Abstract
We present a general diagrammatic approach to the construction of efficient algorithms for computing the Fourier transform of a function on a finite group. By extending work which connects Bratteli diagrams to the construction of Fast Fourier Transform algorithms we make explicit use of the path algebra connection to the construction of Gel’fand–Tsetlin bases and work in the setting of quivers. We relate this framework to the construction of a configuration space derived from a Bratteli diagram. In this setting the complexity of an algorithm for computing a Fourier transform reduces to the calculation of the dimension of the associated configuration space. Our methods give improved upper bounds for computing the Fourier transform for the general linear groups over finite fields, the classical Weyl groups, and homogeneous spaces of finite groups, while also recovering the best known algorithms for the symmetric group.
Similar content being viewed by others
Notes
A straight-line program is a list of instructions for performing the operations \(\times , \div , +, -\) on inputs and precomputed values [4].
References
Auslander, L., Tolimieri, R.: Is computing with the finite Fourier transform pure or applied mathematics? Bull. Am. Math. Soc. (N.S) 1(6), 847–897 (1979)
Barros, D., Wilson, S., Kahn, J.: Comparison of orthogonal frequency-division multiplexing and pulse-amplitude modulation in indoor optical wireless links. IEEE Trans. Commun. 60(1), 153–163 (2012)
Bratteli, O.: Inductive limits of finite dimensional \(C^*\)-algebras. Trans. Am. Math. Soc. 171, 195–234 (1972)
Bürgisser, P., Clausen, M., Shokrollahi, M.: Algebraic Complexity Theory, Volume 315 of Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences]. Springer, Berlin (1997) (With the collaboration of Thomas Lickteig)
Cannon, K., Cariou, R., Chapman, A., Crispin-Ortuzar, M., Fotopoulos, N., Frei, M., Hanna, C., Kara, E., Keppel, D., Liao, L., Privitera, S., Searle, A., Singer, L., Weinstein, A.: Toward early-warning detection of gravitational waves from compact binary coalescence. Astrophys. J. 748(2), 136 (2012)
Clausen, M.: Fast generalized Fourier transforms. Theoret. Comput. Sci. 67(1), 55–63 (1989)
Cooley, J., Tukey, J.: An algorithm for the machine calculation of complex Fourier series. Math. Comput. 19, 297–301 (1965)
Danelakis, A., Mitrouli, M., Triantafyllou, D.: Blind image deconvolution using a banded matrix method. Numer. Algorithms 64(1), 43–72 (2013)
Daugherty, Z., Orellana, R.: The quasi-partition algebra. J. Algebra 404, 124–151 (2014)
Diaconis, P.: Average running time of the fast Fourier transform. J. Algorithms 1, 187–208 (1980)
Diaconis, P.: Group Representations in Probability and Statistics. In: Lecture Notes—Monograph Series, 11. Institute of Mathematical Statistics, Hayward, CA (1988)
Diaconis, P.: A generalization of spectral analysis with application to ranked data. Ann. Stat. 17(3), 949–979 (1989)
Diaconis, P., Rockmore, D.: Efficient computation of the Fourier transform on finite groups. J. Am. Math. Soc. 3(2), 297–332 (1990)
Diaconis, P., Rockmore, D.: Efficient computation of isotypic projections for the symmetric group. In: Groups and computation (New Brunswick, NJ, 1991), Volume 11 of DIMACS Ser. Discrete Math. Theoret. Comput. Sci., pp. 87–104. Amer. Math. Soc., Providence, RI (1993)
Elliott, D., Rao, K.: Fast Transforms: Algorithms, Analyses, Applications. Academic Press Inc. (Harcourt Brace Jovanovich Publishers), New York (1982)
Elliott, G.: On the classification of inductive limits of sequences of semisimple finite-dimensional algebras. J. Algebra 38(1), 29–44 (1976)
Gabriel, P.: Unzerlegbare darstellungen I. Manuscr. Math. 6(1), 71–103 (1972)
Gel’fand, I., Cetlin, M.: Finite-dimensional representations of the group of unimodular matrices. Doklady Akad. Nauk SSSR (N.S.) 71, 825–828 (1950)
Goodman, F., de la Harpe, P., Jones, V.: Coxeter Graphs and Towers of Algebras, Mathematical Sciences Research Institute Publications, vol. 14. Springer, New York (1989)
Grood, C.: The rook partition algebra. J. Comb. Theory Ser. A 113(2), 325–351 (2006)
Halverson, T., del Mas, E.: Representations of the Rook–Brauer algebra. Commun. Algebra 42(1), 423–443 (2014)
Heideman, M., Johnson, D., Burrus, C.: Gauss and the history of the fast Fourier transform. Arch. Hist. Exact Sci. 34(6), 265–277 (1985)
Humphreys, J.: Reflection Groups and Coxeter Groups. Cambridge Studies in Advanced Mathematics, vol. 29. Cambridge University Press, Cambridge (1990)
James, G., Kerber, A.: The Representation Theory of the Symmetric Group, Volume 16 of Encyclopedia of Mathematics and its Applications. Addison-Wesley Publishing Co., Reading, MA (1981)
Johnson, S., Frigo, M.: A modified split-radix FFT with fewer arithmetic operations. IEEE Trans. Signal Process. 55(1), 111–119 (2007)
Leduc, R., Ram, A.: A ribbon Hopf algebra approach to the irreducible representations of centralizer algebras: the Brauer, Birman-Wenzl, and type A Iwahori-Hecke algebras. Adv. Math. 125, 1–94 (1997)
Lundy, T., Van Buskirk, J.: A new matrix approach to real FFTs and convolutions of length \(2^k\). Computing 80(1), 23–45 (2007)
Maslen, D.: The efficient computation of Fourier transforms on the symmetric group. Math. Comp. 67(223), 1121–1147 (1998)
Maslen, D., Orrison, M., Rockmore, D.: Computing isotypic projections with the Lanczos iteration. SIAM J. Matrix Anal. Appl. 25(3), 784–803 (2003)
