betaFeedback

WorksMathematics

Definably isometric cuboids over Hensel minimal fields

Abstract

We show that, over a valued field KK with 11-h-minimal theory Th⁡(K)\operatorname{Th}(K), two Cartesian products of balls must be equal up to translation and coordinate permutation in order for there to exist a definable surjective isometry between them. For this purpose, we introduce a notion of relative dimension for cuboids in KnK^n that is computed combinatorially and relates to the notion of dimension for definable sets in Hensel minimal theories when KK has infinite residue field.

Introduction

Let KK be a valued field with infinite residue field kk. Then any bijection from kk to k2k^2 lifts to an isometric bijection from O×M\mathcal{O} \times\mathcal{M} to O×O\mathcal{O} \times\mathcal{O}, where O\mathcal{O} and M\mathcal{M} denote the valuation ring and valuation ideal of KK, respectively. Precisely when such a map is first-order definable is not clear. In general, it is not clear when there exists a definable surjective isometry between two Cartesian products of balls in KK. We show that, when Th⁡(K)\operatorname{Th}(K) is 1-h-minimal, two Cartesian products of balls in KK must be equal up to translation and coordinate permutation in order to be definably isometric. This question was asked by Immanuel Halupczok and our presented answer will be used in an upcoming article by Cluckers, Halupczok and Vermeulen.

Main results

Let KK be a non-trivially valued field, viewed as a structure in a language that contains the language of valued fields Lval={+,⋅,O}\mathcal{L}_{\mathrm{val}} = \{+,\cdot,\mathcal{O}\}. Let ∣⋅∣|\cdot| denote the valuation on KK, multiplicatively written, that is extended to KnK^n by the sup norm. We call a subset of KnK^n definable if it is definable with any parameters from KK. Our main result is the following.

Theorem 1.1 (Theorem 3.13). Assume that Th⁡(K)\operatorname{Th}(K) is 1-h-minimal and let P,P′⊆KnP, P' \subseteq K^n be Cartesian products of balls in KK. If PP and P′P' are definably isometric, then PP and P′P' are equal up to translation and coordinate permutation.

In particular, this holds true regardless of the residue characteristic of KK. Our notion of 1-h-minimality therefore refers to the general case described in [2 Theorem 2.2.8]. Cartesian products of balls in KK are the main object of study in this article and will be called cuboids throughout. A precise definition of cuboids and isometries is given in § 2. For the purpose of proving Theorem 1.1, we introduce a notion of relative dimension dim⁡(P/Q)\dim(P/Q) for cuboids P,Q⊆KnP,Q \subseteq K^n. This notion of dimension is defined purely combinatorially is consistent with the notion of dimension for definable subsets of KnK^n in the sense of [1 Definition 5.3.1]. More precisely, for a cuboid P⊆KnP \subseteq K^n around 00 and a ball B⊆KnB \subseteq K^n around 00, we have the following characterization.

Proposition 1.2 (Proposition 3.15). Assume that Th⁡(K)\operatorname{Th}(K) is 1-h-minimal and that KK has infinite residue field. Then dim⁡(P/B)\dim(P/B) is the minimal dimension of a definable set X⊆KnX \subseteq K^n such that vox⁡B(X)\operatorname{vox}_B(X) covers vox⁡B(P)\operatorname{vox}_B(P).

Here, vox⁡B\operatorname{vox}_{B} denotes the natural projection from KnK^n to Kn/BK^n/B. This map is essential for our proof of Theorem 1.1 and will be used all throughout § 3. The proof of Theorem 1.1 is divided into two parts. When KK has finite residue field, the claim of Theorem 1.1 applies to all isometric cuboids, regardless of definability. The techniques used in this part are elementary. When KK has infinite residue field, the key ingredient to the proof is the following result.

Proposition 1.3 (Proposition 3.9). Assume that Th⁡(K)\operatorname{Th}(K) is 11-hh-minimal. For every definable set X⊆KnX \subseteq K^n there exists some N∈NN \in\mathbb{N} such that for every ball B⊆KB \subseteq K around 00 and every finite subset A⊆KA \subseteq K, one has

#(vox⁡Bn(X)∩vox⁡Bn(An))≤N⋅(#vox⁡B(A))dim⁡(X).\#\left(\operatorname{vox}_{B^n}(X) \cap\operatorname{vox}_{B^n}(A^n)\right) \leq N \cdot\left(\#\operatorname{vox}_{B}(A)\right)^{\dim(X)}.

