Discrete Bessel functions and partial difference equations

Similar documents
HILLE-KNESER TYPE CRITERIA FOR SECOND-ORDER DYNAMIC EQUATIONS ON TIME SCALES

Hankel Optimal Model Order Reduction 1

SECOND HANKEL DETERMINANT PROBLEM FOR SOME ANALYTIC FUNCTION CLASSES WITH CONNECTED K-FIBONACCI NUMBERS

Nonreversibility of Multiple Unicast Networks

Research Article Approximation of Analytic Functions by Solutions of Cauchy-Euler Equation

(q) -convergence. Comenius University, Bratislava, Slovakia

arxiv:math/ v4 [math.ca] 29 Jul 2006

Stability of alternate dual frames

Ordered fields and the ultrafilter theorem

Complexity of Regularization RBF Networks

Integration of the Finite Toda Lattice with Complex-Valued Initial Data

ON THE GENERAL QUADRATIC FUNCTIONAL EQUATION

Generating Functions For Two-Variable Polynomials Related To a Family of Fibonacci Type Polynomials and Numbers

The First Integral Method for Solving a System of Nonlinear Partial Differential Equations

A xed point approach to the stability of a nonlinear volterra integrodierential equation with delay

Journal of Inequalities in Pure and Applied Mathematics

A Characterization of Wavelet Convergence in Sobolev Spaces

Where as discussed previously we interpret solutions to this partial differential equation in the weak sense: b

arxiv:math/ v1 [math.ca] 27 Nov 2003

REFINED UPPER BOUNDS FOR THE LINEAR DIOPHANTINE PROBLEM OF FROBENIUS. 1. Introduction

Coefficients of the Inverse of Strongly Starlike Functions

General solution to a higher-order linear difference equation and existence of bounded solutions

Math 220A - Fall 2002 Homework 8 Solutions

Maximum Entropy and Exponential Families

A Variational Definition for Limit and Derivative

SURFACE WAVES OF NON-RAYLEIGH TYPE

KAMILLA OLIVER AND HELMUT PRODINGER

RIEMANN S FIRST PROOF OF THE ANALYTIC CONTINUATION OF ζ(s) AND L(s, χ)

the following action R of T on T n+1 : for each θ T, R θ : T n+1 T n+1 is defined by stated, we assume that all the curves in this paper are defined

ON A PROCESS DERIVED FROM A FILTERED POISSON PROCESS

Geometry of Transformations of Random Variables

The Hanging Chain. John McCuan. January 19, 2006

RATIONALITY OF SECANT ZETA VALUES

SQUARE ROOTS AND AND DIRECTIONS

SERIJA III

Lecture 3 - Lorentz Transformations

arxiv: v1 [math.co] 16 May 2016

arxiv:gr-qc/ v2 6 Feb 2004

Millennium Relativity Acceleration Composition. The Relativistic Relationship between Acceleration and Uniform Motion

Rigorous prediction of quadratic hyperchaotic attractors of the plane

ON THE MOVING BOUNDARY HITTING PROBABILITY FOR THE BROWNIAN MOTION. Dobromir P. Kralchev

Differential Equations 8/24/2010

Exact Solution of Space-Time Fractional Coupled EW and Coupled MEW Equations Using Modified Kudryashov Method

Developing Excel Macros for Solving Heat Diffusion Problems

Fuzzy inner product space and its properties 1

Derivation of Non-Einsteinian Relativistic Equations from Momentum Conservation Law

The law of the iterated logarithm for c k f(n k x)

A Queueing Model for Call Blending in Call Centers

Counting Idempotent Relations

a n z n, (1.1) As usual, we denote by S the subclass of A consisting of functions which are also univalent in U.