Maslen, D., Rockmore, D.: Adapted diameters and FFTs on groups. In: Proceedings of the 6th ACM-SIAM SODA, pp. 253–262. ACM, New York (1995)
Maslen, D., Rockmore, D.: Generalized FFTs—a survey of some recent results. In: Groups and computation, II (New Brunswick, NJ, 1995), Volume 28 of DIMACS Ser. Discrete Math. Theoret. Comput. Sci., pp. 183–237. Amer. Math. Soc., Providence, RI (1997)
Maslen, D., Rockmore, D.: Separation of variables and the computation of Fourier transforms on finite groups. I. J. Am. Math. Soc. 10(1), 169–214 (1997)
Maslen, D., Rockmore, D.: Double coset decompositions and computational harmonic analysis on groups. J. Fourier Anal. Appl. 6(4), 349–388 (2000)
Maslen, D., Rockmore, D.: The Cooley–Tukey FFT and group theory. Not. Am. Math. Soc. 48(10), 1151–1160 (2001)
Maslen, D., Rockmore, D., Wolff, S.: Separation of variables and the computation of Fourier transforms on semisimple algebras (submitted)
Ram, A.: Seminormal representations of Weyl groups and Iwahori–Hecke algebras. Proc. Lond. Math. Soc. (3) 75(1), 99–133 (1997)
Rockmore, D.: Some applications of generalized FFTs. In: Groups and computation, II (New Brunswick, NJ, 1995), Volume 28 of DIMACS Ser. Discrete Math. Theoret. Comput. Sci., pp. 329–369. Amer. Math. Soc., Providence, RI (1997)
Rockmore, D.: The FFT: an algorithm the whole family can use. Comput. Sci. Eng. 2(1), 60–64 (2000)
Rørdam, M., Størmer, E.: Classification of Nuclear \(C^*\)-algebras. In: Entropy in Operator Algebras, Encyclopaedia of Mathematical Sciences, vol. 126. Springer, Berlin (2002)
Serre, J.: Linear Representations of Finite Groups. Springer, New York (1977) (Translated from the second French edition by Leonard L. Scott, Graduate Texts in Mathematics, vol. 42)
Stanley, R.: Differential posets. J. Am. Math. Soc. 1(4), 919–961 (1988)
Stanley, R.: Variations on differential posets. In: Invariant Theory and Tableaux (Minneapolis, MN, 1988), vol. 19 of IMA Vol. Math. Appl., pp. 145–165. Springer, New York (1990)
Tolimieri, R., An, M., Lu, C.: Algorithms for Discrete Fourier Transform and Convolution. Signal Processing and Digital Filtering, 2nd edn. Springer, New York (1997)
Van Loan, C.: Computational Frameworks for the Fast Fourier Transform, vol. 10 of Frontiers in Applied Mathematics. Society for Industrial and Applied Mathematics (SIAM), Philadelphia, PA (1992)
Yavne, R.: An economical method for calculating the discrete Fourier transform. Proc. AFIPS Fall Jt. Comput. Conf. 33, 115–125 (1968)
Young, A.: On quantitative substitutional analysis. Proc. Lond. Math. Soc. 31(2), 273–288 (1929)
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Thomas Strohmer.
Daniel N. Rockmore was partially supported by AFOSR Award FA9550-11-1-0166 and the Neukom Institute for Computational Science at Dartmouth College, Sarah Wolff was partially supported by by the NSF GRFP under Grant No. DGE-1313911.
Appendices
Appendix 1: Gel’fand–Tsetlin Bases and Adapted Representations
In Sect. 2 we introduce adapted bases and systems of Gel’fand Tsetlin bases. Here we make the formal connection between adapted bases of a group algebra chain and systems of Gel’fand Tsetlin bases for the corresponding chain of path algebras.
Definition 7.1
Given a Bratteli diagram \(\mathcal {B}\), a representation of \(\mathbf {\mathcal {B}}\) assigns to each \(\alpha \in V(\mathcal {B})\) a linear space \(V_\alpha \) and to each edge \(e\in E(\mathcal {B})\) a linear map \(L_e:V_{s(e)}\rightarrow V_{t(e)}\). Given two representations \((\{V_\alpha \}_{\alpha \in V(\mathcal {B})}, \{L_e\}_{e\in E(\mathcal {B})})\), \((\{W_\alpha \}_{\alpha \in V(\mathcal {B})}, \{S_e\}_{e\in E(\mathcal {B})})\), of \(\mathcal {B}\), a morphism \(m:V\rightarrow W\) is a family of linear maps \(\{m_\alpha :V_\alpha \rightarrow W_\alpha \}_{\alpha \in V(\mathcal {B})}\) such that the diagram
commutes for all \(e\in E(\mathcal {B})\).
A model representation of \(\mathbf {\mathcal {B}}\) is a representation of \(\mathcal {B}\) such that for all \(e\in E(\mathcal {B})\), \(L_e\) is injective, and for all nonroot vertices \(\alpha \in V(\mathcal {B})\),
Definition 7.2
Given a chain of group algebras \(\{\mathbb {C}[G_i]\}\), a model representation for \(\{\mathbb {C}[G_i]\}\) is a model representation of the corresponding Bratteli diagram \(\mathcal {B}\) such that:
-
(i)
for each \(\alpha \in V(\mathcal {B})\) at level i, \(V_\alpha \) is the representation space of the representation of \(\mathbb {C}[G_i]\) corresponding to \(\alpha \),
-
(ii)
for each \(e\in E(\mathcal {B})\) from level i to level \(i+1\), \(L_e\) is \(\mathbb {C}[G_i]\)-equivariant, i.e., for \(\rho _{s(e)}\) the representation of \(\mathbb {C}[G_i]\) corresponding to s(e) and \(\rho _{t(e)}\) the representation of \(\mathbb {C}[G_{i+1}]\) corresponding to t(e), the diagram
$$\begin{aligned} \begin{array}{llll} &{}V_{s(e)}&{}\xrightarrow {\;L_e\;} &{}V_{t(e)}\\ \rho _{s(e)}&{}\downarrow &{}&{}\downarrow \;\;\rho _{t(e)}\\ &{}V_{s(e)}&{}\xrightarrow {\;L_e\;}&{}V_{t(e)} \end{array} \end{aligned}$$commutes for all \(e\in E(\mathcal {B})\).
A model representation of an algebra chain has a natural basis of paths:
Lemma 7.3
Given a model representation of a chain of subalgebras with Bratteli diagram \(\mathcal {B}\), the collection of distinct paths in \(\mathcal {B}\) from the root to a vertex \(\alpha \in V(\mathcal {B})\) corresponds to a choice of basis for \(V_\alpha \).
