Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Eulerian polynomials for subarrangements of Weyl arrangements

Eulerian polynomials for subarrangements of Weyl arrangements EULERIAN POLYNOMIALS FOR SUBARRANGEMENTS OF WEYL ARRANGEMENTS AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA ABSTRACT. Let A be a Weyl arrangement. We introduce and study the notion of A-Eulerian polynomial producing an Eulerian-like polynomial for any subarrangement of A. This polyno- mial together with shift operator describe how the characteristic quasi-polynomial of a new class of arrangements containing ideal subarrangements of A can be expressed in terms of the Ehrhart quasi-polynomial of the fundamental alcove. The method can also be extended to define two types of deformed Weyl subarrangements containing the families of the extended Shi, Catalan, Linial arrangements and to compute their characteristic quasi-polynomials. We obtain several known re- sults in the literature as specializations, including the formula of the characteristic polynomial of A via Ehrhart theory due to Athanasiadis (1996), Blass-Sagan (1998), Suter (1998) and Kamiya- Takemura-Terao (2010); and the formula relating the number of coweight lattice points in the fun- damental parallelepiped with the Lam-Postnikov Eulerian polynomial due to the third author. 1. I NTRODUCTION Motivation. One of typical problems in enumerative combinatorics is to count the sizes of sets depending upon a positive integer q. This often gives rise to polynomials. For instance, the chromatic polynomial of an undirected graph, going back to Birkhoff and Whitney, encodes the number of ways of coloring the vertices with q colors so that adjacent vertices get different colors. However, it may happen that enumerating the cardinalities of sets leads to quasi-polynomials. Generally speaking, a quasi-polynomial is a refinement of polynomials, of which the coefficients may not come from a ring but instead are periodic functions with integral periods. Thus a quasi- polynomial is made of a bunch of polynomials, the constituents of the quasi-polynomial. One of the most classical examples in the theory is that the number of integral points in the q-fold dilation of a rational polytope agrees with a quasi-polynomial in q, broadly known as the Ehrhart quasi-polynomial. We are interested in the connection between the counting problems and the arrangement the- ory. A finite list (multiset) A of vectors in Z determines an arrangement A(R) of hyperplanes ` 1 1 ` in the vector space R , an arrangement A(S ) of subtori in the torus (S ) , and especially an arrangement A(Z ) of subgroups in the finite abelian group Z . Enumerating the cardinality of the complement of A(Z ) produces a quasi-polynomial, the characteristic quasi-polynomial quasi (q) of A [KTT08]. This single quasi-polynomial encodes a number of combinatorial and topological information of several types of arrangements and has generated increasing interest re- quasi cently (e.g., [CW12, BM14, Yos18a, Yos18b, TY19, Tra19]). Among the others,  (q) has the first constituent identical with the characteristic polynomial  (t) of A(R) which justified its A(R) name (e.g., [Ath96, KTT08]), and the last constituent identical with the characteristic polynomial Date: May 1, 2020. 2010 Mathematics Subject Classification. Primary 52C35, Secondary 17B22. Key words and phrases. Characteristic quasi-polynomial, Eulerian polynomial, Ehrhart quasi-polynomial, Wor- pitzky partition, root system, ideal subarrangement. arXiv:1911.01650v3 [math.CO] 30 Apr 2020 2 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA quasi (t) of A(S ) [LTY, TY19]. One of the methods used in [KTT08] for showing that  (q) A(S ) is indeed a quasi-polynomial is to express it as a sum of the Ehrhart quasi-polynomials of rational polytopes, or in the sense of [BZ06], as the Ehrhart quasi-polynomial of an “inside-out” polytope. Such an expression is certainly interesting as it reveals the connection between two seemingly un- related quasi-polynomials, one would hope for a more explicit expression if the list A was chosen to be a more special vector configuration. Objective. A particularly well-behaved class of the hyperplane arrangements is that of Weyl arrangements. More precisely, if A =  is a positive system of an irreducible root system , quasi thenA(R) is called the Weyl arrangement ofA. It is proved that  (q) is expressed in terms of the Ehrhart quasi-polynomial L (q) of the fundamental alcove A , the Weyl group, and the index of connection of  (e.g., [Ath96, BS98, Sut98, KTT10]). Thus we arrive at a natural and essential quasi problem that for which subset   ,  (q) can be computed by using the fundamental invariants of , and more importantly, by means of the Ehrhart quasi-polynomials. Some partial quasi results are known. If the root system  is of classical type and is an ideal of  , then  (q) can be computed from information of the signed graph associated with [Tra19]. A result due to quasi [Yos18b] applied to any root system, asserts that  (q) (or simply q ) can be written in terms of the Lam-Postnikov Eulerian polynomial [LP18], shift operator, and L (q). Results. Inspired by the works of [LP18] and [Yos18b], we introduce the notion of A-Eulerian polynomial E (t) - an arrangement theoretical generalization of the classical Eulerian polynomial, and the notion of compatible subsets of  w.r.t. the Worpitzky partition. The first main result quasi in our paper is a formula of  (q) in terms of E (t), shift operator, and L (q) when is compatible. The formula specializes correctly to the two formulas in the extreme cases ( =  and = ;) mentioned above. In addition, we prove that the formula characterizes the compatibility. The second main result in our paper is that the class of compatible subsets contains the ideals of the root system. Using the similar method, we further define two types of deformed Weyl subarrangements containing the families of the extended Shi, Catalan, Linial arrangements and compute their characteristic quasi-polynomials. Organization of the paper. The remainder of the paper is organized as follows. In Section 2, we recall definitions and basic facts of irreducible root systems, their (affine) Weyl groups and the Worpitzky partition. In Section 3, we recall the definitions of the characteristic and Ehrhart quasi-polynomials and specify the choices of lattices for these quasi-polynomials (Remarks 3.4 and 3.6). We also recall the formula between the quasi-polynomials in the extreme case = (Theorem 3.8), and derive a more general formula (Proposition 3.11). In Section 4, we intro- duce the notion of A-Eulerian polynomial (Definition 4.2), of which the main specialization is the Lam-Postnikov Eulerian polynomial (Remark 4.3). We then define the notion of compatible sets (Definition 4.8) which interpolates between the two extreme cases =  and = ;. We quasi prove that the A-Eulerian polynomial is an essential tool to compute  (q) and its generating function for any compatible subset   (Theorems 4.11 and 4.24). We also prove that every ideal is compatible (Theorem 4.16). In Section 5, we define two types of the deformed Weyl subar- rangements (Definition 5.1), of which the main examples are the truncated affine Weyl and deleted Shi arrangements (Remark 5.2). Then we compute the characteristic quasi-polynomials of these deformed arrangements according to two special choices of intervals (Theorems 5.6 and 5.8). 2. ROOT SYSTEMS AND WORPITZKY PARTITION Our standard reference for root systems and their Weyl groups is [Hum90]. Assume that V = R with the standard inner product (;). Let  be an irreducible (crystallographic) root system in V EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 3 with the Coxeter number h and the Weyl group W . Fix a positive system    and let  := + + f ; : : : ; g be the set of simple roots (base) of  associated with  . For = d 2  , 1 ` i i i=1 the height of is defined by ht( ) := d . Define the partial order on  such that i 1 2 i=1 if and only if = n with all n 2 Z . The highest root (w.r.t. ), denoted by 1 2 i i i 0 i=1 ~ 2  , can be written uniquely as a linear combination of the simple roots ~ = c i i i=1 (c 2 Z ). Set := ~, c := 1, and  := [f g. Then we have the linear relation i >0 0 0 0 c + c + + c = 0: 0 0 1 1 ` ` The coefficients c are important in our study and will appear frequently throughout the paper. _ _ For 2 V nf0g, denote := . The root lattice Q(), coroot lattice Q( ), weight lattice ( ; ) Z (), and coweight lattice Z ( ) are defined as follows: Q() := Z ; i=1 _ _ Q( ) := Z ; i=1 Z () := fx 2 V j ( ; x) 2 Z (1  i  `)g; Z ( ) := fx 2 V j ( ; x) 2 Z (1  i  `)g: _ _ Then Q() is a subgroup of Z () of finite index f , and similarly Q( ) is a subgroup of Z ( ) of _ _ _ the index f . The number f is called the index of connection. Let f$ ; : : : ; $ g  Z ( ) be the 1 ` _ _ _ _ dual basis of the base , namely, ( ; $ ) =  . Then Z ( ) = Z$ and c = ($ ; ~). i ij i j i=1 i i For m 2 Z and 2 , the affine hyperplane H is defined by ;m H := fx 2 V j ( ; x) = mg: ;m A connected component of V n H is called an alcove. The fundamental alcove A ;m 2 ;m2Z is defined by ( ; x) > 0 (1  i  `); A := x 2 V : ( ; x) > 1 The closure A = fx 2 V j ( ; x)  0 (1  i  `); ( ; x)  1g is a simplex, which is i 0 _ _ the convex hull of f0; $ =c ; : : : ; $ =c g. The supporting hyperplanes of the facets of A are 1 ` H ; : : : ; H ; H . The affine Weyl group W := W n Q( ) acts simply transitively on ;0 ;0 ;1 a 1 0 the set of alcoves and admits A as a fundamental domain for its action on V . } _ The fundamental domain P of the coweight lattice Z ( ), called the fundamental paral- lelepiped, is defined by } _ P := (0; 1]$ i=1 = fx 2 V j 0 < ( ; x)  1 (1  i  `)g: #W Let N := , and denote [N ] := f1; 2; : : : ; Ng. Then the cardinality of the set  of all alcoves contained in P equals N (see, e.g., [Hum90, Theorem 4.9]). Let us write = fA  P j i 2 [N ]g; i 4 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA where each alcove A is written uniquely as ( ; x) > k ( 2 I ); A = x 2 V ; ( ; x) < k ( 2 J ) where k = k (A ) 2 Z , k = k (A ) 2 Z and the sets I; J   with #(I t J ) = ` + 1 0 >0 i i indicate the constraints on x 2 V according to the inequality symbols >, <, respectively. Definition 2.1. For each A 2 , the partial closure A of A is defined by i i i ( ; x) > k ( 2 I ); A := x 2 V : ( ; x)  k ( 2 J ) Theorem 2.2 (Worpitzky partition). } } P = A : i2[N ] Proof. See, e.g., [Yos18b, Proposition 2.5], [Hum90, Exercise 4.3]. 3. C HARACTERISTIC AND E HRHART QUASI- POLYNOMIALS A function g : Z ! C is called a quasi-polynomial if there exist  2 Z and polynomials >0 f (t) 2 Z[t] (1  k  ) such that for any q 2 Z with q  k mod , >0 g(q) = f (q): The number  is called a period and the polynomial f (t) is called the k-constituent of the quasi- polynomial g. Let := Z ' Z be a lattice with a basis f ; : : : ; g. Let L be a finite list (multiset) i 1 ` i=1 of elements in . Let q 2 Z and denote Z := Z=qZ. For = a 2 L and m 2 Z, >0 q i i i=1 define a subset H of Z by ;m ;Z q q n o H := z 2 Z j a z  m : ;m ;Z i i q q i=1 Given a vector m = (m ) 2 Z , we can define the Z -plexification (or q-reduced) arrangement 2L q of (L; m) by (L; m)(Z ) := fH j 2 Lg: q ;m ;Z The complement of (L; m)(Z ) is defined by M((L; m); ;Z ) := Z n H : q ;m ;Z q q 2L For a sublist S  L, write d for the largest invariant factor of the torsion subgroup of the S;n finitely generated abelian group =hSi (see, e.g., [Nor12, Definition 3.8]). The LCM-period  of L is defined by := lcm(d j S  L): L S;n Theorem 3.1. #M((L; m); ;Z ) is a monic quasi-polynomial in q for which  is a period. The q L quasi-polynomial is called the characteristic quasi-polynomial of (L; m) w.r.t. the lattice , and quasi denoted by  (q). (L;m) Proof. See [KTT08, Theorem 2.4] and [KTT11, Theorem 3.1].  EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 5 We can also define the R-plexification (in fact, the real hyperplane arrangement) of (L; m) as follows: (L; m)(R) := fH j 2 Lg; ;m ;R n o where H := x 2 R j a x = m . For a real hyperplane arrangement A, denote by ;m ;R i i i=1 (t) the characteristic polynomial (e.g., [OT92, Definition 2.52]) ofA. quasi Theorem 3.2. The first constituent of  (q) coincides with the characteristic polynomial of (L;m) (L; m)(R), i.e., f (t) =  (t): (L;m)(R) (L;m) Proof. See, e.g., [KTT08, Theorem 2.5] and [KTT11, Remark 3.3]. Convention: When m = (0) the zero vector, we simply write L, H , H instead of (L; m), ;Z ;R H , H , respectively. ;0;Z ;0;R Remark 3.3. L(Z ) andL(R) are examples of a more abstract concept, the G-plexifications (G is an abelian group) introduced in [LTY] by viewing G = Z ;R, respectively (see also [TY19] for more information on their combinatorial properties). In addition, the characteristic quasi-polynomial can be generalized to chromatic quasi-polynomial by replacing the lattice by a finitely generated abelian group, see, e.g., [Tra18] for more details. Remark 3.4. Throughout the paper, for every   ( Q()), the characteristic quasi-polynomial quasi (q) is always defined w.r.t. the root lattice = Q(). For each   , define the Weyl arrangement of by A := fH j 2 g, where H = fx 2 V j ( ; x) = 0g is the hyperplane orthogonal to . It is not hard to see that H ' H (as vector spaces), thus we can viewA as the R-plexification of , i.e.,A = (R). ;R In standard terminology,A + is known with the name Weyl (or Coxeter) arrangement, and clearly A is a (Weyl) subarrangement ofA + . Remark 3.5. Sometimes, when we say “the” characteristic quasi-polynomial of (L; m)(Z ) or of quasi (L; m)(R), we mean the characteristic quasi-polynomial  (q) of (L; m). For example, the (L;m) quasi quasi characteristic quasi-polynomial  (q) of A is referred to as  (q). We will use this term later in Section 5 when we deal with deformed Weyl arrangements of . Let be a lattice. For a polytopeP with vertices in the rational vector space generated by , the Ehrhart quasi-polynomial L (q) ofP w.r.t. is defined by L (q) := #(qP \ ): We denote byP the relative interior ofP . Similarly, we can define L  (q) := #(qP \ ): For q > 0, the following reciprocity law holds: dimP L (q) = (1) L  (q): P P Remark 3.6. Throughout the paper, the Ehrhart quasi-polynomials L (q), L  (q) are defined w.r.t. the coweight lattice = Z ( ). Let F := H ; F := H (1  i  `) denote the supporting hyperplanes of the facets 0 ;q i ;0 0 i of qA . Then the number of coweight lattice points in qA after removing some facets can be computed by L with the scale factor of dilation being reduced. A 6 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA Proposition 3.7. Let fi ; : : : ; i g  f0; 1; : : : ; `g. Suppose that q > c + + c . Then 1 k i i #(qA \ Z ( )n (F [[ F )) = L (q (c + + c )): i i i i 1 k A 1 k In particular, for q 2 Z, L  (q) = L (q h): Proof. See [Yos18b, Corollaries 3.4 and 3.5]. In general, it is not easy to find explicit formulas which involve both characteristic and Ehrhart quasi-polynomials. With regards to root systems, there is an interesting relation between these quasi-polynomials. Theorem 3.8. #W quasi (q) = L (q): Proof. See, e.g., [KTT10], [Yos18b, Proposition 3.7]. quasi Theorem 3.9. The minimum period of  (q) is equal to lcm(c ; : : : ; c ). Furthermore, by a + 1 ` case-by-case argument, it can be checked that the minimum period coincides with the LCM-period + . Proof. See [KTT10, Corollary 3.2 and Remark 3.3]. quasi Corollary 3.10.  (q) > 0 (equivalently, L  (q) > 0) if and only if q  h. + A Proof. See, e.g., [KTT10, Corollary 3.4]. quasi We can extend Theorem 3.8 to a formula of  (q) for any   in terms of lattice point counting functions. In the proposition below, we view as a list with possible repetitions of elements. Along the proof, we will see why the choice of lattices for the characteristic and Ehrhart quasi-polynomials is important. Proposition 3.11. Let m = (m ) be a vector in Z . Set } _ X (q) := qP \ Z ( )n H ; ( ;m) ;kq+m 2 ;k2Z _ _ Y (q) := fx 2 Z ( )=qZ ( ) j ( ; x) 6 m mod q;8 2 g: ( ;m) We have bijections between sets X (q) ' Y (q) ' M(( ; m);Z ;Z ): ( ;m) ( ;m) q As a result, quasi (q) = #X (q) = #Y (q): ( ;m) ( ;m) ( ;m) + + Proof. The bijection X (q) ' Y (q) is proved in [Yos18b,x3.3]. We can use exactly the ( ;m) ( ;m) same argument applied to every subset . The proof of Y (q) ' M( ;Z ;Z ) for an arbitrary q EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 7 runs as follows: Y (q) = fx = z $ j ( ; x) 6 m mod q;8 2 g ( ;m) i i=1 ` ` ` X X X _ _ = fx = z $ j ( S ; z $ ) 6 m mod q; (1  j  # )g i ij i i i i i=1 i=1 i=1 ' fz = (z ; : : : ; z ) 2 Z j z S 6 m mod q; (1  j  # )g 1 ` i ij i=1 = M(( ; m);Z ;Z ): + ` Remark 3.12. The bijection X + (q) ' M( ;Z ;Z ) appeared (without proof) in [KTT10, quasi Proof of Theorem 3.1]. Theorem 3.8 is a special case of Proposition 3.11 because  (q) = #W #X + (q) = L  (q) [Yos18b,x3.3]. 4. EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEM ENTS + c + Let   , and set :=  n . Definition 4.1. The descent dsc w.r.t. is the function dsc : W ! Z defined by dsc (w) := c : 0i`; w( )2 Definition 4.2. The (arrangement theoretical Eulerian or) A-Eulerian polynomial of (or of the Weyl subarrangementA ) is defined by hdsc (w) E (t) := t : w2W #W + h + + Remark 4.3. (a) If =  , then dsc (w) = 0 for all w 2 W , and E (t) = t . (b) If = ;, then dsc = dsc = cdes, the circular descent statistic, e.g., [LP18, Definition 6.2], [Yos18b, Definition 4.1]. Then E (t) = R (t), the generalized Eulerian polynomial, e.g., [Yos18b, Definition 4.4]. In this paper, we often call it the Lam-Postnikov Eulerian polynomial. Note that if  is of type A , then R (t) = A (t), the classical `-th Eulerian polynomial [LP18, `  ` Theorem 10.1]. It should also be noted that dsc coincides with the notion of affine descents by Dilks-Petersen-Stembridge [DPS09,x2.5] only in type A case. Lemma 4.4. For all w 2 W , 0  dsc (w) < h. In particular, E (0) = 0. + + Proof. If =  , then the statements are clear by Remark 4.3(a). Assume that 6=  . If w( ) 2= for some 1  i  `, then dsc (w) < h. Otherwise, we have w( ) = i 0 + c c w( ) 2  . Thus w( ) 2= , and hence dsc (w) < h. i i 0 i=1 Lemma 4.5. (i) Let w 2 W . Suppose that w induces a permutation on  = f ; ; : : : ; g. 0 1 ` If w( ) = , then c = c . i p i p i i (ii) Let w ; w 2 W . If there exists 2 V such that w (A ) = w (A ) + , then dsc (w ) = 1 2 1 2 1 dsc (w ). 2 8 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA Proof. (i) is exactly [Yos18b, Lemma 4.3(1)]. (ii) is similar to [Yos18b, Lemma 4.3(2)]: since the supporting hyperplanes coincide, w ( ) = w ( ) for 0  i  `. By Definition 4.1 and (i), we 1 i 2 p have X X dsc (w ) = c = c = dsc (w ): 1 i p 2 c c 0i`; w ( )2 0p `; w ( )2 1 i i 2 p 0 0  _ Let A be an arbitrary alcove. We can write A = w(A ) + for some w 2 W and 2 Q( ). Recall the the Worpitzky partition from Theorem 2.2 that P = A . By Lemma 4.5, we i2[N ] can extend dsc to a function on the set of all alcoves (in particular, on the set  of alcoves A contained in P ) as follows: Definition 4.6. dsc (A ) := dsc (w): Theorem 4.7. hdsc (A ) E (t) = t : i2[N ] Proof. Similar to [Yos18b, Theorem 4.7]: there are exactly f elements of the group W := W n _  } Z ( ) that map A to a given alcove in P . In what follows, we shall sometimes abuse terminology and call a face of A that is contained } } in the partial closure A , a face of A . i i Definition 4.8. A subset   is said to be compatible (with the Worpitzky partition) if for } } each A  P the following condition holds: A \ H for 2 ; m 2 Z is either empty, or ;m i i } } } contained in A \ H for 2 ; m 2 Z so that A \ H is a facet of A . ;m ;m i i i In other words, is compatible if and only if the intersection of each A and H ;k i 2 ;k2Z consisting of a facets K ; : : : ; K , and b faces G ; : : : ; G of dimension  ` 2 of A , either 1 a 1 b is empty, or satisfies the condition that for every non-facet G there exists a facet K such that j s G  K . Intuitively, every “non-empty face intersection” of A and the affine hyperplanes w.r.t. j s roots in can be lifted to a “facet intersection”. + + + Example 4.9. (a) Clearly,; and  are compatible. Let  2  , then =  nfg is compatible. } } It is because if A \H is a non-empty face but not a facet of A for i 2 [N ], 2 ; m 2 ;m i i } } n + Z, then A \ H = \ H \ A , where H ( 2  ) are the supporting ;m ;m ;m j i j=1 j i j j j hyperplanes of the facets of A for all 1  j  n with n  2. Thus, there exists 1  k  n such that 6= , i.e., 2 . k k (b) If = f g with = d 2  n , then is not compatible because H \ 1 i i 1 ;ht( ) i=1 n o n o P P P ` ` ` } _ _ P = p $ j p d = ht( ); p 2 (0; 1] = $ . There exists a non- i i i i i=1 i i=1 i=1 i compatible set consisting of more than one elements. For example, let  = G as in Figure 1. Then it is easily seen from Figure 2 that = f ; + ; 3 + 2 g is not compatible 1 1 2 1 2 (regardless of the value of q) because there exist partially closed alcoves (in green) such that the affine hyperplanes (in pink) w.r.t. roots in intersect them only at their vertices. EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 9 FIGURE 1. Root system of type G . Definition 4.10. Let f : Z ! C be a function and let P (S) = a S be a polynomial in S. k=0 The shift operator via P (S) acting on f is defined by (P (S)f )(t) := a f (t k): k=0 Now let us prove the first main result in the paper. Theorem 4.11. A subset   is compatible if and only if for every positive integer q, quasi (q) = (E (S)L  )(q): Proof. The proof of “)” is similar in spirit to [Yos18b, Proof of Theorem 4.8], but requires a more careful analysis of each partially closed alcove in the Worpitzky partition (Theorem 2.2). Let us proceed the proof with as general as possible to see how the compatibility is crucial. Since both sides are quasi-polynomials, it suffices to prove the formula for q  h. In which case, quasi quasi (q)   (q) > 0 (Corollary 3.10), and q h + dsc (A )  0 (Lemma 4.4) for all i 2 [N ] whence (E (S)L )(q) > 0. Fix i 2 [N ]. The intersection of A and the affine hyperplanes w.r.t. A i roots in consists of a facets (0  a  ` + 1), say K ; : : : ; K , and b faces of dimension ` 2 1 a a b (b  0), say G ; : : : ; G of A . Set A := [ K and B := [ G . The following elementary 1 b s j i s=1 j=1 identity } } A n (A[ B) = (A n A)n (B n A) i i implies that } } _ _ _ qA \ Z ( )n H = q(A n A)\ Z ( ) n (q(B n A)\ Z ( )) : ;kq i i 2 ;k2Z 10 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA } _ F IGURE 2. The Worpitzky partition of qP \ Z ( ) in a type G root system (q = 7). } _ } DenoteR(A ; q) := # (q(B n A)\ Z ( )). In other words,R(A ; q) is the number of coweight i i lattice points in (the union of) the non-facets qG but not in any facet qK of qA . By Definition j s 2.1, we may write (4.1) 8 9 < ( ; x) > qk ( 2 I ) = } _ _ } # qA \ Z ( )n H = # x 2 Z ( ) ( ; x) < qk ( 2 J \ ) R(A ; q); ;kq i i : ; 2 ;k2Z (; x)  qk ( 2 J \ ) where each facet qK is supported by one of the hyperplanes of the form H . s ;qk Since A = w(A ) + for some w 2 W and 2 Z ( ) ( is uniquely is determined by w, e.g., [Hum90, Theorem 4.9]), the dilation qA can be written as (w( ); x) < q((w( ); ) + 1); 0 0 qA = x 2 V : (w( ); x) < q(w( ); ); (1  i  `) i i Thus the half-spaces defined by (; x)  qk ( 2 J \ ) correspond exactly to the roots 2 satisfyingw( ) 2 . Applying Proposition 3.7 and Definition 4.6, we obtain 8 9 < ( ; x) > qk ( 2 I ) = (4.2) # x 2 Z ( ) ( ; x) < qk ( 2 J \ ) = L (q h + dsc (A )): A i : ; (; x)  qk ( 2 J \ ) EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 11 Moreover, by the Worpitzky partition and Proposition 3.11, we have X [ quasi } _ (4.3)  (q) = # qA \ Z ( )n H : ;kq i2[N ] 2 ;k2Z Using Theorem 4.7 together with the shift operator (Definition 4.10), we have (4.4) (E (S)L )(q) = L (q h + dsc (A )): A A i i2[N ] Summarizing Formulas (4.1)-(4.4), it follows that for every   we have quasi (q) = (E (S)L )(q) R(A ; q): A i i2[N ] The compatibility of forcesR(A ; q) = 0 for all i 2 [N ]. This completes the proof of “)”. Now suppose that is not compatible. Then there exist A and its non-facet G such that G * K for all facets K . To prove “(”, it suffices to show that there exists (sufficiently large) j s s q 2 Z such that the relative interior of q(G n A) contains a coweight lattice point, in which >0 j case, R(A ; q) > 0. Since W preserves the lattice points of qG , we can transform the problem a j to (the relative interior of) a face of qA . Since a face of qA is the convex hull of a nonempty _ _ subset of f0; q$ =c ; : : : ; q$ =c g, the relative interior of that face contains a lattice point for 1 ` 1 ` sufficiently large q (e.g., the center of mass of any face is a lattice point when q = lcm(1; : : : ; ` + 1) lcm(c ; : : : ; c )). The conclusion follows. 