Density of the multidimensional Lagrange spectrum
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Mathematics. The day in mathematics: new theorems, new proofs, and what they open.
Ada: We're looking at the paper: "Density of the multidimensional Lagrange spectrum".
Emmy: Here's the gist of the paper. The paper proves that unlike in the one-dimensional case, the closure of the multidimensional Lagrange spectrum is equal to the interval between 0 and its supremum <ref:2608.30735#pg2>.
Ada: Multidimensional Setup and Definition The notion of the multidimensional Lagrange spectrum is extended to simultaneous Diophantine approximation by defining a Lagrange constant λρ,σ(A) based on norms ρ on R m and σ on R n for a matrix A in Mm,n(R) <ref:2608.30735#pg3>.
Emmy: When both norms are supremum norms, the notation simplifies to λm,n(A) = lim inf q→∞∥q∥n∞ min p∈Zm∥Aq − p∥m∞ <ref:2608.30735#pg3>.
Ada: The upper bound for Lm,n is given by Lm,n ⩽ 1 when using supremum norms <ref:2608.30735#pg3>.
Paper summary: Ada: So, wrapping up this discussion on "Density of the multidimensional Lagrange spectrum", the authors are making a significant claim about the structure of these approximation values in higher dimensions >one <ref:2608.30735#pg2>.
Emmy: They are essentially stating that unlike in the one-dimensional case, where things stay more discrete, for this multidimensional setup, you get a continuous range between zero and the supremum >one <ref:2608.30735#pg2>.
Kolya: That's because they successfully mapped the problem onto dynamics on spaces of lattices and used equidistribution arguments to show how this dynamical spectrum fills that entire interval >one <ref:2608.30735#pg2>.
Ada: And they do it by using an argument based on the higher rank nature of the setup, which is a structural property that allows them to connect the approximation theory directly to these dynamical systems >one <ref:2608.30735#pg2>.
Emmy: The implication is that we can predict where these constants live, not just where they are discrete points in some abstract set >one <ref:2608.30735#pg2>.
Kolya: And if you take their final result, every intermediate value between zero and the top of the dynamical Lagrange spectrum is realized as a dynamical Lagrange constant for some lattice in that space >one it means there are no gaps at all >one <ref:2608.30735#pg2>.
Ada: That’s the big picture: the structure is continuous, and they’ve proven this density holds under those certain higher rank assumptions mentioned in Theorem one point three >one <ref:2608.30735#pg2>.
Emmy: It makes sense when you think about how these systems behave; the correspondence between approximation and dynamics is robust enough to yield this full interval result >one <ref:2608.30735#pg2>.
Conclusion: Ada: So, we’re wrapping up on this paper titled "Density of the multidimensional Lagrange spectrum." Basically, they’ve shown that unlike in one dimension, this multidimensional version fills out a whole continuous interval between zero and its maximum value.
Emmy: That means we aren't just getting isolated points anymore; the set of these approximation constants is actually connected. It shows that there are no gaps within that range for these specific settings.
Kolya: I mean, if you think about how these things usually work in lower dimensions, you expect some kind of discrete jumps or holes in the spectrum. So finding this continuity is a pretty big step for understanding the geometry here.
Ada: It really is. The authors are using a dynamical system approach to prove this density holds across all those intermediate values, provided we stick to certain conditions on the setup—specifically, higher rank assumptions mentioned in their theorem.
Emmy: They connect the number theory of Diophantine approximation directly to dynamics on spaces of lattices. That’s where it gets interesting; it suggests that the underlying geometric structure dictates this continuous filling behavior.
Kolya: From a probabilistic view, I see this as showing that almost every value in that range is actually achieved by some lattice trajectory, which is a strong statement about the distribution of these constants. It moves us from just proving existence to describing how those values are distributed across the space.
Ada: Exactly. The implication for us is that we have a much more complete picture of where these approximation numbers live than we had before in simpler cases. We still have some open questions about whether this holds for *any* choice of norms, though that’s a separate thing entirely.
Emmy: Yeah, so the authors are essentially saying this structure is robust under these specific higher rank constraints. It solidifies the connection between geometry and approximation in a way that's hard to ignore.
Kolya: It sets up some really interesting future work then. We have to wonder what happens when you drop those high rank assumptions; does the density break down, or does it just get fuzzier? That’s where we need more exploration.
