.

Lean
Deduction.landau_window_key
File
Development/AutomorphicArithmetic/DedekindLFunctions/LandauTheorems/landauWindow.lean:40

Analytic number theory

High derivatives near an isolated pole

statement read

Let $z,\rho,c\in\mathbb C$ and $r>0$, and write $B(z,r)$ for the open disk of radius $r$ about $z$. Assume $\rho\ne z$, $|\rho-z|<r$, and suppose $\Phi$ is holomorphic on $B(z,r)\setminus\{\rho\}$. Suppose $\Phi(s)=c/(s-\rho)+h(s)$ on a punctured neighborhood of $\rho$, with $h$ holomorphic there and at $\rho$. For every $0<r'<r$ there is a real $M$ such that, for all integers $k\ge0$, $$\left|\frac{(-1)^k\Phi^{(k)}(z)}{k!}-\frac{c}{(z-\rho)^{k+1}}\right|\le\frac{M}{(r')^k}.$$

An isolated pole determines the leading term of high derivatives. Cauchy estimates bound the derivatives of the holomorphic remainder.

Lean
Deduction.landau_pole_asym
File
Development/AutomorphicArithmetic/DedekindLFunctions/LandauTheorems/LandauPoleAsym.lean:19

Fields and the prime count

Analytic number theory

Auxiliary primes with controlled twists

statement checked against Lean

Use the smoothed sums $g_\chi$ of Landau theorem on intervals. Fix an integer $M\ge1$ and real numbers $0<\varepsilon<1/52$ and $A\ge1$. For every sufficiently large $U$, put $z=U^{\varepsilon/5}$. There are primes $q\equiv3\pmod 8$ and $d\equiv7\pmod{16}$ such that $$\tfrac z2<\log q\le z,\qquad \tfrac z2(\log U)^2<\log d\le z(\log U)^2,$$ $$q\mid d+1,\qquad p\mid d+1\quad\text{for every odd prime }p\le z.$$ For every Dirichlet character $\xi$ modulo $M$ and every pair $(i,j)$ in $\{0,1\}^2$ other than $(0,0)$, let $\chi$ be the primitive character inducing $\xi\,(\cdot/q)^i(\cdot/d)^j$. The parenthesized factors are quadratic residue characters. Simultaneously for all these characters, $$|g_\chi(y)|\le y^{1/2+\varepsilon/2},\qquad \left|\frac{L'}{L}(1,\chi)\right|\le4\log y\qquad(U\le y\le U^A).$$

These primes control the extra characters introduced by the auxiliary quadratic fields. They are chosen by combining a mean-square estimate over primes with the zero-density bound near one.

Lean
QH.pairEx, UpperTheory.Top.PairGood, QH.meanSqK, zero_density_box
File
Development/AutomorphicArithmetic/MeanSquareBounds/pairEx.lean:112

Analytic number theory

Prime-power Fourier identity

statement checked against Lean

Use the smoothed character sums $g_\chi$ of Landau theorem on intervals. Let $\psi$ be the chosen Dirichlet character and let $K_\psi$ be the field it cuts out. For the chosen integer $\nu$ and auxiliary primes $q,d$, set $$E=\mathbb Q(\zeta_{2^{\nu+2}})^+\cdot K_\psi\cdot\mathbb Q(\sqrt{-q},\sqrt{-d}),$$ where $\zeta_h$ is a primitive $h$-th root of unity and $+$ denotes the maximal real subfield. Write the given Galois decomposition additively as $$\operatorname{Gal}(E/\mathbb Q)\simeq C_2^2\times H,\qquad g=((\alpha,\beta),h),$$ where $C_2$ is cyclic of order two and $H$ is the remaining finite abelian factor. Put $$\Phi=\{\alpha=0\},\qquad\chi_C(g)=(-1)^\alpha,\qquad c=((1,1),0),\qquad P_t(g)=c+g+(0,t).$$ Here $\chi_C$ is also viewed as the quadratic Dirichlet character attached to $\mathbb Q(\sqrt{-d})$.

Let $T\subset H$ be the selected finite set, $a_t$ the prescribed parameters for $t\in T$, and $w_j$ the positive weight functions on the Galois group. Fix a cutoff $N$ with $N\ge a_t/w_j(\rho)$ for every $\rho\in\Phi$. Let $F_p$ denote Frobenius at an unramified prime $p$, let $I_d$ be inertia at $d$, and choose the indicated Frobenius lift $F_d$ modulo $I_d$. Write $(z)_+=\max(z,0)$ and let $\gamma$ be Euler's constant. Define $$A_{t,j}=2\!\!\sum_{\substack{p\le N,\ p\notin\{2,q,d\}\\ F_p\notin\Phi}}\!\!\log p\sum_{\substack{1\le i\le N\\ iF_p\notin\Phi}}\Big(\frac{a_t}{p^i}-w_j(P_t(iF_p))\Big)_+ +\log d\sum_{i=1}^N\sum_{\substack{g-iF_d\in I_d\\ g\notin\Phi}}\Big(\frac{a_t}{d^i}-w_j(P_t(g))\Big)_+,$$ $$M_{t,j}=a_t\Big(\log a_t-1-\gamma+\Re\frac{L'}{L}(1,\chi_C)-\frac1{|\Phi|}\sum_{\rho\in\Phi}\log w_j(\rho)\Big).$$ Then $A_{t,j}-M_{t,j}$ equals a finite Fourier sum of $g_\chi$ at rescaled arguments, with $\chi$ ranging over characters of $\operatorname{Gal}(E/\mathbb Q)$.

The condition $F_p\notin\Phi$ says that $p$ is inert in $\mathbb Q(\sqrt{-d})$; then $iF_p\notin\Phi$ says that $i$ is odd. Fourier expansion connects this nonnegative prime-power count to the smoothed character sums.

Lean
characterSumData.ASK, MAIN, Deduction.ch_balance
File
Development/AutomorphicArithmetic/DedekindLFunctions/LandauTheorems/ChBalance.lean:200

Analytic number theory

Prime-power count bound

statement checked against Lean

Use the counts $A_{t,j}$, main terms $M_{t,j}$, weights, and finite set $T$ of Prime-power Fourier identity. Let $T_2\subset H$ be the second selected set in the construction. Here $x$ is the scale, $\varepsilon$ is its exponent parameter, $m$ is the number of lifts, $k$ is the initial unitary rank, and $n$ is the degree parameter. Set $$\mathcal A=\sum_{t\in T}\sum_{0\le j<m}A_{t,j},\qquad\mathcal M=\sum_{t\in T}\sum_{0\le j<m}M_{t,j},\qquad b=\exp(x^\varepsilon),\qquad a=xb.$$ With the construction's exponents $e_D,e_L$ and integer scales $$n\asymp x^{2+8\varepsilon},\qquad m\asymp x^{1/2-\varepsilon/16},\qquad k\asymp x^{9/2+10\varepsilon},$$ one has $$\mathcal A-\mathcal M\le x^{(e_D+e_L+3)\varepsilon}\Big[|T|\,a\Big(1+\frac{m^2}{k}\Big)+m\Big(2nmb\,x^{\varepsilon}+\frac{|T_2|\,k\,b\,x^{\varepsilon}}{n^2x^{-\varepsilon}}\Big)\Big].$$

Dividing by $|T|mb|T_2|/(2n)$ leaves competing terms of sizes $x/m$ and $m$. Their balance at $m$ of size $x^{1/2}$ gives the exponent $1/2$.

Lean
Deduction.ch_top
File
Development/AutomorphicArithmetic/DedekindLFunctions/LandauTheorems/ChTop.lean:37

Analytic number theory

Mean square over auxiliary quadratic twists

statement read

There is an absolute $C>0$ with the following property. Fix $0<\varepsilon\le1$ and $A\ge1$. For all sufficiently large $Y$, let $m>0$, $8\mid m$, $(a,m)=1$, and $U>0$ satisfy $$\tfrac14(\log Y)^2\log m\le\log U,\qquad Y^{\varepsilon/6}\le\log U\le Y^{\varepsilon/5}(\log Y)^2.$$ Assume every prime divisor $p$ of $m$ satisfies either $p\le Y^{\varepsilon/5}$ or $p>(Y^A)^{400}$. Set $Z=(Y^A)^{400}$ and $P=\{q\text{ prime}:\sqrt U<q\le U,\ q\equiv a\pmod m\}$. For $Y\le y\le Y^A$ and any $\kappa:\mathbb N\to\mathbb C$ with $|\kappa(n)|\le1$, $$\sum_{q\in P}\left|\sum_{1\le n\le Z}\Lambda(n)\min(1,y/n)\kappa(n)\left(\frac n q\right)\right|^2\le C y\log y\,|P|.$$ Here $\Lambda$ is the von Mangoldt function and the parenthesized symbol is the Jacobi symbol.

This controls averages over the auxiliary primes used to choose the quadratic twists. The threshold for $Y$ may depend on $\varepsilon$ and $A$, and is uniform in the progression and bounded coefficients.

Lean
QH.meanSqK, QH.MeanSq, QH.admitted
File
Development/AutomorphicArithmetic/MeanSquareBounds/meanSqK.lean:51

Analytic number theory

Zeros near one and exceptional zeros

statement read

There are absolute constants $C,c>0$ such that, for $Q\ge3$, the zeros of primitive Dirichlet $L$-functions of conductor at most $Q$ in $$\Re\rho\ge399/400,\qquad |\Im\rho|\le Q$$ have total multiplicity at most $CQ^{1/4}$, with multiplicities given by analytic orders. Equivalently, the bound holds for every finite set of character–zero pairs in this region. For each primitive character $\chi$ of conductor $N\ge3$, a zero satisfying $$\Re\rho\ge1-\frac{c}{\log\max(N,N|\Im\rho|)}$$ forces $\chi$ to be real; that zero is real, simple, and the only zero of this character satisfying the same inequality. Every zeta zero with $0<\Re\rho<1$ satisfies $$\Re\rho\le1-\frac{c}{\log(|\Im\rho|+3)}.$$

The zero-density estimate excludes bad characters when choosing auxiliary primes. A separate bound for the number of primitive characters with a zero strictly to the right of $399/400$ in the same box supplies the version used there.