Long time stability of regularized PML wave equations

An Integrated Architecture of Adaptive Neural Network Control for Dynamic Systems

Advanced Computational Fluid Dynamics AA215A Lecture 4

On Component Order Edge Reliability and the Existence of Uniformly Most Reliable Unicycles

A Functional Representation of Fuzzy Preferences

arxiv:physics/ v1 [physics.class-ph] 8 Aug 2003

Bäcklund Transformations: Some Old and New Perspectives

Control Theory association of mathematics and engineering

EXACT TRAVELLING WAVE SOLUTIONS FOR THE GENERALIZED KURAMOTO-SIVASHINSKY EQUATION

Chapter 8 Hypothesis Testing

Exercise 3: Quadratic sequences

Collinear Equilibrium Points in the Relativistic R3BP when the Bigger Primary is a Triaxial Rigid Body Nakone Bello 1,a and Aminu Abubakar Hussain 2,b

On maximal inequalities via comparison principle

Sensitivity analysis for linear optimization problem with fuzzy data in the objective function

The gravitational phenomena without the curved spacetime

Convergence of the Logarithmic Means of Two-Dimensional Trigonometric Fourier Series

arxiv: v2 [cs.dm] 4 May 2018

arxiv:cond-mat/ v1 [cond-mat.stat-mech] 16 Aug 2004

Asymptotic non-degeneracy of the solution to the Liouville Gel fand problem in two dimensions

Concerning the Numbers 22p + 1, p Prime

The tanh - coth Method for Soliton and Exact Solutions of the Sawada - Kotera Equation

On the density of languages representing finite set partitions

Two Points Hybrid Block Method for Solving First Order Fuzzy Differential Equations

COMBINED PROBE FOR MACH NUMBER, TEMPERATURE AND INCIDENCE INDICATION

Journal of Mathematical Analysis and Applications

7 Max-Flow Problems. Business Computing and Operations Research 608

Lyapunov Exponents of Second Order Linear Systems

Some GIS Topological Concepts via Neutrosophic Crisp Set Theory

Sensitivity Analysis in Markov Networks

Berry s phase for coherent states of Landau levels

22.54 Neutron Interactions and Applications (Spring 2004) Chapter 6 (2/24/04) Energy Transfer Kernel F(E E')

NUMERICALLY SATISFACTORY SOLUTIONS OF HYPERGEOMETRIC RECURSIONS

Taste for variety and optimum product diversity in an open economy

HYPERSTABILITY OF THE GENERAL LINEAR FUNCTIONAL EQUATION

Development of Fuzzy Extreme Value Theory. Populations

The Electromagnetic Radiation and Gravity

Wuming Li and Fan Yang

Error Bounds for Context Reduction and Feature Omission

Sufficient Conditions for a Flexible Manufacturing System to be Deadlocked

Anomaly cancellation and modularity, II: The E 8 E 8 case

An Integer Solution of Fractional Programming Problem

Searching All Approximate Covers and Their Distance using Finite Automata

Some examples of generated fuzzy implicators

Quasi-Monte Carlo Algorithms for unbounded, weighted integration problems

Strauss PDEs 2e: Section Exercise 3 Page 1 of 13. u tt c 2 u xx = cos x. ( 2 t c 2 2 x)u = cos x. v = ( t c x )u

G-subsets and G-orbits of Q ( n) under action of the Modular Group.

arxiv:math.co/ v1 2 Aug 2006

arxiv: v1 [math.gt] 22 Nov 2018

Estimating the probability law of the codelength as a function of the approximation error in image compression

A Recursive Approach to the Kauffman Bracket

Transcription:

Disrete Bessel funtions and partial differene equations Antonín Slavík Charles University, Faulty of Mathematis and Physis, Sokolovská 83, 186 75 Praha 8, Czeh Republi E-mail: slavik@karlin.mff.uni.z Abstrat We introdue a new lass of disrete Bessel funtions and disrete modified Bessel funtions of integer order. After obtaining some of their basi properties, we show that these funtions lead to fundamental solutions of the disrete wave equation and disrete diffusion equation. Keywords: Bessel funtion; modified Bessel funtion; Bessel differene equation; disrete wave equation; disrete diffusion equation; fundamental solution MSC 010 subjet lassifiation: 39A1, 39A14, 33C10, 33C05, 39A06, 39A10, 34A33 1 Introdution In their reent paper [], M. Bohner and T. Cuhta have proposed a new definition of the disrete Bessel funtion J n t = 1n t n n t n F n!, n t + 1 ; n + 1; 1, t N 0, 1.1 where F is the hypergeometri funtion and x k denotes the Pohhammer symbol also known as the rising fatorial given by xx + 1 x + k 1 for k N, x k = 1 for k = 0. The disrete Bessel funtion given by 1.1 is different from the one studied in earlier papers [4, 5], and its advantage is that it shares many properties with the lassial Bessel funtion. For example, it satisfies the differene equation tt 1 yt + t yt 1 + tt 1yt n yt = 0 where ft = ft + 1 ft is the forward differene, whih is a disrete analogue of the Bessel differential equation t y t + ty t + t n yt = 0. The goal of this paper is to introdue a new lass of disrete Bessel funtions denoted by J n, where n N 0 is the order and is a parameter, and to show that these disrete Bessel funtions provide fundamental solutions to the disrete wave equation ux, t = ux + 1, t ux, t + ux 1, t, x Z, t N 0 1. with ux, t being the seond-order forward differene of u with respet to t. The disrete Bessel funtion J n given by 1.1 is a speial ase of J n orresponding to = 1. 1