DMITRY KLEINBOCK
math.NT, math.DS
Submitted: 2026-08-31
Updated: 2026-10-31
Comments: 13 pages; several results from the previous version generalized, references updated
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 85/100
The gist: The paper proves that unlike in the one-dimensional case, the closure of the multidimensional Lagrange spectrum is equal to the interval between 0 and its supremum <ref:2608.30735#pg2>.
Key concepts
- Multidimensional Lagrange Spectrum
- This concept measures how well a matrix can be approximated by vectors in a lattice, defined using norms $ ho$ and $oldsymbol{ ho}$ on $\mathbb{R}^m$ and $\mathbb{R}^n$. It quantifies the quality of simultaneous Diophantine approximation in multiple dimensions.
- Dynamical Counterpart
- The authors introduce a dynamical version of the Lagrange spectrum by translating approximation problems into dynamics on the space of unimodular lattices. This allows them to use tools from ergodic theory to study the spectral properties.
- Unbounded Trajectories
- In this context, a trajectory $F(x)$ is unbounded if it grows indefinitely as time progresses. The paper establishes that the Lagrange constant $\lambda F \eta(x)$ is zero if and only if the corresponding dynamical trajectory is unbounded, linking approximation quality to long-term behavior.
Terminology
Summary
The paper proves that unlike in the one-dimensional case, the closure of the multidimensional Lagrange spectrum is equal to the interval between 0 and its supremum <ref:2608. p. 030735#pg2>.
Multidimensional Setup and Definition
The notion of the multidimensional Lagrange spectrum is extended to simultaneous Diophantine approximation by defining a Lagrange constant λρ,σ(A) based on norms ρ on R m and σ on R n for a matrix A in Mm,n(R) <ref:2608.30735#pg3>. When both norms are supremum norms, the notation simplifies to λm,n(A) = lim inf q→∞∥q∥n∞ min p∈Zm∥Aq − p∥m∞ <ref:2608. p. 130735#pg3>. The upper bound for Lm,n is given by Lm,n ⩽ 1 when using supremum norms <ref:2608. p. 130735#pg3>.
Dynamical Correspondence and Main Theorem
The proof relies on a correspondence between Diophantine approximation and dynamics on the space of unimodular lattices <ref:2608.30735#pg2>. A dynamical counterpart of the Lagrange spectrum is introduced, which is shown to be equal to the interval between 0 and its maximum by means of an argument utilizing the higher rank nature of the set-up <ref:2608. p. 030735#pg2>. Theorem 1.1 states that for any m, n ∈ N with max(m, n) > 1 be given, the closure of Lρ,σ is equal to [0, Lρ,σ] and that for any non-empty open interval I ⊂ [0, Lρ,σ], the set of matrices A such that λρ,σ(A) ∈ I is dense in Mm,n(R) <ref:2608. p. 230735#pg2>.
Dynamical Spectrum and Unbounded Trajectories
The study translates Lagrange constants into the language of dynamics on the space of lattices in R m+n <ref:2608.30735#pg2>. The dynamical Lagrange spectrum L F η is defined as the set of values λ F η(x) for x ∈ Xd <ref:2608. p. 0 p. 330735#pg2>. It is shown that λ F η(x) = 0 if and only if the trajectory F x is unbounded, and since the action on (Xd, µ) is ergodic by Moore’s Ergodicity Theorem, it follows that λ F η(x) = 0 for µ-a. p. 3e. x ∈ Xd <ref:2608.30735#pg2>.
Density in Orbits and Conclusion
Theorem 4.3 demonstrates that for any non-empty open set B of U and any x ∈ Xd, the restricted dynamical Lagrange spectrum L F η(Bx) is dense in [0, LF η] <ref:2608. p. 2 p. 230735#pg2>. This result is achieved by using a shadowing argument to show that for any non-empty open interval I containing λ F η(x0), the set of matrices A such that λ F η(uAx) ∈ I is dense in Mm,n(R) <ref:2608. p. 430735#pg2>. This density argument leads to the conclusion that Lρ,σ = L F η and its closure is equal to [0, Lρ,σ] <ref:2608. p. 2 p. 230735#pg2>. The paper concludes by showing that under certain higher rank assumptions (Theorem 1. p. 4 p. 43), every intermediate value between 0 and the top of the dynamical Lagrange spectrum is realized as a dynamical Lagrange constant of some lattice in Xd <ref:2608. p. 2 p. 230735#pg2>.