1 ` quasi Remark 4.12. (a) When is not compatible, although  (q) and (E (S)L )(q) are never equal as quasi-polynomials, the equality may occur for particular constituents (see Example 4.18(b)). Proof of Theorem 4.11 hints that in order to find a formula of the characteristic quasi- polynomial of a non-compatible subset, we need to count the lattices points in d-simplices with d  `. (b) When the rank of the root system is at most 3, it is possible to examine the (non-)compatibility by using the pictures. The drawing is rather difficult in higher-dimensional cases, in this regard, Theorem 4.11 is useful to check the (non-)compatibility (see Example 4.19). + + + Definition 4.13. A subset   is an ideal of  (or a lower set of the root poset ( ;)) if for ; 2  ,  ; 2 implies 2 . 1 2 1 2 1 2 Our second main result in the paper is: every ideal is compatible. We remark that if `  2 and + + 2  , then the compatible subset  nfg is not an ideal unless  is the highest root. We recall some basic facts on root systems. For any M  , set  (F) := \ FM where F = Z;Q or R. Note that  (Z)   (Q) = M M M (R) and these root subsystems (in the real vector space that they span) have the same rank. In particular, if M  , then these three subsystems are identical. Lemma 4.14. LetJ ( . ThenJ is a base for  (Z). Proof. See, e.g., [Som97, Lemma 4.3]. Lemma 4.15. For w 2 W and J ( , define D := w(J ) = fw( ) j 2 Jg  . Then D is a base for  (Z). D 12 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA Proof. D is linearly independent because (w( );w( )) = ( ; ) for all ; 2 J . Let j k j k j k 2  (Z). We can write = d  where d 2 Z. We need to show that all d  0 or all 2D d  0. Note that J and D are isomorphic via w. Rewriting, we have = w( d ). w( ) 2J Thus, d 2  (Z). Lemma 4.14 completes the proof. w( ) J 2J Theorem 4.16. If   is an ideal, then is compatible. } } Proof. If A \ H is a non-empty face but not a facet of A for i 2 [N ], 2 ; m 2 Z, then ;m i i } } n + we can write A \H = \ H \A , where H ( 2  nf g) are the supporting ;m ;m ;m j i j=1 j i j j j hyperplanes of the facets of A for all 1  j  n with 2  n  `. Set D := f j 1  j  ng. We need to show that there is 2 \ D. We have the following facts. Fact 1. D is a base for  (Z). Let w 2 W so that A = w(A ) + ( 2 Z ( )). It follows D w w from the definition of A that there existsJ ( , such that D = w(J ). The rest follows from Lemma 4.15. Fact 2. 2  (Q). Choose z 2 A \ H , the translation t : V ! V via t (x) = z + x D 0 ;m z z 0 i o o yieldsz + anehull(A \ H ) = \ H  H . Thus, 2  (R) =  (Q). 0 ;m ;0 ;0 D D i j=1 j Claim. 2  (Z). By Fact 2, D [ f g is linearly dependent. Then there exist integers d ; d ; : : : ; d 2 Z not all zero with d 6= 0, gcdfjd j;jd j; : : : ;jd jg = 1 such that d + 1 n 1 n d = 0 (see, e.g., [BV07, Lemma 17]). The claim is proved once we prove d 2 f1; 1g. j=1 j 0 0 Let us transform the situation in A to A . Set := w( ) with 2 . We can rewrite the relation above as (4.5) d + d = 0; 2J where d := d with 2 D such that = w( ). We consider two cases. 0 0 Case 1. J ( . Since 2 Z, the linearly independence of  and Relation (4.5) yield 2 ZJ . Additionally, d divides d for all 2 J . Thus, d 2 f1; 1g. Case 2. J = I [f g with ; =6 I ( . If d = 0, then the same argument as in Case 1 yields 0 + d 2 f1; 1g. Assume that d 6= 0 and 2  . The transformation by w implies that A \ H = \ H \ H \ A  P : ;1 2I ;0 ;1 j j 0 Let x = r $ 2 P (r 2 [0; 1]) be a typical element in both sets. Since x 2 H for ;0 s s j 2 s all 2 I , we can write x = r $ . Thus, 2nI s s (x; d ) = (x;d ) = d : j 0 0 0 2J It follows from Relation (4.5) and the above equation that d = (x; d ) = (x; d ) = d : j 0 2J Thus we can rewrite Relation (4.5) as d ( + ) + d = 0: 0 j 2I 0 1 0 0 Note that + = w ( + ) 6= 0 where := w( ) 2 D. A similar argument as in Case 0 0 0 0 + 1 yields d 2 f1; 1g. Repeat for if 2  . EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 13 By Fact 1 and Claim, we can write = p with at least two p  1, p  1. Thus, j j k k 1 2 j=1 there is 2 D such that  . Since is an ideal, we must have 2 . This completes k  k k the proof. Remark 4.17. There exists a compatible subset that is not an ideal w.r.t. any positive system of quasi (see Example 4.18(d)). The characteristic quasi-polynomial  (q) when is an ideal and is of classical type has been computed by using information of the signed graph associated with [Tra19]. It would be interesting to compare the mentioned computation with Theorem 4.11. Example 4.18. Let  = G as in Figure 1. (a) Let =  nf ~g. We can see that is compatible either by Example 4.9(a) or by Theorem 4.16. By definition of characteristic quasi-polynomial (Theorem 3.1), quasi 2 (q) = # z 2 Z j z ; z ; z + z ; 2z + z ; 3z + z 6= 0 1 2 1 2 1 2 1 2 (q 1)(q 4) if q  1; 5 mod 6; (q 2)(q 3) if q  2; 3; 4 mod 6; q 5q + 8 if q  0 mod 6: Note that the Ehrhart quasi-polynomial of the fundamental alcove (w.r.t. the coweight lattice) of every root system has been completely computed, e.g., in [Sut98]. > (q + 1)(q + 5) if q  1; 5 mod 6; > 12 (q + 2)(q + 4) if q  2; 4 mod 6; L (q) = 1 2 (q + 3) if q  3 mod 6; 1 2 (q + 6q + 12) if q  0 mod 6: 6 5 4 Moreover, from Table 1, we have E (t) = 8t + 2t + 2t . Thus, (E (S)L )(q) = 8L (q 6) + 2L (q 5) + 2L (q 4); A A A A quasi which coincides with the computation of  (q) above, and justifies Theorem 4.11. dsc (w) w 2 s s s ; (s s ) 2 1 2 1 2 1 1; s 2 2 2 3 0 s ; s s ; s s ; s s s ; (s s ) ; s (s s ) ; s (s s ) ; (s s ) 2 1 2 2 1 1 2 1 2 1 1 2 1 2 1 2 2 1 TABLE 1. Computation of dsc (W ) when =  (G )nf3 + 2 g. Here s ; s 2 1 2 1 2 are the reflections w.r.t. ; , respectively. 1 2 14 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA 0 0 (b) Let = f ; + ; 3 + 2 g. Then is not compatible by Example 4.9(b). We may 1 1 2 1 2 compute (q 1)(q 2) if q  1 mod 2; quasi (q) = q 3q + 3 if q  0 mod 2: 6 5 4 3 2 E (t) = 5t + t + 2t + 3t + t ; >(q 1)(q 2) if q  1; 5 mod 6; q 3q + 3 if q  2; 4 mod 6; (E 0 (S)L )(q) = q 3q + 4 if q  3 mod 6; q 3q + 5 if q  0 mod 6: quasi quasi Thus  (q) = (E 0 (S)L )(q) if q  1; 2; 4; 5 mod 6, and  (q) < (E 0 (S)L )(q) 0  0 A A quasi otherwise. In particular, we can easily check on Figure 2 that  (7) = (E (S)L )(7) = (c) In addition to the example above, both sides of the formula in Theorem 4.11 have different values at each constituent when is not compatible. For example, let = f + g as one 1 1 2 of the non-compatible singleton subsets mentioned in Example 4.9(b). Then quasi (q) = q(q 1) for all q; 6 5 4 3 2 E (t) = t + 2t + 3t + 3t + 2t + t; (E (S)L )(q) = q(q 1) + 1 for all q: 1 A quasi Thus  (q) < (E (S)L )(q) for all q. 1 A (d) Let = f ; 3 + g. Then is compatible (e.g., checked by Figure 2) but not an ideal of 2 2 1 2 2 any positive system of  = G . Otherwise, must be the associated base w.r.t. that positive 2 2 system since it has only two elements. However, it is a contradiction because the length of and 3 + are equal. 1 2 Example 4.19. Let  = A and let = [f ~g = f ; ; ; ; + + + g. We may 4 1 2 3 4 1 2 3 4 compute quasi 4 3 2 (q) = q 5q + 10q 10q + 4 for all q; 5 4 3 2 E (t) = 5t + 9t + 9t + t ; 4 3 2 (E (S)L )(q) = q 5q + 11q 12q + 5 for all q: quasi Thus,  (q) < (E (S)L )(q) for all q 6= 1. It follows from Theorem 4.11 that is not compatible. In general, it is not easy to compute the A-Eulerian polynomial of an arbitrary (compatible) subset by using its primary definition (Definition 4.2). For some special cases, e.g., the subsets obtained from the positive system by removing one element as mentioned in Example 4.9(a), we can compute the corresponding Eulerian polynomial in terms of the root system invariants. Let  (resp.,  ) denote the set of all long (resp., short) roots of . We will use the notation l s when we refer to either  or  . If  is simply-laced, we agree that  = ;  = ;. Denote l;s l s l s e e :=  \  and  := \  . l;s l;s l;s l;s EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 15 + c + Proposition 4.20. Let  2  and set  :=  nfg. Define l;s := fw 2 W j  = w( ) for some 2  g: i i l;s Then (i) = fw 2 W j dsc c (w) = c for some 2  g = fw 2 W j dsc c (w) > 0g. i i l;s (ii) , where := fw 2 W j  = w( )g for each 2  . e i i i i l;s i l;s (iii) ' stab() (as sets), where stab() := fw 2 W j  = w()g is the stablizer of  in W . In #W particular, # = . l;s # #W l;s (iv) # = . l;s Proof. (i) and (ii) are trivial, and (iv) is an immediate consequence of (ii) and (iii). Let us prove (iii). Fix w 2 W so that w () = . Thus, if w 2 , then ws w 2 stab(). ! ! i i ! i i i i Consider the set map ! stab(); defined by w 7! ws w : i ! i i Clearly,  is well-defined and injective. Moreover, for any  2 stab(), we have w = w s 2 ! i such that (w) =  . Therefore  is surjective and hence bijective. To prove the second state- #W ment, it suffices to show that #stab() = . This can be done by proving that the surjective l;s map ' : W !  defined by w 7! w() induces a bijection W=stab() !  . l;s l;s Theorem 4.21. If  2  , then l;s #W #W (# # ) l;s l;s hc h E (t) = t + t : f  # f  # l;s l;s i l;s Proof. It follows directly from Definition 4.2 and Proposition 4.20. An example of Theorem 4.21 is already mentioned in Example 4.18(a). In addition, if  is any + 6 3 short root in  (G ), then Theorem 4.21 implies that E c (t) = 10t + 2t . Next, we discuss the generating function of the characteristic quasi-polynomial. Proposition 4.22. L (q)t = : A Q (1 t ) q1 i=0 Proof. See, e.g., [KTT10, Proof of Theorem 3.1]. Theorem 4.23. (i) #W quasi q (q)t = : (1 t ) q1 i=0 (ii) Let R (t) be the generalized Eulerian polynomial (see Remark 4.3(b)). Then R (t) ` q q t = : (1 t ) q1 i=0 Proof. For a proof of (i), see, e.g., [Ath96], [BS98, Theorem 4.1], [KTT10, Theorem 3.1]. (ii) follows from [LP18, Theorem 10.1].  