Here, vox⁡Bn\operatorname{vox}_{B^n} and vox⁡B\operatorname{vox}_{B} denote the natural projections from KnK^n to Kn/BnK^n/B^n and from KK to K/BK/B, respectively. For a precise definition, see § 3.

Acknowledgements

The author would like to thank Immanuel Halupczok for giving them this question and for many helpful discussions throughout all stages of writing this article.

AI statement

During the creation of this article, the author used a free plan of ChatGPT for broad mathematical discussion. The proofs of Lemma 2.6 and Lemma 3.2 were suggested by ChatGPT. The entire article, including the aforementioned proofs, was written by the author.

Relative dimension of cuboids

Throughout, let πi:Kn→K\pi_i: K^n \to K denote the projection to the iith coordinate for 1≤i≤n1 \leq i \leq n.

Definition 2.1. A ball in KnK^n is a proper and infinite subset BB of KnK^n such that for all x,x′∈Bx,x' \in B and y∈Kny \in K^n, ∣x−y∣≤∣x−x′∣\lvert x-y\rvert\leq\lvert x-x'\rvert implies y∈By \in B.

Note that this definition includes balls derived from cuts in the value group. Further note that whether or not a given set is a ball depends on the ambient space. For example, if L/KL/K is an extension of valued fields and BB is a ball in KK, then BB is not necessarily a ball in LL. Since the ambient space will always be apparent, we often omit to mention it.

Definition 2.2. A cuboid in KnK^n is a Cartesian product of nn balls in KK.

Clearly, every ball in KnK^n is a cuboid in KnK^n. Like balls, any cuboid in KnK^n that contains 00 is a subgroup of KnK^n and the non-empty intersection of two cuboids is another cuboid.

Definition 2.3. An isometry between subsets X,Y⊆KnX,Y \subseteq K^n is a bijection f:X→Yf:X \to Y such that

∣f(x)−f(x′)∣=∣x−x′∣for all x,x′∈X.\lvert f(x)-f(x')\rvert= \lvert x-x'\rvert\quad\text{for all } x,x' \in X.

We call XX and YY isometric if there exists an isometry between them. Moreover, we call XX and YY definably isometric if there exists a definable isometry between them.

As shown in the introduction, whenever KK has infinite residue field, there exists an isometry between O×M\mathcal{O}\times\mathcal{M} and O×O\mathcal{O}\times\mathcal{O}, where O\mathcal{O} and M\mathcal{M} denote the valuation ring and valuation ideal of KK, respectively. We will see that such an isometry cannot be definable. By showing that definable isometries between cuboids preserve a certain sense of size and shape. To make this precise, we introduce the following notion of relative dimension for cuboids.

Definition 2.4. Let P,Q⊂KnP,Q \subset K^n be cuboids around 00.

(i) The dimension type of PP over QQ is the tuple t∈{0,1}nt \in\{0,1\}^n with entries ti=0t_i=0 if and only if πi(P)⊆πi(Q)\pi_i(P) \subseteq\pi_i(Q) for 1≤i≤n1 \leq i \leq n. We denote it by t(P/Q)t(P/Q).

(ii) The dimension of PP over QQ is the sum of entries of t(P/Q)t(P/Q). We denote it by dim⁡(P/Q)\dim(P/Q).

Clearly, translation does not affect whether two cuboids are definably isometric. Hence, we may assume without loss of generality that all cuboids in discussion contain 00.

We will see that definably isometric cuboids have the same dimension over certain balls in the sense of the above definition. In particular, it will follow from Proposition 3.12 that the cuboids O×M\mathcal{O} \times\mathcal{M} and O×O\mathcal{O} \times\mathcal{O} must have the same dimension over every ball in K2K^2 in order to be definably isometric. This fails over M×M\mathcal{M} \times\mathcal{M}, as dim⁡(O×M/M×M)=1\dim(\mathcal{O} \times\mathcal{M}/\mathcal{M} \times\mathcal{M}) = 1 and dim⁡(O×O/M×M)=2\dim(\mathcal{O} \times\mathcal{O}/\mathcal{M} \times\mathcal{M}) = 2.