We also introdue a new lass of disrete modified Bessel funtions denoted by I n, whih an be used to onstrut fundamental solutions of the disrete diffusion equation ux, t = ux + 1, t ux, t + ux 1, t, x Z, t N 0 1.3 where ux, t is the forward differene of u with respet to t. Our motivation omes from the theory of lattie differential equations, i.e., equations with disrete spae and ontinuous time. In this ontext, it is known that the fundamental solutions of the lattie wave equation u t x, t = ux + 1, t ux, t + ux 1, t, x Z, t R + 0, 1.4 whih is a semidisrete analogue of 1., have the form u 1 x, t = J x t and u x, t = t 0 J xs ds, where J x is the lassial Bessel funtion see [8, Example 3.3]. Similarly, the fundamental solution of the lattie diffusion equation u t x, t = ux + 1, t ux, t + ux 1, t, x Z, t R + 0, 1.5 whih is a semidisrete analogue of 1.3, has the form ux, t = e t I x t, where I x is the lassial modified Bessel funtion see [10, Example 3.1]. The orresponding formulas for fundamental solutions of the partial differene equations 1. and 1.3 will be obtained in Setions 3 and 4, respetively. To ahieve this goal, we need the disrete analogues of the funtions t J n t and t I n t for an arbitrary > 0, whih are preisely the funtions J n and I n mentioned earlier. While the funtions t J n t and t I n t satisfy the differential equations t y t + ty t + ± t n yt = 0, we will show that their disrete ounterparts J n and I n are solutions of the differene equations tt 1 yt + t yt 1 ± tt 1yt n yt = 0. 1.6 By expanding the differenes, we obtain the equivalent form t n yt tt 1yt 1 + 1 ± tt 1yt = 0. 1.7 We remark that the fundamental solutions of the partial differene equations 1. and 1.3 are already available in the existing literature [8, 10], but they are expressed in a different form than we obtain in Setions 3 and 4. Expressing them in terms of the disrete Bessel funtions an simplify the study of their properties. For example, following the method from [], we prove that the funtion J n is osillatory. This fat implies that for eah fixed x, the first fundamental solution to 1. is osillatory as a funtion of t; this result is new and would be diffiult to obtain by different methods. Disrete Bessel funtions Both types of the Bessel funtions, J n and I n, will be defined in terms of the hypergeometri series F α, β; γ; z = k=0 α k β k z k. γ k k! Definition.1. For eah C, we define the disrete Bessel funtion Jnt = /n t n F n!, n t + 1 ; n + 1;, t N 0, n N 0,.1