16 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA Theorem 4.24. A subset   is compatible if and only if E (t) quasi (4.6)  (q)t = : (1 t ) q1 i=0 Proof. Using Corollary 3.10, Propositions 4.22 and 3.7, we compute X X X q  q (E (S)L )(q)t = L  (q + dsc (A ))t A i q1 q1 i2[N ] X X dsc (A )  q+dsc (A ) i i = t L (q + dsc (A ))t i2[N ] q1 X X dsc (A ) n = t L (n )t A i i2[N ] n h hdsc (A ) E (t) i2[N ] = = : Q Q ` ` c c i i (1 t ) (1 t ) i=0 i=0 If   is compatible, then Formula (4.6) follows from Theorem 4.11 and the calculation above. Assume that Formula (4.6) holds but is not compatible. By Proof of Theorem 4.11, we can write quasi (q) = (E (S)L )(q) R(q); for some nonzero function R(q) (actually quasi-polynomial) in q. Thus, R(q)t = 0, which q1 is a contradiction. Remark 4.25. By Remark 4.3, Theorems 4.11 and 4.24, if (a) =  , then we recover Theorems 3.8 and 4.23(i). (b) = ;, then we recover [Yos18b, Theorem 4.8] and Theorem 4.23(ii). Corollary 4.26. If  2  , then l;s #W #W (# # ) l;s l;s quasi (q) = L (q h + c ) + L (q h); A A f  # f  # l;s l;s i l;s hc h i e #W t + #W (# # )t e l;s l;s quasi i l;s (q)t = : c Q f  #  (1 t ) l;s i=0 q1 Proof. It follows from Theorems 4.11, 4.21 and 4.24. 5. DEFORMATIONS OF W EYL SUBARRANGEM ENTS + c + Let   , and recall the notation =  n . Let a  b be integers, and denote [a; b] := fm 2 Z j a  m  bg. Also, if b  1, then write [b] instead of [1; b]. Definition 5.1. Let a  b, c  d be integers. Define two types of the deformed Weyl arrangements of as follows: [a;b] (Type I) I := fH j 2 ; m 2 [a; b]g. ;m [a;b];[c;d] [a;b] [c;d] (Type II) II := I t I . c c ; EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 17 [a;b];[c;d] [c;d];[a;b] Remark 5.2. (a) There is an obvious duality II = II . We can list some special- c c ; ; [a;b] [a;b];[c;d] [c;d] izations: I = ? the empty arrangement, II = A the truncated affine Weyl ar- + + ; ;; rangement, including the extended Shi, Catalan, Linial arrangements, see, e.g., [SP00, x9]. In [a;b] [a;b];[a;b] [a;b] addition, I = II = A . We refer the reader to [Ath99], [Ath04], [Yos18b] and + c + [a;b] [Yos18a] for more details on the characteristic (quasi-)polynomials ofA . (b) The deformed Weyl arrangements of an arbitrary are less well-known. When  is of type A, the deleted (or graphical) Shi arrangement, see, e.g., [Ath96,x3] or [AR12], is the product [0;1];[0;0] (e.g., [OT92, Definition 2.13]) of the 1-dimensional empty arrangenment and II . Definition 5.3. For w 2 W , define dsc (w) := c ; 0i`; w( )2 asc (w) := c ; 0i`; w( )2 asc (w) := c : 0i`; w( )2 Obviously, asc (w) + asc (w) + dsc (w) + dsc (w) = h for all w 2 W . Similar to Lemma 4.4, each function defined above takes values in [0; h 1]. Furthermore, asc (w) = asc + (w) = asc (w) + asc (w); dsc (w) = dsc + (w) = dsc (w) + dsc (w): Similar to Definition 4.6, we can extend the functions above to functions on the set of all alcoves. Now let us formulate a deformed version of Proposition 3.7. Set [a ;b ] i i F := H (1  i  `); ;m i i m2[a ;b ] i i [a ;b ] 0 0 F := H : ;q+m 0 0 m2[a ;b ] 0 0 Proposition 5.4. Let fi ; : : : ; i g  [0; `], and let b  0 for all 1  j  k. Suppose that 1 k i q > (b + 1)c . Then i i j2[k] j j 0 1 [ X [0;b ] _ j @ A # qA \ Z ( )n F = L (q (b + 1)c ): i i i A j j j2[k] j2[k] Proof. The formula was implicitly used in [Yos18b, x5] and its proof is very similar to the non- deformed case. Note that if i 2 [`], then 8 9 ( ; x)  b + 1 < = i i [0;b ] _ _ # qA \ Z ( )n F = # x 2 Z ( ) ( ; x)  0 (j 2 [`]nfig) : ; ( ; x)  q = L (q (b + 1)c ): i i Here the last equality follows from the bijection x 7! x + (b + 1)$ . The proof for i = 0 is similar. Then apply the formula above repeatedly.  18 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA Remark 5.5. If we replace the interval [0; b ] in Proposition 5.4 by [a; b ] with a  1, there might i i j j be a large change in the right-hand side of the formula above. For example, if 1  a  b , then 1 1 [a ;b ] _ 1 1 # qA \ Z ( )n F = L  (q (b + 1)c ) + L  (q) L  (q a c ): 1 1 1 1 1 A A A Theorem 5.6. Let be a compatible subset of  . (i) If a; b  0, then quasi (q) = L (q (b + 1)asc (A ) asc (A ) (a + 1)dsc (A )): [a;b] A i i i i2[N ] (ii) If b  1, then quasi (q) = L (q (b + 1)asc (A ) asc (A )): [1;b] A i i i2[N ] Proof. Proofs of (i) and (ii) are similar, and both are similar in spirit to the proof of [Yos18b, Theorem 5.1]. See also Proof of Theorem 4.11 in this paper. First, we give a proof for (i). Since both sides are quasi-polynomials, it is sufficient to prove the equality for q  0 (actually, q > (a + b + 3)h is sufficient). By Proposition 5.4, for i 2 [N ], 0 1 } _ @ A # qA \ Z ( )n H ;kq+m 2 ;k2Z;m2[a;b] 8 9 ( ; x)  qk + b + 1 ( 2 I \ ) > > > > < = (; x) > qk ( 2 I \ ) = # x 2 Z ( ) ( ; x)  qk a 1 ( 2 J \ ) > > > > : ; (; x)  qk ( 2 J \ ) = L (q (b + 1)asc (A ) asc (A ) (a + 1)dsc (A )): A i i i By Proposition 3.11, we have quasi (q) = L (q (b + 1)asc (A ) asc (A ) (a + 1)dsc (A )): [a;b] A i i i i2[N ] For (ii), note that for i 2 [N ], 0 1 } _ @ A # qA \ Z ( )n H ;kq+m 2 ;k2Z;m2[1;b] 8 9 > ( ; x)  qk + b + 1 ( 2 I \ ) > > > < = (; x) > qk ( 2 I \ ) = # x 2 Z ( ) ( ; x)  qk ( 2 J \ ) > > > > : ; (; x)  qk ( 2 J \ ) = L (q (b + 1)asc (A ) asc (A )): A i i Remark 5.7. Theorem 5.6 is a generalization of several known results. (a) When =  , we obtain [Ath04, Theorem 1.2] (a = b  0), and Theorem 5.1 (a = b 1 0), Theorem 5.2 (b  1), Theorem 5.3 (b = n + k, a = k 1, n; k  1) in [Yos18b]. When = ;, we obtain [Yos18b, Theorem 4.8]. EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 19 (b) When a = b = 0 (and   is compatible), we obtain Theorem 4.11. By using the same method, we have the following result for the arrangements of type II. Theorem 5.8. Let be a compatible subset of  . (i) If a; b; c; d  0, then quasi (q) = L (q (b + 1)asc (A ) (d + 1)asc (A ) [a;b];[c;d] A i i II i2[N ] (a + 1)dsc (A ) (c + 1)dsc (A )): i i (ii) If a; b  0; d  1, then quasi (q) = L (q (b + 1)asc (A ) (d + 1)asc (A ) [a;b];[1;d] i i II i2[N ] (a + 1)dsc (A )): (iii) If b; d  1, then quasi (q) = L (q (b + 1)asc (A ) (d + 1)asc (A )): [1;b];[1;d] A i i II i2[N ] Remark 5.9. (a) One can work with other intervals [a; b] but the computation may become more complicated (see Remark 5.5). F F n [a ;b ] n k k + (b) One can define and study the arrangement I where  = with n  3. k=1 k=1 See, e.g., [Ath96, Theorem 3.11] for an example when n = 3. We choose not to develop this direction here. It would be interesting to characterize the compatibility in the case of type A (e.g., in terms of graphs) and compare the following result with [AR12, Theorem 3.2] and [Ath96, Theorem 3.9] (see Remark 5.2 for the notation). h+asc (w) Corollary 5.10. Define M (t) := t . If is compatible, then w2W quasi (q) = (M (S)L )(q): [0;1];[0;0] II Acknowledgements. The first author was partially supported by Mitacs Canada Globalink Re- search Award to visit Japan and carry out the collaboration. The second author is partially sup- ported by JSPS Research Fellowship for Young Scientists Grant Number 19J12024. The third author is partially supported by JSPS KAKENHI Grant Number JP18H01115. The second author would like to thank Akiyoshi Tsuchiya for pointing out an error in the proof of the main result of the manuscript (Theorem 4.11) in a previous version. R EFERENCES [AR12] D. Armstrong and B. Rhoades. The Shi arrangement and the Ish arrangement. Trans. Amer. Math. Soc., 364(3):1509–1528, 2012. [Ath96] C. A. Athanasiadis. Characteristic polynomials of subspace arrangements and finite fields. Adv. Math., 122(2):193–233, 1996. [Ath99] C. A. Athanasiadis. Extended Linial hyperplane arrangements for root systems and a conjecture of Postnikov and Stanley. J. Algebr. Comb., 10:207–225, 1999. [Ath04] C. A. Athanasiadis. Generalized Catalan numbers, Weyl groups and arrangements of hyperplanes. Bull. Lond. Math. Soc., 36:294–302, 2004. 20 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA [BM14] P. Brand ¨ en ´ and L. Moci. The multivariate arithmetic Tutte polynomial. Trans. Amer. Math. Soc., 366(10):5523–5540, 2014. [BS98] A. Blass and B. Sagan. Characteristic and Ehrhart polynomials. J. Algebr. Comb., 7(2):115–126, 1998. [BV07] A. Bhattacharya and G. R. Vijayakumar. An integrality theorem of root systems. European J. Combin., 28(6):1854–1862, 2007. [BZ06] M. Beck and T. Zaslavsky. Inside-out polytopes. Adv. Math., 205(1):134–162, 2006. [CW12] B. Chen and S. Wang. Comparison on the coefficients of characteristic quasi-polynomials of integral ar- rangements. J. Combin. Theory Ser. A, 119:271–281, 2012. [DPS09] K. Dilks, T. K. Petersen, and R. Stembridge. Affine descents and the Steinberg torus. Adv. Appl. Math., 42:423–444, 2009. [Hum90] J. E. Humphreys. Reflection Groups and Coxeter Groups. Cambridge University Press, 1990. [KTT08] H. Kamiya, A. Takemura, and H. Terao. Periodicity of hyperplane arrangements with integral coefficients modulo positive integers. J. Algebr. Comb., 27(3):317–330, 2008. [KTT10] H. Kamiya, A. Takemura, and H. Terao. The characteristic quasi-polynomials of the arrangements of root systems and mid-hyperplane arrangements. Arrangements, local systems and singularities, pages 177–190, Progr. Math., 283, Birkhauser ¨ Verlag, Basel, 2010. [KTT11] H. Kamiya, A. Takemura, and H. Terao. Periodicity of non-central integral arrangements modulo positive integers. Ann. Comb., 15(3):449–464, 2011. [LP18] T. Lam and A. Postnikov. Alcoved polytopes II. In: Kac V., Popov V. (eds) Lie Groups, Geometry, and Representation Theory, pages 253–272, Progr. Math., 326, Birkhauser ¨ Verlag, Cham, 2018. [LTY] Y. Liu, T. N. Tran, and M. Yoshinaga. G-Tutte polynomials and abelian Lie group arrangements. Int. Math. Res. Not. IMRN, to appear. [Nor12] C. Norman. Finitely Generated Abelian Groups and Similarity of Matrices over a Field. Springer Under- graduate Mathematics Series, Springer-Verlag London, 2012. [OT92] P. Orlik and H. Terao. Arrangements of hyperplanes. Grundlehren der Mathematischen Wissenschaften 300, Springer-Verlag, Berlin, 1992. [Som97] E. Sommers. A family of affine Weyl group representations. Transform. Groups, 2(4):375–390, 1997. [SP00] R.P. Stanley and A. Postnikov. Deformations of Coxeter hyperplane arrangements. J. Combin. Theory Ser. A, 91:544–597, 2000. [Sut98] R. Suter. The number of lattice points in alcoves and the exponents of the finite Weyl groups. Math. Comp., 67(222):751–758, 1998. [Tra18] T. N. Tran. An equivalent formulation of chromatic quasi-polynomials. arXiv preprint, 2018. https:// arxiv.org/abs/1803.08649. [Tra19] T. N. Tran. Characteristic quasi-polynomials of ideals and signed graphs of classical root systems. European J. Combin., 79:179–192, 2019. [TY19] T. N. Tran and M. Yoshinaga. Combinatorics of certain abelian Lie group arrangements and chromatic quasi-polynomials. J. Combin. Theory Ser. A, 165:258–272, 2019. [Yos18a] M. Yoshinaga. Characteristic polynomials of Linial arrangements for exceptional root systems. J. Combin. Theory Ser. A, 157:267–286, 2018. [Yos18b] M. Yoshinaga. Worpitzky partitions for root systems and characteristic quasi-polynomials. Tohoku Math. J., 70(1):39–63, 2018. AHMED UMER A SHRAF, U NIVERSITY OF W ESTERN ONTARIO, C ANADA. E-mail address: aashra9@uwo.ca TAN NHAT TRAN, D EPARTMENT OF MATHEM ATICS, H OKKAIDO UNIVERSITY, K ITA 10, N ISHI 8, K ITA-K U, SAPPORO 060-0810, JAPAN. E-mail address: trannhattan@math.sci.hokudai.ac.jp M ASAHIKO YOSHINAGA, DEPARTMENT OF M ATHEMATICS, HOKKAIDO UNIVERSITY, KITA 10, NISHI 8, KITA-K U, S APPORO 060-0810, JAPAN. E-mail address: yoshinaga@math.sci.hokudai.ac.jp http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Mathematics arXiv (Cornell University)

Eulerian polynomials for subarrangements of Weyl arrangements

Ashraf, Ahmed Umer; Tran, Tan Nhat; Yoshinaga, Masahiko
Mathematics , Volume 2020 (1911) – Nov 5, 2019
Download PDF
20 pages

Loading next page...