Lean
GlobalTheory.ZerosNearOne.density_and_regions, Formalism.HermitianLattices.Ext.PrimitiveCharacterCasts.ZD.zero_density_box
File
Development/AbelianVarietyTheory/Proofs/GlobalInterpolation/DensityAndRegions.lean:10

Theta maps and divisibility

p-adic Hodge theory and formal groups

Valuations of theta-kernel maps

statement checked against Lean

Use the field, weights, and selected set $T$ of Prime-power Fourier identity. Let $p$ be inert in $\mathbb Q(\sqrt{-d})$. At $p$, the $j$-th theta-kernel map applies a product of minors as a differential operator to a theta series at the origin. The minors have the shape of the target weight. Let $\alpha_s$ be the difference of its first two coordinates at embedding $s$, and let $\operatorname{cap}_j(s)$ be the number of remaining columns of its Young diagram. Write $s+e$ for the embedding obtained by applying $\operatorname{Frob}_p^e$. Let $D(s)$ be the excluded odd distances, $\mathrm{de}$ the digit-sum loss, and $\deg$ the degree of the differential operator. Write $(z)_+=\max(z,0)$.

In coordinates scaled by $\varpi$ with $\varpi^{p^2-1}=p$, every coefficient of the output has $p$-adic valuation at least $$\sum_{s\in T}\ \sum_{\substack{e\ \mathrm{odd}\\ e\notin D(s)}}\Big(\frac{\alpha_s}{p^{e}}-\operatorname{cap}_j(s+e)\Big)_+-\mathrm{de}+\frac{\deg}{p^2-1}.$$

For the global consequence, let $K$ be the coefficient field, $B_u$ the period maps, $N_u^0$ their initial integral source lattices, and $M_{u,0}$ the full initial lattices of modular forms. Here $u$ indexes the chosen source forms. At a finite place $v$ of $K$, let $\mathfrak p_v$ be its prime ideal, $e_v$ its ramification index, and $\sigma_u(v)$ the prescribed local valuation bound. Then $B_u$ is integral on $$N'_u=\Big(\prod_v\mathfrak p_v^{-\lfloor e_v\sigma_u(v)\rfloor}\Big)N_u^0.$$ Let $\mathcal G$, $\mathcal F$, and $\mathcal E_u$ be the totals of local gains, signed finite-place corrections, and errors, respectively. Let $\lambda_u$ be the lattice-index loss and $L_P$ the prime-count error allowance. With $\mu$ the squared-norm arithmetic slope and $\mathcal A=\sum_{t,j}A_{t,j}$, $$\mu(N'_u)\ge\mu(M_{u,0})+\mathcal G+\mathcal F-\mathcal E_u-\lambda_u,\qquad\mathcal A\le\mathcal G+L_P.$$

Up to small weight shifts, $\alpha_s$ and $\operatorname{cap}_j(s)$ are $a_s$ and $w_j(s)$, so the valuation sum gives the prime-power count. Dwork's criterion controls denominators in the theta series built from the height-two Lubin–Tate group of a supersingular CM elliptic curve.

Lean
FrameJetExcessBound, ExpansionEquality, off_level_bound, LinkG_at
File
Formalism/ThetaLogCoordinates/UniformThetaSubstitution/StatementG.lean:33

p-adic Hodge theory and formal groups

Dwork integrality criterion

statement read

Let $p$ be an odd prime, $W=W(\overline{\mathbb F}_p)$ its Witt-vector ring, $K=\operatorname{Frac}W$, and $\sigma$ the Witt Frobenius on $K$. For $\Psi\in K[[X_1,\ldots,X_n]]$ with zero constant term, $$\exp\Psi\in W[[X_1,\ldots,X_n]]\quad\Longleftrightarrow\quad\Psi^\sigma(X_1^p,\ldots,X_n^p)-p\Psi(X_1,\ldots,X_n)\in pW[[X_1,\ldots,X_n]].$$ The superscript means applying $\sigma$ to the coefficients.

This criterion controls the denominators of the exponential occurring in the local theta series. Frobenius turns exponential integrality into a divisibility condition on its exponent.

Lean
Formalism.MainNotions.FormalGroupSystem.Pf.sx_dwork, Formalism.MainNotions.FormalGroupSystem.DworkLemma
File
Formalism/SharpRestriction/GrowthBounds/MultisetValuation/PositivityConditionHolds.lean:107

Algebra and combinatorics

Jets and directional Hasse derivatives

statement read

Let $K$ be a field, $X=(X_1,\ldots,X_n)$, $Q\in K[X]$, $G\in K[[X]]$, and $b_j\in K^n$ for a finite index set $J$. Put $L_j(X)=\sum_a b_{j,a}X_a$. For a multi-index $\beta\in\mathbb N^J$, define $$D_b^{[\beta]}Q=[T^\beta]Q\!\left(X+\sum_{j\in J}T_jb_j\right).$$ The notation $[X^\alpha]$ and $[T^\beta]$ denotes coefficient extraction, and $\alpha!=\prod_a\alpha_a!$. Writing $Q(\partial)G(0)=\sum_\alpha\alpha!\,[X^\alpha]Q\,[X^\alpha]G$, one has $$Q(\partial)\left(\prod_{j\in J}L_j^{\beta_j}G\right)(0)=\left(\prod_{j\in J}\beta_j!\right)(D_b^{[\beta]}Q)(\partial)G(0).$$

This identity transfers linear factors in a theta series to Hasse derivatives of the coefficient polynomial. The coefficient definition makes the formula valid in every characteristic.

Lean
Formalism.InvariantFactorPaths.G7.Top.e_one_identity, Formalism.InvariantFactorPaths.G7.JetPairingHasseIdentity
File
Formalism/ThetaLogCoordinates/UniformThetaSubstitution/KOneConsequence.lean:25

Algebra and combinatorics

Directional derivatives of products of minors

statement read

Over a field $K$, let $X_s$ be independent $r_0\times r_1$ matrices indexed by positions $s$. For $c=1,\ldots,C$, choose a position $s_c$, a matrix $A_c\in M_{l_c\times r_0}(K)$ and a list of $l_c$ columns $I_c$, and put $$Q(X)=\prod_{c=1}^C\det\big((A_cX_{s_c})_{I_c}\big).$$ Let $J$ index directions that add a vector $\xi_j\in K^{r_0}$ to column $b_j$ at position $s_j$. Let $\beta\in\mathbb N^J$, fix a position $s_0$, and let $S\subseteq J$. Suppose $s_j=s_0$ and $\xi_j=\xi_{j'}$ for all $j,j'\in S$. If $$\sum_{j\in S}\beta_j>\#\{c:s_c=s_0\},$$ then $$D^{[\beta]}Q=0.$$ The derivative is the coefficient of $T^\beta$ after adding all the indicated $T_j$-directions, including those outside $S$. The column lists are arbitrary; distinctness and nonvanishing of the minors are not assumed.

A determinant cannot absorb two column derivatives in the same vector direction. Counting the available minor factors gives the derivative bound used in the theta valuation estimate.

Lean
Formalism.InvariantFactorPaths.G7.Top.hitsCapPos, Formalism.InvariantFactorPaths.G7.HitsCapPos
File
Formalism/ThetaLogCoordinates/UniformThetaSubstitution/KOneConsequence.lean:33

Theta correspondence

Gaussian recovery of Taylor coefficients

statement read

Let $F:\mathbb C^N\to\mathbb C$ be continuous and have a power-series expansion $F(z)=\sum_{\alpha\in\mathbb N^N}c_\alpha z^\alpha$ converging at every $z$. Suppose $|F(z)|\le S e^{\|z\|^2/2}$ for some real $S$. For every multi-index $\alpha$, the integrand below is integrable and $$\int_{\mathbb C^N}F(z)\overline z^{\,\alpha}e^{-\|z\|^2}\,dz=\pi^N\alpha!c_\alpha,$$ where $dz$ is ordinary Lebesgue measure, $\|z\|^2=\sum_j|z_j|^2$ and $\alpha!=\prod_j\alpha_j!$.

Gaussian integration recovers a Taylor coefficient with its factorial normalization. This converts growth bounds for theta kernels into bounds for their derivatives.

Lean
Formalism.DimensionVariation.gaussianCoeffIntegral, Formalism.DimensionVariation.FockJet
File
Formalism/ThetaSupremumBounds/GaussianShiftBounds/FockJet.lean:318

Algebra and combinatorics

Fischer norm of a product of minors

statement read

Fix the project data, including their finite CM type. For each embedding $\tau$ in that type, take an independent $m\times n$ matrix $X_\tau$, integers $p_\tau,q_\tau\ge0$ with $p_\tau+q_\tau\le m$ and $p_\tau,q_\tau\le n$, and decreasing sequences of nonnegative integers $a^+_{\tau,i}$, $a^-_{\tau,i}$. Retain their first $p_\tau$ and $q_\tau$ entries as partitions $\lambda^+_\tau,\lambda^-_\tau$. Let $\Delta_{\tau,i}^+$ be the determinant in the first $i$ rows and columns, and $\Delta_{\tau,i}^-$ the determinant in rows $p_\tau+1,\ldots,p_\tau+i$ and the last $i$ columns. Set the entry following each partition equal to zero and put $$\Delta=\prod_\tau\prod_{i=1}^{p_\tau}(\Delta_{\tau,i}^+)^{a^+_{\tau,i}-a^+_{\tau,i+1}}\prod_{i=1}^{q_\tau}(\Delta_{\tau,i}^-)^{a^-_{\tau,i}-a^-_{\tau,i+1}}.$$ Rows and columns are numbered from one. For the Fischer pairing in which distinct monomials are orthogonal and $\|X^\alpha\|_F^2=\alpha!$, $$\|\Delta\|_F^2=\prod_\tau\mathcal H(\lambda^+_\tau)\mathcal H(\lambda^-_\tau),$$ where $\alpha$ is the multi-index of a monomial, $\alpha!=\prod_a\alpha_a!$, and $\mathcal H(\lambda)$ is the product of the hook lengths of the partition $\lambda$.