and the disrete modified Bessel funtion Int = /n t n F n!, n t + 1 ; n + 1;, t N 0, n N 0.. Note that if n > t, then t n = 0 and therefore Jnt = Int = 0. Otherwise, if n t, then one of the frations n t and n t + 1 is a nonpositive integer, whih means that the hypergeometri series ourring in.1 and. have only finitely many nonzero terms. As in [], the definition of Jnt an be extended to all t Z, but the same extension is not always possible for Int. Similarly, it would be possible to onsider Bessel funtions of non-integer orders n. However, for simpliity, we restrit ourselves to nonnegative integer values of t and n; this ase is the most interesting one for appliations to partial differene equations. For = 1, the funtion J n oinides with the disrete Bessel funtion 1.1 introdued in []. For appliations in partial differene equations, the most useful ase is when is a positive real number. One advantage of allowing to be omplex is the onnetion formula I nt = i n J i n t, whih is a straightforward onsequene of the definitions. Our first goal is to prove that J n and I n satisfy the differene equations 1.6 1.7. The next result generalizes [, Theorem 1]. Theorem.. If C and n N 0, then the funtion satisfies the differene equation or equivalently B n t = /n t n F n!, n t + 1 ; n + 1; ±, t N 0, tt 1 B n t + t B n t 1 tt 1B n t n B n t = 0, t, t n B n t tt 1B n t 1 + 1 tt 1B n t = 0, t. Proof. We use the ontiguous relation see [7, formula 15.5.13] γ α βf α, β; γ; z γ αf α 1, β; γ; z + β1 zf α, β + 1; γ; z = 0 with α = n t + 1, β = n t + 1, γ = n + 1, z = ± to get t 1 F + 1, n t + 1 ; n + 1; ± + n t + 1 1 F n + t + 1, n t F, n t + 1 ; n + 1; ± + 3 ; n + 1; ± = 0. By multiplying the equation with /n n! t n+1, using the definition of B n and the symmetry of F in the first two arguments, we obtain tt 1B n t 1 n + t t + nb n t + 1 tt 1B n t = 0. Corollary.3. For eah C and n N 0, the funtion J n is a solution of the differene equation tt 1 yt + t yt 1 + tt 1yt n yt = 0, t, and the funtion I n is a solution of the differene equation tt 1 yt + t yt 1 tt 1yt n yt = 0, t. 3

The next task is to obtain expressions for differenes of the disrete Bessel funtions. The following result generalizes Theorems 5, 6 and Corollary 7 from []. Our proof is simpler than in [] and relies on the ontiguous relations for the hypergeometri funtion. Theorem.4. Assume that C. For eah n N 0, onsider the funtion B n t = /n t n F n! Then we have the following identities:, n t + 1 ; n + 1; ±, t N 0. t B n t 1 = nb n t ± tb n+1 t 1, n 0, t 1,.3 t B n t 1 = nb n t + tb n 1 t 1, n 1, t 1,.4 nb n t = t Bn 1 t 1 B n+1 t 1, n 1, t 1,.5 Proof. To prove.3, we need to show that ± n + 1 tf B n t = Bn 1 t ± B n+1 t, n 1, t 0..6 0 = tb n t 1 + n tb n t ± tb n+1 t 1. Using the definition of B n and dividing by / n t n+1 /n + 1!, we see it is enough to show that 0 = n + 1F + 1, n t + 1; n + 1; ± + n + 1F + 1, n t ; n + 1; ± + 1; n + ; ±. + 3, n t To prove this, we use the ontiguous relations see [7, formulas 15.5.13 and 15.5.16] αγ1 zf α + 1, β + 1; γ; z = γ γ β 1F α, β; γ; z γ α β 1F α, β + 1; γ; z, αγ1 zf α + 1, β + 1; γ; z = αγf α, β + 1; γ; z αγ β 1zF α + 1, β + 1; γ + 1; z. By equating the right-hand sides and dividing by γ β 1, we get γf α, β; γ; z γf α, β + 1; γ; z = αzf α + 1, β + 1; γ + 1; z. The desired relation now follows by letting α = n t + 1, β = n t, γ = n + 1, z = ±. To prove.4, we have to show that 0 = tb n t 1 n + tb n t + tb n 1 t 1. Using the definition of B n and dividing by / n t n /n!, we see it is enough to show that 0 = t + nf + 1, n t + 1; n + 1; ± n + tf, n t + 1 ; n + 1; ± + nf, n t + 1 ; n; ±. To prove this, we use the ontiguous relation see [7, formula 15.5.15] γ αf α, β; γ + 1; z + αf α + 1, β; γ + 1; z γf α, β; γ; z = 0. The desired relation now follows by letting α = n t, β = n t + 1, γ = n, z = ±. Identity.5 is obtained by subtrating.3 from.4. To get the identity.6, add.3 and.4, divide by t, and replae t by t + 1. 4