 
/lp/arxiv-cornell-university/eulerian-polynomials-for-subarrangements-of-weyl-arrangements-SqK6Ck28Qp
ISSN
0196-8858
eISSN
ARCH-3343
DOI
10.1016/j.aam.2020.102064
Publisher site
See Article on Publisher Site

Abstract

EULERIAN POLYNOMIALS FOR SUBARRANGEMENTS OF WEYL ARRANGEMENTS AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA ABSTRACT. Let A be a Weyl arrangement. We introduce and study the notion of A-Eulerian polynomial producing an Eulerian-like polynomial for any subarrangement of A. This polyno- mial together with shift operator describe how the characteristic quasi-polynomial of a new class of arrangements containing ideal subarrangements of A can be expressed in terms of the Ehrhart quasi-polynomial of the fundamental alcove. The method can also be extended to define two types of deformed Weyl subarrangements containing the families of the extended Shi, Catalan, Linial arrangements and to compute their characteristic quasi-polynomials. We obtain several known re- sults in the literature as specializations, including the formula of the characteristic polynomial of A via Ehrhart theory due to Athanasiadis (1996), Blass-Sagan (1998), Suter (1998) and Kamiya- Takemura-Terao (2010); and the formula relating the number of coweight lattice points in the fun- damental parallelepiped with the Lam-Postnikov Eulerian polynomial due to the third author. 1. I NTRODUCTION Motivation. One of typical problems in enumerative combinatorics is to count the sizes of sets depending upon a positive integer q. This often gives rise to polynomials. For instance, the chromatic polynomial of an undirected graph, going back to Birkhoff and Whitney, encodes the number of ways of coloring the vertices with q colors so that adjacent vertices get different colors. However, it may happen that enumerating the cardinalities of sets leads to quasi-polynomials. Generally speaking, a quasi-polynomial is a refinement of polynomials, of which the coefficients may not come from a ring but instead are periodic functions with integral periods. Thus a quasi- polynomial is made of a bunch of polynomials, the constituents of the quasi-polynomial. One of the most classical examples in the theory is that the number of integral points in the q-fold dilation of a rational polytope agrees with a quasi-polynomial in q, broadly known as the Ehrhart quasi-polynomial. We are interested in the connection between the counting problems and the arrangement the- ory. A finite list (multiset) A of vectors in Z determines an arrangement A(R) of hyperplanes ` 1 1 ` in the vector space R , an arrangement A(S ) of subtori in the torus (S ) , and especially an arrangement A(Z ) of subgroups in the finite abelian group Z . Enumerating the cardinality of the complement of A(Z ) produces a quasi-polynomial, the characteristic quasi-polynomial quasi (q) of A [KTT08]. This single quasi-polynomial encodes a number of combinatorial and topological information of several types of arrangements and has generated increasing interest re- quasi cently (e.g., [CW12, BM14, Yos18a, Yos18b, TY19, Tra19]). Among the others,  (q) has the first constituent identical with the characteristic polynomial  (t) of A(R) which justified its A(R) name (e.g., [Ath96, KTT08]), and the last constituent identical with the characteristic polynomial Date: May 1, 2020. 2010 Mathematics Subject Classification. Primary 52C35, Secondary 17B22. Key words and phrases. Characteristic quasi-polynomial, Eulerian polynomial, Ehrhart quasi-polynomial, Wor- pitzky partition, root system, ideal subarrangement. arXiv:1911.01650v3 [math.CO] 30 Apr 2020 2 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA quasi (t) of A(S ) [LTY, TY19]. One of the methods used in [KTT08] for showing that  (q) A(S ) is indeed a quasi-polynomial is to express it as a sum of the Ehrhart quasi-polynomials of rational polytopes, or in the sense of [BZ06], as the Ehrhart quasi-polynomial of an “inside-out” polytope. Such an expression is certainly interesting as it reveals the connection between two seemingly un- related quasi-polynomials, one would hope for a more explicit expression if the list A was chosen to be a more special vector configuration. Objective. A particularly well-behaved class of the hyperplane arrangements is that of Weyl arrangements. More precisely, if A =  is a positive system of an irreducible root system , quasi thenA(R) is called the Weyl arrangement ofA. It is proved that  (q) is expressed in terms of the Ehrhart quasi-polynomial L (q) of the fundamental alcove A , the Weyl group, and the index of connection of  (e.g., [Ath96, BS98, Sut98, KTT10]). Thus we arrive at a natural and essential quasi problem that for which subset   ,  (q) can be computed by using the fundamental invariants of , and more importantly, by means of the Ehrhart quasi-polynomials. Some partial quasi results are known. If the root system  is of classical type and is an ideal of  , then  (q) can be computed from information of the signed graph associated with [Tra19]. A result due to quasi [Yos18b] applied to any root system, asserts that  (q) (or simply q ) can be written in terms of the Lam-Postnikov Eulerian polynomial [LP18], shift operator, and L (q). Results. Inspired by the works of [LP18] and [Yos18b], we introduce the notion of A-Eulerian polynomial E (t) - an arrangement theoretical generalization of the classical Eulerian polynomial, and the notion of compatible subsets of  w.r.t. the Worpitzky partition. The first main result quasi in our paper is a formula of  (q) in terms of E (t), shift operator, and L (q) when is compatible. The formula specializes correctly to the two formulas in the extreme cases ( =  and = ;) mentioned above. In addition, we prove that the formula characterizes the compatibility. The second main result in our paper is that the class of compatible subsets contains the ideals of the root system. Using the similar method, we further define two types of deformed Weyl subarrangements containing the families of the extended Shi, Catalan, Linial arrangements and compute their characteristic quasi-polynomials. Organization of the paper. The remainder of the paper is organized as follows. In Section 2, we recall definitions and basic facts of irreducible root systems, their (affine) Weyl groups and the Worpitzky partition. In Section 3, we recall the definitions of the characteristic and Ehrhart quasi-polynomials and specify the choices of lattices for these quasi-polynomials (Remarks 3.4 and 3.6). We also recall the formula between the quasi-polynomials in the extreme case = (Theorem 3.8), and derive a more general formula (Proposition 3.11). In Section 4, we intro- duce the notion of A-Eulerian polynomial (Definition 4.2), of which the main specialization is the Lam-Postnikov Eulerian polynomial (Remark 4.3). We then define the notion of compatible sets (Definition 4.8) which interpolates between the two extreme cases =  and = ;. We quasi prove that the A-Eulerian polynomial is an essential tool to compute  (q) and its generating function for any compatible subset   (Theorems 4.11 and 4.24). We also prove that every ideal is compatible (Theorem 4.16). In Section 5, we define two types of the deformed Weyl subar- rangements (Definition 5.1), of which the main examples are the truncated affine Weyl and deleted Shi arrangements (Remark 5.2). Then we compute the characteristic quasi-polynomials of these deformed arrangements according to two special choices of intervals (Theorems 5.6 and 5.8). 2. ROOT SYSTEMS AND WORPITZKY PARTITION Our standard reference for root systems and their Weyl groups is [Hum90]. Assume that V = R with the standard inner product (;). Let  be an irreducible (crystallographic) root system in V EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 3 with the Coxeter number h and the Weyl group W . Fix a positive system    and let  := + + f ; : : : ; g be the set of simple roots (base) of  associated with  . For = d 2  , 1 ` i i i=1 the height of is defined by ht( ) := d . Define the partial order on  such that i 1 2 i=1 if and only if = n with all n 2 Z . The highest root (w.r.t. ), denoted by 1 2 i i i 0 i=1 ~ 2  , can be written uniquely as a linear combination of the simple roots ~ = c i i i=1 (c 2 Z ). Set := ~, c := 1, and  := [f g. Then we have the linear relation i >0 0 0 0 c + c + + c = 0: 0 0 1 1 ` ` The coefficients c are important in our study and will appear frequently throughout the paper. _ _ For 2 V nf0g, denote := . The root lattice Q(), coroot lattice Q( ), weight lattice ( ; ) Z (), and coweight lattice Z ( ) are defined as follows: Q() := Z ; i=1 _ _ Q( ) := Z ; i=1 Z () := fx 2 V j ( ; x) 2 Z (1  i  `)g; Z ( ) := fx 2 V j ( ; x) 2 Z (1  i  `)g: _ _ Then Q() is a subgroup of Z () of finite index f , and similarly Q( ) is a subgroup of Z ( ) of _ _ _ the index f . The number f is called the index of connection. Let f$ ; : : : ; $ g  Z ( ) be the 1 ` _ _ _ _ dual basis of the base , namely, ( ; $ ) =  . Then Z ( ) = Z$ and c = ($ ; ~). i ij i j i=1 i i For m 2 Z and 2 , the affine hyperplane H is defined by ;m H := fx 2 V j ( ; x) = mg: ;m A connected component of V n H is called an alcove. The fundamental alcove A ;m 2 ;m2Z is defined by ( ; x) > 0 (1  i  `); A := x 2 V : ( ; x) > 1 The closure A = fx 2 V j ( ; x)  0 (1  i  `); ( ; x)  1g is a simplex, which is i 0 _ _ the convex hull of f0; $ =c ; : : : ; $ =c g. The supporting hyperplanes of the facets of A are 1 ` H ; : : : ; H ; H . The affine Weyl group W := W n Q( ) acts simply transitively on ;0 ;0 ;1 a 1 0 the set of alcoves and admits A as a fundamental domain for its action on V . } _ The fundamental domain P of the coweight lattice Z ( ), called the fundamental paral- lelepiped, is defined by } _ P := (0; 1]$ i=1 = fx 2 V j 0 < ( ; x)  1 (1  i  `)g: #W Let N := , and denote [N ] := f1; 2; : : : ; Ng. Then the cardinality of the set  of all alcoves contained in P equals N (see, e.g., [Hum90, Theorem 4.9]). Let us write = fA  P j i 2 [N ]g; i 4 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA where each alcove A is written uniquely as ( ; x) > k ( 2 I ); A = x 2 V ; ( ; x) < k ( 2 J ) where k = k (A ) 2 Z , k = k (A ) 2 Z and the sets I; J   with #(I t J ) = ` + 1 0 >0 i i indicate the constraints on x 2 V according to the inequality symbols >, <, respectively. Definition 2.1. For each A 2 , the partial closure A of A is defined by i i i ( ; x) > k ( 2 I ); A := x 2 V : ( ; x)  k ( 2 J ) Theorem 2.2 (Worpitzky partition). } } P = A : i2[N ] Proof. See, e.g., [Yos18b, Proposition 2.5], [Hum90, Exercise 4.3]. 3. C HARACTERISTIC AND E HRHART QUASI- POLYNOMIALS A function g : Z ! C is called a quasi-polynomial if there exist  2 Z and polynomials >0 f (t) 2 Z[t] (1  k  ) such that for any q 2 Z with q  k mod , >0 g(q) = f (q): The number  is called a period and the polynomial f (t) is called the k-constituent of the quasi- polynomial g. Let := Z ' Z be a lattice with a basis f ; : : : ; g. Let L be a finite list (multiset) i 1 ` i=1 of elements in . Let q 2 Z and denote Z := Z=qZ. For = a 2 L and m 2 Z, >0 q i i i=1 define a subset H of Z by ;m ;Z q q n o H := z 2 Z j a z  m : ;m ;Z i i q q i=1 Given a vector m = (m ) 2 Z , we can define the Z -plexification (or q-reduced) arrangement 2L q of (L; m) by (L; m)(Z ) := fH j 2 Lg: q ;m ;Z The complement of (L; m)(Z ) is defined by M((L; m); ;Z ) := Z n H : q ;m ;Z q q 2L For a sublist S  L, write d for the largest invariant factor of the torsion subgroup of the S;n finitely generated abelian group =hSi (see, e.g., [Nor12, Definition 3.8]). The LCM-period  of L is defined by := lcm(d j S  L): L S;n Theorem 3.1. #M((L; m); ;Z ) is a monic quasi-polynomial in q for which  is a period. The q L quasi-polynomial is called the characteristic quasi-polynomial of (L; m) w.r.t. the lattice , and quasi denoted by  (q). (L;m) Proof. See [KTT08, Theorem 2.4] and [KTT11, Theorem 3.1].  EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 5 We can also define the R-plexification (in fact, the real hyperplane arrangement) of (L; m) as follows: (L; m)(R) := fH j 2 Lg; ;m ;R n o where H := x 2 R j a x = m . For a real hyperplane arrangement A, denote by ;m ;R i i i=1 (t) the characteristic polynomial (e.g., [OT92, Definition 2.52]) ofA. quasi Theorem 3.2. The first constituent of  (q) coincides with the characteristic polynomial of (L;m) (L; m)(R), i.e., f (t) =  (t): (L;m)(R) (L;m) Proof. See, e.g., [KTT08, Theorem 2.5] and [KTT11, Remark 3.3]. Convention: When m = (0) the zero vector, we simply write L, H , H instead of (L; m), ;Z ;R H , H , respectively. ;0;Z ;0;R Remark 3.3. L(Z ) andL(R) are examples of a more abstract concept, the G-plexifications (G is an abelian group) introduced in [LTY] by viewing G = Z ;R, respectively (see also [TY19] for more information on their combinatorial properties). In addition, the characteristic quasi-polynomial can be generalized to chromatic quasi-polynomial by replacing the lattice by a finitely generated abelian group, see, e.g., [Tra18] for more details. Remark 3.4. Throughout the paper, for every   ( Q()), the characteristic quasi-polynomial quasi (q) is always defined w.r.t. the root lattice = Q(). For each   , define the Weyl arrangement of by A := fH j 2 g, where H = fx 2 V j ( ; x) = 0g is the hyperplane orthogonal to . It is not hard to see that H ' H (as vector spaces), thus we can viewA as the R-plexification of , i.e.,A = (R). ;R In standard terminology,A + is known with the name Weyl (or Coxeter) arrangement, and clearly A is a (Weyl) subarrangement ofA + . Remark 3.5. Sometimes, when we say “the” characteristic quasi-polynomial of (L; m)(Z ) or of quasi (L; m)(R), we mean the characteristic quasi-polynomial  (q) of (L; m). For example, the (L;m) quasi quasi characteristic quasi-polynomial  (q) of A is referred to as  (q). We will use this term later in Section 5 when we deal with deformed Weyl arrangements of . Let be a lattice. For a polytopeP with vertices in the rational vector space generated by , the Ehrhart quasi-polynomial L (q) ofP w.r.t. is defined by L (q) := #(qP \ ): We denote byP the relative interior ofP . Similarly, we can define L  (q) := #(qP \ ): For q > 0, the following reciprocity law holds: dimP L (q) = (1) L  (q): P P Remark 3.6. Throughout the paper, the Ehrhart quasi-polynomials L (q), L  (q) are defined w.r.t. the coweight lattice = Z ( ). Let F := H ; F := H (1  i  `) denote the supporting hyperplanes of the facets 0 ;q i ;0 0 i of qA . Then the number of coweight lattice points in qA after removing some facets can be computed by L with the scale factor of dilation being reduced. A 6 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA Proposition 3.7. Let fi ; : : : ; i g  f0; 1; : : : ; `g. Suppose that q > c + + c . Then 1 k i i #(qA \ Z ( )n (F [[ F )) = L (q (c + + c )): i i i i 1 k A 1 k In particular, for q 2 Z, L  (q) = L (q h): Proof. See [Yos18b, Corollaries 3.4 and 3.5]. In general, it is not easy to find explicit formulas which involve both characteristic and Ehrhart quasi-polynomials. With regards to root systems, there is an interesting relation between these quasi-polynomials. Theorem 3.8. #W quasi (q) = L (q): Proof. See, e.g., [KTT10], [Yos18b, Proposition 3.7]. quasi Theorem 3.9. The minimum period of  (q) is equal to lcm(c ; : : : ; c ). Furthermore, by a + 1 ` case-by-case argument, it can be checked that the minimum period coincides with the LCM-period + . Proof. See [KTT10, Corollary 3.2 and Remark 3.3]. quasi Corollary 3.10.  (q) > 0 (equivalently, L  (q) > 0) if and only if q  h. + A Proof. See, e.g., [KTT10, Corollary 3.4]. quasi We can extend Theorem 3.8 to a formula of  (q) for any   in terms of lattice point counting functions. In the proposition below, we view as a list with possible repetitions of elements. Along the proof, we will see why the choice of lattices for the characteristic and Ehrhart quasi-polynomials is important. Proposition 3.11. Let m = (m ) be a vector in Z . Set } _ X (q) := qP \ Z ( )n H ; ( ;m) ;kq+m 2 ;k2Z _ _ Y (q) := fx 2 Z ( )=qZ ( ) j ( ; x) 6 m mod q;8 2 g: ( ;m) We have bijections between sets X (q) ' Y (q) ' M(( ; m);Z ;Z ): ( ;m) ( ;m) q As a result, quasi (q) = #X (q) = #Y (q): ( ;m) ( ;m) ( ;m) + + Proof. The bijection X (q) ' Y (q) is proved in [Yos18b,x3.3]. We can use exactly the ( ;m) ( ;m) same argument applied to every subset . The proof of Y (q) ' M( ;Z ;Z ) for an arbitrary q EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 7 runs as follows: Y (q) = fx = z $ j ( ; x) 6 m mod q;8 2 g ( ;m) i i=1 ` ` ` X X X _ _ = fx = z $ j ( S ; z $ ) 6 m mod q; (1  j  # )g i ij i i i i i=1 i=1 i=1 ' fz = (z ; : : : ; z ) 2 Z j z S 6 m mod q; (1  j  # )g 1 ` i ij i=1 = M(( ; m);Z ;Z ): + ` Remark 3.12. The bijection X + (q) ' M( ;Z ;Z ) appeared (without proof) in [KTT10, quasi Proof of Theorem 3.1]. Theorem 3.8 is a special case of Proposition 3.11 because  (q) = #W #X + (q) = L  (q) [Yos18b,x3.3]. 4. EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEM ENTS + c + Let   , and set :=  n . Definition 4.1. The descent dsc w.r.t. is the function dsc : W ! Z defined by dsc (w) := c : 0i`; w( )2 Definition 4.2. The (arrangement theoretical Eulerian or) A-Eulerian polynomial of (or of the Weyl subarrangementA ) is defined by hdsc (w) E (t) := t : w2W #W + h + + Remark 4.3. (a) If =  , then dsc (w) = 0 for all w 2 W , and E (t) = t . (b) If = ;, then dsc = dsc = cdes, the circular descent statistic, e.g., [LP18, Definition 6.2], [Yos18b, Definition 4.1]. Then E (t) = R (t), the generalized Eulerian polynomial, e.g., [Yos18b, Definition 4.4]. In this paper, we often call it the Lam-Postnikov Eulerian polynomial. Note that if  is of type A , then R (t) = A (t), the classical `-th Eulerian polynomial [LP18, `  ` Theorem 10.1]. It should also be noted that dsc coincides with the notion of affine descents by Dilks-Petersen-Stembridge [DPS09,x2.5] only in type A case. Lemma 4.4. For all w 2 W , 0  dsc (w) < h. In particular, E (0) = 0. + + Proof. If =  , then the statements are clear by Remark 4.3(a). Assume that 6=  . If w( ) 2= for some 1  i  `, then dsc (w) < h. Otherwise, we have w( ) = i 0 + c c w( ) 2  . Thus w( ) 2= , and hence dsc (w) < h. i i 0 i=1 Lemma 4.5. (i) Let w 2 W . Suppose that w induces a permutation on  = f ; ; : : : ; g. 0 1 ` If w( ) = , then c = c . i p i p i i (ii) Let w ; w 2 W . If there exists 2 V such that w (A ) = w (A ) + , then dsc (w ) = 1 2 1 2 1 dsc (w ). 2 8 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA Proof. (i) is exactly [Yos18b, Lemma 4.3(1)]. (ii) is similar to [Yos18b, Lemma 4.3(2)]: since the supporting hyperplanes coincide, w ( ) = w ( ) for 0  i  `. By Definition 4.1 and (i), we 1 i 2 p have X X dsc (w ) = c = c = dsc (w ): 1 i p 2 c c 0i`; w ( )2 0p `; w ( )2 1 i i 2 p 0 0  _ Let A be an arbitrary alcove. We can write A = w(A ) + for some w 2 W and 2 Q( ). Recall the the Worpitzky partition from Theorem 2.2 that P = A . By Lemma 4.5, we i2[N ] can extend dsc to a function on the set of all alcoves (in particular, on the set  of alcoves A contained in P ) as follows: Definition 4.6. dsc (A ) := dsc (w): Theorem 4.7. hdsc (A ) E (t) = t : i2[N ] Proof. Similar to [Yos18b, Theorem 4.7]: there are exactly f elements of the group W := W n _  } Z ( ) that map A to a given alcove in P . In what follows, we shall sometimes abuse terminology and call a face of A that is contained } } in the partial closure A , a face of A . i i Definition 4.8. A subset   is said to be compatible (with the Worpitzky partition) if for } } each A  P the following condition holds: A \ H for 2 ; m 2 Z is either empty, or ;m i i } } } contained in A \ H for 2 ; m 2 Z so that A \ H is a facet of A . ;m ;m i i i In other words, is compatible if and only if the intersection of each A and H ;k i 2 ;k2Z consisting of a facets K ; : : : ; K , and b faces G ; : : : ; G of dimension  ` 2 of A , either 1 a 1 b is empty, or satisfies the condition that for every non-facet G there exists a facet K such that j s G  K . Intuitively, every “non-empty face intersection” of A and the affine hyperplanes w.r.t. j s roots in can be lifted to a “facet intersection”. + + + Example 4.9. (a) Clearly,; and  are compatible. Let  2  , then =  nfg is compatible. } } It is because if A \H is a non-empty face but not a facet of A for i 2 [N ], 2 ; m 2 ;m i i } } n + Z, then A \ H = \ H \ A , where H ( 2  ) are the supporting ;m ;m ;m j i j=1 j i j j j hyperplanes of the facets of A for all 1  j  n with n  2. Thus, there exists 1  k  n such that 6= , i.e., 2 . k k (b) If = f g with = d 2  n , then is not compatible because H \ 1 i i 1 ;ht( ) i=1 n o n o P P P ` ` ` } _ _ P = p $ j p d = ht( ); p 2 (0; 1] = $ . There exists a non- i i i i i=1 i i=1 i=1 i compatible set consisting of more than one elements. For example, let  = G as in Figure 1. Then it is easily seen from Figure 2 that = f ; + ; 3 + 2 g is not compatible 1 1 2 1 2 (regardless of the value of q) because there exist partially closed alcoves (in green) such that the affine hyperplanes (in pink) w.r.t. roots in intersect them only at their vertices. EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 9 FIGURE 1. Root system of type G . Definition 4.10. Let f : Z ! C be a function and let P (S) = a S be a polynomial in S. k=0 The shift operator via P (S) acting on f is defined by (P (S)f )(t) := a f (t k): k=0 Now let us prove the first main result in the paper. Theorem 4.11. A subset   is compatible if and only if for every positive integer q, quasi (q) = (E (S)L  )(q): Proof. The proof of “)” is similar in spirit to [Yos18b, Proof of Theorem 4.8], but requires a more careful analysis of each partially closed alcove in the Worpitzky partition (Theorem 2.2). Let us proceed the proof with as general as possible to see how the compatibility is crucial. Since both sides are quasi-polynomials, it suffices to prove the formula for q  h. In which case, quasi quasi (q)   (q) > 0 (Corollary 3.10), and q h + dsc (A )  0 (Lemma 4.4) for all i 2 [N ] whence (E (S)L )(q) > 0. Fix i 2 [N ]. The intersection of A and the affine hyperplanes w.r.t. A i roots in consists of a facets (0  a  ` + 1), say K ; : : : ; K , and b faces of dimension ` 2 1 a a b (b  0), say G ; : : : ; G of A . Set A := [ K and B := [ G . The following elementary 1 b s j i s=1 j=1 identity } } A n (A[ B) = (A n A)n (B n A) i i implies that } } _ _ _ qA \ Z ( )n H = q(A n A)\ Z ( ) n (q(B n A)\ Z ( )) : ;kq i i 2 ;k2Z 10 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA } _ F IGURE 2. The Worpitzky partition of qP \ Z ( ) in a type G root system (q = 7). } _ } DenoteR(A ; q) := # (q(B n A)\ Z ( )). In other words,R(A ; q) is the number of coweight i i lattice points in (the union of) the non-facets qG but not in any facet qK of qA . By Definition j s 2.1, we may write (4.1) 8 9 < ( ; x) > qk ( 2 I ) = } _ _ } # qA \ Z ( )n H = # x 2 Z ( ) ( ; x) < qk ( 2 J \ ) R(A ; q); ;kq i i : ; 2 ;k2Z (; x)  qk ( 2 J \ ) where each facet qK is supported by one of the hyperplanes of the form H . s ;qk Since A = w(A ) + for some w 2 W and 2 Z ( ) ( is uniquely is determined by w, e.g., [Hum90, Theorem 4.9]), the dilation qA can be written as (w( ); x) < q((w( ); ) + 1); 0 0 qA = x 2 V : (w( ); x) < q(w( ); ); (1  i  `) i i Thus the half-spaces defined by (; x)  qk ( 2 J \ ) correspond exactly to the roots 2 satisfyingw( ) 2 . Applying Proposition 3.7 and Definition 4.6, we obtain 8 9 < ( ; x) > qk ( 2 I ) = (4.2) # x 2 Z ( ) ( ; x) < qk ( 2 J \ ) = L (q h + dsc (A )): A i : ; (; x)  qk ( 2 J \ ) EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 11 Moreover, by the Worpitzky partition and Proposition 3.11, we have X [ quasi } _ (4.3)  (q) = # qA \ Z ( )n H : ;kq i2[N ] 2 ;k2Z Using Theorem 4.7 together with the shift operator (Definition 4.10), we have (4.4) (E (S)L )(q) = L (q h + dsc (A )): A A i i2[N ] Summarizing Formulas (4.1)-(4.4), it follows that for every   we have quasi (q) = (E (S)L )(q) R(A ; q): A i i2[N ] The compatibility of forcesR(A ; q) = 0 for all i 2 [N ]. This completes the proof of “)”. Now suppose that is not compatible. Then there exist A and its non-facet G such that G * K for all facets K . To prove “(”, it suffices to show that there exists (sufficiently large) j s s q 2 Z such that the relative interior of q(G n A) contains a coweight lattice point, in which >0 j case, R(A ; q) > 0. Since W preserves the lattice points of qG , we can transform the problem a j to (the relative interior of) a face of qA . Since a face of qA is the convex hull of a nonempty _ _ subset of f0; q$ =c ; : : : ; q$ =c g, the relative interior of that face contains a lattice point for 1 ` 1 ` sufficiently large q (e.g., the center of mass of any face is a lattice point when q = lcm(1; : : : ; ` + 1) lcm(c ; : : : ; c )). The conclusion follows. 