Article Search

##  eISSN 0454-8124
pISSN 1225-6951

### Article

Kyungpook Mathematical Journal 2021; 61(3): 455-459

Published online September 30, 2021

### Oh's 8-Universality Criterion is Unique

Scott Duke Kominers

Harvard Business School, Department of Economics, and Center of Mathematical Sciences and Applications, Harvard University, Rock Center 219, Soldiers Field, Boston, MA 02163, USA
e-mail : kominers@fas.harvard.edu

Received: June 8, 2020; Revised: August 30, 2020; Accepted: November 23, 2020

We partially characterize criteria for the n-universality of positive-definite integer-matrix quadratic forms. We then obtain the uniqueness of Oh's 8-universality criterion  as a corollary.

A degree-two homogeneous polynomial in n independent variables is called a quadratic form (or just form) of rank n. For a rank-n quadratic form Q(x1,,xn)= i,jaijxixj (where aij=aji), the matrix given by L=(aij) is the Gram Matrix of a -lattice L equipped with a symmetric bilinear form B(,) such that B(L,L). Then, Q(x)=xTLx=B(Lx,x) for xn.

A rank-n quadratic form Q is said to represent an integer k if there exists an xn such that Q(x)=k. More generally, a -lattice L represents another -lattice ℓ if there exists a -linear, bilinear form-preserving injection lL. A quadratic form is called universal if it represents all positive integers. Analogously, a lattice is called n-universal if it represents all rank-n positive-definite integer-matrix -lattices. Connecting the two notions of universality, we observe that a rank-n quadratic form Q is universal if and only if it is 1-universal, as for an integer k,

k=Q(x1,,xn)Q(x1x,,xnx)=kx2.

In 1993, Conway and Schneeberger announced their celebrated Fifteen Theorem, giving a criterion characterizing the universal positive-definite integer-matrix quadratic forms. Specifically, they showed that any positive-definite integer-matrix form that represents the set of nine critical numbers

{1,2,3,5,6,7,10,14,15}

is universal (see [1, 3]). Kim, Kim, and Oh  presented an analogous criterion for 2-universality, showing that a positive-definite integer-matrix lattice is 2-universal if and only if it represents the set of forms

S2=1001,2003,3003,2112,2113,2114.

Oh  gave a similar criterion for 8-universality, which we state in Theorem 4.1 of Section 4.

A set S of rank-n lattices having the property that a lattice L is n-universal if and only if L represents every lattice in S is called an n-criterion set. Thus, for example, the set S2 obtained by Kim, Kim, and Oh  is a 2-criterion set and the set of integers found by Conway  naturally gives the 1-criterion set

S1=x2,2x2,3x2,5x2,6x2,7x2,10x2,14x2,15x2.

The set S1 is known to be the unique minimal 1-criterion set (see ), in the sense that if S1 is a 1-criterion set, then S1S 1. The author  obtained an analogous uniqueness result for the 2-criterion set S2.

Kim, Kim, and Oh  have proven that n-criterion sets exist for all positive integers n. However, the problems of finding and determining the uniqueness of these sets have proven to be difficult (see the discussion in ). Here, we advance both problems: We obtain two simple (partial) characterization results for arbitrary n-criterion sets, from which we obtain the uniqueness of Oh's 8-universality criterion as a corollary.

Since we first circulated this paper, there has been renewed attention in characterizing criterion sets: Elkies, Kane, and the author  identified several families of lattices for which there exist multiple universality criteria of different sizes, including one based on the n and E8 lattices that builds on our work here. More recently, Lee  and Kim, Lee, and Oh  showed that the minimal n-criterion sets are not unique for n≥ 9, and introduced an elegant theory of recoverable lattices that substantially generalizes . (See also recent work of Chan and Oh  characterizing classes of exceptional sets for rank-n quadratic forms, which in some sense can be thought of as building blocks for criterion sets.)

### 2. Notation and Terminology

We use the lattice-theoretic language of quadratic form theory. A complete introduction to this approach may be found in . In addition, we use the lattice notation of , under which In is the rank-n lattice of the form 1,,1 and E8 is the unique even unimodular lattice of rank 8.

For a -lattice (or hereafter, just lattice) L with basis {x1,,xn}, we write Lx1++xn. If L is of the form L=L1L2 for sublattices L1 and L2 of L with B(L1,L2)=0, then we write LL1L2 and say that L1 and L2 are orthogonal.

For a sublattice ℓ of L1L2 that can be expressed in the form

l(x1,1+x2,1)++(x1,n+x2,n)

with xi,jLi, we denote l(Li):=xi,1++xi,n. We naturally extend this notation to lattices ℓ represented by L1L2. We then say that a lattice is additively indecomposable if either l(L1)0 or l(L2)0 whenever L1L2 represents ℓ. Otherwise, we say that ℓ is additively decomposable.

### 3. Partial Characterization of n-Criterion Sets

In this section, we prove two results that partially characterize the contents of arbitrary n-criterion sets.

Proposition 3.1. Any n-criterion set must include the lattice In.

Proof. If T is a finite, nonempty set of rank-n lattices not containing In, then every lattice TT may be written in the form TIkT, where 0k<n, the sublattice T' is of rank n-k, and the first minimum of T' is larger than 1. Indeed, any Ik-sublattice of T is unimodular and therefore splits T; the condition on T' follows from Minkowski reduction.

