Difference Equation for Quintic 3-Fold

In this paper, we use the Mellin-Barnes-Watson method to relate solutions of a certain type of -difference equations at = 0 and = ∞. We consider two special cases; the first is the -difference equation of the -theoretic -function of the quintic, which is degree 25; we use Adams' method...

Full description

Saved in:
Bibliographic Details
Published in:Symmetry, Integrability and Geometry: Methods and Applications
Date:2022
Main Author: Wen, Yaoxinog
Format: Article
Language:English
Published: Інститут математики НАН України 2022
Online Access:https://nasplib.isofts.kiev.ua/handle/123456789/211625
Tags: Add Tag
No Tags, Be the first to tag this record!
Journal Title:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Cite this:Difference Equation for Quintic 3-Fold. Yaoxinog Wen. SIGMA 18 (2022), 043, 25 pages

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1859943049424011264
author Wen, Yaoxinog
author_facet Wen, Yaoxinog
citation_txt Difference Equation for Quintic 3-Fold. Yaoxinog Wen. SIGMA 18 (2022), 043, 25 pages
collection DSpace DC
container_title Symmetry, Integrability and Geometry: Methods and Applications
description In this paper, we use the Mellin-Barnes-Watson method to relate solutions of a certain type of -difference equations at = 0 and = ∞. We consider two special cases; the first is the -difference equation of the -theoretic -function of the quintic, which is degree 25; we use Adams' method to find the extra 20 solutions at = 0. The second special case is a Fuchsian case, which is confluent to the differential equation of the cohomological -function of the quintic. We compute the connection matrix and study the confluence of the -difference structure.
first_indexed 2026-03-17T20:38:36Z
format Article
fulltext Symmetry, Integrability and Geometry: Methods and Applications SIGMA 18 (2022), 043, 25 pages Difference Equation for Quintic 3-Fold Yaoxinog WEN Korea Institute for Advanced Study, Seoul, 02455, Republic of Korea E-mail: y.x.wen.math@gmail.com Received September 28, 2021, in final form June 04, 2022; Published online June 14, 2022 https://doi.org/10.3842/SIGMA.2022.043 Abstract. In this paper, we use the Mellin–Barnes–Watson method to relate solutions of a certain type of q-difference equations at Q = 0 and Q = ∞. We consider two special cases; the first is the q-difference equation of K-theoretic I-function of the quintic, which is degree 25; we use Adams’ method to find the extra 20 solutions at Q = 0. The second special case is a fuchsian case, which is confluent to the differential equation of the cohomological I-function of the quintic. We compute the connection matrix and study the confluence of the q-difference structure. Key words: q-difference equation; quantum K-theory; Fermat quintic 2020 Mathematics Subject Classification: 14N35; 33D90; 39A13 1 Introduction Since the 1990s, the development of mirror symmetry has changed how people work on enumer- ative geometry and has made some surprising predictions in algebraic geometry. Calabi–Yau manifolds are essential in mirror symmetry. Among them, the quintic threefold was the first example for which mirror symmetry was used to make enumerative predictions [3]. There are several ways to state mirror symmetry. In Givental’s approach to mirror symmetry, two cohomology valued formal functions play a crucial role, i.e., the so-called J-function and I-function. The J-function, by definition, encodes all the genus zero Gromov–Witten invariants, so it is essential. However, it is pretty hard to obtain an explicit formula. On the other hand, the I-function given by the oscillatory integral is computable. In [7] Givental proved that I-function lies on the range of big J-function, and up to a change of coordinate, we can obtain J-function from I-function. Let X be the Fermat quintic, considered as a degree 5 hypersurface in P4, the cohomological I-function of X is as follows IcohX (ℏ, et) = ∞∑ d=0 ∏5d k=1(5H + kℏ)∏d k=1(H + kℏ)5 et( H ℏ +d), where H is the hyperplane class of P4, and ℏ is the equivariant parameter, since H4 = 0 in the cohomology of X, the I-function of quintic satisfies the following degree 4 differential equation which is called the Picard–Fuchs equation[( ℏ d dt )4 − 55et ( ℏ d dt + 1 5 ℏ )( ℏ d dt + 2 5 ℏ )( ℏ d dt + 3 5 ℏ )( ℏ d dt + 4 5 ℏ )] IcohX ( ℏ, et ) = 0. Let Q = 55et, the above differential equation becomes the following form[( Q d dQ )4 −Q ( Q d dQ + 1 5 )( Q d dQ + 2 5 )( Q d dQ + 3 5 )( Q d dQ + 4 5 )] IcohX (ℏ, Q) = 0. (1.1) mailto:y.x.wen.math@gmail.com https://doi.org/10.3842/SIGMA.2022.043 2 Y. Wen The fundamental solutions at Q = 0 are given by the expansion of IcohX (ℏ, Q), i.e., IcohX (ℏ, Q) = I0 · 1 + I1 ·H + I2 ·H2 + I3 ·H3 mod ( H4 ) . More preciously, the coefficients of the I-function relative to the cohomology basis give the fundamental solutions of (1.1). Moreover, the fundamental solutions at Q = ∞ are related to FJRW theory [4] and could be constructed explicitly via the Frobenius method. Around 2000, Givental [8] and Lee [11] introduced the K-theoretic Gromov–Witten (GW) invariants, these invariants are defined by replacing cohomological definitions by their K-theore- tical analogs. The K-theoretic J-function and I-function are also defined and studied; unlike cohomological GW theory, the K-theoretic I-function satisfies a difference equation instead of a differential equation. For example, let us still consider quintic X, denote by qQ∂Q the difference operator shifting Qk by qkQk, the K-theoretic I-function of X is as follows IKX (q,Q) = P lq(Q) ∞∑ d=0 ∏5d k=1 ( 1− P 5qk )∏d k=1 ( 1− Pqk )5Qd, (1.2) where P = O(−1) on P4, and lq(Q) = −Q θ′q(Q) θq(Q) is the q-logarithm function. Here θq(Q) is the Jacobi’s theta function and the q-logarithm function satisfies qQ∂Q(lq(Q)) = lq(Q) + 1. Since (1 − P )5 = 0 in K ( P4 ) , the K-theoretic I-function of X satisfies the following difference equation[( 1− qQ∂Q )5 −Q 5∏ k=1 ( 1− q5Q∂Q+k )] IKX (q,Q) = 0. (1.3) It is a degree 25 difference equation but not fuchsian (Definition 2.15). The characteristic equation (see (3.1) for definition) at Q = ∞ is( 1− q−1x5 )( 1− q−2x5 )( 1− q−3x5 )( 1− q−4x5 )( 1− q−5x5 ) = 0, with 25 distinct roots. Using Frobenius method, we obtain 25 solutions (at Q = ∞) of (1.3) given by Wl,m(1/Q) = e q,q l 5 ξm (1/Q) ∑ d≥0 ∏d−1 k=0 ( 1− ξ−mq−k− l 5 )5∏5d−1 k=0 ( 1− q−k−l ) Q−d, where ξ5 = 1, l = 1, . . . , 5, m = 0, . . . , 4, and eq,λq(Q) = θq(Q) θq(λqQ) ∈ M(C∗). The function eq,λq satisfies the q-difference equation qQ∂Qeq,λq(Q) = λqeq,λq(Q). Difference Equation for Quintic 3-Fold 3 