---
title: Schmidt’s Random Counting Theorem
url: https://www.emergentmind.com/topics/schmidt-s-random-counting-theorem
type: topic
---

# Schmidt’s Random Counting Theorem

Schmidt’s random counting theorem is a lattice-point counting result in the geometry of numbers and homogeneous dynamics. In the formulation used as a template in later work, one studies the primitive Siegel transform
\[
\widehat f(gSL_2(\mathbb Z))=\sum_{v\in P(\mathbb Z^2)} f(gv), \qquad P(\mathbb Z^2)=\{v\in\mathbb Z^2:\gcd(v)=1\},
\]
and uses first and second moment formulas to show that, for almost every unimodular lattice \(g\mathbb Z^2\), the number of primitive lattice points in a family of sets \(A_T\) with \(\operatorname{vol}(A_T)\to\infty\) satisfies an asymptotic law [2507.05905]. In this sense, the theorem is “random” because the lattice is sampled from the moduli space with respect to Haar measure, while the counted quantity is a lattice-point count in expanding regions. A recent two-dimensional analog with congruence conditions makes the structure explicit and isolates the moment identities, orbit decompositions, and almost-sure error term that drive Schmidt’s argument [2507.05905].

## 1. Classical setting and probabilistic interpretation

The classical theorem is organized around primitive lattice points rather than all lattice points. The basic counting functional is the primitive Siegel transform
\[
\widehat f(gSL_2(\mathbb Z))=\sum_{v\in P(\mathbb Z^2)} f(gv),
\]
where \(f\) is a test function on \(\mathbb R^2\) and \(gSL_2(\mathbb Z)\) parametrizes a unimodular lattice in \(SL_2(\mathbb R)/SL_2(\mathbb Z)\) [2507.05905]. The theorem then concerns the number of primitive lattice points in expanding sets \(A_T\), with the lattice varying over the homogeneous space.

The probabilistic content is explicit in the modern description: the random variable is the lattice point count in expanding sets, viewed across the moduli space of lattices [2507.05905]. The asymptotic law is an almost-everywhere statement. Thus the theorem does not assert a deterministic asymptotic for every lattice; it asserts that the normalized count behaves as predicted for almost every unimodular lattice.

A key feature is that the count is compared to volume. In the congruence-restricted analog, this comparison becomes completely explicit: the count is asymptotic to volume multiplied by an arithmetic density, and the proof is driven by a first-moment identity, a second-moment identity, and a Borel–Cantelli argument [2507.05905]. This suggests that the classical theorem should be understood as a prototype of almost-sure lattice counting derived from variance control on a homogeneous space.

## 2. Congruence-restricted formulation in dimension \(2\)

A precise analog fixes a modulus \(N\in\mathbb N\) and a primitive residue class \(\sigma=(v_0,N)\) with \(v_0\in\mathbb Z^2\) and \(\gcd(v_0,N)=1\). One then defines
\[
P^{(\sigma)}=\{v\in P(\mathbb Z^2): v\equiv v_0 \pmod N\},
\]
which is invariant under the principal congruence subgroup
\[
\Gamma(N)=\{g\in SL_2(\mathbb Z): g\equiv I_2\pmod N\}.
\]
The corresponding Siegel transform is
\[
\widehat f^{(\sigma)}(g\Gamma(N)) = \sum_{\substack{v\in P(\mathbb Z^2)\\ v\equiv v_0\;(\mathrm{mod}\,N)}} f(gv), \qquad g\Gamma(N)\in SL_2(\mathbb R)/\Gamma(N)
\]
[2507.05905].

This is the basic counting function for primitive lattice points in one fixed congruence class. The ambient space is no longer \(SL_2(\mathbb R)/SL_2(\mathbb Z)\) but \(SL_2(\mathbb R)/\Gamma(N)\), and the density is modified accordingly. Compared to the classical theorem, the new features are that primitive vectors are still counted, but now only those lying in a fixed residue class \(v\equiv v_0\pmod N\), and the second moment must reflect the arithmetic restriction imposed by the congruence condition [2507.05905].

The density factor is expressed through
\[
\zeta_N(d)=\prod_{p\nmid N}\left(1-\frac1{p^d}\right)^{-1}.
\]
For \(d=2\), the first-moment density becomes \(\frac{1}{\zeta_N(2)N^2}\), described as the primitive density factor \(\zeta_N(2)^{-1}\) times the congruence frequency \(N^{-2}\) up to the Euler factor correction from primes dividing \(N\) [2507.05905].