The exact Fischer norm supplies the hook products in the lower bound for the theta scalar. The formula uses the two partitions at each CM embedding.

Lean
GlobalTheory.LowRowFock.fock_delta
File
Development/AbelianVarietyTheory/Proofs/DivisorProofs/FockFormMinorProductEqProdHookProducts.lean:189

Algebra and combinatorics

Cancellation for interlacing weights

statement read

Put $\phi(t)=t\log t-t$ for $t>0$ and $\phi(0)=0$. Let $k\ge0$ be an integer and $S$ a real number. Let real numbers $c_i$ satisfy $0\le c_i\le S$ for $0\le i\le k$. Let real numbers $c'_i$ satisfy $c_i\ge c'_i\ge c_{i+1}$ for $0\le i<k$. For every $w\ge1$, $$\left|\sum_{i=0}^{k}\phi(w+c_i)-\sum_{i=0}^{k-1}\phi(w+c'_i)\right|\le(w+2S)\log(w+S+2).$$

Interlacing bounds the contribution of matched weight coordinates. In the theta-kernel estimate this leaves the large unmatched coordinate.

Lean
Formalism.LatticeIndexTer.sumInterlacedBound, Formalism.LatticeIndexTer.Interlace
File
Formalism/LogFactorialIndices/SignedSubspaceIndex/SumInterlacedBoundFails.lean:74

Algebra and combinatorics

Integral branching under scaling

statement read

Assume the full integral branching hypothesis for every interlacing pair: a nonzero bounded integer normalization, a nonzero coefficient polynomial with the prescribed degree identity, and mutually transpose inclusion and branching maps satisfying the specified normalization identities. Let $K$ be a number field, $I$ finite, and $N=n+1$. For $\tau\in I$ let dominant integral weights $\lambda_\tau\in\mathbb Z^N$, $\sigma_\tau\in\mathbb Z^n$ satisfy $\lambda_{\tau,i}\ge\sigma_{\tau,i}\ge\lambda_{\tau,i+1}$ and have equal determinant shifts. Put $b_\tau=\sum_i\lambda_{\tau,i}-\sum_i\sigma_{\tau,i}$. Let $M'_\tau\subset K^n$ be $\mathcal O_K$-submodules and $t_\tau\in K^\times$. For the normalized standard branching map $\operatorname{br}_0$, put $$\operatorname{br}_t=\left(\prod_\tau t_\tau^{-b_\tau}\right)\operatorname{br}_0.$$ If $\Delta_\lambda$ denotes the integral coefficient lattice realized by Weyl polynomials, then $$\operatorname{br}_t\big(\Delta_\lambda((M'_\tau\oplus\mathcal O_Kt_\tau)_\tau)\big)\subseteq\Delta_\sigma((M'_\tau)_\tau).$$ The determinant shift is the common determinant twist in the polynomial realization.

Scaling the last basis vector contributes the powers that cancel the inverse factors in the normalized map. This gives an integral restriction map between the rescaled Weyl lattices.

Lean
GlobalTheory.BranchFormula.integrality, Formalism.L2.Col.Br.BrIntegral
File
Development/AbelianVarietyTheory/Proofs/GlobalInterpolation/StatementWithTransposeOfBranching.lean:467

Algebra and combinatorics

Highest-weight branching rule

statement read

Let $n\ge0$ be an integer and put $N=n+1$. Let $V_\lambda$ be the irreducible rational representation of $\mathrm{GL}_N(\mathbb C)$ with dominant integral highest weight $\lambda$. For the upper-left block inclusion $\mathrm{GL}_n\hookrightarrow\mathrm{GL}_N$, $$V_\lambda\big|_{\mathrm{GL}_n}\simeq\bigoplus_{\lambda_i\ge\mu_i\ge\lambda_{i+1}}V_\mu.$$ Here the sum runs over dominant integral $n$-tuples $\mu$, and $V_\mu$ is the irreducible representation of highest weight $\mu$. Each summand occurs once. Weights failing these inequalities occur with multiplicity zero. The same assertions hold for restriction along $U(n)\hookrightarrow U(N)$. At rank zero, weights are empty tuples, as in the source.

This multiplicity-one rule describes the coefficient representations used in the last restriction map. Interlacing determines exactly which smaller weights occur.

Lean
GlobalTheory.Branching.branchesAlong_all
File
Development/AbelianVarietyTheory/Proofs/FoundationProofs/BranchesAlongOfIrreducible.lean:78

Theta correspondence

Lower bound for the theta scalar

statement read

Fix the matched theta-lift data: representations, Schwartz vectors, measures, coefficient map, and level. Let $\Theta:X\to H$ be the prescribed map, where $X$ is a finite-dimensional complex Hilbert space and $H$ is its Hilbert-space target. Let $H_0\subset H$ be the selected nonzero Hecke subspace. Assume the full input conditions on the lift and measurability of the rational quotient, the inner-product and Fock-coefficient formulas, the norm of the minor polynomial, the finite zeta factors, and the level-index normalization. Also assume the mass lower bound, formal-degree upper bound, weight and level size conditions, and nonvanishing of the specified compact-type component.

Let $k$ be the initial rank, $f$ the CM-type size, $n_V$ the definite lifting-space rank, $b$ the weight scale, $g_0,g_1$ the two level primes, and $\mathrm{rd}_E$ the root discriminant of the CM field $E$. The size conditions require positive lift weights at most $2b$, source weight parameters between zero and $b$, the prescribed weight differences at most $b$, and both level primes at least $(kb)^8$.

Assume the following adjoint-pairing identity for every permitted choice of covolume-one oscillator measure, fundamental domain, compatible frames and Haar measures, unit lowest-type vector $v$, vectors $U_1,U_2\in H$, and orthonormal basis $(e_j)$ of the coefficient representation: $$\langle\Theta^*U_1,\Theta^*U_2\rangle=\left(\prod_\tau d_\tau\right)^2\sum_j\langle T_{v,e_j}U_1,T_{v,e_j}U_2\rangle_{L^2}.$$ Here $\tau$ runs over the CM type, $d_\tau$ is the dimension of the prescribed lowest compact type, and $T_{v,e_j}$ is the lift operator formed from the prescribed polynomial, Schwartz vector, and coefficient map, with the prescribed measures.

Let $\Theta_{\mathrm{lev}}^*$ be the adjoint with the prescribed level projection. There is $\kappa_\Theta>0$ such that $$\Theta\Theta_{\mathrm{lev}}^*v=\kappa_\Theta v\qquad(v\in H_0).$$ For the two partitions $\lambda^\pm_\tau$ at each CM embedding $\tau$, $$\log\kappa_\Theta\ge\sum_\tau\log\big(\mathcal H(\lambda^+_\tau)\mathcal H(\lambda^-_\tau)\big)-8fn_V^2\log(b\,\mathrm{rd}_E g_0g_1n_V),$$ where $\mathcal H(\lambda)$ is the product of the hook lengths of $\lambda$.

The inner-product formula reduces the scalar estimate to the exact Fischer norm and formal-degree bounds. The hook products give the main term and the remaining factors give the logarithmic loss.

Lean
GlobalTheory.LowRow.scalar_of_form_key_lowC
File
Development/AbelianVarietyTheory/Proofs/DivisorProofs/AdelCOfInnerProductFormula.lean:214

Slope inequality

Arakelov geometry and slopes

Slope inequality for a family of maps

statement checked against Lean

Let $K$ be a number field and $W$ a finite index set. Let $N'_u$ for $u\in W$ and $C$ be Hermitian lattices over $\mathcal O_K$. Write $\mu$, $\mu_{\max}$, and $\mu_{\min}$ for the mean and extreme Harder–Narasimhan slopes, with $$\mu(A)=\frac{-2\log\operatorname{covol}(A)}{[K:\mathbb Q]\operatorname{rk}A},$$ using discriminant-normalized covolume. Let $B_u:N'_u\to C$ have height $h(B_u)\le h_u$, where height is the mean logarithm of the squared archimedean operator norms. Set $d_u=\operatorname{rk}N'_u$ and define $$s=\frac{\sum_ud_u\big(\mu(N'_u)-h_u\big)}{\sum_ud_u},\qquad U_0=\max_u\big(\mu_{\max}(N'_u)-h_u\big)-s.$$ Suppose $s-\mu_{\min}(C)\le S$ for a real $S>0$.

Suppose maps $\widetilde B_u$ into a Hermitian space of dimension $D$ project to $B_u$, and $\dim C\ge(1-\delta)D$. Assume $0\le\delta<1$, $0\le e_F\le1/4$, and $\eta_F\ge0$. For positive scalars $\gamma_u$, put $Q=\sum_u\gamma_u\widetilde B_u\widetilde B_u^*$ and suppose $$\sum_u\big\|\gamma_u\widetilde B_u^*\widetilde B_u-I\big\|_{\mathrm{HS}}^2\le e_F\sum_ud_u,\qquad\operatorname{tr}Q^2\le(1+\eta_F)\frac{(\operatorname{tr}Q)^2}{D}.$$ Here $I$ is the identity on each source and $\|T\|_{\mathrm{HS}}^2=\operatorname{tr}(T^*T)$. Then $$s-\mu(C)\le\log|W|+\frac{4U_0\log(1+S/U_0)+(\eta_F+4\sqrt{e_F})\,S}{1-\delta}.$$

The moment bounds control how much of the target is spanned by subspaces of the sources. Applying the one-map slope inequality to their Harder–Narasimhan filtrations gives the mean-slope bound.