Proof
Consider the space \(V_\beta \) corresponding to the root \(\beta \), i.e., \(V_\beta \) is the representation space of \(\mathbb {C}[\mathcal {B}_0]=\mathbb {C}\), so \(V_\beta \) is one-dimensional. Now let \(\alpha \) be a vertex in \(\mathcal {B}\) with \(gr(\alpha )=1\). Then \(V_\alpha =\displaystyle \displaystyle \bigoplus _{t(e)=\alpha } Im(L_e)\cong \displaystyle \bigoplus _{t(e)=\alpha }V_\beta \) since \(L_e\) is injective. Induction gives the result. \(\square \)
Thus, given an irreducible representation \(\rho \) of \(\mathbb {C}[G_i]\) corresponding to a vertex \(\alpha \) in the Bratteli diagram associated to the chain of group algebras, there is a basis for the representation space of \(\rho \) indexed by the paths from the root to \(\alpha \). We call such a basis a Gel’fand–Tsetlin basis, as in Definition 2.18. Given a model representation of a Bratteli diagram \(\mathcal {B}\), Lemma 7.3 gives a system of Gel’fand–Tsetlin bases for \(\mathcal {B}\). In fact, these are equivalent concepts:
Theorem 7.4
A system of Gel’fand–Tsetlin bases for a Bratteli diagram \(\mathcal {B}\) uniquely determines a model representation for \(\mathcal {B}\). Conversely, a model representation uniquely determines a system of Gel’fand–Tsetlin bases for \(\mathcal {B}\).
Proof
Both require a choice of vector space for each vertex of \(\mathcal {B}\), so we need only show how a choice of basis corresponds with linear maps \(L_e\) for each edge e.
Given a system of bases and an edge \(e\in \mathcal {B}\), a basis vector for \(V_{s(e)}\) corresponds to a path P from the root to s(e). Then \(e\circ P\) is a path from the root to t(e), which corresponds to a basis vector for \(V_{t(e)}\). In other words, we have an injection of \(V_{s(e)}\) into \(V_{t(e)}\).
Conversely, given a model representation and a vertex \(\alpha \), every path from the root to \(\alpha \) corresponds to an injection of \(\mathbb {C}\) into \(V_\alpha \). Since
the union of the distinct images of \(1\in \mathbb {C}\) over the collection of injections gives a basis for \(V_\alpha \) as we vary over all possible paths from the root to \(\alpha \). \(\square \)
Remark 7.5
The equivalent definitions of Gel’fand–Tsetlin bases and model representations coincide with the notion of a complete set of adapted representations for chains of groups. Clearly a model representation for the group algebra chain gives rise to an adapted basis since the isomorphism
describes how the representation space \(V_{\alpha }\) decomposes at level \(i-1\). Equivariance of the maps \(L_{e}\) then gives the decomposition of the representation \(\rho _\alpha \).
Further, a complete set R of inequivalent irreducible representations adapted to a chain of subgroups \(G_n>G_{n-1}>\cdots >G_0\) determines the paths in the Bratteli diagram \(\mathcal {B}\) of the group algebra chain by drawing \(M(\rho , \gamma )\) arrows from a representation \(\gamma \in R\) of \(G_i\) to a representation \(\rho \in R\) of \(G_{i+1}\). Then a set of bases for the representation spaces of the representations in R determines a system of Gel-fand Tsetlin bases for the group algebra chain, and so by Theorem 7.4 a model representation.
Appendix 2: Restricted Product Lemmas
In this Appendix, we prove the associativity of the restricted product defined in Sect. 5.
Lemma 7.6
Let B be a locally finite graded quiver and R a graded quiver with finite subquivers \(Q_1, Q_2,\dots ,Q_m\) such that \(Q_i\cap Q_j\cap Q_k\) has no edges for all \(i\ne j\ne k\). Let \(Q_i^\triangle \) denote the quiver \(Q_1\triangle \cdots \triangle Q_i\) and let \(Q_i^\cup \) denote the quiver \(Q_1\cup \cdots \cup Q_i\). Then for \(f_i\in A(Q_i;B)\), \(f_1*f_2*\cdots *f_m\) is independent of bracketing. Moreover, for \(\tau \in {\text {Hom}}(Q_m^\triangle ;B)\) and \(\iota _k\) the natural injection \(Q_k\hookrightarrow R\),
Proof
We first prove (9) inductively, as associativity clearly follows. For \(n=2\), (9) is the definition of the restricted product \(f_1*f_2\).
Now suppose (9) holds for \(n-1\). Since \(Q_i\cap Q_j\cap Q_k=\emptyset \),
and
By the induction hypothesis,
By (10) and (11), each choice of \(\mu \) and \(\eta \) which agree on their intersection, the subquiver \(Q_{n-1}^\triangle \), uniquely determines a morphism \(\gamma \in {\text {Hom}}(Q_n^\cup \uparrow R; B)\) such that
\(\square \)
Appendix 3: Quiver Counts
In Sect. 6., we rewrite morphism counts in terms of multiplicities of representations and dimensions of subgroup algebras (Corollary 6.18). Here we give the details needed for the proofs of Sect. 6.
1.1 Smoothing Quivers
An important simplification in morphism counts is to remove superfluous vertices from quivers, i.e., ‘smooth’ them.
Definition 7.7
A quiver B factors at level \(\mathbf {i}\) if there are no arrows from a vertex \(\alpha \in V(B)\) with \(gr(\alpha )<i\) to a vertex \(\beta \in V(B)\) with \(gr(\beta )>i\).
Example 7.8
Let \(\mathcal {B}\) be a Bratteli diagram with highest grading n. Then for all \(0\le i \le n\), \(\mathcal {B}\) factors at level i.
Definition 7.9
Let Q be a quiver with a vertex v that is the target of exactly one arrow, \(e_1\), and the source of exactly one arrow, \(e_2\). To smooth Q at v, remove v and replace \(e_1\) and \(e_2\) with an arrow from the source of \(e_1\) to the target of \(e_2\). To smooth Q, smooth Q at all possible v.
Example 7.10
The quiver \(Q'\) of Fig. 31 results from smoothing the quiver Q.