## 3. First and second moment formulas

The first moment formula is the direct analog of Siegel’s mean value theorem. For bounded compactly supported \(f:\mathbb R^2\to\mathbb R\),
\[
\int_{SL_2(\mathbb R)/\Gamma(N)} \widehat f^{(\sigma)}(g\Gamma(N))\,d\nu_N(g) = \frac{1}{\zeta_N(2)N^2}\int_{\mathbb R^2} f(v)\,dv
\]
[2507.05905]. Equivalently, the average number of primitive points in the class \(v\equiv v_0\pmod N\) is the volume times this density.

The second moment introduces a two-variable transform. For bounded compactly supported \(F:\mathbb R^2\times\mathbb R^2\to\mathbb R\),
\[
\widehat F^{(\sigma)}(g\Gamma(N)) = \sum_{v_1,v_2\in P^{(\sigma)}} F(gv_1,gv_2).
\]
The corresponding second moment is
\[
\begin{aligned}
\int_{SL_2(\mathbb R)/\Gamma(N)} \widehat F^{(\sigma)}(g\Gamma(N))\,d\nu_N(g)
&= \sum_{n\in N\mathbb Z\setminus\{0\}} \frac{\varphi(n)}{\zeta_N(2)N^3\varphi(N)} \int_{SL_2(\mathbb R)} F(gJ_n)\,d\eta(g) \\
&\quad+\frac{1}{\zeta_N(2)N^2} \int_{\mathbb R^2}\bigl(F(v,v)+F(v,-v)\bigr)\,dv,
\end{aligned}
\]
where
\[
J_n=\begin{pmatrix}1&0\\0&n\end{pmatrix}
\]
[2507.05905].

This formula separates the diagonal or degenerate contributions \((v,v)\) and \((v,-v)\) from the non-degenerate pairs with determinant \(n\neq 0\). The arithmetic restriction is sharp: for each fixed congruence class \(\sigma\), the set
\[
D_n^{(\sigma)}=\{(v_1,v_2)\in P^{(\sigma)}\times P^{(\sigma)}:\det(v_1,v_2)=n\}
\]
is empty unless \(n\in N\mathbb Z\setminus\{0\}\), and when nonempty it decomposes into exactly
\[
\frac{N\,\varphi(n)}{\varphi(N)}
\]
\(\Gamma(N)\)-orbits [2507.05905]. This orbit count is the structural input behind the coefficient in the second moment formula.

## 4. Cone normalization and the almost-sure counting law

To adapt Schmidt’s random counting theorem, one works over the cone
\[
\mathfrak C = \{\, t^{1/2}g : t\in(0,1],\, g\in \mathfrak F_N \,\},
\]
where \(\mathfrak F_N\) is a fundamental domain for \(SL_2(\mathbb R)/\Gamma(N)\). The normalized counts are
\[
t\,\widehat f^{(\sigma)}(t^{1/2}g) = \sum_{\substack{v\in P(\mathbb Z^2)\\ v\equiv v_0\;(\mathrm{mod}\,N)}} f(t^{1/2}gv)
\]
[2507.05905].

Over this cone, the first moment remains
\[
\int_{\mathfrak C} t\,\widehat f^{(\sigma)}(t^{1/2}g)\,d\nu_N\,dt = \frac{1}{\zeta_N(2)N^2}\int_{\mathbb R^2} f(v)\,dv.
\]
The second moment becomes
\[
\begin{aligned}
\int_{\mathfrak C} \bigl(t\,\widehat F^{(\sigma)}(t^{1/2}g)\bigr)\,d\nu_N\,dt
&= \frac{1}{\zeta_N(2)N^3} \int_{\mathbb R^2\times\mathbb R^2} \Phi_N(\det(v_1,v_2))\,F(v_1,v_2)\,dv_1\,dv_2 \\
&\quad+\frac{1}{2\zeta_N(2)N^2} \int_{\mathbb R^2}\bigl(F(v,v)+F(v,-v)\bigr)\,dv,
\end{aligned}
\]
where
\[
\Phi_N(x)=|x|\sum_{\substack{n\in N\mathbb Z\\ n\ge |x|}} \frac{N\varphi(n)}{\varphi(N)n^3}.
\]
Moreover,
\[
\Phi_N(x)=\frac{1}{\zeta_N(2)N}+O_N(|x|^{-1}\log|x|)\qquad (|x|\to\infty)
\]
[2507.05905]. This asymptotic shows that the second moment has the same leading density as the first moment squared, with an explicit error kernel.