Lean
Formalism.HN.TheoremF, FrameLA1
File
Formalism/HarderNarasimhan/StatementFrameF.lean:31

Automorphic forms and Galois representations

Second moments of period operators

statement checked against Lean

Use the parameters and multiplicity term $D_*$ of Discrete-series multiplicity estimate. For each chosen source index $u$, let $P_u$ be the Bessel–Poincaré period operator, let $d^f_u$ be the dimension of its full source space of forms, and let $c_u>0$ be its normalization. Put $$K_{uu'}=(c_uc_{u'})^{-1/2}P_u^*P_{u'},\qquad L=x^\varepsilon,\qquad b=e^L,\qquad\Gamma=\exp(c_0m^5fL),$$ where $x,\varepsilon$ are the scale parameters and $c_0$ is the estimate's constant. Then $$\|K_{uu'}\|_{\mathrm{HS}}^2\le(1+b^{-6})\frac{d^f_ud^f_{u'}}{D_*}\qquad(u\ne u'),$$ $$\|K_{uu}-I\|_{\mathrm{HS}}^2\lesssim\frac{d^f_u}{\Gamma}+2d^f_ue^{-kf\log k}.$$ Here $I$ is the identity of the source space and $\|T\|_{\mathrm{HS}}^2=\operatorname{tr}(T^*T)$. For the frame parameters $\eta_F,e_F,\delta$ of Slope inequality for a family of maps, these give $$\eta_F\le5x^{-C_H}+3e^{-x^{\varepsilon/8}},\qquad4\sqrt{e_F}\le x^{-C_H},\qquad\delta\le e^{-x^c},$$ with the construction's constants $C_H,c$.

Unfolding the Gram operator gives a sum over double cosets, and a second expansion over pairs bounds its second moment. Siegel–Weil and uniqueness of Bessel functionals identify the resulting terms.

Lean
AdmissibleBoundsFromSource.Conclusion1, Frame_at
File
Formalism/SharpRestriction/CuspSpanNormBounds/AdmissibleBoundsFromSource.lean:212

Automorphic forms and Galois representations

Discrete-series multiplicity estimate

from a worker report, not rechecked

Let $E$ be the CM field, put $f=[E:\mathbb Q]/2$, and let $k$ and $m$ be the initial rank and number of theta lifts. Let $D_{\mathrm{all}}$ be the dimension of the multiplicity space of the selected discrete series at the prescribed level on the unitary group in $r=k+m-1$ variables. Let $D_*$ be formal degree times adelic volume times the value at the identity of the level idempotent. With the estimate's constant $c$, one has $$|D_{\mathrm{all}}-D_*|\le e^{-c\,rf\log r}\,D_*.$$

This compares the target multiplicity with the main term in the period second moments. A trace formula uses the level idempotent and a matrix coefficient of the lowest compact type of the discrete series.

Lean
PlaceDegreeData.ConclB, BridgeLemma.CData.count_of_steps
File
Formalism/SharpRestriction/CuspSpanNormBounds/PlaceDegreeData.lean

Arakelov geometry and slopes

Slope bound for the image of a map

statement read

Let $K$ be a totally complex number field. Let $A$ be a finitely generated $\mathcal O_K$-submodule of a finite-dimensional Hermitian $K$-space and let $\varphi$ be a $K$-linear map to another such space, with $\varphi(A)\ne0$. Write $\mu$, $\mu_{\min}$, and $\mu_{\max}$ for the mean and extreme Harder–Narasimhan slopes. Use $\mu(A)=-2\log\operatorname{covol}(A)/([K:\mathbb Q]\operatorname{rk}A)$, with discriminant-normalized covolume and the same squared-norm normalization for the extreme slopes. Let $r_2(K)$ be the number of complex places and let $\|\varphi\|_v$ be the operator norm at $v$. Define the height by $$h(\varphi)=\frac1{r_2(K)}\sum_{v\mid\infty}\log\|\varphi\|_v^2.$$ Then $$\mu_{\min}(\varphi(A))\ge\mu_{\min}(A)-h(\varphi).$$ For any finitely generated target submodule $B\supseteq\varphi(A)$, $$\mu_{\min}(A)\le\mu_{\max}(B)+h(\varphi).$$ If $\varphi$ is injective on $A$, also $\mu_{\max}(A)\le\mu_{\max}(B)+h(\varphi)$.

The operator norms bound the change of slope under one map. Applying this to the source filtrations is the arithmetic input to the slope inequality for a family.

Lean
Formalism.HN.slope_inequality_mean, Formalism.HN.SlopeIneqMean
File
Formalism/HarderNarasimhan/HarderNarasimhanProofs/SlopeIneqMean.lean:249

Arakelov geometry and slopes

Codimension bound from second moments

statement read

Let $\mathcal I$ be a finite index set, let $V_i$ for $i\in\mathcal I$ be finite-dimensional complex Hilbert spaces, and let $H$ be another such space. Let $B_i:V_i\to H$, $c_i>0$. Put $d_i=\dim V_i$, $D=\dim H$, $R=\sum_i d_i>0$, and $Q=\sum_i c_iB_iB_i^*$. Suppose $0\le e\le1/4$, $\eta\ge0$, and $$\sum_i\|c_iB_i^*B_i-I\|_{\mathrm{HS}}^2\le eR,\qquad\operatorname{tr}(Q^2)\le(1+\eta)\frac{(\operatorname{tr}Q)^2}{D}.$$ For a real number $\varepsilon$ and subspaces $V'_i\subseteq V_i$ with $\sum_i(d_i-\dim V'_i)\le\varepsilon R$, $$\operatorname{codim}_H\sum_iB_i(V'_i)\le(\eta+4\varepsilon+4\sqrt e)D.$$ Here $I$ denotes the identity on each source. The Hilbert–Schmidt norm is $\|T\|_{\mathrm{HS}}^2=\operatorname{tr}(T^*T)$. Division by zero, if $D=0$, has the source's value zero; the conclusion then reads $0\le0$.

This estimate shows how much of the target remains spanned after subspaces are removed from the sources. It is the linear-algebra input to the slope inequality.

Lean
Formalism.HN.frame_linear_algebra_first, Formalism.HN.FrameLA1
File
Formalism/HarderNarasimhan/HarderNarasimhanProofs/FrameLinearAlgebraFirst.lean:77

Arakelov geometry and slopes

Descent of a slope bound through several maps

statement read

Let $K$ be a totally complex number field and $A\ne0$, $B$ be finitely generated $\mathcal O_K$-modules in finite-dimensional $K$-spaces with Hermitian norms at every embedding. Assume the norm families arise from invertible Hermitian Gram matrices. Fix real numbers $P,t$ and $\Lambda,\varepsilon\ge0$. Let $\varphi_h:A\otimes K\to B\otimes K$ for $1\le h\le n$ be jointly injective and integral on $A$. Suppose $$\sum_{h=1}^n\|\varphi_hx\|_\sigma^2\le e^{2P}\|x\|_\sigma^2$$ at every embedding $\sigma$, for every $x$, and suppose $B$ is spanned over $K$ by elements of $B$ of norm at most $e^\Lambda$ at all embeddings. Assume $$n\operatorname{rk}B\le(1+\varepsilon)\operatorname{rk}A,\qquad t\le\mu(A)-2P.$$ Then $$\mu(B)\ge t-\varepsilon(|t|+2\Lambda).$$ Here $\mu(A)=-2\log\operatorname{covol}(A)/([K:\mathbb Q]\operatorname{rk}A)$, with discriminant-normalized covolume.

The joint norm bound transfers a slope bound from $A$ to a target spanned by small elements. Spanning is over $K$.

Lean
GlobalTheory.DescentFamily.descent
File
Development/AbelianVarietyTheory/Proofs/DivisorProofs/HtMapLeOfPoints.lean:69

Arakelov geometry and slopes

Arithmetic degree and sublattice index

statement read

Let $K$ be a number field, $V$ a finite-dimensional $K$-space with a Hermitian norm at each embedding, and $N\subseteq L\subset V$ be $\mathcal O_K$-submodules. Suppose $L$ is finitely generated and $\operatorname{rk}N=\operatorname{rk}L=m$. Their lattice-index ideal $\mathfrak i(L,N)$ is nonzero and $$\widehat{\deg}L=\widehat{\deg}N+\frac{\log N_{K/\mathbb Q}\mathfrak i(L,N)}{[K:\mathbb Q]}.$$ Here a subscript denotes localization, and the index ideal is the ideal whose localization at every nonzero prime $\mathfrak p$ is $\mathfrak p^{\operatorname{length}_{\mathcal O_{K,\mathfrak p}}(L_\mathfrak p/N_\mathfrak p)}$, and $N_{K/\mathbb Q}$ denotes its absolute norm. Arithmetic degree uses the induced determinant metric and the source's field-degree normalization.

The formula converts a finite lattice index into an arithmetic degree correction. This degree is half the squared-norm degree used for slopes.

Lean
GlobalTheory.T.lattice_degree_eq_sublattice_degree_add_log_index
File
Development/AbelianVarietyTheory/Proofs/DivisorProofs/LatticeDegreeEqSublatticeDegreeAddLogIndexStatement.lean:7

Geometric estimate

Arakelov geometry and slopes

Target slope and map heights

statement checked against Lean

Use the field, character, and weights of Prime-power Fourier identity and the scales of Prime-power count bound. Let $C$ be the target Hermitian lattice and let $h_u$ be the height bound for the period map indexed by $u$, as in Slope inequality for a family of maps. The index $u$ specifies an initial source form and its auxiliary prime $T_u$. Let $\mathcal F$ be the signed finite-place correction from the theta lifts. Let $\mathrm{rd}_E$ be the root discriminant, let $\gamma$ be Euler's constant, and put $$\log\Omega=\gamma+\log\frac{2\pi}{\mathrm{rd}_E}-\Re\frac{L'}{L}(1,\chi_C)=\frac{L'}{L}(0,\chi_C)+\log d-\log\mathrm{rd}_E.$$ Set $B_0=2nmbx^\varepsilon$ and let $N_c$ be the congruence-error budget. With the construction's exponent $e_D$ and constant $c$, define $$E_{\mathrm{Div}}=c\,|T|\,a\log(a\,\mathrm{rd}_E)+c\,m\,x^{e_D\varepsilon}(B_0+N_c).$$ For every index $u$, $$\mu(C)+h_u-\mathcal F\le\sum_{t\in T}\Big[m\big(a_t\log a_t-a_t-a_t\log\Omega\big)-a_t\log T_u\Big]+E_{\mathrm{Div}}.$$

The estimate combines the archimedean heights of the maps with the slope of the target. Congruence defects enter through $N_c$.