Remark 2.5. Let P,Q⊆KnP,Q \subseteq K^n be cuboids around 00. Then t(P/Q∩P)=t(P/Q)t(P/Q \cap P) = t(P/Q).

Lemma 2.6. Let P⊆KnP \subseteq K^n be a cuboid around 00 and let B⊊B′⊆KnB \subsetneq B' \subseteq K^n be strictly nested balls around 00. Then t(P∩B′/B)=t(P/B)t(P \cap B'/B) = t(P/B).

Proof. Fix some index ii. Clearly, the inclusion πi(P)⊆πi(B)\pi_i(P) \subseteq\pi_i(B) implies πi(P∩B′)⊆πi(B)\pi_i(P \cap B') \subseteq\pi_i(B). Now suppose πi(P)⊈πi(B)\pi_i(P) \nsubseteq\pi_i(B). Since PP and B′B' have non-empty intersection, the balls πi(P)\pi_i(P) and πi(B′)\pi_i(B') are nested. Hence, πi(P∩B′)\pi_i(P \cap B') equals πi(P)\pi_i(P) or πi(B′)\pi_i(B') and in either case one has πi(P∩B′)⊈πi(B)\pi_i(P \cap B') \nsubseteq\pi_i(B). □

:::

Definable isometries between cuboids

In this section, we go about proving Theorem 1.1. We treat KK having finite and infinite residue field separately. For a ball B⊆KnB \subseteq K^n around 00, we often make use of the natural projection

vox⁡B:Kn→Kn/B.\operatorname{vox}_{B}: K^n \to K^n/B.

The name vox⁡B\operatorname{vox}_{B} is short for voxelization with voxels the size of BB.

Remark 3.1. Let P⊆KnP \subseteq K^n be a cuboid and B⊆KnB \subseteq K^n a ball, each around 00. Then t(P/B)t(P/B) is the unique tuple t∈{0,1}nt \in\{0,1\}^n with the least sum of entries such that vox⁡B(∏i=1ntiK)\operatorname{vox}_{B}\left(\prod_{i=1}^{n} t_iK\right) covers vox⁡B(P)\operatorname{vox}_{B}(P). Here, 0⋅K={0}0 \cdot K = \{0\} and 1⋅K=K1 \cdot K = K.

Lemma 3.2. Let P,P′⊆KnP,P' \subseteq K^n be isometric cuboids and B⊆KnB \subseteq K^n a ball, each around $0. Then

(i) the cuboids P∩BP \cap B and P′∩BP' \cap B are isometric, and

(ii) there exists a bijection between vox⁡B(P)\operatorname{vox}_{B}(P) and vox⁡B(P′)\operatorname{vox}_{B}(P').

Proof. Let f:P→P′f:P \to P' be an isometry. For (i), see that f(P∩B)=P′∩(f(0)+B)f(P \cap B) = P' \cap(f(0) + B). Hence, the mapping x↦f(x)−f(0)x \mapsto f(x) - f(0) gives an isometry from P∩BP \cap B to P′∩BP' \cap B. For (ii), all x,x′∈Knx,x' \in K^n satisfy f(x)−f(x′)∈Bf(x) - f(x') \in B if and only if x−x′∈Bx - x' \in B. Hence, ff descends to a well-defined bijection from vox⁡B(P)\operatorname{vox}_{B}(P) to vox⁡B(P′)\operatorname{vox}_{B}(P'). □

Finite residue field

When KK has finite residue field, we can use Lemma 3.2 to find a finite bijection between vox⁡B(P)\operatorname{vox}_{B}(P) and vox⁡B(P′)\operatorname{vox}_{B}(P') for suitable isometric cuboids P,P′⊆KnP,P' \subseteq K^n and balls B⊆KnB \subseteq K^n.