The characteristic equation at Q = 0 is (1− x)5 = 0, with 5 multiple roots. One may obtain explicit formulas for solutions via the Frobenius method. However, we could obtain 5 solutions at Q = 0 from IKX (q,Q) as mentioned above. We use Adams’ method to obtain the rest solutions. Proposition 1.1. The exceptional 20 solutions of (1.3) are given as follows: for each 20th root of unity ξ, we have a solution of form ep,ξp−1/2(z)F (z) = ep,ξp−1/2(z) ∑ n≥0 fnz n, where z = Q 1 20 , p = q 1 20 . Let σp = pz∂z , then F (z) satisfies[( z − ξp− 9 2σp )( z − ξp− 7 2σp )( z − ξp− 5 2σp )( z − ξp− 3 2σp )( z − ξp− 1 2σp ) − ( z5−ξ5p− 25 2 σ5 p )( z5−ξ5p− 15 2 σ5 p )( z5−ξ5p− 5 2σ5 p )( z5−ξ5p 5 2σ5 p )( z5−ξ5p 15 2 σ5 p )] F (z)=0. To the best of the author’s knowledge, the connection matrix is critical in classifying fuchsian difference equations, and very little is known about the non-fuchsian case. Even in the fuchsian case, the connection matrix is hard to obtain if the characteristic equation has multiple roots. We construct the following K ( Pn−1 ) valued q-series motived by [4] to obtain connection matrix Fm,n(Q) = P lq(Q) ∞∑ d=0 ∏m i=1(Pαi; q)d (Pq; q)nd Qd. (1.4) Here we use q-Pochhammer symbol notation: (a; q)d := (1− a)(1− qa) · · · ( 1− qd−1a ) for d > 0. Since (1− P )n = 0 in K ( Pn−1 ) , then (1.4) satisfies the following difference equation[( 1− qQ∂Q )n −Q m∏ i=1 ( 1− αiq Q∂Q )] Fm,n(Q) = 0 mod ( (1− P )n ) . (1.5) Suppose αi /∈ αjq Z\{0} then we could find the explicit formula for m solutions at Q = ∞ denoted by {Wk(1/Q)}mk=1 and we use Mellin–Barnes–Watson method to related solutions at Q = 0 and Q = ∞. Theorem 1.2. For m ≥ n, the K ( Pn−1 ) valued q-series has the following analytic continuation: P lq(Q) ∞∑ d=0 ∏m i=1(Pαi; q)d (Pq; q)nd Qd = P lq(Q) ∏m i=1(Pαi; q)∞ (Pq; q)n∞ m∑ j=1 (q, q, PαjQ, q/(PαjQ); q)∞ (Pαj , q/(Pαj), Q, q/Q; q)∞ × ( α−1 j q; q )n ∞e−1 q,αj (1/Q)∏m i=1,i ̸=j(αi/αj ; q)∞(q; q)∞ Wj(1/Q). As for applications, if we take n = 5, m = 25 and {αi}25i=1 = { ξlq k 5 | k, l = 1, 2, 3, 4, 5 } , then (1.4) becomes (1.2). And if we take n = 4, m = 4 and {αi}4i=1 = { q i 5 }4 i=1 , then (1.5) becomes[( 1− qQ∂Q )4 −Q 4∏ i=1 ( 1− q i 5 qQ∂Q )] F (Q) = 0. (1.6) 4 Y. Wen This difference equation is a lift of the differential equation (1.1), i.e., if we let q → 1, then (1.6) becomes (1.1). This phenomenon is called confluence which was studied first by J. Sauloy in 2000 [15]. Under the above specific choice, the formula in Theorem 1.2 becomes P lq(Q) ∞∑ d=0 ∏4 i=1 ( Pq i 5 ; q ) d (Pq; q)4d Qd = P lq(Q) ∏4 i=1 ( Pq i 5 ; q ) ∞ (Pq; q)4∞ 4∑ j=1 ( q, q, Pq j 5Q, q/ ( Pq j 5Q ) ; q ) ∞( Pq j 5 , q/ ( Pq j 5 ) , Q, q/Q; q ) ∞ × ( q− j 5 q; q )4 ∞e−1 q,qj/5 (1/Q)∏4 i=1,i ̸=j ( q i−j 5 ; q ) ∞(q; q)∞ Wj(1/Q), where {Wj(1/Q)}4j=1 are the fundamental solutions at Q = ∞. If we expand two sides with respect to K-group basis (1 − P )k, k = 0, 1, 2, 3, we obtain the connection matrix, for more details, see Section 5.1. Besides, the fundamental solutions of (1.6) at Q = 0 and Q = ∞ are confluent to the solutions of (1.1), finally, we compute the confluence of the connection matrix. The paper is arranged as follows. Section 2 reviews some basic definitions and concepts of difference equations and introduces some special functions. In Section 3, we use the difference equation of quintic as an example, and we use Adams’ method and Frobenius method to solve the degree 25 difference equation at Q = 0 and Q = ∞ respectively. In Section 4, we generalize the difference equation for the quintic and construct a K-group valued series, and then we use the Mellin–Barnes–Watson method to relate solutions at Q = 0 and Q = ∞. In Section 5, we apply the results in Section 4 to a particular fuchsian case, and we expand the formula with respect to theK-group basis (1−P )k to find the connection matrix. Since the particular fuchsian case is confluent to the differential equation of quintic. In Section 6, we study the confluence of the connection matrix. 2 Preliminaries In this section, we define some basic notions in the theory of q-difference equations. The main references are [13, 15, 16]. Notations 2.1. Here are some standard notations of general use: – Q and q are complex variables and |q| < 1, q ̸= 0, – C({Q}) is the field of meromorphic germs at 0, is the quotient field of C{Q}, – M(C∗) is the field of meromorphic functions on C∗, – M(C∗, 0) is the ring of germs at punctured neighborhood of Q = 0, – M (Eq) is the field of meromorphic functions on elliptic curve Eq = C∗/qZ, i.e, the field of elliptic functions. – (a; q)d = (1− a)(1− qa) · · · ( 1− qd−1a ) for d ∈ N ∪ {+∞} is the q-Pochhammer symbol. Definition 2.2. A difference field is a pair (K,σ), where K is a field, and σ is a field automor- phism of K. Example 2.3. We will focus on the fields in the above notations, M(C∗) ⊂ M(C∗, 0), they are all endowed with the q-shift operator σq := qQ∂Q : f(Q) 7→ f(qQ). Let K = M(C∗) or M(C∗, 0). Usually, we denote the field of constants of the difference field (K,σq) as K σq . For example, M(C∗)σq = M(C∗)σq = M(Eq). This is the main reason that the modular form such as elliptic function appears naturally in the theory of q-difference equation. Difference Equation for Quintic 3-Fold 5 2.1 Regular singular q-difference equations Definition 2.4. Let (Eq) : q Q∂QXq(Q) = Aq(Q)Xq(Q) be a q-difference system, with Aq ∈ GLn(K). We define the solution space of this q-difference equation by Sol(Eq) = { Xq ∈ Kn | qQ∂QXq(Q) = Aq(Q)Xq(Q) } . Remark 2.5. From now on, we will focus on the local solutions at Q = 0, and the results will also hold for Q = ∞. The reason why we don’t consider solutions at other singular points is that: if a function f(Q) is a solution of a q-difference equation qQ∂Qf(Q) = a(Q)f(Q) and has a singularity at some Q0 ̸= 0,∞, then f(Q) has a singularity at any complex number Q0q k. Proposition 2.6 ([16, Theorem 2.3.1, p. 118]). Let (Eq) : q Q∂QXq(Q) = Aq(Q)Xq(Q) be a q- difference system. Then, we have dimM(Eq) ( Sol(Eq) ) ≤ rank(Aq). Definition 2.7. Let qQ∂QXq(Q) = Aq(Q)Xq(Q) be a q-difference system. A fundamental solu- tion of this system is an invertible matricial solution Xq ∈ GLn(K) such that qQ∂QXq(Q) = Aq(Q)Xq(Q). Definition 2.8. Let qQ∂QXq(Q) = Aq(Q)Xq(Q) be a q-difference system. Consider a matrix Pq ∈ GLn(K). The gauge transform of the matrix Aq by the gauge transformation Pq is the matrix Pq · [Aq] := ( qQ∂QPq ) AqP −1 q . A second q-difference system qQ∂QXq(Q) = Bq(Q)Xq(Q) is said to be equivalent (over K) by gauge transform to the first one if there exists a matrix Pq ∈ GLn(K) such that Bq = Pq · [Aq]. Let us define