Corollary.5. For eah C, the following relations hold: t J nt 1 = nj nt tj n+1t 1, n 0, t 1,.7 t Jnt 1 = njnt + tjn 1t 1, n 1, t 1,.8 njnt = t J n 1 t 1 + Jn+1t 1, n 1, t 1,.9 J nt = J n 1 t J n+1t, n 1, t 0,.10 t I nt 1 = ni nt + ti n+1t 1, n 0, t 1,.11 t Int 1 = nint + tin 1t 1, n 1, t 1,.1 nint = t I n 1 t 1 In+1t 1, n 1, t 1,.13 I nt = I n 1 t + I n+1t, n 1, t 0..14 The next theorem provides additional information about the values and differenes of J n and I n. Theorem.6. For eah C, the funtions J n and I n have the following properties: J 00 = I 00 = 1. J nt = I nt = 0 for all t N 0 and n N suh that n > t. J n0 = I n0 = 0 for all n N 0 \ 1}, and J 10 = I 10 = /. Proof. The first two statements follow from the definitions of Jn and In; note that t n = 0 for all t N 0 and n N suh that n > t. Using.10 and.14, we get Jn0 = J n 1 0 Jn+10 and In0 = I n 1 0 + In+10 for all n N. Both expressions are equal to / if n = 1, and zero for all n N \ 1}. For n = 0, the relations.7 and.11 with t = 1 imply J00 = I00 = 0. The remaining results in this setion are onerned with the sign of J n and I n if is a real number. The first statement generalizes [, Theorem 1]. Theorem.7. For eah R \ 0} and n N 0, the funtion J n is osillatory i.e., J nt hanges sign or vanishes for infinitely many values of t N 0. Proof. To simplify notation, we denote yt = J nt. Aording to Corollary.3 with t replaed by t +, we see that y satisfies the differene equation t + t + 1 yt + t + yt + 1 + t + t + 1yt n yt + = 0, t N 0. Using the formulas yt = yt + 1 yt + 1 + yt and yt + = yt + 1 + yt + 1, we obtain yt t + t + 11 + + yt + 1 t + t + t + 1 n + yt + 1 t + 1t + n = 0, and therefore yt = t + t + 1 + n t t + t + 11 + yt + 1 t + 1t + n t + t + 11 + yt + 1, t N 0. Let vt = t 1+ t/ t t n t Γ t n + 1Γ t+n + 1 =, t n, n + 1, n +,...}. 1 + t/ Γt + 1 5

One an verify using a omputer system suh as Mathematia or by a hand alulation similar to [, Lemma 11] that Let vt + 1 + vt + vt t + 1t + + n t t + t + 11 + = 0, t n, n + 1, n +,...},.15 vt lim t vt + 1 = 1 +..16 Using the produt rule twie, we get ut = vtyt, t n, n + 1, n +,...}. ut = yt + 1 vt + vt yt, ut = yt + 1 vt + vt + 1 + vt yt + 1 + vt yt ut + 1 = vt + 1 vt + vt + 1 + vt yt + 1 t + t + 1 + n t +vt t + t + 11 + yt + 1 t + 1t + n t + t + 11 + vt = ut + 1 vt + 1 t + 1t + n vt t + t + 11 + vt + 1 + yt + 1 vt + 1 + vt + vt t + t + 1 + n t t + t + 11 + The last term vanishes thanks to.15, and therefore t + 1t + n vt ut + ut + 1 t + t + 11 + vt + 1 vt = 0. vt + 1 This is a seond-order differene equation of the form ut + qtut + 1 = 0, where qt = t + 1t + n vt t + t + 11 + vt + 1 ut + 1 vt + 1. vt + vt + 1 + vt. vt + 1 By Wintner s theorem see [3, Theorem 4.45], suh equation is osillatory if t=n qt =. To verify this fat, it is enough to show that lim t qt > 0. Using.16, we alulate lim qt = t + 1 lim vt vt + t vt + 1 lim t vt + 1 + vt vt + 1 = + 1 1 + 1 + 1 + = + 1 > 0. This shows that u is osillatory. Sine v is positive, y is osillatory. Theorem.8. For eah 0 and n N 0, the funtion I n is nonnegative. For eah < 0, the funtion I n is nonnegative if n is even and nonpositive if n is odd. Proof. The first statement where 0 is easily proved by indution with respet to t. For t = 0, it follows from the first and seond part of Theorem.6 that I n0 0 for all n N 0. Suppose that I nt 0 for all n N 0. By the relation.14, we have I nt 0 for all n N. If n = 0, then the relation.11 with t replaed by t + 1 implies I 0t = I 1t 0. Consequently, we have I nt + 1 = I nt + I nt 0 for all n N 0. The seond statement where < 0 is a onsequene of the first part and the identity whih follows immediately from the definition. I nt = 1 n I n t, 6