Lemma 7.11
Let B be a graded quiver that factors at level i, R a graded quiver with subquiver Q, and v a vertex of Q at level i such that Q can be smoothed at v. Let \(Q'\) (respectively \(R'\)) be the quiver obtained by smoothing Q (respectively R) at v. Then \(\#{\text {Hom}}(Q\uparrow R; B)=\#{\text {Hom}}(Q'\uparrow R'; B).\)
Proof
Let \(\phi \in {\text {Hom}}(Q'\uparrow R';B)\) and let f be the arrow in \(Q'\) resulting from smoothing Q at v. Then f replaced two arrows, \(e_1, e_2\) in Q, with \(t(e_1)=v, s(e_2)=v\). Further, \(s(e_1)=s(f), t(e_2)=t(f)\), so \(\phi (f)\) is a path in B from a vertex \(\alpha \) with \(gr(\alpha )<i\) to a vertex \(\beta \) with \(gr(\beta )>i\). Since B factors at level i, this path contains a vertex, \(v'\), with \(gr(v')=i\). Let \(e_1'\) be the subpath of f starting at the source of f and ending at \(v'\). Similarly, let \(e_2'\) be the subpath of f starting at \(v'\) and ending at the target of f.
Denote by \(\tilde{\phi }\) the morphism in \({\text {Hom}}(Q\uparrow R;B)\) such that:
Clearly \(\phi \rightarrow \tilde{\phi }\) is a bijection. \(\square \)
Corollary 7.12
Let \(\mathcal {B}\) be a Bratteli diagram, R a graded quiver with subquiver Q, and \(Q'\) (respectively \(R'\)) the quiver obtained by smoothing Q (respectively R). Then
1.2 Properties of Locally Free Quivers
We give the details of the proofs of Proposition 6.9 and Theorem 6.13 of Sect. 6.
Proposition 7.13
(Proposition 6.9) Let \(\mathcal {B}\) be a Bratteli diagram. Then the following properties are equivalent:
-
(i)
\(\mathcal {B}\) is locally free.
-
(ii)
For each i and all \(\beta \in \mathcal {B}^{i-1}\), there exists \(\lambda _i\in \mathbb {C}\) such that
$$\begin{aligned} \sum _{\alpha \in \mathcal {B}^i} M_\mathcal {B}(\alpha ,\beta )d_\alpha =\lambda _i d_\beta . \end{aligned}$$ -
(iii)
For each i, \(d_i\) is an eigenvector of DU.
-
(iv)
For each i there exists \(\lambda _i\in \mathbb {C}\) with \(DU^i\hat{0}=\lambda _iU^{i-1}\hat{0}\).
Proof
Statements (ii), (iii), and (iv) are equivalent by definition and Lemma 6.8. For example:
(iii)\(\Rightarrow \) (iv) \(DU^i\hat{0}=DUd_{i-1}=\lambda _id_{i-1}=\lambda _iU^{i-1}\hat{0}\)
(iv)\(\Rightarrow \) (iii) \(DUd_i=DU^{i+1}\hat{0}=\lambda _{i+1}U^i\hat{0}=\lambda _{i+1}d_i\) We leave the remaining equivalences of (ii), (iii), and (iv) to the reader.
To show the equivalence of (i) and (iv), recall from Note 6.6 that elements of \(\mathbb {C}[\mathcal {B}^i]\) correspond to representations of \(\mathbb {C}[\mathcal {B}_i]\), i.e. \(\mathbb {C}[\mathcal {B}_i]\)-modules. Under this identification, the regular representation of \(\mathbb {C}[\mathcal {B}_i]\) corresponds to the sum
Since D is restriction ([19, Proposition 2.3.12]), the restriction of the regular representation of \(\mathbb {C}[\mathcal {B}_i]\) to \(\mathbb {C}[\mathcal {B}_{i-1}]\) corresponds to \(Dd_i=DUd_{i-1}\).
(i)\(\Rightarrow \)(iv) For \(\mathcal {B}\) locally free, \(\mathbb {C}[\mathcal {B}_i]\) is free as a module over \(\mathbb {C}[\mathcal {B}_{i-1}]\) with rank \(\lambda _i\in \mathbb {C}\). Then the regular representation of \(\mathbb {C}[\mathcal {B}_i]\) decomposes in \(\mathbb {C}[\mathcal {B}_{i-1}]\) as \(\lambda _i\) copies of the regular representation of \(\mathbb {C}[\mathcal {B}_{i-1}]\). Thus,
(iv)\(\Rightarrow \)(i) \(DUd_{i-1}=Dd_i=\lambda _id_{i-1}\) and so \(\dim _\mathbb {C}\mathbb {C}[\mathcal {B}_i]=\lambda _i\dim _\mathbb {C}\mathbb {C}[\mathcal {B}_{i-1}]\); hence \(\lambda _i\) is rational and positive. To show \(\lambda _i\) integral, let
Let \(m=\gcd (d_\beta )\) over all \(\beta \in \mathcal {B}^{i-1}\), so \(\frac{d_\beta }{m}\) an integer for all \(\beta \in \mathcal {B}^{i-1}\). Then
Then the coefficient of \(\beta \) is an integer and thus \(q\vert \frac{d_\beta }{m}\) for all \(\beta \in \mathcal {B}^i\). But \(m=\gcd (d_\beta )\), and thus \(q=1\), making \(\lambda _i\) an integer. \(\square \)
Theorem 7.14
(cf. Theorem 6.13) Let \(\mathcal {B}\) be a locally free Bratteli diagram and \(w=D^{d_n}U^{u_n}\cdots D^{d_1}U^{u_1}\) an admissible word in U and D. Then for \(s=\sum _{i=1}^n u_i-d_i\) and \(\alpha \in \mathcal {B}^{s}\),
Proof
To be admissible, \(d_{i},u_{i}> 0\) for all \(1\le i\le n\) and
Let \(\mathcal {S}_k=\{i\in \mathcal {S}\vert i\le \sum _{j=1}^k (d_j+u_j)\}\) let \(s_k=\sum _{i=1}^k u_i-d_i\), and let \(w_k=D^{d_k}U^{u_k}\cdots D^{d_1}U^{u_1}\). We prove inductively that
Note that \(w_1=D^{d_1}U^{u_1}\). Then \(\mathcal {S}_1=\{u_1+1,u_1+2\dots ,u_1+d_1\}\) and for all \(i\in \mathcal {S}_1\), \(b_i=u_1\) and \(a_i=i-u_1-1\). By Proposition 6.9, Lemma 6.8, and induction,
Now suppose true for \(n-1\). Then
and the same argument as in the base case gives the result. \(\square \)
Appendix 4: Combinatorial Lemmas for the Weyl Groups
The SOV approach reduces Theorem 1.1 (respectively, Theorem 1.2) to counting the number of morphisms of the quivers of Fig. 14 into the Bratteli diagram of \(B_n\) (respectively, \(D_n\)). In this section we consider the Bratteli diagrams associated to \(B_n\) and \(D_n\) to provide the bounds used in the proofs of Theorems 1.1 and 1.2. Note that Lemmas 7.16, 7.17, 7.20, 7.21 and Corollaries 7.19 and 7.22 all hold for \(n\ge 2\), \(i\ge 2\).