the regular singular q-difference equation. We shall start from the local analytic study, i.e., taking field C({Q}), and then look for solutions in the field K = M(C∗) or M(C∗, 0). Definition 2.9. Let Aq ∈ GLn(C({Q})), a system qQ∂QXq(Q) = Aq(Q)Xq(Q) is said to be regular singular at Q = 0 if there exists a q-gauge transform Pq ∈ GLn(C({Q})) such that the matrix (Pq · [Aq])(0) is well-defined and invertible: Pq · [Aq](0) ∈ GLn(C). Definition 2.10. Consider a regular singular q-difference system qQ∂QXq(Q) = Aq(Q)Xq(Q). Suppose Aq(0) ∈ GLn(C) and denote by (λi) the eigenvalues of the matrix Aq(0). This q- difference system is said to be non q-resonant if for every i ̸= j, we have λi λj /∈ qZ\{0}, where qZ\{0} := { qk | k ∈ Z\{0} } ⊂ C. Let’s introduce some special functions which are needed to solve regular singular q-difference equations. We define Jacobi’s theta function by θq(Q) = ∑ d∈Z q d(d−1) 2 Qd. This function satisfies the q-difference equation qQ∂Qθq(Q) = 1 Qθq(Q). And it has a famous Jacobi’s triple identity θq(Q) = (q; q)∞(−Q; q)∞(−q/Q; q)∞. In the following, we define two special functions which are essential in solving regular singular (irregular) q-difference equations. 6 Y. Wen Definition 2.11. Let λq ∈ C∗. The q-character associated to λ is the function eq,λq ∈ M (C∗) defined by eq,λq(Q) = θq(Q) θq(λqQ) ∈ M(C∗). The function eq,λq satisfies the q-difference equation qQ∂Qeq,λq(Q) = λqeq,λq(Q). Definition 2.12. The q-logarithm is the function ℓq ∈ M (C∗) defined by ℓq(Q) = −Q θ′q(Q) θq(Q) . By a little computation, one could know that the function ℓq satisfies the following q-difference equation qQ∂Qℓq(Q) = ℓq(Q) + 1. Now we can state the existence of a fundamental solution for regular singular q-difference equations under certain conditions. For a q-difference system qQ∂QXq(Q) = Aq(Q)Xq(Q), without loss of generality, we assume Aq(0) ∈ GLn(C) and moreover that it is non-resonant. We can recursively build a gauge trans- form Fq ∈ GLn(C({Q})) which sends the matrix Aq(0) to the constant matrix Aq(Q), for details, see [16, Corollary 3.2.4]. Then we take the Jordan–Chevalley decomposition of Aq(0) = AsAu, where As is semi-simple, Au is unipotent and As, Au commute. Since N = Au − In is nilpotent, we can define A ℓq u := (In +N)ℓq := ∑ k≥0 ( ℓq k ) Nk, (2.1) where( ℓq k ) := ℓq(ℓq − 1) · · · (ℓq − (k − 1)) k! . Note that (2.1) is actually a finite sum and A ℓq u is unipotent, and we have qQ∂QA ℓq u = AuA ℓq u = A ℓq u Au. Thus we set eq,Au := A ℓq u . Take a basis change P to diagonalise As = P−1 diag(λi)P . We define eq,As := P−1 diag(eq,λi (Q))P, (2.2) which satisfies qQ∂Qeq,As = Aseq,As = eq,AsAs. Then one can check that the product Fqeq,Aseq,Au =: Xq(Q) is a fundamental solution of the q-difference system qQ∂QXq(Q) = Aq(Q)Xq(Q). We arrive at the following theorem. Proposition 2.13 ([16, Theorem 3.3.1]). The q-difference system σqXq(Q) = Aq(Q)Xq(Q), regular singular at Q = 0, admits a fundamental matricial solution X := Meq,C ∈ GLn(M(C∗, 0)), where C ∈ GLn(C) and where M ∈ GLn(C({Q})). The eq,C is defined by Jordan–Chevalley decomposition of C as above. Remark 2.14. Let A,P ∈ GLn(C), one can check that eq,PAP−1 = Peq,AP −1. Thus, (2.2) is independent of the choice of P . Difference Equation for Quintic 3-Fold 7 2.2 Monodromy of regular singular q-difference equations Definition 2.15. A q-difference system qQ∂QXq(Q) = Aq(Q)Xq(Q) is called fuchsian if it is regular singular both at Q = 0 and Q = ∞. It is easy to see the difference equation (1.3) is not fuchsian since it is not regular singular at Q = 0. But we will see it is regular singular at Q = ∞ (see (3.5)). Definition 2.16. Let qQ∂QXq(Q) = Aq(Q)Xq(Q) be a fuchsian q-difference system. This q- difference system admits a fundamental solution X0(Q) at Q = 0 and a second one X∞(1/Q) at Q = ∞. Birkhoff’s connection matrix (or q-monodromy) Pq is the ratio Mq(Q) = (X∞(1/Q))−1X0(Q). Since the connection matrix relates two fundamental matrix solutions. It is invariant by difference operator qQ∂Q , i.e., Mq(Q) ∈ GLn(M(Eq)). However, it is not well defined: it depends on the choice of fundamental matrix solutions. To get rid of this dependence, we need to consider the following triple. Definition 2.17. A Birkhoff connection triple is a triple( A(0),Mq, A (∞) ) ∈ GLn(C)×GLn(Eq)×GLn(C) up to certain equivalent. Where A(0) and A(∞) are related to the fundamental solutions at Q = 0 and Q = ∞ respectively, for more details, see [16, p. 133]. The data of Birkhoff’s connection triples classifies fuchsian q-difference systems up to gauge transformations. Proposition 2.18 ([16, Theorem 3.4.9]). Rational classes (under rational equivalence, i.e., over field C(Q)) of fuchsian rational systems are in bijection with equivalence classes of Birkhoff connection triples. 2.3 Confluence of regular singular q-difference equations First, let us introduce some interesting formulas we needed when considering the confluence of difference equations. We fix τ0 such that Im τ0 > 0 and q0 := e−2iπτ0 and |q0| < 1. This defines a discrete logarithmic spiral qZ0 := { qk0 | k ∈ Z } ⊂ C and a continuous spiral qR0 := { qk0 | k ∈ R } ⊂ C. Let Ω = C∗−qR0 . Denote by log(Q) the logarithm on Ω such that 1 7→ 0. LetQµ := eµ log(Q). Lemma 2.19 ([15, Section 3.1.7, Corollaire 1]). Let q(t) = qt0, t ∈ (0, 1]. Assume there exist complex numbers α0, α1 ∈ C so that Qi(q(t)) = Q0q αit+o(t) 0 , Q0 ∈ Ω. Then, on Ω, we have the uniform convergence when t → 0 lim q→1 θq(t)(Q1(q(t))) θq(t)(Q2(q(t))) = Qα2−α1 0 . Proposition 2.20 ([15, Sections 3.1.3 and 3.1.4]). As the above notation, consider λq(t), µ ∈ C∗ such that λq(t)−1 q−1 → µ. Then we have the asymptotics: 1. We have the uniform convergence on any compact of Ω lim t→0 (q(t)− 1)ℓq(t)(−Q) = log(Q). 8 Y. Wen 2. We have the uniform convergence on any compact of Ω lim t→0 eq(t),λq(t) (−Q) = Qµ. Now, let’s introduce the definition of confluence. Definition 2.21 ([15, Section 3.2]). Let q(t) = qt0, for t ∈ (0, 1]. A regular singular, non q- resonant difference system qQ∂QXq(Q) = Aq(Q)Xq(Q) is said to be confluent if it satisfies four conditions below. Set Bq(Q) = Aq(Q)−Id q−1 , whose coefficients have poles Q1(q), . . . , Qk(q) in the input Q. We require that 1. The q-spirals satisfy ⋂k i=1Qi(q0)q R 0 = ∅. 2. There exists a matrix B̃ ∈ GLn(C(Q)) such that lim t→0 Bq(t) = B̃, uniformly in Q on any compact of C∗ − ⋃k i=0Qiq R 0 , set Q0 = 1. 3. This limit defines a regular singular, non resonant differential system Q d dQ X̃ = B̃X̃. 