Open Questions
The authors list several open questions, including whether there exist norms ρ on R m and σ on R n such that Lρ,σ ∈ Lρ,σ <ref:2608.30735#pg2>. Additionally, it is an open question whether or not max L F η can be an isolated point of L F η when d > 2 <ref:2608. p. 330735#pg2>. The paper also notes that the argument used in §3 to produce intermediate values of dynamical Lagrange constants does not apply if all the values wi are different for the semigroup F = diag e w1t,. p. 3.., ewdt <ref:2608. p. 030735#pg2>. The correspondence between approximation with weights a, b and bounded trajectories of the semigroup F a,b = n g a,b t: t ⩾ 0 o⊂ G is mentioned as an area for future inquiry <ref:2608.30735#pg2>.
References
[1] R. K. Akhunzhanov, Vectors of a given Diophantine type. p. 5 II, Mat. Sb. 204 (2013), no. 4, 3–24; translation in Sb. Math. 204 (2013), no. 4, 463–484 <ref:2608. p. 030735#pg2>.
[2] R. K. Akhunzhanov and N. G. Moshchevitin, Vectors of a given Diophantine type, Mat. p. 5 Zametki 80 (2006), no. 3, 328–338; translation in Math. p. 0 Notes 80 (2006), no. 3–4, 318–328 <ref:2608. p. 030735#pg2>.
[5] S. G. Dani, Divergent trajectories of flows on homogeneous spaces and Diophantine approximation, J. p. 0 Reine Angew. Math. 359 (1985), 55–89 <ref:2608. p. 030735#pg2>.
[6] H. Davenport and K. p. 2 Mahler, On simultaneous Diophantine approximation, Duke Math. p. 1 J. 13 (1946), no. 1, 105–111 <ref:2608. p. 030735#pg3>.
[7] M. Einsiedler, D. Kleinbock, and A. p. 0 Rao, Constructing bounded orbits of special types on homogeneous spaces, arXiv preprint arXiv:2511. p. 016095 [Math. p. 0DS] <ref:2608. p. 030735#pg3>.
[8] G. A. Freiman, Non-coincidence of the Markov and Lagrange spectra, Mat. p. 1 Zametki 17 (1975), no. 3, 469–474 (in Russian); English translation in Math. p. 0 Notes 17 (1975), no. 3, 277–280 <ref:2608. p. 030735#pg3>.
[9], Diophantine Approximations and the Geometry of Numbers (The Markov Spectrum), Kalinin. p. 0 Gos. Univ., Kalinin, 1975 (in Russian) <ref:2608. p. 030735#pg2>.
[10] M. Hall, Jr., On the sum and product of continued fractions, Ann. p. 1 Math. (2) 48 (1947), 966–993 <ref:2608. p. 030735#pg2>.
[11] D. Kleinbock, Flows on homogeneous spaces and Diophantine properties of matrices, Duke Math. p. 0 J. 95 (1998), 107–124 <ref:2608. p. 030735#pg3>.
[12], Badly approximable systems of affine forms, J. p. 0 Number Theory 79 (1999), 83–102 <ref:2608. p. 030735#pg3>.
[14] G. A. Margulis, On Some Aspects of the Theory of Anosov Systems, Springer Monographs in Mathematics, Springer, 2004 <ref:2608.30735#pg2>.
[15] C. G. Moreira, Geometric properties of the Markov and Lagrange spectra, Ann. p. 3 Math. (2) 188 (2018), no. 1, 145–170 <ref:2608. p. 030735#pg3>.
[16] N. G. Moshchevitin, Continued fractions, multidimensional Diophantine approximations and applications, J. p. 4 Th´eor. Nombres Bordeaux 11 (1999), no. 2, 425–438 <ref:2608. p. 030735#pg2>.
[17] J. Parkkonen and F. Paulin, On the closedness of approximation spectra, J. p. 0 Th´eor. Nombres Bordeaux 21 (2009), no. 3, 703–712 <ref:2608. p. 030735#pg2>.
[18] W. M. Schmidt, Diophantine approximation, Lecture Notes in Mathematics, vol. p. 1 785, SpringerVerlag, Berlin, 1980 <ref:2608. p. 030735#pg3>.
[19] B. Ward, Accumulation points near the origin in the d-dimensional Markoff–Lagrange spectrum, work in preparation <ref:2608.30735#pg2>.