1.1 The Weyl Group \(B_n\)
The Bratteli diagram \(\mathcal {B}\) associated to the chain \(\mathbb {C}[B_n]>\mathbb {C}[B_{n-1}]>\cdots > \mathbb {C}\) is a generalization of Young’s diagram — inequivalent irreducible representations of \(\mathbb {C}[B_i]\) are indexed by pairs of partitions \((\lambda _1,\lambda _2)\), of k and l, respectively, with \(k+l=i\). Pairs \((\lambda _1,\lambda _2)\), \((\mu _1,\mu _2)\) are connected by an edge if either \(\lambda _1\) may be obtained from \(\mu _1\) by adding a box, or if \(\lambda _2\) may be obtained from \(\mu _2\) by adding a box [36] (see Fig. 32). Note that this is a multiplicity-free diagram.
Theorem 7.15
For \(\mathcal {J}_i^n\) as in Fig. 14 and \(\mathcal {B}\) the Bratteli diagram associated to the Weyl group \(B_n\), \(\# {\text {Hom}}(\mathcal {J}_i^n\uparrow \mathcal {Q};\mathcal {B})\le 2\vert B_n\vert \).
Proof
By Theorem 6.1, we see that \(\# {\text {Hom}}(\mathcal {J}_i^n\uparrow \mathcal {Q};\mathcal {B})\) is equal to the sum
for \(M_\mathcal {B}(B_i,B_j):=\max M_\mathcal {B}(\alpha _i,\alpha _j)\) over all \(\alpha _i\in \mathcal {B}^i, \alpha _j\in \mathcal {B}^j\). By Corollary 6.18,
Lemma 7.16 below shows that \(M_{\mathcal {B}}(B_{i},B_{i-2})\le 2.\) Thus
\(\square \)
Lemma 7.16
\(M_\mathcal {B}(B_i,B_{i-2})\le 2\)
Proof
Suppose not. Then since \(\mathcal {B}\) is multiplicity-free, we must have distinct pairs of partitions \(\mathbf {\kappa }=(\kappa _1,\kappa _2),\), \(\mathbf {\rho }=(\rho _1,\rho _2),\), \(\mathbf {\gamma }=(\gamma _1,\gamma _2),\) \(\mathbf {\eta }=(\eta _1,\eta _2),\) and \(\mathbf {\lambda }=(\lambda _1,\lambda _2)\) as in Fig. 33.
Pairs of partitions are adjacent in \(\mathcal {B}\) if one is acquired from the other by adding a single box; hence either \(\rho _1=\lambda _1\) or \(\rho _2=\lambda _2\). The same holds for \(\eta ,\gamma \). Similarly, \(\kappa _1=\rho _1\) or \(\kappa _2=\rho _2\) and the same holds for \(\eta , \gamma \).
Without loss of generality, we need only consider the following two cases:
Case 1: \(\rho _1,\eta _1,\gamma _1=\lambda _1\).
If \(\kappa _1\ne \lambda _1\), then \(\kappa _2=\rho _2,\gamma _2,\eta _2\), but then \(\mathbf {\rho }=\mathbf {\eta }=\mathbf {\gamma }\), a contradiction.
If \(\kappa _1=\lambda _1\) then \(\kappa _2\) is obtained from \(\lambda _2\) by adding two boxes, which may be done in at most two ways, so \(\eta ,\gamma ,\rho \) are not all distinct, a contradiction.
Case 2: \(\rho _1=\lambda _1=\eta _1\) and \(\gamma _2=\lambda _2\).
If \(\kappa _1=\lambda _1\) then since \(\gamma _1\ne \lambda _1\), we see that \( \kappa _1\ne \gamma _1\). Thus, \(\kappa _2=\gamma _2\). But then \((\kappa _1,\kappa _2)=(\lambda _1,\lambda _2)\), a contradiction.
Now if \(\kappa _1\ne \lambda _1\), then \(\kappa _2=\rho _2,\eta _2\), but then \(\rho =\eta \), a contradiction. \(\square \)
The following two lemmas provide a bound for \(\dim A(\mathcal {H}_i^n\uparrow G;\mathcal {B})\), for \(\mathcal {H}_i^n\) as in Fig. 14.
Lemma 7.17
-
(1)
\(\displaystyle \#{\text {Hom}}(\mathcal {H}_i^n\uparrow G;\mathcal {B})=\frac{\vert B_{n-1}\vert }{\vert B_{i-1}\vert }\#{\text {Hom}}(\mathcal {H}_i^i\uparrow G;\mathcal {B})\)
-
(2)
$$\begin{aligned} \#{\text {Hom}}(\mathcal {H}_i^i\uparrow G;\mathcal {B})= & {} \quad 2(i-1)\vert B_{i-1}\vert + \sum _{\begin{array}{c} (\beta _{i-1}^1,\beta _{i-1}^2)=\\ \mathbf {\beta }_{i-1}\in \mathcal {B}^{i-1} \end{array}} ({\text {jmp}}(\beta _{i-1}^1)\\&+{\text {jmp}}(\beta _{i-1}^2))({\text {jmp}}(\beta _{i-1}^1)+{\text {jmp}}(\beta _{i-1}^2)+1)d_{\beta _{i-1}}^2, \end{aligned}$$
where \({\text {jmp}}\) denotes the jump of a partition, i.e, the number of ways to remove a single box to form a new partition.