4. There exists, for each t, a Jordan decompositions Bq(t)(0) = P−1 q(t)Jq(t)Pq(t) as well as B̃(0) = P̃−1J̃ P̃ . We ask that lim t→0 Pq(t) = P̃ . If the difference system is confluent, then there is a confluence of the solutions. Proposition 2.22 ([13, Theorem V.2.4.7]). Let q(t) = qt0, for t ∈ [0, 1]. Consider a regu- lar singular confluent q-difference system qQ∂QXq(Q) = Aq(Q)Xq(Q), whose limit system is Q∂QX̃(Q) = B̃(Q)X̃(Q). Assume that there exists a vector X0 ∈ Cn\{0}, independent of q, such that Aq(t)X0 = X0 for all t ∈ (0, 1]. We also assume that we have a solution Xq(Q) of the q-difference system satisfying the initial condition Xq(0) = X0. Let X̃ (Q) be the unique solution of Q∂QX(Q) = B̃(Q)X(Q) satisfying the initial condition X̃ (0) = X0. We have lim t→0 Xq(t)(Q) = X̃ (Q) uniformly in Q on any compact of C∗ − ⋃k i=0Qiq R 0 . 3 The difference equation for quintic 3.1 General technique: Newton polygon Let’s consider the equation n∑ i=0 ai(Q)(σq) if(Q) = 0, Difference Equation for Quintic 3-Fold 9 with ai(Q) = ai,0 + ai,1Q+ ai,2Q 2 + · · · . We call the following equation the characteristic equation an,0x n + an−1,0x n−1 + · · ·+ a1,0x+ a0,0 = 0, (3.1) which plays an important role in constructing solutions. Denote by ai,ji the first nonzero coefficient in ai(Q), and choosing i- and j-axes as horizontal and vertical axes respectively, plot the points (n− i, ji). Construct a broken line, convex down- ward, such that both ends of each segment of the line are points of the set (n− i, ji). Then we obtain a Newton polygon as follows Note that the horizontal segment corresponds to the characteristic equation ak,0x k + ak−1,0x k−1 + · · ·+ ad,0x d = 0. The degree of the above characteristic equation is 1 less than the number of points on or above that segment. Example 3.1. Consider the following equation:[( Q4 + 2Q7 ) σ6 q + ( Q+ 3Q5 ) σ5 q + ( 3 + 2Q3 ) σ4 q + 2σ3 q + 3Qσ2 q +Q2σq ] f(Q) = 0. Then the associated Newton polygon is (0,4) (1,1) (2,0) (3,0) (4,0) (5,2) The general technique to construct solutions is as follows: 10 Y. Wen � Horizontal segment: As mentioned above, it corresponds to characteristic equation. Using the non-zero roots, we could construct the associated solutions as regular singular cases. � Non-horizontal segment: For each non-horizontal segment of slope µ, a rational number. – If µ = r is an integer, we consider a formal series solution of the form θrq(Q) ∞∑ n=0 fn(q)Q n. – If µ = t/s is a rational number with s positive, then we consider a formal series solution of the form θqt/s ( Qt/s ) ∞∑ n=0 fn(q)Q n/s. For more details, see [1, 2, 14]. Remark 3.2. In Adams’ works, he used q µ 2 (t2−t), where t = lnQ ln q . 3.2 Solutions at Q = 0 The K-theoretic I-function of quintic is as follows [9] IK = P lq(Q) ∑ d≥0 ∏5d k=1 ( 1− P 5qk )∏d k=1 ( 1− Pqk )5Qd, it satisfies the following degree 25 difference equation[( 1− qQ∂Q )5 −Q 5∏ k=1 ( 1− qkq5Q∂Q )] IK = 0 mod ( (1− P )5 ) . (3.2) Remark 3.3. See [5] for additional discussion on q-deformed Picard–Fuchs equation. The characteristic equation at Q = 0 is (1− x)5 = 0. We can construct only 5 solutions by expanding IKX (q,Q) with respect to the K-group basis (1 − P )k, k = 0, 1, 2, 3, 4. Next we use Adams’ method to find other solutions at Q = 0. The Newton’s polygon of the above difference equation (3.2) is as follows (0,1) (20,0) Then we have solutions at Q = 0 of the form θ q 1 20 ( Q 1 20 ) G ( Q 1 20 ) = θ q 1 20 ( Q 1 20 ) ∞∑ d=0 gdQ d 20 , Difference Equation for Quintic 3-Fold 11 where θ q 1 20 ( Q 1 20 ) satisfies qQ∂Qθ q 1 20 ( Q 1 20 ) = ( 1 Q ) 1 20 θ q 1 20 ( Q 1 20 ) , then q5Q∂Qθ q 1 20 ( Q 1 20 ) = q− 1 2 ( 1 Q ) 1 4 θ q 1 20 ( Q 1 20 ) . Substituting into (3.2), we find that G ( Q 1 20 ) satisfies the following difference equation:[( Q 1 20 − q− 4 20 qQ∂Q )( Q 1 20 − q− 3 20 qQ∂Q )( Q 1 20 − q− 2 20 qQ∂Q )( Q 1 20 − q− 1 20 qQ∂Q )( Q 1 20 − qQ∂Q ) − ( Q 1 4 − q− 2 4 q5Q∂Q )( Q 1 4 − q− 1 4 q5Q∂Q )( Q 1 4 − q5Q∂Q )( Q 1 4 − q 1 4 q5Q∂Q )( Q 1 4 − q 1 2 q5Q∂Q )] ×G ( Q 1 20 ) = 0. Let z = Q 1 20 , p = q 1 20 and σp = pz∂z , then the above difference equation takes the following form[( z − p−4σp )( z − p−3σp )( z − p−2σp )( z − p−1σp ) (z − σp) − ( z5 − p−10σ5 p )( z5 − p−5σ5 p )( z5 − σ5 p )( z5 − p5σ5 p )( z5 − p10σ5 p )] G(z) = 0. (3.3) For z = 0, we obtain the characteristic equation p−10x5 − x25 = 0, i.e., x = ξp− 1 2 , for ξ20 = 1. Consider a solution of the form ep,ξp−1/2(z)F (z) = ep,ξp−1/2(z) ∑ n≥0 fnz n, then F (z) satisfies[( z − ξp− 9 2σp )( z − ξp− 7 2σp )( z − ξp− 5 2σp )( z − ξp− 3 2σp )( z − ξp− 1 2σp ) − ( z5 − ξ5p− 25 2 σ5 p )( z5 − ξ5p− 15 2 σ5 p )( z5 − ξ5p− 5 2σ5 p )( z5 − ξ5p 5 2σ5 p )( z5 − ξ5p 15 2 σ5 p )] × F (z) = 0. After a short computation, we expand the above difference equation as follows[ ξ5p− 25 2 ( σ25 p − σ5 p ) + 5ξ4p−8zσ4 p − 10ξ3p− 9 2 z2σ3 p + 10ξ2p−2z3σ2 p − 5ξp− 1 2 z4σp + z5 ( 1− ( 1 + p 15 2 + p 55 2 + p 95 2 + p 135 2 ) σ20 p ) + ξ15 ( p 15 2 + p 55 2 + 2p 95 2 + 2p 135 2 + 2p 175 2 + p 215 2 + p 255 2 ) z10σ15 p − ξ10 ( p10 + p30 + 2p50 + 2p70 + 2p90 + p110 + p130 ) z15σ10 p + ξ5 ( p 15 2 + p 55 2 + p 95 2 + p 135 2 + p 175 2 ) z20σ5 p − z25 ] F (z) = 0. (3.4) Thus we arrive at the following proposition. 12 Y. Wen Proposition 3.4. The exceptional 20 solutions of (3.3) are given as follows: for each 20th root of unity ξ, we have a solution of form ep,ξp−1/2(z)F (z) = ep,ξp−1/2(z) ∑ n≥0 fnz n, where F (z) satisfies[( z − ξp− 9 2σp )( z − ξp− 7 2σp )( z − ξp− 5 2σp )( z − ξp− 3 2σp )( z − ξp− 1 2σp ) − ( z5 − ξ5p− 25 2 σ5 p )( z5 − ξ5p− 15 2 σ5 p )( z5 − ξ5p− 5 2σ5 p )( z5 − ξ5p 5 2σ5 p )( z5 − ξ5p 15 2 σ5 p )] × F (z) = 0. Proof. From (3.4), only f0 is a free variable, and fn is determined by {fk}k≤n. Then we obtain 20 solutions. ■ Remark 3.5. The exceptional solutions are linearly independent over M(Eq), since their Wron- skian matrix (see [16, Lemma 2.3.3]) does not equal to 0. Actually, it is sufficient to see the Wronskian matrix for {ep,ξkp−1/2}19k=0 and it turns out to be the Vandermonde matrix. 3.3 Solutions at Q = ∞ Let w = 1/Q, (3.2) becomes[ 5∏ k=1 ( 1− q−kq5w∂w ) − q10wq20w∂w ( 1− qw∂w )5] F (w) = 0, (3.5) which is regular singular. The characteristic equation is as follows 5∏ k=1 ( 1− q−kx5 ) = 0, with 25 distinct roots q l 5 ξm, l = 1, . . . , 5, m = 0, . . . , 4. Here ξ is the fifth root of unity. For each root q l 5 ξm, we construct the following solution Wl,m(w) = e q,q l 5 ξm (w) ∑ d≥0 fdw d. Substituting the above formula into (3.5), since qw∂we q,q l 5 ξm (w) = q l 5 ξme q,q l 5 ξm (w), then ∑ d≥0 fdw d satisfies the following difference equation[ 5∏ k=1 ( 1− ql−kq5w∂w ) − q10+4lwq20w∂w ( 1− q l 5 ξmqw∂w )5] G(w) = 0. Difference Equation for Quintic 3-Fold 13 Then one obtain 25 solutions at Q = ∞ as follows Wl,m(w) = e q,q l 5 ξm (w) ∑ d≥0 ∏d−1 k=0 ( 1− ξ−mq−k− l 5 )5∏5d−1 k=0 ( 1− q−k−l ) wd = e q,q l 5 ξm (1/Q) ∑ d≥0 ∏d−1 k=0 ( 1− ξ−mq−k− l 5 )5∏5d−1 k=0 ( 1− q−k−l ) Q−d. Since we require |q| < 1 and lim n→∞ ∏(n+1)−1 k=0 ( 1− ξ−mq−k− l 5 )5∏5(n+1)−1 k=0 ( 1− q−k−l ) ∏5n−1 k=0 ( 1− q−k−l )∏n−1 k=0 ( 1− ξ−mq−k− l 5 )5 = lim n→∞ ( 1− ξ−mq−n− l 5 )5∏5n+4 k=5n ( 1− q−k−l ) = 0. The 25 solutions are convergent. Remark 3.6. These 25 solutions may relate to K-theoretic FJRW theory, for hints, see [10]. 