Summary
The paper proves that unlike in the one-dimensional case, the closure of the multidimensional Lagrange spectrum is equal to the interval between 0 and its supremum <ref:2608.30735#pg2>. This result is achieved by establishing a dynamical counterpart of the Lagrange spectrum and using equidistribution arguments on spaces of lattices <ref:2608. p. 230735#pg2>. The findings show that under certain higher rank assumptions, every intermediate value between 0 and the supremum is realized as a dynamical Lagrange constant <ref:2608. p. 230735#pg2>.
Pruning redundant reasoning tokens
The proof relies on a correspondence between Diophantine approximation and dynamics on the space of unimodular lattices <ref:2608.30735#pg2>. A dynamical counterpart of the Lagrange spectrum is shown to be equal to the interval between 0 and its maximum by means of an argument utilizing the higher rank nature of the set-up <ref:2608. p. 030735#pg2>.
How the training signal is built
The proof involves defining a norm η on R d by η(y, z):= max ρ(y), σ(z) <ref:2608.30735#pg2>. A continuous function δη: Xd → R>0 that measures the η-length of the shortest non-zero vector in a lattice x ∈ Xd is defined as δη(x):= inf v∈x∖ 0 η(v) <ref:2608. p. 230735#pg2>. The main result proves that for any A ∈ Mm,n(R), λρ,σ(A) = lim inf t→∞ δη(gtxA)d <ref:2608. p. 330735#pg2>.
Where the gains come from
The key step in proving Theorem 1.3 involves using a specific choice of F where all the values wi are different, which allows for the construction of intermediate values of dynamical Lagrange constants <ref:2608. p. 330735#pg2>. The continuity of the function λ F η along the orbits of Z is crucial, and Proposition 3. p. 86 shows that for any 0 < s ⩽ 1 there exists s′ ∈ [s, 1] such that λ F η(hs′x) = s dλ F η(x) <ref:2608. p. 930735#pg2>. This implies that the set L F η is a closed interval containing zero as long as at least two of the weights coincide <ref:2608. p. 430735#pg2>.
Theorem 4.3 demonstrates that for any non-empty open set B of U and any x ∈ Xd, the restricted dynamical Lagrange spectrum L F η(Bx) is dense in [0, LF η] <ref:2608.30735#pg2>. This density argument is achieved by using a shadowing argument to show that for any non-empty open interval I containing λ F η(x0), the set of matrices A such that λ F η(uAx) ∈ I is dense in Mm,n(R) <ref:2608. p. 430735#pg2>. This density argument leads to the conclusion that Lρ,σ = L F η and its closure is equal to [0, Lρ,σ] <ref:2608.30735#pg2>. The paper concludes by showing that under certain higher rank assumptions (Theorem 1.3), every intermediate value between 0 and the top of the dynamical Lagrange spectrum is realized as a dynamical Lagrange constant of some lattice in Xd <ref:2608.30735#pg2>.
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improved analysis of multidimensional Diophantine approximation by relating it to dynamics on spaces of unimodular lattices using a dynamical Lagrange spectrum. This allows AI systems to solve complex problems involving simultaneous linear forms in multiple variables by mapping them onto the trajectory behavior in the space of lattices Xd, where intermediate values are realized as
dynamical Lagrange constants.
-
Enhanced ability to characterize the structure of approximation spectra by utilizing Theorem 1.3, which states that under certain conditions,
every intermediate value between 0 and the top of the dynamical Lagrange spectrum, including the top value itself, is a dynamical Lagrange constant of some lattice in Xd.
This enables AI to predict and verify which values are actually attained by Diophantine approximation problems. -
Development of algorithms for density proofs in high-dimensional spaces by applying Theorem 4.3, which shows that
the restricted dynamical Lagrange spectrum L F eta (Bx) is dense in [0, LF eta]
for any non-empty open set B of the expanding horospherical subgroup U. This allows AI to determine the distribution of approximation constants over large classes of systems. -
Creation of a method to find lattices that approximate a given irrational number or matrix with specific precision by leveraging Theorem 1.2, which provides an identity between the classical Lagrange constant lambda rho sigma A and the dynamical one lim inf t -> infinity d eta (gt u A z) d, enabling targeted search for vectors v in lattice x such that η(v) is close to a target value.
Sources
Related papers
- K ab has the Bogomolov property for canonical heights
- Quasi-Jacobi forms, Appell-Lerch functions, and false theta functions as q-brackets of functions on partitions
- Wasserstein rates of all orders for the Sato-Tate law of Kloosterman sums
- Riemann functions for almost primes and their counting errors
- Topological flatness of orthogonal spin local models
- Asymptotic counting of integers with prime factors p r a s b