Proof
To prove (1), first note by Theorem 6.1, \(\# {\text {Hom}}(\mathcal {H}_i^n\uparrow G;\mathcal {B})\) equals
By Corollary 6.18,
Then \(\# {\text {Hom}}(\mathcal {H}_i^n\uparrow G;\mathcal {B})\) equals
To prove (2),
for \(\displaystyle \sum _{\alpha _{i-1}\ne \beta _{i-1}}=\sum _{\begin{array}{c} \alpha _j,\beta _j\in \mathcal {B}^j\\ \alpha _{i-1}\ne \beta _{i-1} \end{array}}M_\mathcal {B}(\beta _{i-1},\beta _{i-2})M_\mathcal {B}(\alpha _{i},\alpha _{i-1}) M_\mathcal {B}(\alpha _{i},\beta _{i-1})M_\mathcal {B}(\alpha _{i-1},\beta _{i-2})d_{\beta _{i-1}}d_{\alpha _{i-1}}\) and
\(\displaystyle \sum _{\alpha _{i-1}=\beta _{i-1}}=\sum _{\begin{array}{c} \alpha _j,\beta _j\in \mathcal {B}^j\\ \alpha _{i-1}=\beta _{i-1} \end{array}}M_\mathcal {B}(\beta _{i-1},\beta _{i-2})^2M_\mathcal {B}(\alpha _{i},\beta _{i-1})^2(d_{\beta _{i-1}})^2.\)
First suppose \(\mathbf {\alpha }_{i-1}=(\alpha _{i-1}^1,\alpha _{i-1}^2),\mathbf {\beta }_{i-1}=(\beta _{i-1}^1,\beta _{i-1}^2)\) are distinct pairs of partitions. Then they jointly determine \(\mathbf {\alpha }_i=(\alpha _i^1,\alpha _i^2)\). Thus, the sum \(\sum \nolimits _{\alpha _{i-1}\ne \beta _{i-1}}\) becomes
and so
Now suppose \(\alpha _{i-1}=\beta _{i-1}\). Then \(\alpha _{i}\) is obtained from \(\beta _{i-1}\) by adding a box to \(\beta _{i-1}^1\) or \(\beta _{i-1}^2\), while \(\beta _{i-2}\) is obtained from \(\beta _{i-1}\) by removing a box from \(\beta _{i-1}^1\) or \(\beta _{i-1}^2\). Thus,
Summing Eqs. (12) and (13) gives (2). \(\square \)
Lemma 7.18
For any pair of partitions \((\beta _i^1,\beta _i^2)\) with \(\vert \beta _i^1\vert +\vert \beta _i^2\vert =i\),
Proof
Let \(k=\vert \beta _i^1\vert \), \(l=\vert \beta _i^2\vert \), \(a_k={\text {jmp}}(\beta _i^1)\), and \(a_l={\text {jmp}}(\beta _i^2)\). Then \(k+l=i\) and by [28, Lemma 5.3],
Then
\(\square \)
Combining Lemmas 7.17 and 7.18 gives the following bound:
Corollary 7.19
\(\#{\text {Hom}}(\mathcal {H}_i^n\uparrow G;\mathcal {B})\le \frac{4(i-1)}{n}\vert B_n\vert .\)
1.2 The Weyl Group \(D_n\)
The Bratteli diagram \(\mathcal {B}\) associated to the chain \(\mathbb {C}[D_n]> \mathbb {C}[D_{n-1}]>\cdots > \mathbb {C}\) is similar to the Bratteli diagram associated to the Weyl group \(B_n\) in that irreducible representations of \(\mathbb {C}[D_i]\) are indexed by pairs of partitions, \((\lambda _1,\lambda _2)\) of k and l, respectively, with \(k+l=i\). However, if \(\lambda _1\ne \lambda _2\), the irreducible representation indexed by \((\lambda _1,\lambda _2)\) is the same as that indexed by \((\lambda _2,\lambda _1)\). If \(\lambda _1=\lambda _2=\lambda \) then two distinct irreducible representations are indexed by the pair \((\lambda ,\lambda )\), and denoted by \((\lambda ,\lambda )^+\) and \((\lambda ,\lambda )^-\) [36] (see Fig. 34). Note that this is a multiplicity-free diagram.
Lemma 7.20
\(M_\mathcal {B}(D_i,D_{i-2})\le 3\).
Proof
Suppose not. Then since \(\mathcal {B}\) is multiplicity-free, there exist pairs of partitions \(\mathbf {\kappa }=(\kappa _1,\kappa _2),\) \(\mathbf {\rho }=(\rho _1,\rho _2),\) \(\mathbf {\gamma }=(\gamma _1,\gamma _2),\) \(\mathbf {\eta }=(\eta _1,\eta _2),\) \(\mathbf {\mu }=(\mu _1,\mu _2),\) and \(\mathbf {\lambda }=(\lambda _1,\lambda _2)\) connected in \(\mathcal {B}\) as in Fig. 35.
However, the proof of Lemma 7.16 dictates that no three of \(\eta ,\mu ,\gamma ,\rho \) are distinct pairs of partitions. Thus, without loss of generality,
for \(\alpha , \beta \) distinct partitions of \(\frac{j-1}{2}\). Then as in the proof of Lemma 7.16, either \(\lambda _1=\eta _1=\alpha \) or \(\lambda _2=\eta _2=\alpha \). Without loss, suppose \(\lambda _1=\alpha \). Then since \(\alpha \ne \beta \), \(\lambda _2\) must be \(\beta \). However, \(\vert \lambda _1\vert +\vert \lambda _2\vert =\vert \alpha \vert +\vert \beta \vert >j-2\), a contradiction. \(\square \)
Lemma 7.20 is used in the proof of Theorem 1.2 to give a bound on \(\dim A(\mathcal {J}_i^n\uparrow G;\mathcal {B})\), for \(\mathcal {J}_i^n\) as in Fig. 14. The following two lemmas provide a bound for \(\dim A(\mathcal {H}_i^n\uparrow G;\mathcal {B})\), for \(\mathcal {H}_i^n\) as in Fig. 14.
Lemma 7.21
-
(1)
\(\displaystyle \#{\text {Hom}}(\mathcal {H}_i^n\uparrow G;\mathcal {B})=\frac{\vert D_{n-1}\vert }{\vert D_{i-1}\vert }\#{\text {Hom}}(\mathcal {H}_i^i\uparrow G;\mathcal {B}),\)
-
(2)
for i odd, \(\#{\text {Hom}}(\mathcal {H}_i^i\uparrow G;\mathcal {B})\) is at most
$$\begin{aligned} \begin{array}{ll} \displaystyle \frac{\vert D_{i-1}\vert ^2}{\vert D_{i-2}\vert }+\sum _{(\alpha ,\alpha )^\pm \in \mathcal {B}^{i-1}} ({\text {jmp}}(\alpha ))({\text {jmp}}(\alpha )+1)(d_{(\alpha ,\alpha )^+})^2\\ \displaystyle +\sum _{\begin{array}{c} (\beta _{i-1}^1,\beta _{i-1}^2)=\\ \mathbf {\beta }_{i-1}\in \mathcal {B}^{i-1} \end{array}} ({\text {jmp}}(\beta _{i-1}^1)+{\text {jmp}}(\beta _{i-1}^2))({\text {jmp}}(\beta _{i-1}^1)+{\text {jmp}}(\beta _{i-1}^2)+1)d_{\beta _{i-1}}^2, \end{array} \end{aligned}$$ -
(3)
for i even, \(\#{\text {Hom}}(\mathcal {H}_i^i\uparrow G;\mathcal {B})\) is at most
where \({\text {jmp}}\) denotes the jump of a partition, i.e., the number of ways to remove a single box to form a new partition.