4 Auxiliary q-series and analytic continuation In this section, we construct a K-group valued q-series, which is a generalization of the se- ries (1.2), and it satisfies a difference equation like (3.2). Besides, we use Mellin–Barnes–Watson method to relate the solutions at Q = 0 and Q = ∞. 4.1 Auxiliary q-series We construct the following K ( Pn−1 ) valued q-series motivated by [4]: Fm,n(α⃗, Q) = P lq(Q) ∞∑ d=0 ∏m i=1(Pαi; q)d (Pq; q)nd Qd, m ≥ n. (4.1) Since (1− P )n = 0 in K ( Pn−1 ) , then it satisfies the following difference equation:[( 1− qQ∂Q )n −Q m∏ i=1 ( 1− αiq Q∂Q )] Fm,n(Q) = 0 mod ( (1− P )n ) . (4.2) Suppose αi /∈ αjq Z\{0}, i.e., the difference equation is (q-)non-resonant, the characteristic equa- tion at Q = ∞ is as follows m∏ i=1 ( 1− α−1 i x ) = 0, with m-distinct roots {αi}mi=1, the same as the discussion in previous section, for each root αi, we construct a solution as follows Wi(1/Q) = eq,αi(1/Q) ∑ d≥0 fdQ −d. Recall qQ∂Qeq,αi(1/Q) = α−1 i eq,αi(1/Q), 14 Y. Wen then ∑ d≥0 fdQ −d satisfies the following difference equation[( 1− α−1 i qQ∂Q )n −Q m∏ j=1 ( 1− αj/αiq Q∂Q )] G ( Q−1 ) = 0. Thus, we obtain Wj(1/Q) = eq,αj (1/Q) ∑ d≥0 ∏d−1 k=0(1− αjq k)nq(m−n)d(d−1)/2∏m i=1 ∏d k=1(1− αj/αiqk) × ( (−1)m−n ( m∏ i=1 αj/αi ) α−n j qm/Q )d = eq,αj (1/Q) ∑ d≥0 ∏d−1 k=0 ( 1− α−1 j q−k )n∏m i=1 ∏d k=1 ( 1− αi/αjq−k )Q−d. Remark 4.1. The above solutions are linearly independent over M(Eq). The same reason as Remark 3.2. Remark 4.2. If we take n = 5, m = 25 and {αi}25i=1 = { ξlq k 5 | k, l = 1, 2, 3, 4, 5 } , then F25,5 ({ ξlq k 5 } , Q ) = P lq(Q) ∑ d≥0 ∏5d k=1 ( 1− P 5qk )∏d k=1 ( 1− Pqk )5Qd, and for αj = q l 5 ξm , we have Wj(1/Q) = e q,q l 5 ξm (1/Q) ∏d−1 k=0 ( 1− q− l 5 ξ−mq−k )n∏d k=1 ( 1− q1−lq−5k ) · · · ( 1− q5−lq−5k )Q−d = e q,q l 5 ξm (1/Q) ∑ d≥0 ∏d−1 k=0 ( 1− ξ−mq−k− l 5 )5∏5d−1 k=0 ( 1− q−k−l ) Q−d = Wl,m(1/Q). 4.2 Analytic continuation For the sake of simplicity, we shall assume in this section that 0 < q < 1 and write q = e−w, w > 0. The results can be extended to complex q in the unit disc using analytic continuation. Consider the following contour integral. We follow the argument of [6, pp. 115–118] to show that this integral is well defined. For |Q| < 1, we can close the contour to the right, it equals to (4.1), P lq(Q) ∏m i=1(Pαi; q)∞ (Pq; q)n∞ ∫ C ( Pqs+1; q )n ∞∏m i=1(Pαiqs; q)∞ π(−Q)s sinπs ds −2πi . (4.3) Here we view P = e−H . Although H is the hyperplane class, we consider it as a formal variable valued in C. C is a curve from −i∞ to +i∞ such that only the non-negative zeros of sinπs lie on the right side of C. Difference Equation for Quintic 3-Fold 15 By the triangle inequality,∣∣1− ∣∣a∣∣e−ωRe(s) ∣∣ ≤ |1− aqs| ≤ 1 + |a|e−ωRe(s), we have∣∣∣∣ ( Pqs+1; q )n ∞∏m i=1(Pαiqs; q)∞ ∣∣∣∣ ≤ ∞∏ k=0 ( 1 + |P |e−(k+1+Re(s)w) )n∏m i=1 ( 1− |Pαi|e−(k+Re(s)w) ) , which is bounded on the contour C. Hence the integral (4.3) converges if | arg(−Q)| < π. Let CR+ be a large clockwise-oriented semicircle of radius R with a center at the origin that lies to the right of C. The semicircle is terminated by C and bounded away from the poles. Now consider the contour integral over CR+ instead of C. Setting s = Reiθ, we have for |s| < 1 that Re [ log (−Q)s sinπs ] = R[cos θ log |Q| − sin θ arg(−Q)− π| sin θ|] +O(1) ≤ −R[sin θ arg(−Q) + π| sin θ|] +O(1). Hence, when |Q| < 1 and | arg(−Q)| < π − δ, 0 < δ < π, we have (−Q)s sinπs = O[exp(−δR| sin θ|)], as R → ∞, then the integral on CR+ tends to zero as R → ∞. Therefore, by applying Cauchy’s theorem, we can prove (4.3) equals to (4.1) through tedious computation. Similarly, if we replace the contour C by a contour CR− consisting of a large counterclockwise- oriented semicircle of radius R with center at the origin that lies to the left of C. From an asymptotic formula Re[log(qs; q)∞] = ω 2 (Re(s))2 + ω 2 Re(s) +O(1), as R → −∞. Without loss of generality, we assume P = qh, αi = qai and let h, ai be real numbers. Then n(1 + h+Re(s))2 + n(1 + h+Re(s))− m∑ i=1 [ (ai + h+Re(s))2 + (ai + h+Re(s)) ] = 2 [ n(1 + h)− m∑ i=1 (ai + h) ] Re(s) + (n−m) ( Re(s) + Re2(s) ) + const. Note that q = e−w, w > 0, then the asymptotic formula for (qs; q)∞ implies that( Pqs+1; q )n ∞∏m i=1(Pαiqs; q)∞ = O (∣∣∣∣qm−n 2 ∏m i=1 Pαi (Pq)n ∣∣∣∣Re(s) ) , when Re(s) → −∞ with s bounded away from the zeros and poles. Similarly, it can be shown that if |Q| is big enough, we can close the contour the left, i.e., the integral (4.3) on CR− tends to zero as R → −∞. Thus, (4.3) equals to the sum of residues at s = w−1(−H + logαj + l · 2πi)− k and s = −1− h, 16 Y. Wen where j = 1, . . . ,m, l ∈ Z, n, h ∈ N. The residue at s = −1− h contains a term (1− P )n from ( Pqs+1; q )n ∞. And the residue at s = w−1(−H + logαj + l · 2πi)− k is Ress=w−1(−H+logαj+l·2πi)−k ( Pqs+1; q )n ∞∏m i=1(Pαiqs; q)∞ π(−Q)s sinπs = ( α−1 j q1−k; q )n ∞∏m i=1,i ̸=j ( αi/αjq−k; q ) ∞ (qk+1; q)∞ (q, q; q)∞ (−1)kq k(k+1) 2 × π(−Q)w −1(−H+logαj)−kw−1 sinπ ( w−1(−H + logαj)− k + w−12lπi ) exp{2lπiw−1 log(−Q) } . If we sum over k, we obtain ∞∑ k=0 ( α−1 j q1−k; q )n ∞∏m i=1,i ̸=j ( αi/αjq−k; q ) ∞ ( qk+1; q ) ∞ (q, q; q)∞ q k(k+1) 2 Q−k = ( α−1 j q; q )n ∞∏m i=1,i ̸=j(αi/αj ; q)∞(q; q)∞ ∞∑ k=0 ( α−1 j ; q−1 )n k∏m i=1 ( αi/αjq−1; q−1 ) k Q−k = ( α−1 j q; q )n ∞∏m i=1,i ̸=j(αi/αj ; q)∞(q; q)∞ e−1 q,αj (1/Q)Wj(1/Q). If we sum over l, we have ∞∑ l=−∞ exp { 2lπiw−1 log(−Q) } sin(w−1 ( −H + logαj)π + 2lπ2iw−1 )(−Q)w −1(−H+logαj) = −w(q, q, PαjQ, q/(PαjQ); q)∞ π(Pαj , q/(Pαj), Q, q/Q; q)∞ , which comes from the following lemma. Lemma 4.3 ([6, equation (4.3.9), p. 119]). ∞∑ m=−∞ csc ( απ − 2mπ2iw−1 ) exp { 2mπiw−1 log(−Q) } (−Q)−α = w(q, q, aQ, q/(aQ); q)∞ π(a, q/a,Q, q/Q; q)∞ , where a = qα = e−wα. Summing up the above discussion, we arrive at the following theorem. Theorem 4.4. Suppose αi /∈ αjq Z\{0}. For m ≥ n, the K ( Pn−1 ) valued q-series has the following analytic continuation: P lq(Q) ∞∑ d=0 ∏m i=1(Pαi; q)d (Pq; q)nd Qd = P lq(Q) ∏m i=1(Pαi; q)∞ (Pq; q)n∞ m∑ j=1 (q, q, PαjQ, q/(PαjQ); q)∞ (Pαj , q/(Pαj), Q, q/Q; q)∞ × ( α−1 j q; q )n ∞ · e−1 q,αj (1/Q)∏m i=1,i ̸=j(αi/αj ; q)∞(q; q)∞ Wj(1/Q). for |Q| < 1 and | arg(−Q)| < π − δ, 