• 検索結果がありません。

Doubly Periodic Traveling Waves in a Cellular Neural Network with Linear Reaction

N/A
N/A
Protected

Academic year: 2022

シェア "Doubly Periodic Traveling Waves in a Cellular Neural Network with Linear Reaction"

Copied!
29
0
0

読み込み中.... (全文を見る)

全文

(1)

Advances in Dierence Equations Volume 2009, Article ID 243245,29pages doi:10.1155/2009/243245

Research Article

Doubly Periodic Traveling Waves in a Cellular Neural Network with Linear Reaction

Jian Jhong Lin and Sui Sun Cheng

Department of Mathematics, Tsing Hua University, Hsinchu 30043, Taiwan

Correspondence should be addressed to Sui Sun Cheng,sscheng@math.nthu.edu.tw Received 4 June 2009; Accepted 13 October 2009

Recommended by Roderick Melnik

Szekeley observed that the dynamic pattern of the locomotion of salamanders can be explained by periodic vector sequences generated by logical neural networks. Such sequences can mathematically be described by “doubly periodic traveling waves” and therefore it is of interest to propose dynamic models that may produce such waves. One such dynamic network model is built here based on reaction-diffusion principles and a complete discussion is given for the existence of doubly periodic waves as outputs. Since there are 2 parameters in our model and 4 a priori unknown parameters involved in our search of solutions, our results are nontrivial. The reaction term in our model is a linear function and hence our results can also be interpreted as existence criteria for solutions of a nontrivial linear problem depending on 6 parameters.

Copyrightq2009 J. J. Lin and S. S. Cheng. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

1. Introduction

Szekely in 1 studied the locomotion of salamanders and showed that a bipolar neural network may generate dynamic rhythms that mimic the “sequential” contraction and relaxation of four muscle pools that govern the movements of these animals. What is interesting is that we may explain the correct sequential rhythm by means of the transition of state values of four differentartificialneurons and the sequential rhythm can be explained in terms of an 8-periodic vector sequence and subsequently in terms of a “doubly periodic traveling wave solution” of the dynamic bipolar cellular neural network.

Similar dynamiclocomotivepatterns can be observed in many animal behaviors and therefore we need not repeat the same description in1. Instead, we may use “simplified”

snorkeling or walking patterns to motivate our study here. When snorkeling, we need to float on water with our faces downward, stretch out our arms forward, and expand our legs backward. Then our legs must move alternatively. More precisely, one leg kicks downward and another moves upward alternatively.

Letv0andv1 be two neuron pools controlling our right and left legs, respectively, so that our leg moves upward if the state value of the corresponding neuron pool is 1,and

(2)

−1 0 1 u

0 2 4 6 8 10 12

i

0 5

10 15

20 25

30 35

40

t

Figure 1: Doubly periodic traveling wave.

downward if the state value of the corresponding neuron pool is −1. Let v0t and v1t be the state values ofv0 andv1 during the time staget,wheretN {0,1,2, . . .}.Then the movements of our legs in terms ofvt0 , vt1 , t∈N,will form a 2-periodic sequential pattern

−1,1−→1,−1−→−1,1 → 1,−1−→ · · · 1.1

or

1,−1−→−1,1−→1,−1−→1,1−→ · · ·. 1.2

If we setvti vitmod 2 for anytNandi∈Z{0,±1,±2, . . .},then it is easy to check that

vt1i vti1 ∀i∈Z, t∈N

temporal-spatial transition condition , vt2i vit ∀i∈Z, t∈N

temporal periodicity condition , vti vti2 ∀i∈Z, t∈N

spatial periodicity condition .

1.3

Such a sequence {vit} may be called a “doubly periodic traveling wave” see Figure 1.

Now we need to face the following important issue as in neuromorphic engineering.

Can we build artificial neural networks which can support dynamic patterns similar to {vt0 , v1t}t∈N? Besides this issue, there are other related questions. For example, can we build nonlogical networks that can support different types of graded dynamic patterns remember an animal can walk, run, jump, and so forth, with different strength?

To this end, in 2, we build a nonlogical neural network and showed the exact conditions such doubly periodic traveling wave solutions may or may not be generated by it. The network in2has a linear “diffusion part” and a nonlinear “reaction part.” However,

(3)

the reaction part consists of a quadratic polynomial so that the investigation is reduced to a linear and homogeneous problem. It is therefore of great interests to build networks with general polynomials as reaction terms. This job is carried out in two stages. The first stage results in the present paper and we consider linear functions as our reaction functions. In a subsequent paper, as a report of the second stage investigation, we consider polynomials with more general formsee the statement after2.11.

2. The Model

We briefly recall the diffusion-reaction network in 2. In the following, we set N {0,1,2, . . .},Z{. . . ,−2,−1,0,1,2, . . .}andZ{1,2,3, . . .}.For anyx∈R,we also usexto denote the greatest integer part ofx.Suppose thatv0, . . . , vΥ−1are neuron pools, whereΥ≥1, placedin a counterclockwise manneron the vertices of a regular polygon such that each neuron poolvihas exactly two neighbors,vi−1andvi1,wherei∈ {0, . . . ,Υ−1}.For the sake of convenience, we have setv0 v−1andv1 vΥto reflect the fact that these neuron pools are placed on the vertices of a regular polygon. For the same reason, we definevi vimodΥ for anyi∈ Zand let eachvti be the state value of theith unitvi in the time periodtN.