Lean
DIV, LinkDIV_at
File
Formalism/RankGapBounds/MeanBoundSecondForm.lean:118

Theta correspondence

Archimedean bounds for theta kernels

statement checked against Lean

Use the field, weights, and selected set $T$ of Prime-power Fourier identity. Put $\phi(y)=y\log y-y$ for $y>0$, with value zero at $y=0$. For the $j$-th lift, let $\nu_{j,\tau,i}$ be the source weight coordinates at embedding $\tau$, shifted by $w_j(\tau)$ and the prescribed rank-dependent half-integer. Set $n_\tau=\sum_i\nu_{j,\tau,i}$. Let $\Omega_\tau$ be the coordinate scale factors, whose mean logarithm over coefficient-field embeddings is $\log\Omega$, with $\Omega$ as in Target slope and map heights.

The squared norm of the $j$-th theta kernel is at most $$\prod_\tau\Omega_\tau^{-n_\tau}\,\exp\sum_{\tau,i}\phi(\nu_{j,\tau,i})$$ times the construction's controlled prefactors. Let $\mathcal H_{j,\tau}$ be the hook-length product in the lower bound for the normalizing theta scalar, and let $e_{j,\tau}$ be the prescribed error. Then $$\sum_i\phi(\nu_{j,\tau,i})-\log\mathcal H_{j,\tau}\le\mathbf 1_{\tau\in T}\,\phi(a_\tau)+e_{j,\tau}.$$ Here the indicator is one for $\tau\in T$ and zero otherwise.

The Fock-model estimate uses $\sup_{s\ge0}s^ye^{-s^2/2}=e^{\phi(y)/2}$ for $y\ge0$, with $s^0=1$. Interlacing cancels the smaller coordinates, leaving $\phi(a_t)$ once per lift.

Lean
SizeAtEmbedding, PaddedLastEntry.BlockBound, RecNums.weight_height_sum_le
File
Formalism/ThetaSupremumBounds/StatementC.lean:13

Arakelov geometry and slopes

Covolume of integral cusp forms

statement checked against Lean

Use the field, weights, and parameters of Prime-power Fourier identity and Prime-power count bound. Let $C$ be the target Hermitian lattice, with mean slope $\mu(C)$ as in Slope inequality for a family of maps. For each $1\le j\le m$, let $\ell_3(j)$ bound the normalized logarithmic covolume of the integral holomorphic cusp forms with Petersson metric on the moduli space of abelian varieties with $\mathcal O_E$-action of signature $(k+j-1,m-j)$.

Let $\overline{\log\kappa}$ be the mean logarithm of the scalar comparing the period-pairing metric with the target metric. Let $\mathcal F$ be the signed correction from the finite places in the theta-lift construction. With the period $\Omega$ of Target slope and map heights, put $B_0=2nmbx^\varepsilon$. Then $$\mu(C)\le\overline{\log\kappa}+\sum_{j=1}^m\ell_3(j)+\mathcal F+\text{errors},$$ $$\ell_3(j)\le\sum_{\rho}\big[(k+j-1)-(m-j)\big]\,w_j(\rho)\log\Omega+c_0B_0\log\log x.$$ The sum is over the chosen CM type, $c_0$ is the estimate's constant, and the first inequality retains the construction's unspecified error terms.

The leading term comes from arithmetic degrees at CM points and the Chowla–Selberg period of $\mathbb Q(\sqrt{-d})$. Arithmetic Hilbert–Samuel and the intersection-volume inequality bound the remainder.

Lean
GlobalTheory.CovTop.CovolumeBoundOfX, IndefData.L3, CMDiffSt
File
Development/AutomorphicArithmetic/ThetaCorrespondence/ThetaPairLifts/ThirdPlaceSeries/LiftLambdaBound.lean:60

Algebraic geometry

Resolution in characteristic zero

statement read

Let $X$ be a reduced separated scheme of finite type over a field $k$ of characteristic zero. There is a finite composition of blow-ups $\pi:X'\to X$ such that $X'$ is smooth over $k$. If $X_{\mathrm{sm}}$ is the smooth locus, then $\pi$ restricts to an isomorphism $$\pi^{-1}(X_{\mathrm{sm}})\xrightarrow{\ \sim\ }X_{\mathrm{sm}}.$$ The reduced inverse image of $X\setminus X_{\mathrm{sm}}$ is a simple normal-crossing divisor.

Resolution provides smooth models with a simple normal-crossing boundary. These models are used in the geometric covolume estimate.

Lean
PositivityTheory.Resplug.ResolutionBlowup.resolutionCharZero, GlobalTheory.CompanionTheory.Hypotheses₂.ResolutionCharZero
File
Development/WeilBound/CoreStatements/ResolutionCharZero/ResolutionCharZero.lean:6

Algebraic geometry

Principalization on a smooth projective variety

statement read

Let $Y$ be an integral smooth projective variety of dimension $d$ over a characteristic-zero field $k$, and let $\mathcal I\subseteq\mathcal O_Y$ be a nonzero ideal sheaf. The ideal may be the unit ideal. There are an integral smooth projective $k$-variety $Y'$ of dimension $d$ and a morphism $\pi:Y'\to Y$, obtained by a finite sequence of blow-ups with smooth centers lying over $\operatorname{Supp}(\mathcal O_Y/\mathcal I)$, such that $\pi$ is an isomorphism off that support. Moreover, $\mathcal I\mathcal O_{Y'}$ is an effective Cartier ideal whose reduced support is a simple normal-crossing divisor: $$\mathcal I\mathcal O_{Y'}=\mathcal O_{Y'}(-D)\quad\text{for an effective Cartier divisor }D.$$

This replaces a section system's base ideal by an invertible ideal. It supports the arithmetic Hilbert–Samuel theorem and the geometric covolume estimate.

Lean
PositivityTheory.ResPlug.WeightsCompactStabilizer.principalisation_closed, GlobalTheory.Published.Principalisation
File
Development/WeilBound/CoreStatements/ResolutionCharZero/PrincipalisationClosed.lean:14

Arakelov geometry and slopes

Arithmetic Hilbert–Samuel asymptotic

statement read

Let $R$ be an integral ring finite over $\mathbb Z$. Let $X$ be an integral projective $R$-scheme, flat over $\mathbb Z$, of dimension $D+1$. Let $\overline L$ and $\overline N$ be Hermitian line bundles on $X$. Assume Principalization on a smooth projective variety. Suppose $\overline L$ is nef: its metric is semipositive and its degrees on vertical and horizontal curves are nonnegative. Suppose its top arithmetic self-intersection exists and equals $a$ in the source's normalization.

Write $\widehat H^0(X,\overline L^{\otimes m}\otimes\overline N)$ for the integral sections of $L^{\otimes m}\otimes N$ with supremum norm at most one at every archimedean embedding. These sets are finite for all sufficiently large integers $m$, and $$\log\#\widehat H^0(X,\overline L^{\otimes m}\otimes\overline N)=\frac{a}{(D+1)!}m^{D+1}+o(m^{D+1}).$$

This asymptotic converts arithmetic intersection numbers into counts of small sections. The base ring $R$ need not be normal.

Lean
GlobalTheory.arithHilbertSamuel_of_principalisation, GlobalTheory.Published.ArithHilbertSamuel
File
Development/AbelianVarietyTheory/Proofs/DivisorProofs/ArithHilbertSamuelOfPrincipalisation.lean:11

Arakelov geometry and slopes

Arithmetic intersection and volume

statement read

Let $X$ be the integral arithmetic variety of dimension $D+1$ defined by an ideal in the given multiprojective coordinate ring. Fix an arithmetic intersection system, a birational coordinate tuple, and an ambient ideal $P_0$ contained in the coordinate ideal of $X$. Let $\overline L$ be the given Hermitian line bundle. Assume relative semipositivity with respect to $P_0$, invariance under conjugation, and nefness on the fibres. Also assume that the geometric degree is nonnegative on every arithmetic surface whose ideal contains the coordinate ideal of $X$.

Write $\widehat H^0(X,\overline L^{\otimes n})$ for the integral sections of supremum norm at most one in the prescribed chart lattice. Define the arithmetic volume by $$\widehat{\operatorname{vol}}(\overline L)=\limsup_{n\to\infty}\frac{\log\#\widehat H^0(X,\overline L^{\otimes n})}{n^{D+1}/(D+1)!}.$$ Then $$\widehat{\deg}\big(\widehat c_1(\overline L)^{D+1}\big)\le\widehat{\operatorname{vol}}(\overline L).$$ Here $\widehat c_1$ is the arithmetic first Chern class and $\widehat{\deg}$ is arithmetic degree, with the prescribed intersection normalization.

The inequality bounds the intersection contribution to the cusp-form covolume. Its proof uses the comparison of supremum and $L^2$ norms for small sections.

Lean
PositivityTheory.Pf1.HilbertSamuelMoriwaki.moriwakiBig_outright, ReciprocityTheory.ChartRestriction.G.ArakelovGeometry.MoriwakiBigStatement
File
Development/WeilBound/CoreStatements/AutomorphicKernels/MoriwakiBigness/MoriwakiBigTheorem/Outright.lean:9

Abelian varieties and moduli

Seminegativity of a semiabelian Lie bundle

statement read

Let $C$ be a smooth projective geometrically connected curve over a characteristic-zero field $k$, let $G\to C$ be a semiabelian scheme, and let $W$ be a vector bundle of rank $r$ with an injective morphism of sheaves $W\hookrightarrow\operatorname{Lie}(G/C)$. Then $$\deg_C\det W\le0.$$ A locally free cokernel is not assumed.

This bounds the determinant degrees of locally free subsheaves of the Lie bundle. The inequality is one of the geometric inputs to the cusp-form covolume estimate.

Lean
GlobalTheory.Nef.semiAbelianLieSeminegative, GlobalTheory.Published.SemiAbelianLieSeminegative
File
Development/AbelianVarietyTheory/Proofs/DivisorProofs/SemiAbelianLieSeminegativeBis.lean:22

Abelian varieties and moduli

CM contribution to degrees on covers

statement read

Use the field $E$ and quadratic character $\chi_C$ of Prime-power Fourier identity. Assume the full uniform hypotheses on the parameter families, points, covers, and flat-section degrees in the construction. These include the closed-family conditions, the prescribed conditions on points and compatible comparison data, the inequality excluding the terminal stage, and the conditions on a good cover. The degree condition is required for every flat section on every component.

