Transcendence theory in Lean 4

2 Tools

General results used by several of the proofs below. Most of them are in the namespace Transcendence, stated for general use rather than for the proof that first needed them. Two facts about exponential polynomials from the four exponentials development, Lemmas 2.6 and 2.8, are here too, because two of the tools rest on them.

The house \(\overline{\left|\alpha \right|}\) of an algebraic number \(\alpha \) is the largest absolute value of its conjugates. The length \(L(P)\) of a polynomial \(P\in \mathbb {Z}[X][Y]\) is the sum of the absolute values of its integer coefficients, and \(M(F)\) is the Mahler measure of \(F\in \mathbb {C}[X]\).

2.1 Algebraic numbers

Lemma 2.1 Liouville’s inequality, house form
✓
#

Let \(K\) be a number field of degree \(d\), let \(\sigma \colon K\to \mathbb {C}\) be a field embedding, and let \(\alpha \in K\), \(\alpha \neq 0\). If \(c\alpha \) is an algebraic integer for some integer \(c\neq 0\), then

\[ 1\le |c|^{d}\, |\sigma (\alpha )|\, \overline{\left|\alpha \right|}^{\, d-1}. \]

Source: Standard (Liouville’s inequality).

Proof ▶

The norm of \(c\alpha \) is a non-zero integer. Its absolute value is \(|c|^{d}|\sigma (\alpha )|\) times the absolute values of the other \(d-1\) conjugates of \(\alpha \), each at most \(\overline{\left|\alpha \right|}\).

Lemma 2.2 A common denominator for monomials
✓

Let \(K\) be a number field and \((\theta _i)_{i\in I}\) a finite family in \(K\). There are a non-zero algebraic integer \(b\in K\) and a real number \(H\ge 1\) such that for every \(e\in \mathbb {N}^{I}\) and every \(E\in \mathbb {N}\) with \(\sum _ie_i\le E\), the number \(b^{E}\prod _i\theta _i^{e_i}\) is an algebraic integer whose house is at most \(H^{E}\).

Source: Standard.

Proof ▶

Take \(b\) with every \(b\theta _i\) an algebraic integer, and \(H=1+\overline{\left|b\right|}+\sum _i\overline{\left|b\theta _i\right|}\); the house is submultiplicative.

Lemma 2.3 Scaled powers of an algebraic number
✓
#

Let \(K\) be a field of characteristic zero and let \(\alpha \in K\) be algebraic over \(\mathbb {Q}\). There are \(\ell ,A\in \mathbb {N}\) and an integer \(L\neq 0\) such that for every \(n\in \mathbb {N}\) there are integers \(r_0,\dots ,r_{\ell -1}\) with \(|r_l|\le A^{n}\) and

\[ L^{n}\alpha ^{n}=\sum _{l\lt \ell }r_l\, \alpha ^{l}. \]

Source: Standard.

Proof ▶

Take for \(L\) the leading coefficient of an integer polynomial vanishing at \(\alpha \). Then \(L\alpha \) is an algebraic integer, and reducing its powers modulo a monic integer polynomial keeps the coefficients geometrically bounded.

Lemma 2.4 Rational multiples of logarithms
✓
#

Let \(u\in \mathbb {C}\) with \(e^{u}\) algebraic, and let \(a\in \mathbb {Q}\). Then \(e^{au}\) is algebraic.

Source: Standard.

Proof ▶

With \(a=p/q\) and \(q\gt 0\), \((e^{au})^{q}=(e^{u})^{p}\) is algebraic, and so is a number whose \(q\)-th power is algebraic.

2.2 Siegel’s lemma

Lemma 2.5 Siegel’s lemma with an entrywise bound
✓
#

Let \(A\) be an \(m\times n\) matrix with integer entries, where \(n\ge 1\) and \(2m\le n\), and suppose that every entry satisfies \(|A_{ab}|\le B\) for some real \(B\ge 1\). Then there is \(t\in \mathbb {Z}^{n}\), \(t\neq 0\), with \(At=0\) and \(|t_b|\le nB\) for every \(b\).