The main almost-sure theorem in the congruence-restricted setting is stated for an increasing family of Borel sets \(\{A_T\}\subset\mathbb R^2\) with
\[
V_T=\operatorname{vol}(A_T)\to\infty,
\]
and a non-decreasing function \(\psi:(0,\infty)\to(0,\infty)\) satisfying
\[
\int^\infty \frac{dx}{\psi(x)}<\infty.
\]
Then for almost every unimodular lattice \(g\mathbb Z^2\),
\[
\#\bigl(gP(\mathbb Z^2)\cap A_T \cap \{v\equiv v_0\ (\mathrm{mod}\ N)\}\bigr)
=
\frac{V_T}{\zeta_N(2)N^2}
+O\!\left(V_T^{1/2}(\log V_T)^2\psi(V_T)^{1/2}\right)
\]
[2507.05905]. This is described as the precise congruence-restricted version of Schmidt’s theorem.

## 5. Proof strategy and Diophantine consequences

The proof follows Schmidt’s strategy closely. The ingredients identified in the modern analog are moment estimates on the homogeneous space \(SL_2(\mathbb R)/\Gamma(N)\), a Borel–Cantelli argument on a cone over a fundamental domain, and the ergodicity and invariance of Haar measure [2507.05905]. At a high level, the method is to establish first and second moment formulas for the congruence-restricted Siegel transform, refine them over the cone \(\mathfrak C\), use the asymptotic control of \(\Phi_N(x)\) to estimate the variance of the centered counting function, apply a Borel–Cantelli argument over dyadic scales and a decomposition into interval chains, and then conclude almost sure asymptotics for counts in \(A_T\) [2507.05905].

The technical heart is the second moment computation, especially the classification of \(\Gamma(N)\)-orbits in pairs of primitive vectors with fixed determinant and fixed congruence class [2507.05905]. This orbit classification is not an auxiliary arithmetic decoration; it is the mechanism that converts pair counting into a computable second-moment coefficient.

The same moment method yields a metric Diophantine approximation theorem with congruence restrictions. The stated corollary is a congruence-restricted quantitative Khintchine theorem:
\[
\#\left\{ \binom{p}{q}\in P(\mathbb Z^2): |qx-p|<\psi(|q|),\ |q|\le T,\ (p,q)^T\equiv (p_0,q_0)^T\!\!\!\pmod N \right\}
\sim \frac{1}{\zeta_N(2)N^2}\sum_{1\le t\le T}\psi(t)
\]
for almost all \(x\) [2507.05905]. This places Schmidt-type random counting within the broader metric theory of rational approximation.

## 6. Terminological scope and nearby Schmidt results

The name “Schmidt” appears in several distinct theories, and these should not be conflated with the random counting theorem.

| Result | Content | arXiv id |
|---|---|---|
| Asmus Schmidt’s conjecture | Integrality of coefficients \(c_k^{(r)}\) defined by a binomial basis expansion | [1203.3731] |
| Schmidt-type theorem for partitions | Periodic counted and uncounted part-positions translated into colored partitions | [2203.05202] |
| Schmidt’s conjecture via Schmidt’s game | \(\mathrm{Bad}(s,t)\) is \(\bigl(24\sqrt2\bigr)^{-1}\)-winning for Schmidt’s game | [1204.3610] |
| Schmidt’s subspace theorem, effective function-field form | Hypersurfaces in \(N\)-subgeneral position on smooth projective varieties | [1509.07397] |

The integrality theorem of Asmus Schmidt concerns the sequence \(\{c_k^{(r)}\}_{k\ge 0}\) defined implicitly by
\[
\sum_{k=0}^n \binom{n}{k}^r \binom{n+k}{k}^r
=
\sum_{k=0}^n \binom{n}{k}\binom{n+k}{k}\, c_k^{(r)},
\qquad n=0,1,2,\dots,
\]
and proves that all \(c_k^{(r)}\) are integers for every \(r\ge 1\) and every \(k\ge 0\) [1203.3731]. That problem is algebraic-recursive and unrelated to random lattice counting.

The partition-theoretic Schmidt-type theorem generalizes the “odd-place counting” phenomenon to periodic subsets of counted and uncounted positions modulo \(m\), using a colored refinement of Stockhofe’s bijection [2203.05202]. The Schmidt-game result concerns weighted badly approximable vectors and proves that \(\mathrm{Bad}(s,t)\) is \(\bigl(24\sqrt2\bigr)^{-1}\)-winning [1204.3610]. The effective subspace theorem is a higher-degree function-field generalization in arithmetic geometry, involving local Weil functions, heights, and \(N\)-subgeneral position [1509.07397]. These results share nomenclature and, in different ways, a counting or largeness perspective, but they belong to distinct mathematical frameworks.

In the random-counting sense, Schmidt’s theorem is best understood as an almost-sure asymptotic for primitive lattice points in expanding planar sets, established through first and second moments of Siegel transforms and sharpened in modern work by congruence-restricted formulas that make the density, orbit structure, and error term explicit [2507.05905].

Source: https://www.emergentmind.com/topics/schmidt-s-random-counting-theorem