1 ` quasi Remark 4.12. (a) When is not compatible, although  (q) and (E (S)L )(q) are never equal as quasi-polynomials, the equality may occur for particular constituents (see Example 4.18(b)). Proof of Theorem 4.11 hints that in order to find a formula of the characteristic quasi- polynomial of a non-compatible subset, we need to count the lattices points in d-simplices with d  `. (b) When the rank of the root system is at most 3, it is possible to examine the (non-)compatibility by using the pictures. The drawing is rather difficult in higher-dimensional cases, in this regard, Theorem 4.11 is useful to check the (non-)compatibility (see Example 4.19). + + + Definition 4.13. A subset   is an ideal of  (or a lower set of the root poset ( ;)) if for ; 2  ,  ; 2 implies 2 . 1 2 1 2 1 2 Our second main result in the paper is: every ideal is compatible. We remark that if `  2 and + + 2  , then the compatible subset  nfg is not an ideal unless  is the highest root. We recall some basic facts on root systems. For any M  , set  (F) := \ FM where F = Z;Q or R. Note that  (Z)   (Q) = M M M (R) and these root subsystems (in the real vector space that they span) have the same rank. In particular, if M  , then these three subsystems are identical. Lemma 4.14. LetJ ( . ThenJ is a base for  (Z). Proof. See, e.g., [Som97, Lemma 4.3]. Lemma 4.15. For w 2 W and J ( , define D := w(J ) = fw( ) j 2 Jg  . Then D is a base for  (Z). D 12 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA Proof. D is linearly independent because (w( );w( )) = ( ; ) for all ; 2 J . Let j k j k j k 2  (Z). We can write = d  where d 2 Z. We need to show that all d  0 or all 2D d  0. Note that J and D are isomorphic via w. Rewriting, we have = w( d ). w( ) 2J Thus, d 2  (Z). Lemma 4.14 completes the proof. w( ) J 2J Theorem 4.16. If   is an ideal, then is compatible. } } Proof. If A \ H is a non-empty face but not a facet of A for i 2 [N ], 2 ; m 2 Z, then ;m i i } } n + we can write A \H = \ H \A , where H ( 2  nf g) are the supporting ;m ;m ;m j i j=1 j i j j j hyperplanes of the facets of A for all 1  j  n with 2  n  `. Set D := f j 1  j  ng. We need to show that there is 2 \ D. We have the following facts. Fact 1. D is a base for  (Z). Let w 2 W so that A = w(A ) + ( 2 Z ( )). It follows D w w from the definition of A that there existsJ ( , such that D = w(J ). The rest follows from Lemma 4.15. Fact 2. 2  (Q). Choose z 2 A \ H , the translation t : V ! V via t (x) = z + x D 0 ;m z z 0 i o o yieldsz + anehull(A \ H ) = \ H  H . Thus, 2  (R) =  (Q). 0 ;m ;0 ;0 D D i j=1 j Claim. 2  (Z). By Fact 2, D [ f g is linearly dependent. Then there exist integers d ; d ; : : : ; d 2 Z not all zero with d 6= 0, gcdfjd j;jd j; : : : ;jd jg = 1 such that d + 1 n 1 n d = 0 (see, e.g., [BV07, Lemma 17]). The claim is proved once we prove d 2 f1; 1g. j=1 j 0 0 Let us transform the situation in A to A . Set := w( ) with 2 . We can rewrite the relation above as (4.5) d + d = 0; 2J where d := d with 2 D such that = w( ). We consider two cases. 0 0 Case 1. J ( . Since 2 Z, the linearly independence of  and Relation (4.5) yield 2 ZJ . Additionally, d divides d for all 2 J . Thus, d 2 f1; 1g. Case 2. J = I [f g with ; =6 I ( . If d = 0, then the same argument as in Case 1 yields 0 + d 2 f1; 1g. Assume that d 6= 0 and 2  . The transformation by w implies that A \ H = \ H \ H \ A  P : ;1 2I ;0 ;1 j j 0 Let x = r $ 2 P (r 2 [0; 1]) be a typical element in both sets. Since x 2 H for ;0 s s j 2 s all 2 I , we can write x = r $ . Thus, 2nI s s (x; d ) = (x;d ) = d : j 0 0 0 2J It follows from Relation (4.5) and the above equation that d = (x; d ) = (x; d ) = d : j 0 2J Thus we can rewrite Relation (4.5) as d ( + ) + d = 0: 0 j 2I 0 1 0 0 Note that + = w ( + ) 6= 0 where := w( ) 2 D. A similar argument as in Case 0 0 0 0 + 1 yields d 2 f1; 1g. Repeat for if 2  . EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 13 By Fact 1 and Claim, we can write = p with at least two p  1, p  1. Thus, j j k k 1 2 j=1 there is 2 D such that  . Since is an ideal, we must have 2 . This completes k  k k the proof. Remark 4.17. There exists a compatible subset that is not an ideal w.r.t. any positive system of quasi (see Example 4.18(d)). The characteristic quasi-polynomial  (q) when is an ideal and is of classical type has been computed by using information of the signed graph associated with [Tra19]. It would be interesting to compare the mentioned computation with Theorem 4.11. Example 4.18. Let  = G as in Figure 1. (a) Let =  nf ~g. We can see that is compatible either by Example 4.9(a) or by Theorem 4.16. By definition of characteristic quasi-polynomial (Theorem 3.1), quasi 2 (q) = # z 2 Z j z ; z ; z + z ; 2z + z ; 3z + z 6= 0 1 2 1 2 1 2 1 2 (q 1)(q 4) if q  1; 5 mod 6; (q 2)(q 3) if q  2; 3; 4 mod 6; q 5q + 8 if q  0 mod 6: Note that the Ehrhart quasi-polynomial of the fundamental alcove (w.r.t. the coweight lattice) of every root system has been completely computed, e.g., in [Sut98]. > (q + 1)(q + 5) if q  1; 5 mod 6; > 12 (q + 2)(q + 4) if q  2; 4 mod 6; L (q) = 1 2 (q + 3) if q  3 mod 6; 1 2 (q + 6q + 12) if q  0 mod 6: 6 5 4 Moreover, from Table 1, we have E (t) = 8t + 2t + 2t . Thus, (E (S)L )(q) = 8L (q 6) + 2L (q 5) + 2L (q 4); A A A A quasi which coincides with the computation of  (q) above, and justifies Theorem 4.11. dsc (w) w 2 s s s ; (s s ) 2 1 2 1 2 1 1; s 2 2 2 3 0 s ; s s ; s s ; s s s ; (s s ) ; s (s s ) ; s (s s ) ; (s s ) 2 1 2 2 1 1 2 1 2 1 1 2 1 2 1 2 2 1 TABLE 1. Computation of dsc (W ) when =  (G )nf3 + 2 g. Here s ; s 2 1 2 1 2 are the reflections w.r.t. ; , respectively. 1 2 14 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA 0 0 (b) Let = f ; + ; 3 + 2 g. Then is not compatible by Example 4.9(b). We may 1 1 2 1 2 compute (q 1)(q 2) if q  1 mod 2; quasi (q) = q 3q + 3 if q  0 mod 2: 6 5 4 3 2 E (t) = 5t + t + 2t + 3t + t ; >(q 1)(q 2) if q  1; 5 mod 6; q 3q + 3 if q  2; 4 mod 6; (E 0 (S)L )(q) = q 3q + 4 if q  3 mod 6; q 3q + 5 if q  0 mod 6: quasi quasi Thus  (q) = (E 0 (S)L )(q) if q  1; 2; 4; 5 mod 6, and  (q) < (E 0 (S)L )(q) 0  0 A A quasi otherwise. In particular, we can easily check on Figure 2 that  (7) = (E (S)L )(7) = (c) In addition to the example above, both sides of the formula in Theorem 4.11 have different values at each constituent when is not compatible. For example, let = f + g as one 1 1 2 of the non-compatible singleton subsets mentioned in Example 4.9(b). Then quasi (q) = q(q 1) for all q; 6 5 4 3 2 E (t) = t + 2t + 3t + 3t + 2t + t; (E (S)L )(q) = q(q 1) + 1 for all q: 1 A quasi Thus  (q) < (E (S)L )(q) for all q. 1 A (d) Let = f ; 3 + g. Then is compatible (e.g., checked by Figure 2) but not an ideal of 2 2 1 2 2 any positive system of  = G . Otherwise, must be the associated base w.r.t. that positive 2 2 system since it has only two elements. However, it is a contradiction because the length of and 3 + are equal. 1 2 Example 4.19. Let  = A and let = [f ~g = f ; ; ; ; + + + g. We may 4 1 2 3 4 1 2 3 4 compute quasi 4 3 2 (q) = q 5q + 10q 10q + 4 for all q; 5 4 3 2 E (t) = 5t + 9t + 9t + t ; 4 3 2 (E (S)L )(q) = q 5q + 11q 12q + 5 for all q: quasi Thus,  (q) < (E (S)L )(q) for all q 6= 1. It follows from Theorem 4.11 that is not compatible. In general, it is not easy to compute the A-Eulerian polynomial of an arbitrary (compatible) subset by using its primary definition (Definition 4.2). For some special cases, e.g., the subsets obtained from the positive system by removing one element as mentioned in Example 4.9(a), we can compute the corresponding Eulerian polynomial in terms of the root system invariants. Let  (resp.,  ) denote the set of all long (resp., short) roots of . We will use the notation l s when we refer to either  or  . If  is simply-laced, we agree that  = ;  = ;. Denote l;s l s l s e e :=  \  and  := \  . l;s l;s l;s l;s EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 15 + c + Proposition 4.20. Let  2  and set  :=  nfg. Define l;s := fw 2 W j  = w( ) for some 2  g: i i l;s Then (i) = fw 2 W j dsc c (w) = c for some 2  g = fw 2 W j dsc c (w) > 0g. i i l;s (ii) , where := fw 2 W j  = w( )g for each 2  . e i i i i l;s i l;s (iii) ' stab() (as sets), where stab() := fw 2 W j  = w()g is the stablizer of  in W . In #W particular, # = . l;s # #W l;s (iv) # = . l;s Proof. (i) and (ii) are trivial, and (iv) is an immediate consequence of (ii) and (iii). Let us prove (iii). Fix w 2 W so that w () = . Thus, if w 2 , then ws w 2 stab(). ! ! i i ! i i i i Consider the set map ! stab(); defined by w 7! ws w : i ! i i Clearly,  is well-defined and injective. Moreover, for any  2 stab(), we have w = w s 2 ! i such that (w) =  . Therefore  is surjective and hence bijective. To prove the second state- #W ment, it suffices to show that #stab() = . This can be done by proving that the surjective l;s map ' : W !  defined by w 7! w() induces a bijection W=stab() !  . l;s l;s Theorem 4.21. If  2  , then l;s #W #W (# # ) l;s l;s hc h E (t) = t + t : f  # f  # l;s l;s i l;s Proof. It follows directly from Definition 4.2 and Proposition 4.20. An example of Theorem 4.21 is already mentioned in Example 4.18(a). In addition, if  is any + 6 3 short root in  (G ), then Theorem 4.21 implies that E c (t) = 10t + 2t . Next, we discuss the generating function of the characteristic quasi-polynomial. Proposition 4.22. L (q)t = : A Q (1 t ) q1 i=0 Proof. See, e.g., [KTT10, Proof of Theorem 3.1]. Theorem 4.23. (i) #W quasi q (q)t = : (1 t ) q1 i=0 (ii) Let R (t) be the generalized Eulerian polynomial (see Remark 4.3(b)). Then R (t) ` q q t = : (1 t ) q1 i=0 Proof. For a proof of (i), see, e.g., [Ath96], [BS98, Theorem 4.1], [KTT10, Theorem 3.1]. (ii) follows from [LP18, Theorem 10.1].  16 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA Theorem 4.24. A subset   is compatible if and only if E (t) quasi (4.6)  (q)t = : (1 t ) q1 i=0 Proof. Using Corollary 3.10, Propositions 4.22 and 3.7, we compute X X X q  q (E (S)L )(q)t = L  (q + dsc (A ))t A i q1 q1 i2[N ] X X dsc (A )  q+dsc (A ) i i = t L (q + dsc (A ))t i2[N ] q1 X X dsc (A ) n = t L (n )t A i i2[N ] n h hdsc (A ) E (t) i2[N ] = = : Q Q ` ` c c i i (1 t ) (1 t ) i=0 i=0 If   is compatible, then Formula (4.6) follows from Theorem 4.11 and the calculation above. Assume that Formula (4.6) holds but is not compatible. By Proof of Theorem 4.11, we can write quasi (q) = (E (S)L )(q) R(q); for some nonzero function R(q) (actually quasi-polynomial) in q. Thus, R(q)t = 0, which q1 is a contradiction. Remark 4.25. By Remark 4.3, Theorems 4.11 and 4.24, if (a) =  , then we recover Theorems 3.8 and 4.23(i). (b) = ;, then we recover [Yos18b, Theorem 4.8] and Theorem 4.23(ii). Corollary 4.26. If  2  , then l;s #W #W (# # ) l;s l;s quasi (q) = L (q h + c ) + L (q h); A A f  # f  # l;s l;s i l;s hc h i e #W t + #W (# # )t e l;s l;s quasi i l;s (q)t = : c Q f  #  (1 t ) l;s i=0 q1 Proof. It follows from Theorems 4.11, 4.21 and 4.24. 