Remark 3.3. Assume that KK has finite residue field kk. Let P⊆KnP \subseteq K^n be a cuboid around 00 and let B={x∈Kn∣∣x∣<γ}B = \{x \in K^n \mid|x| < \gamma\} and B′={x∈Kn∣∣x∣≤γ}B' = \{x \in K^n \mid|x| \leq\gamma\} for some γ\gamma in the value group of KK. If PP is contained in B′B', then #vox⁡B(P)=(#k)dim⁡(P/B)\#\operatorname{vox}_{B}(P) = (\#k)^{\dim(P/B)}.

Lemma 3.4. Assume that KK has finite residue field kk. Let P,P′⊆KnP,P' \subseteq K^n be isometric cuboids around 00 and let B={x∈Kn∣∣x∣<γ}B = \{x \in K^n \mid|x| < \gamma\} for some γ\gamma in the value group of KK. Then dim⁡(P/B)=dim⁡(P′/B)\dim(P/B) = \dim(P'/B).

Proof. Let B′={x∈Kn∣∣x∣≤γ}B' = \{x \in K^n \mid|x| \leq\gamma\}. By Lemma 3.2, there exists a bijection between vox⁡B(P∩B′)\operatorname{vox}_{B}(P \cap B') and vox⁡B(P′∩B′)\operatorname{vox}_{B}(P' \cap B'). Hence, Remark 3.3 yields dim⁡(P∩B′/B)=dim⁡(P′∩B′/B)\dim(P \cap B'/B) = \dim(P' \cap B'/B). Now apply Lemma 2.6. □

:::

Proposition 3.5. Assume that KK has finite residue field and let P,P′⊆KnP, P' \subseteq K^n be cuboids. If PP and P′P' are isometric, then PP and P′P' are equal up to translation and coordinate permutation.

Proof. Assume that PP and P′P' are isometric. Since translation and coordinate permutation do not affect whether PP and P′P' are isometric, we may assume without loss of generality that PP and P′P' each contain 00 and that πi(P)⊆πi+1(P)\pi_i(P) \subseteq\pi_{i+1}(P) and πi(P′)⊆πi+1(P′)\pi_i(P') \subseteq\pi_{i+1}(P') for 1≤i<n1 \leq i < n. Suppose that PP and P′P' are not equal. Then there is a largest index ii such that πi(P)\pi_i(P) and πi(P′)\pi_i(P') differ. Without loss, assume πi(P)⊊πi(P′)\pi_i(P) \subsetneq\pi_i(P'). By choice of ii, this implies πi(P)⊆πi+1(P)\pi_i(P) \subseteq\pi_{i+1}(P). Pick some a∈πi(P′)∖πi(P)a \in\pi_i(P') \setminus\pi_i(P) and let I={x∈K∣∣x∣<∣a∣}I = \{x \in K \mid|x| < |a|\}. Then πi(P)⊆I⊆πi(P′)\pi_i(P) \subseteq I \subseteq\pi_i(P'), hence dim⁡(P/In)=n−i\dim(P/I^n) = n-i and dim⁡(P′/In)=n−(i−1)\dim(P'/I^n) = n-(i-1). But this contradicts Lemma 3.4.

Infinite residue field

When KK has infinite residue field, the previous strategy fails, since vox⁡B(P)\operatorname{vox}_B(P) is either a singleton or infinite for any cuboid P⊆KnP \subseteq K^n and ball B⊆KnB \subseteq K^n. To bypass this issue, we assume that the theory of KK is 1-h-minimal. This way, we are able to get finite approximations of vox⁡B(X)\operatorname{vox}_B(X) for any definable set X⊆KnX \subseteq K^n and ball B⊆KnB \subseteq K^n in the form of Proposition 1.3.

Definition 3.6. A map f:X→Knf : X \to K^n on a subset X⊆KnX \subseteq K^n is called 1-Lipschitz if

∣f(x)−f(x′)∣≤∣x−x′∣for all x,x′∈X.|f(x) - f(x')| \leq|x - x'| \quad\text{for all } x, x' \in X.