0 < δ < π. Remark 4.5. In general, the above formula only contains a part of solutions at Q = 0. Difference Equation for Quintic 3-Fold 17 5 A special fuchsian case Consider the following difference equation:[( 1− qQ∂Q )4 −Q ( 1− q 1 5 qQ∂Q )( 1− q 2 5 qQ∂Q )( 1− q 3 5 qQ∂Q )( 1− q 4 5 qQ∂Q )] F (Q) = 0. (5.1) By definition, it is fuchsian. One could easily construct the solutions at Q = ∞, indeed, let w = 1/Q, then (5.1) becomes[( 1− q− 1 5 qw∂w )( 1− q− 2 5 qw∂w )( 1− q− 3 5 qw∂w )( 1− q− 4 5 qw∂w ) − q2w ( 1− qw∂w )4] G(w) = 0. The characteristic equation of the above difference equation at w = 0 is( 1− q− 1 5x )( 1− q− 2 5x )( 1− q− 3 5x )( 1− q− 4 5x ) = 0, with 4 different roots x = q i 5 , i = 1, 2, 3, 4. So the difference equation is non-resonant. By using Frobenius method, we could construct solutions of the form Wi(w) = e q,q i 5 (w)W̃i(w) = e q,q i 5 (w) ∞∑ n=0 ginw n. After a short computation one obtain four solutions as follows W1(1/Q) = e q,q 1 5 (1/Q)4ϕ3 ( q 1 5 , q 1 5 , q 1 5 , q 1 5 ; q 4 5 , q 3 5 , q 2 5 ; q; q2/Q ) , (5.2) W2(1/Q) = e q,q 2 5 (1/Q)4ϕ3 ( q 2 5 , q 2 5 , q 2 5 , q 2 5 ; q 6 5 , q 4 5 , q 3 5 ; q; q2/Q ) , (5.3) W3(1/Q) = e q,q 3 5 (1/Q)4ϕ3 ( q 3 5 , q 3 5 , q 3 5 , q 3 5 ; q 7 5 , q 6 5 , q 4 5 ; q; q2/Q ) , (5.4) W4(1/Q) = e q,q 4 5 (1/Q)4ϕ3 ( q 4 5 , q 4 5 , q 4 5 , q 4 5 ; q 8 5 , q 7 5 , q 6 5 ; q; q2/Q ) . (5.5) Remark 5.1. By a small trick, we could find the formula for W̃i(w) easily. Note that, for α /∈ N, we have( 1− qw∂w )k eq,qα(w)W̃ (w) = eq,qα(w) ( 1− qα · qw∂w )k W̃ (w). For example, W̃1(w) satisfies the following difference equation[( 1− qw∂w )( 1− q− 1 5 qw∂w )( 1− q− 2 5 qw∂w )( 1− q− 3 5 qw∂w ) − q2w ( 1− q 1 5 qw∂w )4] W̃1(w) = 0. From the above explicit form, it’s quite easy to find a q-hypergeometric series representation for W̃1(w). Let’s consider the solutions of (5.1) at Q = 0, the characteristic equation is as follows (1− x)4 = 0. From the general theorem (see [2, 14] for details), we have solutions of the form Gi(Q) = lq(Q)Gi−1(Q) + gi(Q), i = 2, 3, 4, (5.6) 18 Y. Wen where G1(Q) = ∞∑ d=0 ∏4 i=1 ( q i 5 ; q ) d (q; q)4d Qd = 4ϕ3 ( q 1 5 , q 2 5 , q 3 5 , q 4 5 ; q, q, q; q;Q ) (5.7) is a solution of (5.1) and gi(Q) are power series. For general q-hypergeometric function 4ϕ3(a1, a2, a3, a4; b1, b2, b3; q; z), we have the following famous transformation formula. Proposition 5.2 ([6, equation (4.5.2), p. 120]). 4ϕ3(a1, a2, a3, a4; b1, b2, b3; q; z) = (a2, a3, a4, b1/a1, b2/a1, b3/a1, a1z, q/a1z; q)∞ (b1, b2, b3, a2/a1, a3/a1, a4/a1, z, q/z; q)∞ × 4ϕ3 ( a1, a1q/b1, a1q/b2, a1q/b3; a1q/a2, a1q/a3, a1q/a4; q; qb1b2b3 za1a2a3a4 ) + idem(a1, a2, a3, a4). (5.8) The symbol “ idem(a1, a2, a3, a4)” after an expression stands for the sum of the 3 expressions obtained from the preceding expression by interchanging a1 with ak, k = 2, 3, 4. So in our special case, (5.7) can be written as a combination of (5.2)–(5.5). As mentioned before, the other solutions has the form of (5.6) which is hard to compute. Thus, it’s very hard to find the connection matrix. 5.1 Connection matrix Notice that (5.1) is a special case of (4.2) with n = m = 4 and αi = q i 5 , i = 1, 2, 3, 4, then F4,4 ({ q i 5 }4 i=1 , Q ) = P lq(Q) ∞∑ d=0 ∏4 i=1 ( Pq i 5 ; q ) d (Pq; q)4d Qd. (5.9) The solutions of difference equation (5.1) atQ = 0 are given by the expansion of F4,4 ({ q i 5 }4 i=1 , Q ) with respect to (1− P )i, i = 0, 1, 2, 3. From Theorem 4.4, then we have Corollary 5.3. P lq(Q) ∞∑ d=0 ∏4 i=1 ( Pq i 5 ; q ) d (Pq; q)4d Qd = P lq(Q) ∏4 i=1 ( Pq i 5 ; q ) ∞ (Pq; q)4∞ 4∑ j=1 ( q, q, Pq j 5Q, q/ ( Pq j 5Q ) ; q ) ∞ × ( Pq j 5 , q/ ( Pq j 5 ) , Q, q/Q; q ) ∞ ( q− j 5 q; q )4 ∞e−1 q,qj/5 (1/Q)∏4 i=1,i ̸=j ( q i−j 5 ; q ) ∞(q; q)∞ ×Wj(1/Q). (5.10) Remark 5.4. Taking P = 1 in the above formula, we obtain ∞∑ d=0 ∏4 i=1 ( q i 5 ; q ) d (q; q)4d Qd = 4∑ k=1 ( q 1 5 · · · k̂ · · · q 4 5 , q1− k 5 , q1− k 5 , q1− k 5 ; q ) ∞( q 1−k 5 , . . . k̂ . . . , q 4−k 5 , q, q, q; q ) ∞ ( q k 5Q; q 5−k 5 /Q ) ∞ (Q, q/Q)∞ × 4ϕ3 ( q k 5 , q k 5 , q k 5 , q k−1 5 +1; q k−2 5 +1, . . . k̂ . . . , q k−4 5 +1; q; q2/Q ) . It agrees with (5.8). Difference Equation for Quintic 3-Fold 19 In order to obtain the connection matrix, we need to expand (5.10) with respect to {(1− P )k}3k=1. Notice that P lq(Q) = (1− (1− P ))lq(Q) = ∑ k≥0 (−1)k ( ℓq(Q) k ) (1− P )k, where( ℓq(Q) k ) = 1 k! k−1∏ r=0 (ℓq(Q)− r). Then P lq(Q) ∞∑ d=0 ∏4 i=1 ( Pq i 5 ; q ) d (Pq; q)4d Qd = 3∑ m=0 ∑ a+b=m (−1)a ( ℓq(Q) a ) Xb(q,Q)(1− P )m, where Xb(q,Q) is the coefficient of ∞∑ d=0 ∏4 i=1 ( Pq i 5 ; q ) d (Pq; q)4d Qd = 3∑ b=0 Xb(q,Q)(1− P )b. (5.11) Let’s consider the expansion of (Pqα; q)∞ and ( P−1qα; q ) ∞. By definition (Pqα; q)∞ = ∞∏ k=0 ( 1− Pqα+k ) . For (1− Pqα+k), we have ( 1− Pqα+k ) = ( 1− qα+k + qα+k(1− P ) ) = ( 1− qα+k )( 1 + qα+k 1− qα+k (1− P ) ) . Then ∞∏ k=0 ( 1− qα+k )( 1 + qα+k 1− qα+k (1− P ) ) = (qα; q)∞ ∞∏ k=0 [ 1 + ∞∑ k=0 qα+k 1− qα+k (1− P ) + ∞∑ i<j q2α+i+j (1− qα+i)(1− qα+j) (1− P )2 + ∞∑ i<j<l q3α+i+j+l (1− qα+i)(1− qα+j)(1− qα+l) (1− P )3 +O ( (1− P )4 )] . Similarly, ( 1− P−1qα+k ) = 1− qα+k 1− (1− P ) = ( 1− qα+k ) − qα+k(1− P )− qα+k(1− P )2 − qα+k(1− P )3 +O ( (1− P )4 ) . Then ∞∏ k=0 (( 1− qα+k ) − qα+k(1− P )− qα+k(1− P )2 − qα+k(1− P )3 +O ( (1− P )4 )) 20 Y. Wen = (qα; q)∞ ∞∏ k=0 [ 1− ∞∑ k=0 qα+k 1− qα+k (1− P ) + ( ∞∑ i<j q2α+i+j (1− qα+i)(1− qα+j) − ∞∑ k=0 qα+k 1− qα+k ) (1− P )2 + ( − ∞∑ i<j<l q3α+i+j+l (1− qα+i)(1− qα+j)(1− qα+l) + 2 ∞∑ i<j q2α+i+j (1− qα+i)(1− qα+j) − ∞∑ k=0 qα+k 1− qα+k ) (1− P )3 ] +O ( (1− P )4 ) . In order to simplify the computation, we introduce the following notations f1(x) = ∞∑ k=0 xqk 1− xqk , f2(x) = ∞∑ i<j x2qi+j (1− xqi)(1− xqj) , f3(x) = ∞∑ i<j<l x3qi+j+l (1− xqi)(1− xqj)(1− xql) and F1(x1, x2, x3, x4) = − 4∑ k=1 f1(xk), F2(x1, x2, x3, x4) = − 4∑ k=1 f2(xk) + 4∑ i<j f1(xi)f1(xj), F3(x1, x2, x3, x4) = − 4∑ k=1 f3(xk) + 4∑ i<j (f1(xi)f2(xj) + f2(xi)f1(xj)) − 4∑ i<j<l f1(xi)f1(xj)f1(xl). With a little computation, we obtain∏4 i=1 ( Pq i 5 ; q ) ∞ (Pq; q)4∞ = ∏4 k=1 ( q k 5 ; q ) ∞ (q; q)4∞ [ 1 + ( F1 ( q • 5 ) − F1(q) ) (1− P ) + ( F1(q) 2 − F1 ( q • 5 ) F1(q)− F2(q) + F2 ( q • 5 )) (1− P )2 + ( F1(q) 3 + F1 ( q • 5 )( F1(q) 2 − F2(q) ) + 2F1(q)F2(q)− F1(q)F2 ( q • 5 ) − F3(q) + F3 ( q • 5 )) (1− P )3 ] +O ( (1− P )4 ) . (5.12) Here we use the notations: Fi ( q • 5 ) := Fi ( q 1 5 , q 2 5 , q 3 5 , q 4 5 ) , Fi(q) := Fi(q, q, q, q). Difference Equation for Quintic 3-Fold 21 For simplicity, we write the above formula as∏4 i=1 ( Pq i 5 ; q ) ∞ (Pq; q)4∞ = ∏4 k=1 ( q k 5 ; q ) ∞ (q; q)4∞ [ 1 + 3∑ k=1 Fk · (1− p)k +O ( (1− P )4 )] , (5.13) where Fk stands for the coefficient of (1− P )k in (5.12). Similarly, we consider (Px; q)∞ ( P−1x−1q; q ) ∞. Then ( 1− Pxqd )( 1− qP−1x−1qd ) = ( 1− xqd + xqd(1− P ) )(( 1− qx−1qd ) − qx−1qd(1− P )− qx−1qd(1− P )2 − qx−1qd(1− P )3 +O ( (1− P )4 )) = ( 1− xqd )( 1− qx−1qd )[ 1 + xqd ( 1− qx−2 )( 1− xqd )( 1− qx−1qd )(1− P ) − qd+1x−1 (1− xqd)(1− qx−1qd) (1− P )2 − qd+1x−1 (1− xqd)(1− qx−1qd) (1− P )3+O ( (1− P )4 )] . Set g1(x) = ∞∑ d=0 xqd ( 1− qx−2 ) (1− xqd) ( 1− qx−1qd ) , g2(x) = ∞∑ i<j ∏ k=i,j ( xqk ( 1− qx−2 ) (1− xqk) ( 1− qx−1qk ))− ∞∑ d=0 qd+1x−1 (1− xqd) ( 1− qx−1qd ) , g3(x) = ∞∑ i<j<l ∏ k=i,j,l ( xqk ( 1− qx−2 ) (1− xqk) ( 1− qx−1qk ))− ∑ i ̸=j qi+j+1 ( 1− qx−2 )∏ k=i,j(1− xqk) ( 1− qx−1qk ) − ∞∑ d=0 qd+1x−1 (1− xqd) ( 1− qx−1qd ) . Then (Px; q)∞ ( P−1x−1q; q ) ∞ = θq(−x) (q; q)∞ ( 1 + g1(x)(1− P ) + g2(x)(1− P )2 + g3(x)(1− P )3 +O ( (1− P )4 )) . So we obtain( Pq k 5Q,P−1Q−1q1− k 5 ; q ) ∞( Pq k 5 , P−1q1− k 5 ; q ) ∞ = θq ( −q k 5Q ) θq ( −q k 5 ) [1 + (g1(q k 5Q ) − g1 ( q k 5 )) (1− P ) + ( −g1 ( q k 5Q ) g1 ( q k 5 ) +g21 ( q k 5 ) +g2 ( q k 5Q ) −g2 ( q k 5 )) (1−P )2 + ( −g31 ( q k 5 ) −g1 ( q k 5 ) g2 ( q k 5Q ) +g1 ( q k 5Q )( g21 ( q k 5 ) −g2 ( q k 5 )) + 2g1 ( q k 5 ) g2 ( q k 5 ) + g3 ( q k 5Q ) − g3 ( q k 5 )) (1− P )3 ] +O ( (1− P )4 ) . (5.14) 22 Y. Wen For simplicity, we write the above formula as follows( Pq k 5Q,P−1Q−1q1− k 5 ; q ) ∞( Pq k 5 , P−1q1− k 5 ; q ) ∞ = θq ( −q k 5Q ) θq ( −q k 5 ) [1 + 3∑ k=1 Gk · (1− P )k +O ( (1− P )4 )] , (5.15) where Gk stands for the coefficient of (1− P )k in (5.14). In conclusion, we arrive at the following corollary. Corollary 5.5. The 4 solutions of (5.1) at Q = 0 are given as the expansion of (5.9), i.e.∑ a+b=m (−1)a ( ℓq(Q) a ) Xb(q,Q), m = 0, 1, 2, 3, where Xb(q,Q) is defined in (5.11). The 4 solutions of (5.1) at Q = ∞ are given explicitly as (5.2)–(5.5). The connection matrix is as follows • X0(q,Q) = ∞∑ d=0 ∏4 i=1 ( q i 5 ; q ) d (q; q)4d Qd = 4∑ k=1 ( q 1 5 · · · k̂ · · · q 4 5 , q1− k 5 , q1− k 5 , q1− k 5 ; q ) ∞( q 1−k 5 , . . . k̂ . . . , q 4−k 5 , q, q, q; q ) ∞ ( q k 5Q; q 5−k 5 /Q ) ∞ (Q, q/Q)∞ × e−1 q,qk/5 (1/Q)Wk(1/Q). • X1(q,Q) = 4∑ k=1 ( q 1 5 · · · k̂ · · · q 4 5 , q1− k 5 , q1− k 5 , q1− k 5 ; q ) ∞( q 1−k 5 , . . . k̂ . . . , q 4−k 5 , q, q, q; q ) ∞ ( q k 5Q; q 5−k 5 /Q ) ∞ (Q, q/Q)∞ × (G1 + F1)e −1 q,qk/5 (1/Q)Wk(1/Q). • X2(q,Q) = 4∑ k=1 ( q 1 5 · · · k̂ · · · q 4 5 , q1− k 5 , q1− k 5 , q1− k 5 ; q ) ∞( q 1−k 5 , . . . k̂ . . . , q 4−k 5 , q, q, q; q ) ∞ ( q k 5Q; q 5−k 5 /Q ) ∞ (Q, q/Q)∞ × (G2 + F2 +G1F1)e −1 q,qk/5 (1/Q)Wk(1/Q). • X3(q,Q) = 4∑ k=1 ( q 1 5 · · · k̂ · · · q 4 5 , q1− k 5 , q1− k 5 , q1− k 5 ; q ) ∞( q 1−k 5 , . . . k̂ . . . , q 4−k 5 , q, q, q; q ) ∞ ( q k 5Q; q 5−k 5 /Q ) ∞ (Q, q/Q)∞ × (G3 + F3 +G2F1 +G1F2)e −1 q,qk/5 (1/Q)Wk(1/Q). Here, for simplicity, we use the notations Gk and Fk defined in (5.13) and (5.15). 6 Confluence of the q-difference structure Notice that lim q→1 1− qQ∂Q 1− q = Q d dQ , then one could easily see that the following difference equation is confluent to (1.1), i.e., lim q→1 [( 1− qQ∂Q )4 −Q ( 1− q 1 5 qQ∂Q )( 1− q 2 5 qQ∂Q )( 1− q 3 5 qQ∂Q )( 1− q 4 5 qQ∂Q )] /(1− q)4 = [( Q d dQ )4 −Q ( Q d dQ + 1 5 )( Q d dQ + 2 5 )( Q d dQ + 3 5 )( Q d dQ + 4 5 )] . In the following, we set q(t) = e−t and P = qH = e−tH . Difference Equation for Quintic 3-Fold 23 Lemma 6.1. lim t→0 P lq(Q) ∞∑ d=0 ∏4 i=1 ( Pq i 5 ; q ) d (Pq; q)4d Qd = QH ∞∑ d=0 ∏4 i=1 ( H + i 5 ) d (H + 1)4d Qd mod ( H4 ) . Proof. Since lim t→0 ∏d k=1 ( 1− Pqk )∏d k=1 ( 1− qk ) = ∏d k=1(H + k) d! , lim t→0 (1− P )k (1− q)k = Hk, and P lq(Q) = ∑ k≥0 (−1)k ( ℓq(Q) k ) (1− P )k = ∑ k≥0 (q − 1)k ( ℓq(Q) k ) (1− P )k (1− q)k . Then from Proposition 2.20, we arrive at the conclusion. ■ The q-Gamma function is defined as follows Γq(x) = (q; q)∞ (qx; q)∞ (1− q)1−x. It has a nice property lim q→1 Γq(x) = Γ(x). Using q-Gamma function, we rewrite (5.10) in the following form P lq(Q) ∞∑ d=0 ∏4 i=1 ( Pq i 5 ; q ) d (Pq; q)4d Qd = P lq(Q) Γq(H + 1)4 Γq ( H + 1 5 ) Γq ( H + 2 5 ) Γq ( H + 3 5 ) Γq ( H + 4 5 ) 4∑ k=1 Γq ( 1−k 5 ) · · · k̂ · · ·Γq ( 4−k 5 ) Γq ( 1− k 5 )4 × Γq ( H + k 5 ) Γq ( −H + 1− k 5 ) θq ( −qH+k/5Q ) θq(−Q) e−1 q,qk/5 (1/Q)Wk(1/Q). (6.1) After taking limit, we arrive at the following proposition. Proposition 6.2. QH ∑ d≥0 ∏5d k=1(5H + k)∏d k=1(H + k)5 ( Q/55 )d = 55HΓ(H + 1)5 Γ(5H + 1) 4∑ k=1 5k−1Γ(5− k)∏4 i=1,i ̸=k(i− k)Γ ( 1− k 5 )5 πe−πi(H+ k 5 ) sin ( π ( H + k 5 ))W̃k(1/Q), (6.2) where W̃k = ∑ d≥0 ∏d−1 l=0 ( k 5 + l )5∏5d−1 l=0 (k + l) ( 55/Q )d . 24 Y. Wen Proof. After taking limit, the left hand side of (6.1) becomes QH ∞∑ d=0 ∏4 i=1 ( H + i 5 ) d (H + 1)4d Qd = QH ∞∑ d=0 4∏ i=1 ( Γ ( H + i 5 + d ) Γ ( H + i 5 ) )( Γ(H + 1) Γ(H + d+ 1) )4 Qd. Recall some formulas of Gamma function: Γ(x)Γ(1− x) = π sin(πx) , Γ(nx)(2π)(n−1)/2 = nnx− 1 2Γ(x)Γ ( x+ 1 n ) · · ·Γ ( x+ n− 1 n ) . Then 4∏ i=1 ( Γ ( H + i 5 + d ) Γ ( H + i 5 ) ) Γ(H + d+ 1) Γ(H + 1) = 1 55d Γ(5H + 5d+ 1) Γ(5H + 1) . Thus, we arrive at the left-hand side of (6.2). After taking limit, the right-hand side of (6.1) becomes QH Γ(H + 1)4 Γ ( H + 1 5 ) · · ·Γ ( H + 4 5 ) 4∑ k=1 Γ ( 1−k 5 ) · · · k̂ · · ·Γ ( 4−k 5 ) Γ ( 1− k 5 )4 Γ ( H + k 5 ) Γ ( −H + 1− k 5 ) × (−Q)−H− k 5 W̃k(1/Q) = Γ(H + 1)4 Γ ( H + 1 5 ) · · ·Γ ( H + 4 5 ) 4∑ k=1 Γ ( 1−k 5 ) · · · k̂ · · ·Γ ( 4−k 5 ) Γ ( 1− k 5 )4 × Γ ( H + k 5 ) Γ ( −H + 1− k 5 ) eπi(−H− k 5 )W̃k(1/Q), similarly, Γ(H + 1)4 Γ ( H + 1 5 ) · · ·Γ ( H + 4 5 ) = Γ(H + 1)5 Γ(5H + 1) 55H+1/2 (2π)2 , and Γ ( 1−k 5 ) · · · k̂ · · ·Γ ( 4−k 5 ) Γ ( 1− k 5 )4 = 53 (1− k) · · · k̂ · · · (4− k) ∏4 i=1 Γ ( 1− k 5 + i 5 ) Γ ( 1− k 5 ) Γ ( 1− k 5 )5 = 5k−1−1/2(2π)2 (1− k) · · · k̂ · · · (4− k) Γ(5− k) Γ ( 1− k 5 )5 . Thus we arrive at the right-hand side of (6.2). ■ Remark 6.3. Recall that in the introduction, we have the change of variables Q = 55et. Under the above change of variables, (6.2) becomes etH ∑ d≥0 ∏5d k=1(5H + k)∏d k=1(H + k)5 etd = Γ(H + 1)5 Γ(5H + 1) 4∑ k=1 5k−1Γ(5− k)∏4 i=1,i ̸=k(i− k)Γ ( 1− k 5 )5 πe−πi(H+ k 5 ) sin ( π ( H + k 5 ))W̃k. Difference Equation for Quintic 3-Fold 25 From [4], we know Γ(H + 1)5 Γ(5H + 1) = 1 + C(2πi)2H2 − E(2πi)3H3 +O ( H4 ) , where C = 5/12 and E = −ξ(3)40/(2πi)3 with ξ(3) equals to Apéry’s constant, i.e., it is related to the intersection theory of the quintic three-fold. We hope the expansion of the above equation on both sides with respect to the basis {H i}3i=0 will match the result in [4, formula (53)] up to the monodromy at 0 and ∞. For additional discussion on confluence, see [13] for projective spaces, and [12] for weak Fano manifolds. Acknowledgements The author would like to thank Professor Yongbin Ruan for suggesting this problem