Source: Siegel’s lemma [ Lang66 , Ch. I, § 2 ] , in the form Mathlib states it for integer matrices.

Proof ▶

Mathlib’s Siegel lemma bounds a solution by \((n\max (1,\| A\| ))^{m/(n-m)}\), and \(2m\le n\) makes the exponent at most one.

2.3 Exponential polynomials

Lemma 2.6 Distinct frequencies
✓
#

Let \(\omega _1,\dots ,\omega _l\in \mathbb {C}\) be pairwise distinct, let \(q_1,\dots ,q_l\in \mathbb {N}\), and let \(b_{j,i}\in \mathbb {C}\) (\(1\le j\le l\), \(0\le i\lt q_j\)) be not all zero. Then the function

\[ f(z)=\sum _{j=1}^{l}\sum _{i=0}^{q_j-1}b_{j,i}\, z^{i}e^{\omega _jz} \]

takes a non-zero value at some point of \(\mathbb {C}\).

Source: Classical; see [ Tij71 ] or [ Wal71 , § 4 ] .

Proof ▶

Induction on \(l\): multiply by \(e^{-\omega _lz}\) and differentiate \(q_l\) times. This removes the last block and multiplies the leading coefficient of every other block by \((\omega _j-\omega _l)^{q_l}\neq 0\).

Lemma 2.7 First non-vanishing derivative
✓
#

Let \(F(z)=\sum _{i\in I}c_ie^{\rho _iz}\) be a finite exponential sum with pairwise distinct \(\rho _i\in \mathbb {C}\) and coefficients \(c_i\in \mathbb {C}\) not all zero. Let \(m\ge 1\) and \(n\in \mathbb {N}\), and suppose that \(F^{(k)}(j)=0\) for all \(j\in \{ 1,\dots ,m\} \) and \(k\lt n\). Then there are \(r\ge n\) and \(l_0\in \{ 1,\dots ,m\} \) with \(F^{(r)}(l_0)\neq 0\) and \(F^{(k)}(j)=0\) for all \(j\in \{ 1,\dots ,m\} \) and \(k\lt r\).

Source: Standard.

Proof ▶

By Lemma 2.6 the function \(F\) is not identically zero, so it vanishes to finite order at each of \(1,\dots ,m\); take for \(r\) the least of these orders.

Lemma 2.8 Cauchy’s estimate with zeros
✓
#

Let \(F\) be an entire function, \(c\in \mathbb {C}\), \(\rho \ge 0\) and \(R\gt \rho +1\). Let \(S\) be a finite set of points \(z\) with \(|z-c|\le \rho \), and put \(\sigma =\sum _{z\in S}\operatorname {ord}_zF\). If \(|F(z)|\le M\) whenever \(|z-c|=R\), then for every \(s\in \mathbb {N}\)

\[ |F^{(s)}(c)|\le s!\, \frac{R}{R-1}\Bigl(\frac{\rho +1}{R-\rho }\Bigr)^{\sigma }M. \]

Source: [ Wal71 , § 4, (4.3) ] .

Proof ▶

Divide \(F\) by \(\prod _{\beta \in S}(z-\beta )^{\operatorname {ord}_\beta F}\). The quotient is bounded on \(|z-c|=R\), hence on \(|z-c|\le 1\) by the maximum principle; Cauchy’s estimate on the unit circle about \(c\) concludes.

Lemma 2.9 Cauchy’s estimate on a grid of zeros
✓
#

Let \(y_1,\dots ,y_l\in \mathbb {C}\) be linearly independent over \(\mathbb {Q}\), and let \(F(z)=\sum _{k\in s}c_kz^{e_k}e^{\omega _kz}\) be a finite sum with \(|\omega _k|\le W\) and \(e_k\le E\) for all \(k\). Let \(t_j\le A_j\) (\(1\le j\le l\)) and \(n\) be natural numbers, and suppose that \(F^{(i)}\bigl(\sum _jm'_jy_j\bigr)=0\) for all \(i\lt n\) and all \(m'\in \mathbb {N}^{l}\) with \(m'_j\lt t_j\) for every \(j\). Let \(u\ge 2\) and \(Z\ge \bigl(\sum _jA_j|y_j|+1\bigr)(u+1)\). Then for every \(m\in \mathbb {N}^{l}\) with \(m_j\lt A_j\) for every \(j\), every \(r\in \mathbb {N}\) and \(w=\sum _jm_jy_j\),