For the rest of this section, let π≤d:Kn→Kd\pi_{\leq d} : K^n \to K^d denote the projection to the first dd coordinates for 0≤d≤n0 \leq d \leq n. The following theorem is a simple corollary of [2 Theorem 3.3.3, Addendum 4].

Theorem 3.7 (1-Lipschitz cell decomposition). Assume that Th⁡(K)\operatorname{Th}(K) is 1-h-minimal. Then for any definable set X⊆KnX \subseteq K^n, there exists a finite partition of XX into definable subsets C1,…,CN⊆XC_1, \ldots, C_N \subseteq X, together with coordinate permutations σ1,…,σN:Kn→Kn\sigma_1, \ldots, \sigma_N : K^n \to K^n, such that each σi(Ci)\sigma_i(C_i) is the graph of a definable 1-Lipschitz map

π≤dim⁡(Ci)(σi(Ci))→Kn−dim⁡(Ci).\pi_{\leq\dim(C_i)}(\sigma_i(C_i)) \to K^{n-\dim(C_i)}.

Remark 3.8. Let X⊆KdX \subseteq K^d be a subset for some integer d≤nd \leq n. Then any 1-Lipschitz map f:X→Kn−df : X \to K^{n-d} satisfies

∣(x,f(x))−(x′,f(x′))∣=∣x−x′∣for all x,x′∈X.|(x, f(x)) - (x', f(x'))| = |x - x'| \quad\text{for all } x, x' \in X.

Hence, for any ball B⊆KB \subseteq K around 00, the projection π≤d:graph⁡(f)→X\pi_{\leq d} : \operatorname{graph}(f) \to X descends to a well-defined bijection π‾≤d:vox⁡Bn(graph⁡(f))→vox⁡Bd(X)\overline{\pi}_{\leq d} : \operatorname{vox}_{B^n}(\operatorname{graph}(f)) \to\operatorname{vox}_{B^d}(X).

With this, we are able to prove Proposition 1.3.

Proposition 3.9 (Counting balls). Assume that Th⁡(K)\operatorname{Th}(K) is 1-h-minimal. For every definable set X⊆KnX \subseteq K^n there exists some N∈NN \in\mathbb{N} such that for every ball B⊆KB \subseteq K around 00 and every finite subset A⊆KA \subseteq K, one has

#(vox⁡Bn(X)∩vox⁡Bn(An))≤N⋅(#vox⁡B(A))dim⁡(X).\#\left(\operatorname{vox}_{B^n}(X) \cap\operatorname{vox}_{B^n}(A^n)\right) \leq N \cdot\left(\#\operatorname{vox}_B(A)\right)^{\dim(X)}.

Proof. Let C1,…,CNC_1, \ldots, C_N and σ1,…,σN\sigma_1, \ldots, \sigma_N be as in Theorem 3.7 and fix some index i∈{1,…,N}i \in\{1, \ldots, N\}. Let B⊆KB \subseteq K be a ball around 00 and let A⊆KA \subseteq K be a finite subset. Since σi(An)=An\sigma_i(A^n) = A^n, one has

#(vox⁡Bn(Ci)∩vox⁡Bn(An))=#(vox⁡Bn(σi(Ci))∩vox⁡Bn(An)).\#\left(\operatorname{vox}_{B^n}(C_i) \cap\operatorname{vox}_{B^n}(A^n)\right) = \#\left(\operatorname{vox}_{B^n}(\sigma_i(C_i)) \cap\operatorname{vox}_{B^n}(A^n)\right).

Since σi(Ci)\sigma_i(C_i) is the graph of a 1-Lipschitz map on π≤di(σi(Ci))\pi_{\leq d_i}(\sigma_i(C_i)), where di=dim⁡(Ci)d_i = \dim(C_i), Remark 3.8 yields the upper bound

#(vox⁡Bn(σi(Ci))∩vox⁡Bn(An))≤#π≤di(vox⁡Bn(An))=(#vox⁡B(A))di.\#\left(\operatorname{vox}_{B^n}(\sigma_i(C_i)) \cap\operatorname{vox}_{B^n}(A^n)\right) \leq\#\pi_{\leq d_i}\left(\operatorname{vox}_{B^n}(A^n)\right) = \left(\#\operatorname{vox}_B(A)\right)^{d_i}.