Proof
Part (1) follows from the proof of Lemma 7.17.
To prove (2), consider note that \(\#{\text {Hom}}(\mathcal {H}_i^i\uparrow G;\mathcal {B})\) equals
For
over partitions \(\alpha _{i-1}\ne \beta _{i-1}\) such that if \(\alpha _{i-1}=(\alpha ,\alpha )^\pm \) then \(\beta _{i-1}\ne (\alpha ,\alpha )^\pm \),
and
As in the proof of Lemma 7.17,
the inequality appearing because if \(\beta _{i-1}=(\alpha , \alpha )\), \({\text {jmp}}(\alpha )+{\text {jmp}}(\alpha )\) is an overestimate since \((\alpha ,\beta )\) represents the same representation as \((\beta , \alpha )\) in \(\mathcal {B}\). Similarly, the proof of Lemma 7.17 gives
Now suppose \(\alpha _{i-1}\ne \beta _{i-1}\) and \(\alpha _{i-1}=(\alpha ,\alpha )^\pm =\beta _{i-1}\). Then
Summing equations (14), (15), and (16) gives part (2).
To prove (3), note that in this case
since \(i-1\) is odd so \((\alpha , \alpha )^\pm \notin \mathcal {B}^{i-1}\). However, pairs of partitions of this form may be found at levels i and \(i-2\).
First suppose \(\alpha _{i-1}\ne \beta _{i-1}\). Then as in the proof of Lemma 7.17 they jointly determine \(\alpha _i=(\alpha _i^1,\alpha _i^2)\). This means that they jointly determine at most two pairs of partitions (if \(\alpha _i^1=\alpha _i^2\)). Thus
as in the proof of Lemma 7.17.
Now suppose \(\alpha _{i-1}=\beta _{i-1}\). As before there are \({\text {jmp}}(\beta _{i-1}^1)\) ways to obtain \(\beta _i^1\) and \({\text {jmp}}(\beta _{i-1}^2)\) ways to obtain \(\beta _i^2\), but to account for when \(\beta _i^1=\beta _i^2\), we overcount by multiplying by 2. The same holds for the number of ways to obtain \(\alpha _{i-2}\) from \(\beta _{i-1}\). Thus,
Summing equations (17) and (18) gives part 3. \(\square \)
Combining Lemma 7.21 with Lemma 7.18 gives the following bound:
Corollary 7.22
\(\#{\text {Hom}}(\mathcal {H}_i^n\uparrow G;\mathcal {B})\le \frac{20(i-1)}{n}\vert D_n\vert \).
Appendix 5: The General Linear Group
The SOV approach reduces Theorem 1.3 to counting the number of morphisms of the quivers of Fig. 19 into the Bratteli diagram of \(Gl_n(q)\). In this section we use known results on the number of conjugacy classes and the multiplicities of representations of \(Gl_n(q)\) to provide the bounds used in the proof of Theorem 1.3.
Theorem 7.23
For \(\mathcal {H}_j^n\) the quiver of Fig. 19 and \(\mathcal {B}\) the Bratteli diagram for the subgroup chain \(Gl_n(q)>Gl_{n-1}(q)>\cdots >\{e\}\),
Proof
By Theorem 6.1, we see that \(\# {\text {Hom}}(\mathcal {H}_j^n\uparrow \mathcal {Q};\mathcal {B})\) is equal to the sum
for \(M_\mathcal {B}(G_i,G_j)=M_\mathcal {B}(G_i(q),G_j(q)):=\max M_\mathcal {B}(\alpha _i,\alpha _j)\) over all \(\alpha _i\in \mathcal {B}^i, \alpha _j\in \mathcal {B}^j\) and \(\vert \hat{G}_i(q)\vert \) the number of conjugacy classes of \(G_i(q).\) By Corollary 6.18,
By [32, Lemma 5.9], \(M(Gl_j(q),Gl_{j-1}(q))\le 2^{j-1}\) and \(\vert \hat{Gl}_{j}(q)\vert \le q^j\). Thus, since \(\dim A(\mathcal {H}_j^n\uparrow \mathcal {Q};\mathcal {B})=\# {\text {Hom}}(\mathcal {H}_j^n\uparrow \mathcal {Q};\mathcal {B})\),
\(\square \)
Theorem 7.24
For \(\mathcal {J}_j^n\) the quiver of Fig. 19,
Proof
By [32, Lemma 5.9], \(M(Gl_{j}(q),Gl_{j-2}(q))\le 2^{2j-3}q^{j-1}\). Thus,
\(\square \)
1.1 Factoring Coset Representatives of \(GL_n(\mathbb {F}_q)\)
In this section we provide the set of coset representatives and their factorizations used in the proof of Theorem 1.3 by developing a correspondence between \(Gl_n(q)/Gl_{n-1}(q)\) and the set \(Z_n=\{\mathbf {z}=(x_1y_1,\dots , x_ny_n)\mid \mathbf {x},\mathbf {y}\in (\mathbb {F}_q)^n,\;y\cdot x=1\}.\)
Define an action of \(Gl_n(q)\) on \(Z_n\) via \(A.\mathbf {z}=(\tilde{x}_1\tilde{y}_1,\dots ,\tilde{x}_n\tilde{y}_n),\) for \(\tilde{\mathbf {y}}=\mathbf {y}A^{-1},\tilde{\mathbf {x}}=(A\mathbf {x}^T)^T \). Note that the action of A preserves \(y\cdot x\). For \(\mathbf {1}=(0,\dots ,0,1)\), we show (Theorem 7.27) that
Note that \(Gl_{n-1}(q)\), viewed as a subgroup of \(Gl_n(q)\), stabilizes \(\mathbf {1},\) so the orbit-stabilizer theorem gives a bijection between \(Z_n={\text {Orb}}(\mathbf {1})\) and \(Gl_n(q)/Gl_{n-1}(q)\) through the correspondence
Thus, writing \(\mathbf {z}=A_1\cdots A_m.\mathbf {1}\) for each \(\mathbf {z}\in Z_n\) gives a factorization of the corresponding coset representative. We find a factorization in which each matrix \(A_i=A\bigoplus I_{n-2}\) for \(A\in Gl_2(q)\).