and for valuable discussions. Thanks are also due to Professor Shuai Guo and Dr. Yizhen Zhao for their helpful discussion. This work was initiated during the author’s stay at the Institute For Advanced Study In Mathematics (IASM) at Zhejiang University. The author would like to express his thanks to IASM, Professor Bohan Fan, Professor Huijun Fan, and Peking University for their helpful support during this visit. The author wants to thank the anonymous referees who help improve the paper a lot. The author is supported by a KIAS Individual Grant (MG083901) at Korea Institute for Advanced Study. References [1] Adams C.R., On the irregular cases of the linear ordinary difference equation, Trans. Amer. Math. Soc. 30 (1928), 507–541. [2] Adams C.R., Linear q-difference equations, Bull. Amer. Math. Soc. 37 (1931), 361–400. [3] Candelas P., de la Ossa X.C., Green P.S., Parkes L., A pair of Calabi–Yau manifolds as an exactly soluble superconformal theory, Nuclear Phys. B 359 (1991), 21–74. [4] Chiodo A., Ruan Y., Landau–Ginzburg/Calabi–Yau correspondence for quintic three-folds via symplectic transformations, Invent. Math. 182 (2010), 117–165, arXiv:0812.4660. [5] Garoufalidis S., Scheidegger E., On the quantum K-theory of the quintic, SIGMA 18 (2022), 021, 20 pages, arXiv:2101.07490. [6] Gasper G., Rahman M., Basic hypergeometric series, 2nd ed., Encyclopedia of Mathematics and its Appli- cations, Vol. 96, Cambridge University Press, Cambridge, 2004. [7] Givental A., A mirror theorem for toric complete intersections, in Topological Field Theory, Primitive Forms and Related Topics (Kyoto, 1996), Progr. Math., Vol. 160, Birkhäuser Boston, Boston, MA, 1998, 141–175. [8] Givental A., On the WDVV equation in quantum K-theory, Michigan Math. J. 48 (2000), 295–304, arXiv:math.AG/0003158. [9] Givental A., Permutation-equivariant quantum K-theory V. Toric q-hypergeometric functions, arXiv:1509.03903. [10] Gu W., Pei D., Zhang M., On phases of 3d N = 2 Chern–Simons-matter theories, Nuclear Phys. B 973 (2021), 115604, 20 pages, arXiv:2105.02247. [11] Lee Y.-P., QuantumK-theory. I. Foundations, Duke Math. J. 121 (2004), 389–424, arXiv:math.AG/0105014. [12] Milanov T., Roquefeuil A., Confluence in quantum K-theory of weak Fano manifolds and q-oscillatory integrals for toric manifolds, arXiv:2108.08620. [13] Roquefeuil A., Confluence of quantum K-theory to quantum cohomology for projective spaces, arXiv:1911.00254. [14] Ruan Y., Wen Y., Quantum K-theory and q-difference equations, arXiv:2109.02218. [15] Sauloy J., Systèmes aux q-différences singuliers réguliers: classification, matrice de connexion et monodromie, Ann. Inst. Fourier (Grenoble) 50 (2000), 1021–1071. [16] Sauloy J., Analytic study of q-difference equations, in Galois Theories of Linear Difference Equations: an Introduction, Math. Surveys Monogr., Vol. 211, Amer. Math. Soc., Providence, RI, 2016, 103–171. https://doi.org/10.2307/1989081 https://doi.org/10.1090/S0002-9904-1931-05162-4 https://doi.org/10.1016/0550-3213(91)90292-6 https://doi.org/10.1007/s00222-010-0260-0 https://arxiv.org/abs/0812.4660 https://doi.org/10.3842/SIGMA.2022.021 https://arxiv.org/abs/2101.07490 https://doi.org/10.1017/CBO9780511526251 https://doi.org/10.1007/978-1-4612-0705-4_5 https://doi.org/10.1307/mmj/1030132720 https://arxiv.org/abs/math.AG/0003158 https://arxiv.org/abs/1509.03903 https://doi.org/10.1016/j.nuclphysb.2021.115604 https://arxiv.org/abs/2105.02247 https://doi.org/10.1215/S0012-7094-04-12131-1 https://arxiv.org/abs/math.AG/0105014 https://arxiv.org/abs/2108.08620 https://arxiv.org/abs/1911.00254 https://arxiv.org/abs/2109.02218 https://doi.org/10.5802/aif.1784 1 Introduction 2 Preliminaries 2.1 Regular singular q-difference equations 2.2 Monodromy of regular singular q-difference equations 2.3 Confluence of regular singular q-difference equations 3 The difference equation for quintic 3.1 General technique: Newton polygon 3.2 Solutions at Q=0 3.3 Solutions at Q=infty 4 Auxiliary q-series and analytic continuation 4.1 Auxiliary q-series 4.2 Analytic continuation 5 A special fuchsian case 5.1 Connection matrix 6 Confluence of the q-difference structure References
id nasplib_isofts_kiev_ua-123456789-211625
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 1815-0659
language English
last_indexed 2026-03-17T20:38:36Z
publishDate 2022
publisher Інститут математики НАН України
record_format dspace
spelling Wen, Yaoxinog
2026-01-07T13:40:24Z
2022
Difference Equation for Quintic 3-Fold. Yaoxinog Wen. SIGMA 18 (2022), 043, 25 pages
1815-0659
2020 Mathematics Subject Classification: 14N35; 33D90; 39A13
arXiv:2011.07527
https://nasplib.isofts.kiev.ua/handle/123456789/211625
https://doi.org/10.3842/SIGMA.2022.043
In this paper, we use the Mellin-Barnes-Watson method to relate solutions of a certain type of -difference equations at = 0 and = ∞. We consider two special cases; the first is the -difference equation of the -theoretic -function of the quintic, which is degree 25; we use Adams' method to find the extra 20 solutions at = 0. The second special case is a Fuchsian case, which is confluent to the differential equation of the cohomological -function of the quintic. We compute the connection matrix and study the confluence of the -difference structure.
The author would like to thank Professor Yongbin Ruan for suggesting this problem and for valuable discussions. Thanks are also due to Professor Shuai Guo and Dr. Yizhen Zhao for their helpful discussion. This work was initiated during the author’s stay at the Institute for Advanced Study in Mathematics (IASM) at Zhejiang University. The author would like to express his thanks to IASM, Professor Bohan Fan, Professor Huijun Fan, and Peking University for their helpful support during this visit. The author wants to thank the anonymous referees who helped improve the paper a lot. The author is supported by a KIAS Individual Grant (MG083901) at the Korea Institute for Advanced Study.
en
Інститут математики НАН України
Symmetry, Integrability and Geometry: Methods and Applications
Difference Equation for Quintic 3-Fold
Article
published earlier
spellingShingle Difference Equation for Quintic 3-Fold
Wen, Yaoxinog
title Difference Equation for Quintic 3-Fold
title_full Difference Equation for Quintic 3-Fold
title_fullStr Difference Equation for Quintic 3-Fold
title_full_unstemmed Difference Equation for Quintic 3-Fold
title_short Difference Equation for Quintic 3-Fold
title_sort difference equation for quintic 3-fold
url https://nasplib.isofts.kiev.ua/handle/123456789/211625
work_keys_str_mv AT wenyaoxinog differenceequationforquintic3fold