Since each did_i is bounded by dim⁡(X)\dim(X), we obtain

#(vox⁡Bn(X)∩vox⁡Bn(An))=∑i=1N#(vox⁡Bn(Ci)∩vox⁡Bn(An))≤N⋅(#vox⁡B(A))dim⁡(X).\#\left(\operatorname{vox}_{B_n}(X)\cap\operatorname{vox}_{B_n}(A^n)\right)=\sum_{i=1}^{N}\#\left(\operatorname{vox}_{B_n}(C_i)\cap\operatorname{vox}_{B_n}(A^n)\right)\leq N\cdot\left(\#\operatorname{vox}_B(A)\right)^{\dim(X)}.

For the rest of this section, let PP, P′P', Q⊆KnQ\subseteq K^n be cuboids around 00 and let BB, B′⊆KnB'\subseteq K^n be balls around 00. For a subset X⊆KX\subseteq K and tuple t∈{0,1}nt\in\{0,1\}^n, we write XtX^t for ∏i=1ntiX\prod_{i=1}^n t_iX, where 0⋅X={0}0\cdot X=\{0\} and 1⋅X=X1\cdot X=X.

Lemma 3.10. If Q⊆PQ\subseteq P, then there exists a ball I⊆KI\subseteq K such that It(P/Q)⊈QI^{t(P/Q)}\not\subseteq Q and It(P/Q)⊆PI^{t(P/Q)}\subseteq P.

Proof. Let t=t(P/Q)t=t(P/Q) and D={i∈{1,…,n}∣ti=1}D=\{i\in\{1,\ldots,n\}\mid t_i=1\}. Since QQ is a proper subset of PP, DD is non-empty. Choose j∈Dj\in D such that the projection πj(P)\pi_j(P) is inclusion-minimal among πi(P)\pi_i(P) for all i∈Di\in D. Set I=πj(P)I=\pi_j(P). Then clearly, ItI^t is contained in PP. Choose some a∈πj(P)∖πj(Q)a\in\pi_j(P)\setminus\pi_j(Q) and consider the point x∈Knx\in K^n with coordinates

xi={0,if ti=0,ai,if ti=1.x_i= \begin{cases} 0, & \text{if }t_i=0,\\ a_i, & \text{if }t_i=1. \end{cases}

Then x∈Itx\in I^t and x∉Qx\notin Q. □\square

For a given sequence (bm)m∈N(b_m)_{m\in\mathbb{N}}, we write {b<m}\{b_{<m}\} for {bi∣0≤i<m}\{b_i\mid0\leq i<m\}.

Lemma 3.11. Assume that KK has infinite residue field. If dim⁡(P/B)>0\dim(P/B)>0, then there is a sequence (bm)m∈N(b_m)_{m\in\mathbb{N}} in KK, starting at 00, such that {b<m}t(P/B)⊆P\{b_{<m}\}^{t(P/B)}\subseteq P and #vox⁡B({b<m}t(P/B))=mdim⁡(P/B)\#\operatorname{vox}_B(\{b_{<m}\}^{t(P/B)})=m^{\dim(P/B)} for all m>0m>0.

Proof. If dim⁡(P/B)>0\dim(P/B)>0, then PP is not a subset of BB. In turn, the intersection B∩PB\cap P is a proper subset of PP. Hence, Lemma 3.10 yields a ball I⊆KI\subseteq K such that It(P/B∩P)⊈B∩PI^{t(P/B\cap P)}\not\subseteq B\cap P and It(P/B∩P)⊆PI^{t(P/B\cap P)}\subseteq P. In particular, there is some index ii such that t(P/B∩P)i=1t(P/B\cap P)_i=1 and πi(B∩P)⊊I\pi_i(B\cap P)\subsetneq I. Since πi(I)⊆πi(P)\pi_i(I)\subseteq\pi_i(P), this leaves πi(B)⊊I\pi_i(B)\subsetneq I. Since BB is a ball, we have B⊊InB\subsetneq I^n.