Eventually in each permitted parameter family, at every permitted point and on every good cover of a nonterminal stage of signature $(p,q)$, the prescribed flat sections exist. At a selected embedding $\tau$, let $c_1(\tau)$ be minus their arithmetic degree, averaged over components. Put $L=x^\varepsilon$ for the scale parameters $x,\varepsilon$, and define $$\log\Omega=\gamma+\log(2\pi/\mathrm{rd}_E)-\Re\frac{L'}L(1,\chi_C),$$ where $\gamma$ is Euler's constant and $\mathrm{rd}_E$ is the root discriminant. Then $$\left|c_1(\tau)+\frac{p-q}{2}\log\Omega\right|\le3qL,\qquad\sum_{r\ \mathrm{bad}}\log r\le40L.$$ The sum runs over rational primes at which the cover's prescribed integral model is bad. The complex comparison data exist, and the indicated local models are good at every rational prime.

This estimates the CM contribution to the covolume on each cover. The error $3qL$ remains in the geometric estimate.

Lean
GlobalTheory.CovTop.FlatVal.flat_val_alt
File
Development/AutomorphicArithmetic/ThetaCorrespondence/ThetaPairLifts/ThirdPlaceSeries/FlatVal.lean:13

Arakelov geometry and slopes

Small sections from comparison of norms

statement read

Fix a characteristic-zero coordinate field and arithmetic multiprojective section data of dimension $D+1$, with $D\ge1$. Assume the distinguished coordinate block gives a birational map. Fix a Hermitian line bundle of unit degree in that block, invariant under conjugation, together with a contained ambient ideal and relative semipositivity. Assume the prescribed analytic resolution data with their Borel measurable structure and a smooth volume measure.

Let $n\ge0$ be an integer. Let $M_n$ be the integral lattice of sections of degree $n$ in the distinguished block, with rank $r$, supremum norm $\|\cdot\|_\infty$, and symmetrized $L^2$ norm $\|\cdot\|_2$. Let $G_n$ be the symmetrized real Gram matrix in the prescribed integral section basis. Put $\chi_n=-\tfrac12\log\det G_n$, using the source's zero convention if no such basis is available. Suppose $c>0$ and $$\|s\|_\infty\le c\|s\|_2$$ for every section represented by an integral multihomogeneous polynomial of that degree. Then $$r\le2c^2,\qquad\chi_n-r\log(c\sqrt r)\le\log\#\{s\in M_n:\|s\|_\infty\le1\}.$$ At $r=0$, the product with the logarithm is interpreted as zero. Logarithms and cardinalities retain the source's conventions; the conclusion does not additionally assert finiteness.

A comparison of supremum and $L^2$ norms gives a lower bound for the count of small sections. This is the finite section estimate used in the arithmetic intersection-volume inequality.

Lean
GroundTheory.Heights₂.Heights.HilbertSamuel.SupBoundCount.full, GroundTheory.Heights₂.Heights.HilbertSamuel.SupBoundCount.Statement
File
GroundTheory/HeightsAndTheta/Heights/HilbertSamuel/AssemblyBase.lean:750

Arakelov geometry and slopes

Gluing lattices from localizations

statement read

Let $K$ be a number field and $L_0\subset K^D$ a full $\mathcal O_K$-lattice. For every nonzero prime ideal $\mathfrak p$ choose a full lattice $L(\mathfrak p)\subset K^D$, and assume $L(\mathfrak p)_\mathfrak p=(L_0)_\mathfrak p$ except at finitely many primes. Then $$L=\bigcap_\mathfrak p L(\mathfrak p)_\mathfrak p\subset K^D$$ is a full $\mathcal O_K$-lattice and satisfies $L_\mathfrak p=L(\mathfrak p)_\mathfrak p$ at every prime. The subscript denotes localization inside $K^D$.

This constructs the target lattice from its prescribed local images. Only finitely many localizations may differ from those of $L_0$.

Lean
Formalism.DivisionValues.lattice_gluing, Formalism.DivisionValues.LatticeGlue
File
Formalism/QuotientMetricDivisors/HermitianTwistDivisors/SecondStepGluing.lean:256

Cohomological input

Galois cohomology, Iwasawa theory and K-theory

Congruence defects and Galois cohomology

statement checked against Lean

Let $E$ be a number field, $\ell$ a prime, and $r\ge1$ an integer. Let $G_{E,S_\ell}$ be the Galois group of the maximal extension of $E$ unramified outside $\ell$. Define the localization kernel $$\mathcal S_{\ell,r}(E)=\ker\Big[H^1\big(G_{E,S_\ell},\mathbb Q_\ell/\mathbb Z_\ell(-r)\big)\to\prod_{v\mid\ell}H^1\big(E_v,\mathbb Q_\ell/\mathbb Z_\ell(-r)\big)\Big].$$ Here $E_v$ is the completion at $v$, and $(-r)$ is the Tate twist.

In the theta-lift construction, let $M$ be a local Hecke lattice, let $Z$ be the selected sum of Hecke eigenspaces, and let $P_Z$ be projection along the other eigenspaces. Let $\eta_j$ be the weighted sum of the local lengths of $P_ZM/(M\cap Z)$ for the $j$-th lift, and let $d_j$ be the dimension of its theta image. Write $f=[E:\mathbb Q]/2$, let $m$ be the number of lifts, and put $L=x^\varepsilon$ and $b=e^L$, where $x$ and $\varepsilon$ are the scale parameters. Let $N_c$ be the congruence-error budget and $t$ the Tate-twist cutoff. Provided the lengths are finite and the series converges, eventually and uniformly in the construction's parameters, $$\eta_j\le d_j\Big[c_0\big\{fmbL^2+(\log x)^2N_c\big\}+\sum_{\ell\ \mathrm{prime}}\log\ell\sum_{1\le r<t}\operatorname{length}_{\mathbb Z_\ell}\mathcal S_{\ell,r}(E)\Big],$$ where $c_0$ is the estimate's constant.

This bounds failures of integral Hecke projections by Galois cohomology. It supplies the congruence contribution to the geometric estimate.

Lean
LinkN_at, GlobalLatticeIndexTheorem
File
Formalism/SharpRestriction/GlobalTheoremInputs.lean:40

Galois cohomology, Iwasawa theory and K-theory

Bound for localization kernels

statement checked against Lean

Use the groups $\mathcal S_{\ell,r}(E)$ of Congruence defects and Galois cohomology. Let $\psi$ be a primitive Dirichlet character, and let $K_\psi$ be the number field it cuts out. Fix $0<\varepsilon<1/52$ and an integer-valued function $\nu$ satisfying $$x^{2+8\varepsilon}\le2^{\nu(x)}\le2x^{2+8\varepsilon}.$$ There are $J\ge1$ and $X_0$ with the following property. For every $X\ge X_0$ and every pair $(q,d)$ satisfying the construction's auxiliary-prime conditions at $U=X^{1/J}$, there is $Y$ with $X^{1/J}\le Y\le\sqrt X$. For every $x\in[Y,Y^2]$, put $$E=\mathbb Q(\zeta_{2^{\nu(x)+2}})^+K_\psi\,\mathbb Q(\sqrt{-q},\sqrt{-d}),$$ where $\zeta_h$ is a primitive $h$-th root of unity and the superscript $+$ denotes the maximal real subfield. All the lengths in the following sum are finite, and $$\sum_{\ell\ \mathrm{prime}}\log\ell\sum_{1\le r<\lfloor x^{401}\rfloor}\operatorname{length}_{\mathbb Z_\ell}\mathcal S_{\ell,r}(E)\le\exp\big(x^{\varepsilon}/2\big).$$

This estimate bounds the cohomological contribution to the congruence defects. The field $E$ varies with $x$, and the right side is $|\operatorname{disc}E|^{o(1)}$.

Lean
UpperTheory.Top.VWin, UpperTheory.One.vwin_of_row
File
Development/AutomorphicArithmetic/CharacterBoxes/ExistsPageN.lean:49

Automorphic forms and Galois representations

Selmer lengths and trace congruences

statement checked against Lean

Let $K_0$ be an imaginary quadratic field, let $\ell$ be an odd prime, and let $\Psi$ be the prescribed Galois character, subject to the parity, coprimality, and residual conditions. Write $G_{K_0}$ for the absolute Galois group. Fix the specified finite set $\mathcal R$ of three-dimensional Galois representations. Each member is irreducible automorphic or has the form $W\oplus1$, with a twist of $W$ attached to a newform. Let $V_\ell$ be the normalized length of the group of cohomology classes with torsion coefficients twisted by $\Psi^{-1}$, unramified everywhere and trivial at the decomposition groups above the level and $\ell$. Let $\sigma\in G_{K_0}$ be such that no representation $\rho\in\mathcal R$ has $\operatorname{tr}\rho(\sigma)=\Psi(\sigma)+1+\Psi(\sigma)^{-1}$. Then $$V_\ell\le2\sum_{\rho\in\mathcal R}v_\ell\Big(\operatorname{tr}\rho(\sigma)-\big(\Psi(\sigma)+1+\Psi(\sigma)^{-1}\big)\Big),$$ where $v_\ell$ is the normalized $\ell$-adic valuation.

Moreover there is a nonzero integer $A$, with $\log|A|$ polynomial in the level, the weight and the orbit size, such that, for all primes $\ell$ outside a fixed finite set, $$|A|_\ell\le\prod_{\rho\in\mathcal R}|\operatorname{tr}\rho(\sigma)-(\Psi(\sigma)+1+\Psi(\sigma)^{-1})|_\ell,$$ where $|\cdot|_\ell$ is the normalized $\ell$-adic absolute value.

Selmer classes give square-zero deformations at $\Psi+1+\Psi^{-1}$, and automorphy and rigidity bound their cotangent length by trace valuations. The integer $A$ gives simultaneous bounds over the primes with weights $\log\ell$.