We may therefore write T in the form

T= k=0 n1 IkT k,ii=1ik,

where 0<|T|= k=0 n1ik and each Tk,i is a rank-(n-k) lattice with first minimum greater than 1. Then, the lattice

In1i=1i0T0,ii=1in1Tn1,i

represents all of T but does not represent In. It follows that T is not an n-criterion set; hence, any n-criterion set must contain In.

Proposition 3.2. it Let E be the set of additively indecomposable unimodular lattices of rank n. If E, then any n-criterion set must include at least one lattice EE.

Proof. Suppose that E. If T={Ti}i=1k is a finite, nonempty set of rank-n lattices with TE=, then every lattice TiT is either additively decomposable or not unimodular (or both). Now, we consider the lattice

T1Tk,

which of course represents all of T by construction.

If T1Tk were to represent some EE, then under any such representation we would have E(Ti)0 for all but one i (with 1ik) because E is additively indecomposable. Then, for some i (again, with 1ik), the lattice Ti would represent E. In that case, as E is unimodular, the associated sublattice of Ti would split Ti as TiET--and since both E and Ti are of rank n, we would have T0; hence, TiE. But this is impossible because Ti is either additively decomposable or not unimodular, whereas EE is both additively indecomposable and unimodular.

Thus, we have found a lattice that represents all of T but cannot represent any EE. As E by hypothesis, we see that T must not be an n-criterion set; the result follows.

Remark 3.3. It is clear that direct analogues of Propositions 3.1 and 3.2 hold in the more general setting of S-universal lattices discussed in . In particular, suppose that S is an infinite set of lattices. Then, if n=maxk:IkS>0, any finite set SSS with the property that a lattice L represents every lS if and only if L represents every lSS must contain In. Similarly, such a set SS must contain an additively indecomposable unimodular lattice if S does.

### 4. Uniqueness of The 8-Criterion Set

Oh  obtained the following 8-criterion set.

Theorem 4.1.([11, remark on Theorem 3.1]) The set S8={I8,E8} is an 8-criterion set.

The set S8 is clearly a minimal 8-criterion set, as for each lS8 there is a lattice that represents S8l but does not represent ℓ. (The single lattice in S8l suffices.) Meanwhile, our characterization results imply the following corollary, which strengthens Theorem 4.1.

Corollary 4.2. Every 8-criterion set must contain S8 as a subset.

Proof. As E8 is the unique additively indecomposable unimodular lattice of rank 8, the result follows directly from Propositions 3.1 and 3.2.

Corollary 4.2, when combined with Theorem 4.1, shows that S8 is the unique minimal 8-criterion set.

The author is grateful to Pablo Azar, Noam D. Elkies, Andrea J. Hawksley, Sonia Jaffe, Paul M. Kominers, and especially Ravi Jagadeesan for helpful comments and suggestions, and particularly thanks an anonymous referee for pointing out a problem with the original form of Proposition 3.2.

1. M. Bhargava, On the Conway-Schneeberger ﬁfteen theorem, Quadratic forms and their applications (Dublin, 1999), 27-37, Contemp. Math. 272, Amer. Math. Soc., Providence, RI, 2000.
2. W. K. Chan and B.-K. Oh, On the exceptional sets of integral quadratic forms, Int. Math. Res. Notices, https://doi.org/10.1093/imrn/rnaa382.
3. J. H. Conway, Universal quadratic forms and the ﬁfteen theorem, Quadratic forms and their applications (Dublin, 1999), 2326, Contemp. Math. 272, Amer. Math. Soc., Providence, RI, 2000.
4. J. H. Conway and N. J. A. Sloane, Low-dimensional lattices I. Quadratic forms of small determinant, Proc. Roy. Soc. London Ser. A, 418(1854)(1988), 17-41.
5. N. D. Elkies, D. M. Kane and S. D. Kominers, Minimal S-universality criteria may vary in size, J. Théor. Nombres Bordeaux, 25(3)(2013), 557-563.
6. B. M. Kim and M.-H. Kim and B.-K. Oh, 2-universal positive deﬁnite integral quinary quadratic forms, Integral quadratic forms and lattices (Seoul, 1998), 5162, Contemp. Math. 249, Amer. Math. Soc., Providence, RI, 1999.
7. B. M. Kim and M.-H. Kim and B.-K. Oh, A ﬁniteness theorem for representability of quadratic forms by forms, J. Reine Angew. Math., 581(2005), 23-30.
8. K. Kim and J. Lee and B.-K. Oh, Minimal universality criterion sets on the representations of quadratic forms, (2020), preprint, arXiv:2009.04050.
9. S. D. Kominers, Uniqueness of the 2-universality criterion, Note Mat., 28(2)(2008), 203-206.
10. J. Lee, Minimal S-universality criterion sets, Seoul National University, Thesis (Ph.D.), 2020.
11. B.-K. Oh, Universal Z-lattices of minimal rank, Proc. Amer. Math. Soc., 128(3)(2000), 683-689.
12. O. T. O’Meara, Introduction to quadratic forms, Springer-Verlag, New York, 2000.