Since KK has infinite residue field, the quotient In/BI^n/B is infinite. Hence, there is a sequence (bm)m∈N(b_m)_{m\in\mathbb{N}} in II, starting at 00, such that each bmb_m lies in a unique coset of π1(B)\pi_1(B). Since t(P/B∩P)=t(P/B)t(P/B\cap P)=t(P/B), the sequence (bm)m∈N(b_m)_{m\in\mathbb{N}} satisfies the necessary conditions. □\square

Proposition 3.12. Assume that Th⁡(K)\operatorname{Th}(K) is 1-h-minimal and that KK has infinite residue field. If PP and P′P' are definably isometric, then dim⁡(P/B)=dim⁡(P′/B)\dim(P/B)=\dim(P'/B).

Proof. Without loss of generality, assume dim⁡(P/B)≤dim⁡(P′/B)\dim(P/B)\leq\dim(P'/B). Let f:P→P′f:P\to P' be a definable isometry. By Remark 3.1, the definable set X=Kt(P/B)∩PX=K^{t(P/B)}\cap P is of dimension dim⁡(P/B)\dim(P/B) and vox⁡B(X)\operatorname{vox}_B(X) covers vox⁡B(P)\operatorname{vox}_B(P). Since ff is an isometry, the image f(X)f(X) is also of dimension dim⁡(P/B)\dim(P/B), and by the proof of Lemma 3.2, vox⁡B(f(X))\operatorname{vox}_B(f(X)) covers vox⁡B(P′)\operatorname{vox}_B(P').

If dim⁡(P′/B)=0\dim(P'/B)=0, then dim⁡(P/B)=0\dim(P/B)=0 and we are done. If dim⁡(P′/B)>0\dim(P'/B)>0, then there exists a sequence (bm)m∈N(b_m)_{m\in\mathbb{N}} in KK as described in Lemma 3.11. Then vox⁡B({b<m}t(P′/B))⊆vox⁡B(f(X))\operatorname{vox}_B(\{b_{<m}\}^{t(P'/B)})\subseteq\operatorname{vox}_B(f(X)), hence applying Proposition 3.9 to f(X)f(X) yields an integer N∈NN\in\mathbb{N} such that

#vox⁡B({b<m}t(P′/B))≤N⋅(#vox⁡π1(B)({b<m}))dim⁡(f(X))for all m>0.\#\operatorname{vox}_B(\{b_{<m}\}^{t(P'/B)})\leq N\cdot\left(\#\operatorname{vox}_{\pi_1(B)}(\{b_{<m}\})\right)^{\dim(f(X))}\quad\text{for all }m>0.

By construction, the left-hand side is equal to mdim⁡(P′/B)m^{\dim(P'/B)} and the right-hand side is bounded by N⋅mdim⁡(f(X))N\cdot m^{\dim(f(X))}. Since dim⁡(f(X))\dim(f(X)) is equal to dim⁡(P/B)\dim(P/B), we obtain

mdim⁡(P′/B)≤N⋅mdim⁡(P/B)for all m>0.m^{\dim(P'/B)}\leq N\cdot m^{\dim(P/B)}\quad\text{for all }m>0.

Hence, dim⁡(P′/B)≤dim⁡(P/B)\dim(P'/B)\leq\dim(P/B). This finishes the proof. □\square

We can now prove our main result Theorem 1.1.

Theorem 3.13. Assume that Th⁡(K)\operatorname{Th}(K) is 11-hh-minimal and let P,P′⊆KnP,P' \subseteq K^n be Cartesian products of balls in KK. If PP and P′P' are definably isometric, then PP and P′P' are equal up to translation and coordinate permutation.

Proof. If KK has finite residue field, apply Proposition 3.5. Assume that KK has infinite residue field and that PP and P′P' are definably isometric. As in the proof of Proposition 3.5, we may assume without loss of generality that PP and P′P' each contain 00 and satisfy πi(P)⊆πi+1(P)\pi_i(P) \subseteq\pi_{i+1}(P) and πi(P′)⊆πi+1(P′)\pi_i(P') \subseteq\pi_{i+1}(P') for 1≤i<n1 \leq i < n. Now suppose PP and P′P' are not equal. Then there is a largest index ii such that πi(P)\pi_i(P) and πi(P′)\pi_i(P') differ. Without loss, assume πi(P)⊆πi(P′)\pi_i(P) \subseteq\pi_i(P') and let B=(πi(P))nB=(\pi_i(P))^n. Then dim⁡(P/B)=n−i\dim(P/B)=n-i and dim⁡(P′/B)=n−(i−1)\dim(P'/B)=n-(i-1). But this contradicts Proposition 3.12. □

Since every ball in KnK^n is the Cartesian product of its coordinate projections, we further obtain the following generalization of Theorem 1.1.