3 Disrete wave equation In this setion, we explore the relation between the disrete Bessel funtion J n and the disrete wave equation ux, t = ux + 1, t ux, t + ux 1, t, x Z, t N 0 the forward differene operator always applies to the time variable t; differenes with respet to the spae variable x are never onsidered in this paper. Suppose that u 1 : Z N 0 R is the solution orresponding to the initial onditions 1 if x = 0, u 1 x, 0 = 0 if x 0, u 1 x, 0 = 0, x Z. Then it is not diffiult to hek that the funtion u : Z N 0 R given by t 1 u x, t = u 1 x, s s=0 where the sum is understood as empty if t = 0 is the solution of the disrete wave equation satisfying the onditions u x, 0 = 0, x Z, 1 if x = 0, u x, 0 = 0 if x 0. In [8, Theorem 3.], it is shown that for arbitrary bounded real sequenes u 0 x} x Z, v 0 x} x Z, the funtion ux, t = k Zu 0 k u 1 x k, t + v 0 k u x k, t, x Z, t N 0, 3.1 is the solution of the disrete wave equation satisfying ux, 0 = u 0 x, ux, 0 = v 0 x, x Z. In fat, it is not diffiult to see use indution with respet to t that we have u 1 x, t = u x, t = 0 whenever x > t. Hene, on the right-hand side of the formula 3.1, the terms orresponding to k Z suh that x k > t do not ontribute to ux, t, and we an write ux, t = x+t k=x t u 0 k u 1 x k, t + v 0 k u x k, t, x Z, t N 0. The solutions u 1, u are referred to as the fundamental solutions of the disrete wave equation. The next theorem shows that u 1 and onsequently also u an be expressed in terms of the disrete Bessel funtion J n. Theorem 3.1. For eah > 0, the solution of the initial-value problem ux, t = ux + 1, t ux, t + ux 1, t, x Z, t N 0, 3. 1 if x = 0, ux, 0 = 3.3 0 if x 0, ux, 0 = 0, x Z, 3.4 7

is given by Moreover, for eah x Z, the funtion t ux, t osillatory. ux, t = J x t, x Z, t N 0. 3.5 Proof. Let u be defined by 3.5. The relations 3.3 and 3.4 follow from Theorem.6. If x 1, we use the identity.10 to alulate ux, t = Jxt = Jx 1t Jx+1t, ux, t = Jx t Jxt + Jx+t = ux 1, t ux, t + ux + 1, t. Similarly, if x 1, we obtain ux, t = J xt = J x 1t J x+1t, ux, t = J x t J xt + J x+t = ux + 1, t ux, t + ux 1, t. Finally, for x = 0, we use the identity.7 with n = 0 and t replaed by t + 1 to get and onsequently by identity.10 u0, t = J 0 t = J 1 t, u0, t = J 0 t + J t = J t J 0 t + J t = u1, t u0, t + u 1, t. Thus, the relation 3. holds for all x Z, t N 0. The fat that t ux, t is osillatory follows from Theorem.7. Remark 3.. The first fundamental solution of the disrete wave equation an be alternatively expressed using the multinomial oeffiients as follows see [8, Example 3.5]: ux, t = t t 1 j j+x j, t j x, j + x j=0 4 Disrete diffusion equation We now turn our attention to the disrete diffusion equation wx, t = d wx + 1, t wx, t + wx 1, t, x Z, t N 0. The solution w : Z N 0 R orresponding to the initial onditions 1 if x = 0, wx, 0 = 0 if x 0, is alled the fundamental solution. In [9, Corollary 3.8], it is shown that for an arbitrary bounded real sequene u 0 x} x Z, the funtion ux, t = k Z u 0 k wx k, t, x Z, t N 0, 4.1 is the solution of the disrete diffusion equation satisfying ux, 0 = u 0 x, x Z. 8