Lean
UpperTheory.V.Z.J.oneCharU_of_katoTwoW_rig_admI_automorphy, UpperTheory.V.U3.CorGK
File
Development/AutomorphicArithmetic/AutomorphicVarieties/UnitaryThreeGroups/KatoUse.lean:44

Galois cohomology, Iwasawa theory and K-theory

Vanishing of two rank-two Selmer groups

from a worker report, not rechecked

Let $g$ be a newform satisfying the prescribed ordinary, residual, and local hypotheses. Let $V_g$ be its two-dimensional Galois representation, and let $\xi$ be the prescribed quadratic character. Then the two Bloch–Kato Selmer groups vanish: $$H^1_f(V_g)=0,\qquad H^1_f(V_g\otimes\xi)=0.$$ The base field, twist, and detailed hypotheses remain those prescribed in the construction.

The vanishing supplies the rank-two input to the trace-congruence estimate. Euler-system classes from Siegel units, an explicit Hodge–Tate formula, and local nonvanishing feed into Selmer duality.

Lean
UpperTheory.V.Z.Kato.WeilSphericalRecords.katoTwoW_of_printsH, GlobalTheory.Ord2.vanishOrdL_of_zetaUr
File
Development/AutomorphicArithmetic/AutomorphicVarieties/ModularityArgument/ThresholdArguments/DenseThresholdCases/KatoTwoWOfPublishedH.lean:48

Galois cohomology, Iwasawa theory and K-theory

Bound for integral second cohomology

statement checked against Lean

Let $\ell$ be a prime, let $\mu_h$ denote the group of $h$-th roots of unity, and let $\Delta_L$ denote the discriminant of a number field $L$. Write $G_{L,S_\ell}$ for the Galois group of the maximal extension of $L$ unramified outside $\ell$. There is an absolute constant $C_K$ such that for every totally complex number field $L$ containing $\mu_\ell$ (or $\mu_4$ if $\ell=2$) and every $j\ge1$, the group $H^2(G_{L,S_\ell},\mathbb Z_\ell(j+1))$ is finite and $$\log\#H^2\big(G_{L,S_\ell},\mathbb Z_\ell(j+1)\big)\le\exp\Big((2j[L:\mathbb Q])^{C_K}\log(2|\Delta_L|)\Big).$$

This controls the integral cohomology terms in the localization-kernel estimate. The proof uses bounds for odd homology, homological stability, torsion estimates, cyclotomic class groups, and Poitou–Tate duality.

Lean
UpperTheory.V.LemmaK, UpperTheory.V.ZK.lemmaK_litN_alt
File
Development/AutomorphicArithmetic/AutomorphicVarieties/GeneratorCountTuples/LemmaK.lean:11

Galois cohomology, Iwasawa theory and K-theory

Local cohomology finiteness and dimension

statement read

Let $\ell$ be prime, $K/\mathbb Q_\ell$ a finite extension in a fixed algebraic closure, $r>0$, and $M$ a finite discrete $\mathbb Z/\ell^r\mathbb Z$-module with continuous action of $G_K$, the absolute Galois group of $K$. Then $H^i(K,M)$ is finite for $i=0,1,2$, and $$H^i(K,M)=0\qquad(i\ge3).$$

This gives finite local cohomology groups for the localization and duality arguments. The vanishing above degree two limits the terms in those arguments.

Lean
CompanionTheory.LocalTateOrthogonality.Gcoh.top_locfin, Formalism.GaloisCohomology.LocalFinitenessK
File
CompanionTheory/AutomorphicTheory/PadicCyclotomicFields/GcohTop.lean:13

Galois cohomology, Iwasawa theory and K-theory

Local Tate duality

statement read

Let $\ell$ be prime, $K/\mathbb Q_\ell$ finite, $r>0$, and $M$ a finite discrete $\mathbb Z/\ell^r\mathbb Z$-module with continuous action of $G_K$, the absolute Galois group of $K$. Put $M^\vee(1)=\operatorname{Hom}_{\mathbb Z/\ell^r\mathbb Z}(M,\mathbb Z/\ell^r\mathbb Z)(1)$, with the contragredient and cyclotomic actions. There is an isomorphism $$H^2(K,\mathbb Z/\ell^r\mathbb Z(1))\simeq\mathbb Z/\ell^r\mathbb Z,$$ and cup products give perfect pairings $$H^i(K,M)\times H^{2-i}(K,M^\vee(1))\longrightarrow H^2(K,\mathbb Z/\ell^r\mathbb Z(1)),\qquad i=0,1,2.$$ Perfectness means that each induced map to the corresponding linear dual is bijective, in both orders also for the degree-one pairing.

Cup products identify each local cohomology group with the dual of its complementary-degree group. These pairings control the local conditions in the global cohomological argument.

Lean
CompanionTheory.LocalTateOrthogonality.Gcoh.top_tate, Formalism.GaloisCohomology.LocalTateDualityK
File
CompanionTheory/AutomorphicTheory/PadicCyclotomicFields/GcohTop.lean:11

Galois cohomology, Iwasawa theory and K-theory

Local Euler characteristic

statement read

Let $\ell$ be prime, $K/\mathbb Q_\ell$ finite, $r>0$, and $M$ a finite discrete $\mathbb Z/\ell^r\mathbb Z$-module with continuous action of the absolute Galois group of $K$. Then $$\#H^1(K,M)=\#H^0(K,M)\,\#H^2(K,M)\,(\#M)^{[K:\mathbb Q_\ell]}.$$

The formula relates the sizes of the three local cohomology groups. It is used with local duality in the localization-kernel estimates.

Lean
CompanionTheory.LocalTateOrthogonality.Gcoh.top_euler, Formalism.GaloisCohomology.LocalEulerCharK
File
CompanionTheory/AutomorphicTheory/PadicCyclotomicFields/GcohTopEuler.lean:6

Galois cohomology, Iwasawa theory and K-theory

Class-group ranks in cyclotomic towers

statement read

Let $K/\mathbb Q$ be an abelian number field and $\ell$ a prime. Suppose $K$ contains a primitive $\ell$-th root of unity if $\ell$ is odd, or a primitive fourth root if $\ell=2$. There is an integer $r\ge0$ such that for every $n\ge0$, with $K_n=K(\mu_{\ell^n})$, $$\#\big(\operatorname{Cl}(\mathcal O_{K_n})/\ell\operatorname{Cl}(\mathcal O_{K_n})\big)\le\ell^r.$$ Here $\mu_m$ denotes the group of $m$-th roots of unity and $\operatorname{Cl}$ the ideal class group.

The bound controls class-group contributions in the cyclotomic tower. It may depend on $K$ and $\ell$, and is uniform in $n$.

Lean
UpperTheory.V.FW.ferreroWashington, UpperTheory.V.Published.FerreroWashington
File
Development/AutomorphicArithmetic/AutomorphicVarieties/GaloisLibrary/FerreroWashingtonIndex/FerreroWashington.lean:104

Galois cohomology, Iwasawa theory and K-theory

Rank bound for selected odd homology classes

statement read

Let $L$ be a totally complex number field and $r_2$ its number of complex places. Let $k\ge0$ be even, $k+1<r_2$, $b\ge1$, and $n\ge2(k+1)+4$. For $x\in H_{k+1}(\mathrm{GL}_n(\mathcal O_L),\mathbb Z)$ define $z_0(x)=x$ and recursively $$z_{t+1}(x)=(\operatorname{diag}_b)_*z_t(x)-b^{t+2}(\operatorname{stab})_*z_t(x).$$ Here $\operatorname{diag}_b$ repeats a matrix in $b$ diagonal blocks, and $\operatorname{stab}$ embeds it as the first block with an identity complement; both maps go from rank $b^tn$ to rank $b^{t+1}n$. Then $$\operatorname{rank}_{\mathbb Z}\left\langle z_k(x):x\in H_{k+1}(\mathrm{GL}_n(\mathcal O_L),\mathbb Z)\right\rangle\le r_2.$$

The estimate bounds the span produced by the displayed operations on odd homology classes. It is used in the bound for integral second cohomology.

Lean
PositivityTheory.KB.bOdd_zk, UpperTheory.V.ZK.BOdd
File
Development/AutomorphicArithmetic/AutomorphicVarieties/LiteratureResults/KTheoryStability/ReductionBoundsTorsion/TorsionIntegers/Bridge.lean:34

Galois cohomology, Iwasawa theory and K-theory

Homological stability for general linear groups

statement read

Let $R$ be a commutative ring of stable range at most $s$. For integers $i,n\ge0$ with $n\ge2i+s+2$, the standard block inclusion induces an isomorphism $$H_i(\mathrm{GL}_n(R),\mathbb Z)\xrightarrow{\ \sim\ }H_i(\mathrm{GL}_{n+1}(R),\mathbb Z).$$ Stable range at most $s$ means that every unimodular row of length $n+1$ with $n\ge s$ can be shortened to length $n$ by adding multiples of its last entry to the preceding entries.

For rings of integers, stable range at most two gives the range $n\ge2i+4$. This supplies the stabilization step in the integral cohomology bound.

Lean
PositivityTheory.v8.KStability.kStability, UpperTheory.V.Published.KStability
File
Development/AutomorphicArithmetic/AutomorphicVarieties/LiteratureResults/KTheoryStability/HomologicalStability/KStability.lean:3

Automorphic and Galois input

Automorphic forms and Galois representations

Galois representations of polarized cusp forms

statement read

Let $F$ be a CM field and let $\pi$ be a regular algebraic polarized cuspidal automorphic representation of $\mathrm{GL}_a(\mathbb A_F)$. Write $G_F$ for the absolute Galois group of $F$. For every prime $\ell$ there is a continuous semisimple representation $$\rho_{\pi,\ell}:G_F\longrightarrow\mathrm{GL}_a(\overline{\mathbb Q}_\ell).$$ Its Hodge–Tate weights are determined by the weight of $\pi$. It is crystalline at the places above $\ell$ where $\pi$ is spherical. At finite places, local-global compatibility holds under the prescribed local hypotheses: exact compatibility for toral parameters, and zero monodromy with matching characteristic polynomials in the tame case.

These representations connect automorphic congruences to the cohomological bounds. Their construction uses étale cohomology of powers of universal abelian schemes over compact unitary moduli spaces, congruences, and theta lifts.