Corollary 3.14. Assume that Th⁡(K)\operatorname{Th}(K) is 11-hh-minimal and let B1,…,Bm⊆KnB_1,\ldots,B_m \subseteq K^n and B1′,…,Bm′⊆KnB'_1,\ldots,B'_m \subseteq K^n be finitely many balls. If ∏i=1mBi\prod_{i=1}^{m} B_i and ∏i=1mBi′\prod_{i=1}^{m} B'_i are definably isometric, then ∏i=1mBi\prod_{i=1}^{m} B_i and ∏i=1mBi′\prod_{i=1}^{m} B'_i are equal up to translation and coordinate permutation.

Proof. The sets ∏i=1mBi\prod_{i=1}^{m} B_i and ∏i=1mBi′\prod_{i=1}^{m} B'_i are cuboids in KmnK^{mn}. □

Lastly, with a similar method as the one used in the proof of Proposition 3.12, we obtain Proposition 1.2.

Proposition 3.15. Assume that Th⁡(K)\operatorname{Th}(K) is 11-hh-minimal and that KK has infinite residue field. Then dim⁡(P/B)\dim(P/B) is the minimal dimension of a definable set X⊆KnX \subseteq K^n such that vox⁡B(X)\operatorname{vox}_B(X) covers vox⁡B(P)\operatorname{vox}_B(P).

Proof. The existence of XX follows from Remark 3.1. Now let X⊆KnX \subseteq K^n be an arbitrary definable set such that vox⁡B(X)\operatorname{vox}_B(X) covers vox⁡B(P)\operatorname{vox}_B(P). If dim⁡(P/B)=0\dim(P/B)=0, then clearly dim⁡(P/B)≤dim⁡(X)\dim(P/B) \leq\dim(X). If dim⁡(P/B)>0\dim(P/B)>0, then there is a sequence (bm)m∈N(b_m)_{m\in\mathbb{N}} in KK as described in Lemma 3.11. Applying Proposition 3.9 to XX yields an integer N∈NN \in\mathbb{N} such that

#vox⁡B({b<m}t(P/B))≤N⋅(vox⁡π1(B)({b<m}))dim⁡(X)for all m>0.\#\operatorname{vox}_B\left(\{b_{<m}\}^{t(P/B)}\right) \leq N\cdot\left(\operatorname{vox}_{\pi_1(B)}\left(\{b_{<m}\}\right)\right)^{\dim(X)} \quad\text{for all }m>0.

Since the left-hand side is equal to mdim⁡(P/B)m^{\dim(P/B)} and the right-hand side is bounded by N⋅mdim⁡(X)N\cdot m^{\dim(X)}, we obtain

mdim⁡(P/B)≤N⋅mdim⁡(X)for all m>0.m^{\dim(P/B)} \leq N\cdot m^{\dim(X)} \quad\text{for all }m>0.

Hence, dim⁡(P/B)≤dim⁡(X)\dim(P/B) \leq\dim(X). This finishes the proof. □

References

References

  1. [1]Raf Cluckers, Immanuel Halupczok, and Silvain Rideau-Kikuchi. Hensel minimality I. Forum of Mathematics, Pi, 10:e11, 2022.
  2. [2]Raf Cluckers, Immanuel Halupczok, Silvain Rideau-Kikuchi, and Floris Vermeulen. Hensel minimality II: Mixed characteristic and a diophantine application. Forum of Mathematics, Sigma, 11:e89, 2023.arxiv.org/abs/2104.09475

Paper details

vlnc:W_bfwvnjeu
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

ChatGPT

View statement
Looking for related papers…
Contents