\[ |F^{(r)}(w)|\le 2\, r!\, (u-1)^{-n\prod _jt_j}\Bigl(\sum _{k\in s}|c_k|\Bigr)Z^{E}e^{WZ}. \]

Source: The analytic step of the extrapolation in [ Wal73 , Lemme 5 ] and in the six exponentials theorem [ Lang66 , Ch. II ] ; it rests on [ Wal71 , § 4, (4.3) ] .

Proof ▶

The \(\prod _jt_j\) grid points are distinct because the \(y_j\) are linearly independent over \(\mathbb {Q}\). Apply Lemma 2.8 on the circle of radius \((\rho +1)u\) about \(w\), where \(\rho =\sum _jA_j|y_j|\).

Lemma 2.10 Derivatives at a point
✓
#

Let \(f(z)=\sum _{j\in J}P_j(z)e^{w_jz}\), over a finite set \(J\), with \(|w_j|\le W\) for some \(W\ge 0\), and each \(P_j\in \mathbb {C}[z]\) zero or of degree less than \(q_j\); put \(N=\sum _jq_j\). Let \(c\in \mathbb {C}\) and \(D\ge 0\). If \(|f^{(s)}(c)|\le D\) for every \(s\lt N\), then

\[ |f^{(n)}(c)|\le D\, (W+1)^{n+N}\qquad \text{for every }n\in \mathbb {N}. \]

The \(w_j\) need not be distinct.

Source: No source is claimed; an elementary induction.

Proof ▶

Induction on \(N\). If \(q_{j_0}\ge 1\), then \(g=f'-w_{j_0}f\) has the same form with \(q_{j_0}\) lowered by one, and \(f^{(n+1)}(c)=g^{(n)}(c)+w_{j_0}f^{(n)}(c)\).

2.4 Integer polynomials

Lemma 2.11 Length of a sum
✓
#

For every finite family \((f_i)_{i\in s}\) in \(\mathbb {Z}[X][Y]\),

\[ L\Bigl(\sum _{i\in s}f_i\Bigr)\le \sum _{i\in s}L(f_i). \]

Source: Standard.

Proof ▶
Lemma 2.12 Length of a product
✓
#

For all \(P,Q\in \mathbb {Z}[X][Y]\), \(L(PQ)\le L(P)\, L(Q)\).

Source: Standard.

Proof ▶
Lemma 2.13 Reduction modulo a polynomial
✓
#

Let \(Q\in \mathbb {Z}[X][Y]\). There is \(C\in \mathbb {N}\) such that for all \(b,e,n\in \mathbb {N}\) and every \(P\in \mathbb {Z}[X][Y]\) with \(L(P)\le b\), every coefficient of \(X\)-degree at most \(e\), and \(Y\)-degree at most \(n\), the remainder \(P\bmod Q\) of the division by \(Q\) in \(Y\) satisfies

\[ L(P\bmod Q)\le C^{n}b, \]

and every coefficient of \(P\bmod Q\) has \(X\)-degree at most \(e+Cn\). (When \(Q\) is not monic in \(Y\), the remainder is \(P\), as in Mathlib’s modByMonic.)

Source: Standard.

Proof ▶
Lemma 2.14 Liouville’s inequality through the resultant
✓

Let \(F,G\in \mathbb {C}[X]\) with \(M(F)\ge 1\) and \(M(G)\ge 1\). Then for every \(\theta \in \mathbb {C}\)

\[ |\operatorname {Res}(F,G)|\le 2^{\deg F\deg G}M(F)^{\deg G}M(G)^{\deg F}\max \bigl(|F(\theta )|,|G(\theta )|\bigr). \]

Source: [ RW97 , Corollary 3.7 ] , in the complex form of their Lemma 3.4; they describe the inequality as well known.