Observing that wx, t = 0 whenever x > t use indution with respet to t, we an simplify the formula 4.1 to ux, t = x+t k=x t u 0 k wx k, t, x Z, t N 0. For d 1/, the next theorem shows that the fundamental solution w an be onstruted using the disrete modified Bessel funtion I n. Theorem 4.1. For eah d 1/, the solution of the initial-value problem is given by wx, t = d wx + 1, t wx, t + wx 1, t, x Z, t N 0, 4. 1 if x = 0, wx, 0 = 4.3 0 if x 0. wx, t = 1 d t I d/1 d x t, x Z, t N 0. 4.4 Moreover, for eah x Z, the funtion t wx, t is nonnegative if d 0, 1/, and osillatory if d > 1/. Proof. Let w be defined by 4.4. To simplify notation, let = d/1 d. The relation 4.3 follows from Theorem.6. Let zx, t = I x x Z, t N 0. If x 1, we use the identity.14 to alulate Similarly, if x 1, we obtain zx, t = I x t = I x+1t + I x 1t = zx + 1, t + zx 1, t. zx, t = I xt = I x+1t + I x 1t = zx 1, t + zx + 1, t. Finally, if x = 0, identity.11 with n = 0 and t replaed by t + 1 implies that Note that Hene, by the produt rule, we have z0, t = I 0 t = I 1 t = z1, t + z 1, t. 1 d t = 1 d t+1 1 d t = d1 d t. wx, t = zx, t1 d t = zx, t1 d t+1 + zx, t 1 d t = d zx + 1, t + zx 1, t 1 d t+1 zx, td1 d t 1 d = d zx + 1, t zx, t + zx 1, t 1 d t = d wx + 1, t wx, t + wx 1, t. The fat that t wx, t is nonnegative if d 0, 1/ and osillatory if d > 1/ follows from the definition of w and Theorem.8. Remark 4.. An alternative form of the fundamental solution to the disrete diffusion equation is see [10, Example 3.3] t t wx, t = d j+x 1 d t j x. j, t j x, j + x j=0 This formula is valid also for d = 1/, when it redues to t 1 t t+x wx, t = if t + x is even, 0 if t + x is odd. 9

5 Conlusion We onlude the paper by pointing out two possible diretions for further researh: The lassial Bessel funtions have their multivariable ounterparts [1], whih found appliations in various areas of physis see, e.g., [6] and the referenes there. Is there a reasonable extension of the disrete Bessel funtions to several variables? If yes, is it related to the higher-dimensional disrete diffusion/wave equations? Note that an expliit formula for the fundamental solution of the n-dimensional disrete diffusion equation, whih does not rely on Bessel funtions, an be found in [10]. In Setions 3 and 4, we were dealing with the disrete diffusion/wave equations whose left-hand sides involve forward differenes of first and seond order with respet to time. In some situations, it might be more appropriate to onsider the bakward first-order differene for the diffusion equation, and the entral or bakward seond-order differene for the wave equation. Is it possible to express their solutions with the help of some Bessel-type funtions? Aknowledgement This paper was supported by the Czeh Siene Foundation, grant no. GA15-07690S. I am grateful to the anonymous referees for their suggestions, whih helped to improve the exposition of the paper. Referenes [1] P. Appell, Sur l inversion approhée de ertaines intégrales réelles et sur l extension de l équation de Kepler et des fontions de Bessel, C. R. Aad. Si. 160 1915, 419 43. [] M. Bohner, T. Cuhta, The Bessel differene equation, Pro. Amer. Math. So. 145 017, 1567 1580. [3] M. Bohner, A. Peterson, Dynami Equations on Time Sales: An Introdution with Appliations, Birkhäuser, Boston, 001. [4] R. H. Boyer, Disrete Bessel funtions, J. Math. Anal. Appl. 1961, 509 54. [5] J. J. Gergen, Bessel differene systems of zero order, J. Math. Anal. Appl. 13 1966, 10 117. [6] H. J. Korsh, A. Klumpp, D. Witthaut, On two-dimensional Bessel funtions, J. Phys. A, Math. Gen. 39 006, 14947 14964. [7] F. W. J. Olver, D. W. Lozier, R. F. Boisvert, C. W. Clark, eds., NIST Handbook of Mathematial Funtions, Cambridge University Press, New York, 010. Online version at http://dlmf.nist.gov/. [8] A. Slavík, Disrete-spae systems of partial dynami equations and disrete-spae wave equation, Qual. Theory Dyn. Syst. 16 017, 99 315. [9] A. Slavík, P. Stehlík, Dynami diffusion-type equations on disrete-spae domains, J. Math. Anal. Appl. 47 015, 55 545. [10] A. Slavík, P. Stehlík, Expliit solutions to dynami diffusion-type equations and their time integrals. Appl. Math. Comput. 34 014, 486 505. 10