5. DEFORMATIONS OF W EYL SUBARRANGEM ENTS + c + Let   , and recall the notation =  n . Let a  b be integers, and denote [a; b] := fm 2 Z j a  m  bg. Also, if b  1, then write [b] instead of [1; b]. Definition 5.1. Let a  b, c  d be integers. Define two types of the deformed Weyl arrangements of as follows: [a;b] (Type I) I := fH j 2 ; m 2 [a; b]g. ;m [a;b];[c;d] [a;b] [c;d] (Type II) II := I t I . c c ; EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 17 [a;b];[c;d] [c;d];[a;b] Remark 5.2. (a) There is an obvious duality II = II . We can list some special- c c ; ; [a;b] [a;b];[c;d] [c;d] izations: I = ? the empty arrangement, II = A the truncated affine Weyl ar- + + ; ;; rangement, including the extended Shi, Catalan, Linial arrangements, see, e.g., [SP00, x9]. In [a;b] [a;b];[a;b] [a;b] addition, I = II = A . We refer the reader to [Ath99], [Ath04], [Yos18b] and + c + [a;b] [Yos18a] for more details on the characteristic (quasi-)polynomials ofA . (b) The deformed Weyl arrangements of an arbitrary are less well-known. When  is of type A, the deleted (or graphical) Shi arrangement, see, e.g., [Ath96,x3] or [AR12], is the product [0;1];[0;0] (e.g., [OT92, Definition 2.13]) of the 1-dimensional empty arrangenment and II . Definition 5.3. For w 2 W , define dsc (w) := c ; 0i`; w( )2 asc (w) := c ; 0i`; w( )2 asc (w) := c : 0i`; w( )2 Obviously, asc (w) + asc (w) + dsc (w) + dsc (w) = h for all w 2 W . Similar to Lemma 4.4, each function defined above takes values in [0; h 1]. Furthermore, asc (w) = asc + (w) = asc (w) + asc (w); dsc (w) = dsc + (w) = dsc (w) + dsc (w): Similar to Definition 4.6, we can extend the functions above to functions on the set of all alcoves. Now let us formulate a deformed version of Proposition 3.7. Set [a ;b ] i i F := H (1  i  `); ;m i i m2[a ;b ] i i [a ;b ] 0 0 F := H : ;q+m 0 0 m2[a ;b ] 0 0 Proposition 5.4. Let fi ; : : : ; i g  [0; `], and let b  0 for all 1  j  k. Suppose that 1 k i q > (b + 1)c . Then i i j2[k] j j 0 1 [ X [0;b ] _ j @ A # qA \ Z ( )n F = L (q (b + 1)c ): i i i A j j j2[k] j2[k] Proof. The formula was implicitly used in [Yos18b, x5] and its proof is very similar to the non- deformed case. Note that if i 2 [`], then 8 9 ( ; x)  b + 1 < = i i [0;b ] _ _ # qA \ Z ( )n F = # x 2 Z ( ) ( ; x)  0 (j 2 [`]nfig) : ; ( ; x)  q = L (q (b + 1)c ): i i Here the last equality follows from the bijection x 7! x + (b + 1)$ . The proof for i = 0 is similar. Then apply the formula above repeatedly.  18 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA Remark 5.5. If we replace the interval [0; b ] in Proposition 5.4 by [a; b ] with a  1, there might i i j j be a large change in the right-hand side of the formula above. For example, if 1  a  b , then 1 1 [a ;b ] _ 1 1 # qA \ Z ( )n F = L  (q (b + 1)c ) + L  (q) L  (q a c ): 1 1 1 1 1 A A A Theorem 5.6. Let be a compatible subset of  . (i) If a; b  0, then quasi (q) = L (q (b + 1)asc (A ) asc (A ) (a + 1)dsc (A )): [a;b] A i i i i2[N ] (ii) If b  1, then quasi (q) = L (q (b + 1)asc (A ) asc (A )): [1;b] A i i i2[N ] Proof. Proofs of (i) and (ii) are similar, and both are similar in spirit to the proof of [Yos18b, Theorem 5.1]. See also Proof of Theorem 4.11 in this paper. First, we give a proof for (i). Since both sides are quasi-polynomials, it is sufficient to prove the equality for q  0 (actually, q > (a + b + 3)h is sufficient). By Proposition 5.4, for i 2 [N ], 0 1 } _ @ A # qA \ Z ( )n H ;kq+m 2 ;k2Z;m2[a;b] 8 9 ( ; x)  qk + b + 1 ( 2 I \ ) > > > > < = (; x) > qk ( 2 I \ ) = # x 2 Z ( ) ( ; x)  qk a 1 ( 2 J \ ) > > > > : ; (; x)  qk ( 2 J \ ) = L (q (b + 1)asc (A ) asc (A ) (a + 1)dsc (A )): A i i i By Proposition 3.11, we have quasi (q) = L (q (b + 1)asc (A ) asc (A ) (a + 1)dsc (A )): [a;b] A i i i i2[N ] For (ii), note that for i 2 [N ], 0 1 } _ @ A # qA \ Z ( )n H ;kq+m 2 ;k2Z;m2[1;b] 8 9 > ( ; x)  qk + b + 1 ( 2 I \ ) > > > < = (; x) > qk ( 2 I \ ) = # x 2 Z ( ) ( ; x)  qk ( 2 J \ ) > > > > : ; (; x)  qk ( 2 J \ ) = L (q (b + 1)asc (A ) asc (A )): A i i Remark 5.7. Theorem 5.6 is a generalization of several known results. (a) When =  , we obtain [Ath04, Theorem 1.2] (a = b  0), and Theorem 5.1 (a = b 1 0), Theorem 5.2 (b  1), Theorem 5.3 (b = n + k, a = k 1, n; k  1) in [Yos18b]. When = ;, we obtain [Yos18b, Theorem 4.8]. EULERIAN POLYNOMIALS FOR WEYL SUBARRANGEMENTS 19 (b) When a = b = 0 (and   is compatible), we obtain Theorem 4.11. By using the same method, we have the following result for the arrangements of type II. Theorem 5.8. Let be a compatible subset of  . (i) If a; b; c; d  0, then quasi (q) = L (q (b + 1)asc (A ) (d + 1)asc (A ) [a;b];[c;d] A i i II i2[N ] (a + 1)dsc (A ) (c + 1)dsc (A )): i i (ii) If a; b  0; d  1, then quasi (q) = L (q (b + 1)asc (A ) (d + 1)asc (A ) [a;b];[1;d] i i II i2[N ] (a + 1)dsc (A )): (iii) If b; d  1, then quasi (q) = L (q (b + 1)asc (A ) (d + 1)asc (A )): [1;b];[1;d] A i i II i2[N ] Remark 5.9. (a) One can work with other intervals [a; b] but the computation may become more complicated (see Remark 5.5). F F n [a ;b ] n k k + (b) One can define and study the arrangement I where  = with n  3. k=1 k=1 See, e.g., [Ath96, Theorem 3.11] for an example when n = 3. We choose not to develop this direction here. It would be interesting to characterize the compatibility in the case of type A (e.g., in terms of graphs) and compare the following result with [AR12, Theorem 3.2] and [Ath96, Theorem 3.9] (see Remark 5.2 for the notation). h+asc (w) Corollary 5.10. Define M (t) := t . If is compatible, then w2W quasi (q) = (M (S)L )(q): [0;1];[0;0] II Acknowledgements. The first author was partially supported by Mitacs Canada Globalink Re- search Award to visit Japan and carry out the collaboration. The second author is partially sup- ported by JSPS Research Fellowship for Young Scientists Grant Number 19J12024. The third author is partially supported by JSPS KAKENHI Grant Number JP18H01115. The second author would like to thank Akiyoshi Tsuchiya for pointing out an error in the proof of the main result of the manuscript (Theorem 4.11) in a previous version. R EFERENCES [AR12] D. Armstrong and B. Rhoades. The Shi arrangement and the Ish arrangement. Trans. Amer. Math. Soc., 364(3):1509–1528, 2012. [Ath96] C. A. Athanasiadis. Characteristic polynomials of subspace arrangements and finite fields. Adv. Math., 122(2):193–233, 1996. [Ath99] C. A. Athanasiadis. Extended Linial hyperplane arrangements for root systems and a conjecture of Postnikov and Stanley. J. Algebr. Comb., 10:207–225, 1999. [Ath04] C. A. Athanasiadis. Generalized Catalan numbers, Weyl groups and arrangements of hyperplanes. Bull. Lond. Math. Soc., 36:294–302, 2004. 20 AHMED UMER ASHRAF, TAN NHAT TRAN, AND MASAHIKO YOSHINAGA [BM14] P. Brand ¨ en ´ and L. Moci. The multivariate arithmetic Tutte polynomial. Trans. Amer. Math. Soc., 366(10):5523–5540, 2014. [BS98] A. Blass and B. Sagan. Characteristic and Ehrhart polynomials. J. Algebr. Comb., 7(2):115–126, 1998. [BV07] A. Bhattacharya and G. R. Vijayakumar. An integrality theorem of root systems. European J. Combin., 28(6):1854–1862, 2007. [BZ06] M. Beck and T. Zaslavsky. Inside-out polytopes. Adv. Math., 205(1):134–162, 2006. [CW12] B. Chen and S. Wang. Comparison on the coefficients of characteristic quasi-polynomials of integral ar- rangements. J. Combin. Theory Ser. A, 119:271–281, 2012. [DPS09] K. Dilks, T. K. Petersen, and R. Stembridge. Affine descents and the Steinberg torus. Adv. Appl. Math., 42:423–444, 2009. [Hum90] J. E. Humphreys. Reflection Groups and Coxeter Groups. Cambridge University Press, 1990. [KTT08] H. Kamiya, A. Takemura, and H. Terao. Periodicity of hyperplane arrangements with integral coefficients modulo positive integers. J. Algebr. Comb., 27(3):317–330, 2008. [KTT10] H. Kamiya, A. Takemura, and H. Terao. The characteristic quasi-polynomials of the arrangements of root systems and mid-hyperplane arrangements. Arrangements, local systems and singularities, pages 177–190, Progr. Math., 283, Birkhauser ¨ Verlag, Basel, 2010. [KTT11] H. Kamiya, A. Takemura, and H. Terao. Periodicity of non-central integral arrangements modulo positive integers. Ann. Comb., 15(3):449–464, 2011. [LP18] T. Lam and A. Postnikov. Alcoved polytopes II. In: Kac V., Popov V. (eds) Lie Groups, Geometry, and Representation Theory, pages 253–272, Progr. Math., 326, Birkhauser ¨ Verlag, Cham, 2018. [LTY] Y. Liu, T. N. Tran, and M. Yoshinaga. G-Tutte polynomials and abelian Lie group arrangements. Int. Math. Res. Not. IMRN, to appear. [Nor12] C. Norman. Finitely Generated Abelian Groups and Similarity of Matrices over a Field. Springer Under- graduate Mathematics Series, Springer-Verlag London, 2012. [OT92] P. Orlik and H. Terao. Arrangements of hyperplanes. Grundlehren der Mathematischen Wissenschaften 300, Springer-Verlag, Berlin, 1992. [Som97] E. Sommers. A family of affine Weyl group representations. Transform. Groups, 2(4):375–390, 1997. [SP00] R.P. Stanley and A. Postnikov. Deformations of Coxeter hyperplane arrangements. J. Combin. Theory Ser. A, 91:544–597, 2000. [Sut98] R. Suter. The number of lattice points in alcoves and the exponents of the finite Weyl groups. Math. Comp., 67(222):751–758, 1998. [Tra18] T. N. Tran. An equivalent formulation of chromatic quasi-polynomials. arXiv preprint, 2018. https:// arxiv.org/abs/1803.08649. [Tra19] T. N. Tran. Characteristic quasi-polynomials of ideals and signed graphs of classical root systems. European J. Combin., 79:179–192, 2019. [TY19] T. N. Tran and M. Yoshinaga. Combinatorics of certain abelian Lie group arrangements and chromatic quasi-polynomials. J. Combin. Theory Ser. A, 165:258–272, 2019. [Yos18a] M. Yoshinaga. Characteristic polynomials of Linial arrangements for exceptional root systems. J. Combin. Theory Ser. A, 157:267–286, 2018. [Yos18b] M. Yoshinaga. Worpitzky partitions for root systems and characteristic quasi-polynomials. Tohoku Math. J., 70(1):39–63, 2018. AHMED UMER A SHRAF, U NIVERSITY OF W ESTERN ONTARIO, C ANADA. E-mail address: aashra9@uwo.ca TAN NHAT TRAN, D EPARTMENT OF MATHEM ATICS, H OKKAIDO UNIVERSITY, K ITA 10, N ISHI 8, K ITA-K U, SAPPORO 060-0810, JAPAN. E-mail address: trannhattan@math.sci.hokudai.ac.jp M ASAHIKO YOSHINAGA, DEPARTMENT OF M ATHEMATICS, HOKKAIDO UNIVERSITY, KITA 10, NISHI 8, KITA-K U, S APPORO 060-0810, JAPAN. E-mail address: yoshinaga@math.sci.hokudai.ac.jp

Journal

Mathematics – arXiv (Cornell University)

Published: Nov 5, 2019

References

  • The Shi arrangement and the Ish arrangement
    Armstrong
  • Characteristic polynomials of subspace arrangements and finite fields
    Athanasiadis
  • Extended Linial hyperplane arrangements for root systems and a conjecture of Postnikov and Stanley
    Athanasiadis
  • Generalized Catalan numbers, Weyl groups and arrangements of hyperplanes
    Athanasiadis
  • Inside-out polytopes
    Beck
  • An integrality theorem of root systems
    Bhattacharya
  • Characteristic and Ehrhart polynomials
    Blass
  • The multivariate arithmetic Tutte polynomial
    Brändén
  • Comparison on the coefficients of characteristic quasi-polynomials of integral arrangements
    Chen
  • Affine descents and the Steinberg torus
    Dilks