Proof ▶
Lemma 2.15 The cofactor of an irreducible factor
✓

Let \(P\in \mathbb {Z}[X]\) have positive degree and let \(\theta \in \mathbb {C}\). Then \(P=Q^{e}R\) with \(Q,R\in \mathbb {Z}[X]\), \(Q\) irreducible of positive degree, \(e\ge 1\), and

\[ 1\le 2^{\deg Q\deg R}M(Q)^{\deg R}M(R)^{\deg Q}|R(\theta )|. \]

Source: The multiplicative step of Gel’fond’s lemma [ Gel52 ] , as in [ Wal71 , § 3, Lemme 2 ] .

Proof ▶

Take for \(Q\) an irreducible factor of positive degree at which \(|Q(\theta )|\) is least, and for \(e\) its multiplicity. Lemma 2.14 bounds every irreducible factor of \(R\) from below at \(\theta \), and the bound is multiplicative.

2.5 Transcendence degree

Lemma 2.16 A monic integral model
✓
#

Let \(\omega \in \mathbb {C}\) be transcendental and let \(\theta \in \mathbb {C}\) be algebraic over \(\mathbb {Q}(\omega )\). There are \(\omega _1\in \mathbb {C}\) and \(Q\in \mathbb {Z}[X][Y]\), monic in \(Y\) of degree \(d\ge 1\), with \(Q(\omega ,\omega _1)=0\) and minimal there: no non-zero \(A\in \mathbb {Z}[X][Y]\) with \(\deg _YA\lt d\) vanishes at \((\omega ,\omega _1)\). Moreover \(\omega _1=v(\omega )\, \theta \) for some \(v\in \mathbb {Z}[X]\) with \(v(\omega )\neq 0\).

Source: Standard; the normalisation at the start of [ Wal73 , § III ] .

Proof ▶

Clear the denominators of the minimal polynomial of \(\theta \) over \(\mathbb {Q}(\omega )\) with one \(v\in \mathbb {Z}[X]\), and rescale \(\theta \).

Lemma 2.17 One model for finitely many numbers
✓

Let \(\omega \in \mathbb {C}\) be transcendental and let \((z_i)_{i\in I}\) be a finite family of complex numbers algebraic over \(\mathbb {Q}(\omega )\). There are \(\omega _1\in \mathbb {C}\) and \(Q\in \mathbb {Z}[X][Y]\), monic in \(Y\) of degree \(d\ge 1\), with \(Q(\omega ,\omega _1)=0\) and minimal there in the sense of Lemma 2.16, and \(D,E_i\in \mathbb {Z}[X][Y]\) with \(D(\omega ,\omega _1)\neq 0\) and

\[ z_i\, D(\omega ,\omega _1)=E_i(\omega ,\omega _1)\qquad \text{for every }i. \]

Source: Standard; the presentation used in [ Wal73 , § III ] .

Proof ▶

By the primitive element theorem the \(z_i\) lie in \(\mathbb {Q}(\omega )(\theta )\) for one \(\theta \). Apply Lemma 2.16 to \(\theta \) and clear denominators.

Lemma 2.18 Transcendence degree at most one
✓

Let \(K\subseteq L\) be fields, \(x\in L\), and \(S\subseteq L\) a set each of whose elements is algebraic over the ring \(K[x]\). Then the \(K\)-algebra \(K[S]\) generated by \(S\) has transcendence degree at most one over \(K\).

Source: Standard.

Proof ▶
Lemma 2.19 A dependent pair
✓

Let \(R\) be a commutative ring, \(A\) a commutative \(R\)-algebra and \(t\in A\) transcendental over \(R\). If \(t\) and \(z\in A\) are not algebraically independent over \(R\), then \(z\) is algebraic over the subalgebra \(R[t]\).

Source: Standard.

Proof ▶