Lean
UpperTheory.Gal.Int.galoisLocalGlobalStrong_recParameter, UpperTheory.GalStar.Row0b
File
Development/AutomorphicArithmetic/Foundations/GaloisRepresentations/TypeConditions/WeightsConjBis.lean:54

Automorphic forms and Galois representations

Base change from a definite unitary group

statement read

Let $E/E^+$ be a CM extension unramified at every finite place and let $n\ge1$. Let $\pi$ be an automorphic representation of the unitary group of the identity Hermitian form in $n$ variables. There is a finite decomposition $n=\sum_jn_j$ into positive ranks. There are discrete automorphic representations $V_j$ of $\mathrm{GL}_{n_j}(\mathbb A_E)$, with local factors $\sigma_{j,w}$ at finite places $w$, and irreducible local representations $\Pi_w$ of $\mathrm{GL}_n(E_w)$. They have the following properties. Each $V_j$ has an injective equivariant realization in the prescribed induced forms built from a unitary normalized conjugate-self-dual cuspidal subspace and absolute-power Hecke twists.

$\Pi_w$ is a subquotient of the normalized parabolic product of the $\sigma_{j,w}$. It is spherical wherever $\pi$ is unramified. At split places it agrees with the local component of $\pi$ under the split unitary identification. At an inert place, assume the prescribed compatibility of local involution, uniformizer, and residue cardinality. If that place admits an unramified unitary Satake parameter $\alpha=(\alpha_1,\ldots,\alpha_{\lfloor n/2\rfloor})$, the Satake multiset of $\Pi_w$ is $$\{\alpha_i,\alpha_i^{-1}:1\le i\le\lfloor n/2\rfloor\}\ \sqcup\ \{1\}^{\,n\bmod2}.$$ If $\pi$ has highest weight $\lambda$ at an embedding, the multiset union of the infinitesimal characters of the $V_j$ there is $$\{\lambda_i+(n+1)/2-i:1\le i\le n\}.$$

Base change transfers the definite unitary forms to general linear groups for the Galois argument. The local conclusions retain the stated restrictions at inert places.

Lean
UpperTheory.BaseChangeIdentities.baseChangeN, PositivityTheory.Rg.BcN.BaseChangeIsobaricN
File
Development/AutomorphicArithmetic/Foundations/BaseChangeIdentities/DefiniteNHolds.lean:5

Representation theory of reductive groups

Irreducibility of unitary parabolic induction

statement read

Let $k$ be a nonarchimedean local field. Let $\pi_1,\pi_2$ be smooth irreducible unitarizable complex representations of $\mathrm{GL}_r(k)$ and $\mathrm{GL}_t(k)$. Their normalized parabolic induction $$\pi_1\times\pi_2=\operatorname{Ind}_{P_{r,t}}^{\mathrm{GL}_{r+t}(k)}(\pi_1\boxtimes\pi_2)$$ is irreducible, where $P_{r,t}$ is the standard parabolic subgroup with Levi $\mathrm{GL}_r\times\mathrm{GL}_t$, and the induction includes the square root of its modulus character.

This supplies the irreducibility input for products of local unitary representations. The local field may have characteristic zero or positive characteristic.

Lean
PositivityTheory.RgUd.All.bernsteinUnitaryProductIrreducible, GlobalTheory.Published.BernsteinUnitaryProductIrreducible
File
Development/WeilBound/RepresentationTheoryGL/UnitaryDualProof/BernsteinExponentBounds/BernAll.lean:27

Representation theory of reductive groups

Bounds for unramified Satake eigenvalues

statement read

Let $E$ be a totally complex number field and $\pi$ a unitary cuspidal automorphic representation of $\mathrm{GL}_a(\mathbb A_E)$. If $\pi$ has the unramified Satake multiset $B_w$ at a finite place $w$, with $q_w$ the residue-field cardinality, then every $\alpha\in B_w$ satisfies $$q_w^{-1/2}<|\alpha|<q_w^{1/2}.$$

These strict bounds use the normalized Satake convention. They control the local automorphic parameters entering the Galois and Selmer arguments.

Lean
PositivityTheory.JacquetShalikaLocal.jacquetShalikaBound, GlobalTheory.Published.JacquetShalikaBound
File
Development/WeilBound/CoreStatements/PublishedTheorems/JacquetShalikaBound/JacquetShalikaBound.lean:77

Representation theory of reductive groups

Formal degree with compact-dual normalization

statement read

For a real unitary group $U(p,q)$, put $N=p+q$. Let $\lambda_1>\cdots>\lambda_N$ with $\lambda_i-(N-1)/2\in\mathbb Z$, and let $\pi$ be a discrete-series representation with the specified sign vector of signature $(p,q)$ and Harish–Chandra parameter $\lambda$. Fix a compact-dual Haar measure system. Assume the positive scalings of the reference measures match this system under every coordinate permutation compatible with the sign vector. Let $dg$ be the resulting Haar measure and let $w\ne0$ be in the Hilbert space of $\pi$. Then $$\frac{\|w\|^4}{\displaystyle\int_{U(p,q)}|\langle w,\pi(g)w\rangle|^2\,dg}=\frac{\displaystyle\prod_{i<j}|\lambda_i-\lambda_j|}{\displaystyle\prod_{i<j}(j-i)}.$$

The formula supplies the formal-degree factor in the theta-scalar estimate and the multiplicity count. Its value depends on the stated compact-dual measure normalization.

Lean
PositivityTheory.RgFd.B.formalDegreeCompactDual, GlobalTheory.Published.FormalDegreeCompactDual
File
Development/WeilBound/RepresentationTheoryGL/DiscreteSeries/DiscreteSeriesCovers/FdTopB.lean:9

Representation theory of reductive groups

Cuspidal characters of finite general linear groups

statement read

Let $p$ be prime, $d\ge1$, and $K=\mathbb F_{p^d}$. A character $\theta:K^\times\to\mathbb C^\times$ is regular if its $d$ Frobenius conjugates are distinct. There are cuspidal characters $\chi_\theta$ of $\mathrm{GL}_d(\mathbb F_p)$ for regular $\theta$, with $$\chi_\theta(1)=\prod_{i=1}^{d-1}(p^i-1).$$ If $g\in\mathrm{GL}_d(\mathbb F_p)$ has irreducible characteristic polynomial and $z\in K^\times$ is a root, then $$\chi_\theta(g)=(-1)^{d-1}\sum_{i=0}^{d-1}\theta(z^{p^i}).$$ Here $\theta^{p^i}(z)=\theta(z^{p^i})$. Two regular parameters give the same character exactly when $\theta'=\theta^{p^i}$ for some $0\le i<d$; every cuspidal character arises in this way.

These characters specify the finite-field cuspidal types used in the automorphic construction. Regular multiplicative characters parametrize them up to Frobenius conjugacy.

Lean
GlobalTheory.greenRegularCharacters_holds, GlobalTheory.Published.GreenRegularCharacters
File
Development/AbelianVarietyTheory/Proofs/InertiaEigenvalues/GreenCuspidalCharsByRegularCharsProof.lean:137

p-adic Hodge theory and formal groups

Hodge jumps under uniform trace approximation

statement read

Let $E$ be a number field, $\ell$ prime, $a\ge1$, and $r:G_E\to\mathrm{GL}_a(\overline{\mathbb Q}_\ell)$ continuous and semisimple. Here $G_E$ is the absolute Galois group. Fix an algebraic closure of $E$, a place $w$ above $\ell$, a valuation place above $w$, and the prescribed local period and integral period-ring data with their compatibility conditions. Assume a topological $\mathbb Z_\ell$-algebra with a compatible local comparison map inducing the prescribed potentially semistable comparison. Use the resulting embeddings of $E_w$ to read the Hodge jumps. Let $\mathcal C$ be a set of continuous representations of the same rank. Assume that for every $\eta>0$ there are a finite $F\subset\mathcal C$ and scalars $c_\rho\in\overline{\mathbb Q}_\ell$ with $$\sup_{g\in G_E}\left|\operatorname{tr}r(g)-\sum_{\rho\in F}c_\rho\operatorname{tr}\rho(g)\right|_\ell\le\eta.$$ Fix multisets of integers $J_\tau$ indexed by embeddings $\tau:E\to\overline{\mathbb Q}_\ell$. If every $\rho\in\mathcal C$ is semistable at the chosen place $w\mid\ell$ and has jumps $J_\tau$ for each embedding induced from $E_w$, then $r$ is de Rham at $w$ with those jumps. If all the approximants are crystalline there, $r$ is crystalline there as well.

This preserves the required Hodge jumps when congruences remove restrictions on automorphic weights. The approximation is uniform on the Galois group and uses finite linear combinations of traces.

Lean
GlobalTheory.LimitTop.closedAtL, GlobalTheory.Cong.ClosedAtL
File
Development/AbelianVarietyTheory/Proofs/SatakeParameters/ClosedAtL.lean:25

Representation theory of reductive groups

Schur orthogonality for square-integrable representations

statement read

Let $G$ be a locally compact Hausdorff topological group with a regular bi-invariant Haar measure $dg$ on its Borel sets. Let $\pi$ be a strongly continuous irreducible unitary representation on a complex Hilbert space $H$. Irreducibility means that the only closed invariant complex subspaces are zero and $H$. Suppose some nonzero vectors $v_0,w_0$ have square-integrable matrix coefficient. Then every matrix coefficient is square integrable and there is $d_\pi>0$ such that $$\int_G\langle w,\pi(g)v\rangle\,\overline{\langle w',\pi(g)v'\rangle}\,dg=\frac{\langle v',v\rangle\langle w,w'\rangle}{d_\pi}.$$ The product integrand is integrable for all $v,w,v',w'\in H$. Inner products are linear in the second argument.

The constant $d_\pi$ is the formal degree for the chosen Haar measure. Orthogonality supplies the matrix-coefficient normalization used in the theta and trace-formula calculations.

Lean
GlobalTheory.Published.schurOrthogonalitySquareIntegrable', GlobalTheory.Published.SchurOrthogonalitySquareIntegrable
File
Development/AbelianVarietyTheory/PublishedTheorems/TwistedTraceFormula/SchurOrthogonalitySquareIntegrable.lean:210