Lemma 7.25
Suppose \(\mathbf {z}=(x_1y_1,x_2y_2)\in Z_2\) with \(x_1y_1+x_2y_2\ne 0\). Then there exists a matrix \(A\in Gl_2(q)\) and \(y_2',x_2'\in \mathbb {F}_q^\times \) such that
Proof
Case 1: \(x_1=0\). Let \(A=\begin{pmatrix} 1&{}0\\ \frac{y_1}{y_2}&{}1 \end{pmatrix}\). Note that for all possible choices of \(\mathbf {z}\in Z_2\), there are q possibilities for A.
Case 2: \(x_1\ne 0\), \(y_1=0\). Let \(A=\begin{pmatrix} 1&{}\frac{-x_1}{x_2}\\ 0&{}1 \end{pmatrix}\). Note there are \(q-1\) possibilities for A.
Case 3: \(x_1\ne 0\), \(y_1\ne 0\). Let \(A=\begin{pmatrix} \frac{-x_2}{x_1}&{}1\\ 1&{}\frac{y_2}{y_1} \end{pmatrix}\). Note there are \(q^2\) possibilities for A. Note further that for \(z_1:=x_1y_1\) and \(z_2:=x_2y_2\) fixed and nonzero,
and there are \(q-1\) possibilities for A. \(\square \)
We use Lemma 7.25 to systematically write \(\mathbf {z}\in Z_n\) in form
with \(A_i\in Gl_n(q)\). Recall that p is the characteristic of \(\mathbb {F}_q\).
Proposition 7.26
Let \(\tilde{\mathbf {z}}\in Z_n\). Then there is a permutation matrix \(\pi \in Gl_n(q)\), \(b\in \mathbb {F}_q^\times \), and \(i\ge 1\) such that \(\pi .\tilde{\mathbf {z}}=\mathbf {z}\) with:
-
(i)
\(z_1+\dots +z_j\ne 0\) for all \(i\le j\le n\),
-
(ii)
\(z_1=\dots =z_i=b\),
-
(iii)
\(p\vert (i-1)\).
Proof
Let \(\tilde{\mathbf {z}}=(\tilde{z}_1,\dots ,\tilde{z}_n)\in Z_n\). Note that \(\tilde{z}_1+\cdots +\tilde{z}_n=1\ne 0\). Let j be an index (if it exists) such that \(\tilde{z}_1+\cdots \tilde{z}_n-\tilde{z}_j\ne 0\). Note that for a permutation matrix \(\pi \),
Permute \(\tilde{\mathbf {z}}\) to make \(\tilde{z}_j\) the last entry, then delete \(\tilde{z}_j\) to produce a vector of length \(n-1\). Repeat until no such index exists, and let i be the length of the resultant vector, \(\mathbf {z}\). Then clearly \(z_1+\cdots +z_j\ne 0\) for all \(i\le j\le n\). Further, \(z_1+\cdots +z_i-z_k=0\) for all \(1\le k\le i\); in particular, \(z_1=\cdots =z_i=b\in \mathbb {F}_q^\times \). Finally note that \(z_1+\cdots +z_{i-1}=(i-1)b=0\) and so \(p\vert (i-1)\). \(\square \)
In light of Proposition 7.26, let
Theorem 7.27
For \(p\ne 2\) and \(\mathbf {z}\in S_i(n)\), there exist invertible matrices \(u_j,u_j', v_j, t_j\in Gl_j(q)\cap {\text {Centralizer}}(Gl_{j-2}(q)),\) such that
Proof
Let \(\mathbf {z}\in S_i(n)\) and let \(i>p\). Note that \(\mathbf {z}=\begin{pmatrix} b,\dots ,b,z_{i+1}, \dots ,z_n \end{pmatrix},\) and since \(z_1+z_2=2b\ne 0\), by Lemma 7.25, there is a matrix \( A\in Gl_2(q)\) such that \(A.(z_1,z_2)=\begin{pmatrix}0,x_2'y_2'\end{pmatrix}\) with \(y_2'x_2'=2b\). Let \(u_2= A\bigoplus I_{n-2}\in Gl_n(q).\) Then
Repeat this process, defining matrices \(u_3,u_4,\dots ,u_{p-1}\) (i.e., find the matrix A guaranteed by Lemma 7.25, and let \(u_j=I_{j-2}\bigoplus A\bigoplus I_{n-j}\)). Note that
Since \(z_{p-1}+z_p=pb=0\), we cannot use Lemma 7.25. Instead, define \((u_{p+1}')\) as above and let \(t_p\) be the permutation matrix of the transposition \((p-1\; p)\). Then
and since now \(z_{p-1}+z_p\ne 0\), define \(u_{p+1}\) as before so that
Repeat this process through definition of the matrix \(u_i\), so that
Since \(z_1+\cdots +z_j\ne 0\) for all \(i\le j\le n\), we use Lemma 7.25 to find the appropriate 2x2 matrix \(A_j\) so that for \(v_j=I_{j-2}\bigoplus A_j\bigoplus I_{n-j}\),
For \(i<p\) analogous arguments apply without needing the matrices \(t_p\). \(\square \)
Remark 7.28
By Lemma 7.25, there are \((q-1)\) possibilities for each \(u_j\) and \(q^2\) possibilities for each \(v_j\).
By Proposition 7.26,
and so by Expression 19 a complete set of coset representatives for \(Gl_n(q)/Gl_{n-1}(q)\) is contained in \(\{\pi s_i\vert \;1\le i\le n, p\mid (i-1), s_i\in S_i(n)\},\) with each \(s_i\) of form:
Finally, we note that similar results hold in the \(p=2\) case.
Theorem 7.29
For \(p=2\), \(i\ge 3\) odd, \((\mathbf {y},\mathbf {x})\in S_i(n)\), there exist invertible matrices
such that
Note that there are \((q-1)\) choices for \(a_j\) and \(b_j\), that \(c_j\) is completely determined by \(a_j\) and \(b_j\), and that there are \(q^2\) choices for \(v_j\).
Rights and permissions
About this article
Cite this article
Maslen, D., Rockmore, D.N. & Wolff, S. Separation of Variables and the Computation of Fourier Transforms on Finite Groups, II. J Fourier Anal Appl 24, 226–284 (2018). https://doi.org/10.1007/s00041-016-9516-4
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00041-016-9516-4