During the time periodt, if the valuevti of theith unit is higher thanvti−1, we assume that

“information” will flow from theith unit to its neighbor. The subsequent change of the state value of theith unit isvit1vit, and it is reasonable to postulate that it is proportional to the differencevtivti−1, say,αvitvi−1t, whereαis a proportionality constant. Similarly, information is assumed to flow from thei1-unit to theith unit if vi1t > vti . Thus, it is reasonable that the total effect is

vt1ivti α

vti−1vti α

vi1tvti α

vi1t −2vitvti−1

, i∈Z, t∈N. 2.1

If we now assume further that a control or reaction mechanism is imposed, a slightly more complicated nonhomogeneous model such as the following

vt1ivti α

vti1vti α

vi−1tvti g

vti

∀i∈Z, t∈N 2.2

may result. In the above model, we assume thatgis a function andα∈R.

The existence and uniqueness ofrealsolutions of2.2is easy to see. Indeed, if the realinitial distribution{v0i }i∈Zis known, then we may calculate successively the sequence

v−11, v10 , v11;v1−2, v−12, v20 , v12, v12 , . . . 2.3

(4)

in a unique manner, which will give rise to a unique solution{vti }t∈N,i∈Zof2.2. Motivated by our example above, we want to find solutions that satisfy

vi vt ∀i∈Z, t∈N, 2.4

vi vit ∀i∈Z, t∈N, 2.5

vit vt ∀i∈Z, t∈N, 2.6

whereτ,Δ,Υ∈Zandδ∈Z.It is clear that equations in1.3are special cases of2.4,2.5, and2.6, respectively.

Suppose that v{vti }is a double sequence satisfying2.4for someτ ∈Zandδ∈Z.

Then it is clear that

vtkτi vtikδ for any i∈Z, t∈N, 2.7

wherek ∈Z.Hence when we want to find any solution{vti }of2.2satisfying2.4, it is sufficient to find the solution of2.2satisfying

vtτ/qi vtiδ/q, 2.8

whereqis the greatest common divisorτ, δofτandδ.For this reason, we will pay attention to the condition thatτ, δ 1.Formally, given anyτ ∈Zandδ ∈Zwithτ, δ 1,a real double sequence{vti }t∈N,i∈Zis called a traveling wave with velocity−δ/τif

vivt, tN, i∈Z. 2.9

In caseδ0 andτ 1,our traveling wave is also called a standing wave.

Next, recall that a positive integer ω is called a period of a sequence ϕm} if ϕϕmfor allm∈Z. Furthermore, ifω∈Zis the least among all periods of a sequence ϕ,then ϕis said to beω-periodic. It is clear that if a sequenceϕ is periodic, then the least number of all itspositiveperiods exists. It is easy to see the following relation between the least period and a period of a periodic sequence.

Lemma 2.1. If y {yi}isω-periodic andω1 is a period of y,thenωis a factor ofω1,orω mod ω10.

We may extend the above concept of periodic sequences to double sequences. Suppose that v {vti }is a real double sequence. Ifξ ∈Z such thatvt vit for alliandt,thenξ is called a spatial period of v.Similarly, ifη∈Zsuch thatvi vitfor alliandt,thenη is called a temporal period of v. Furthermore, ifξis the least among all spatial periods of v, then v is called spatialξ-periodic, and ifηis the least among all temporal periods of v,then v is called temporalη-periodic.

In seeking solutions of2.2that satisfy2.5and2.6, in view ofLemma 2.1, there is no loss of generality to assume that the numbers Δand Υare the least spatial and the

(5)

least temporal periods of the sought solution. Therefore, from here onward, we will seek such doubly-periodic traveling wave solutions of2.2. More precisely, given any function g, α ∈R, δ ∈ZandΔ,Υ, τ ∈Z withτ, δ 1,in this paper, we will mainly be concerned with the traveling wave solutions of2.2with velocity−δ/τwhich are also spatialΥ-periodic and temporalΔ-periodic. For convenience, we call such solutionsΔ,Υ-periodic traveling wave solutions of2.2with velocity−δ/τ.

In general, the control functiongin2.2can be selected in many different ways. But naturally, we should start with the trivial polynomial and general polynomials of the form

gx κfx:κxr1x−r2· · ·x−rn, 2.10

wherer1, r2, . . . , rn are real numbers, andκis a real parameter. In2, the trivial polynomial and the quadratic polynomialfx x2 are considered. In this paper, we will consider the linear case, namely,

fx 1 for x∈R or fx xr forx∈R, wherer∈R, 2.11

while the cases wherer1, r2, . . . , rn are mutually distinct andn ≥ 2 will be considered in a subsequent paperfor the important reason that quite distinct techniques are needed.

Since the trivial polynomial is considered in2, we may avoid the case whereκ0.

A further simplification of2.11is possible in view of the following translation invariance.

Lemma 2.2. Letτ,Δ,Υ∈Z, δ∈Zwithτ, δ 1 andα, κ, r∈Rwithκ /0.Then v{vit}is a Δ,Υ-periodic traveling wave solution with velocity−δ/τfor the following equation:

vt1ivti α

vi1t −2vitvti−1 κ

vtir

, i∈Z, t∈N, 2.12

if, and only if, y{yit}{vitr}is aΔ,Υ-periodic traveling wave solution with velocity−δ/τ for the following equation

yit1yti α

yi1t −2yti yti−1

κyit, i∈Z, t∈N. 2.13

Therefore, from now on, we assume in2.2that

α∈R, gκf, 2.14

where

κ /0,

fx 1 forx∈R, or, fx x forx∈R. 2.15

As for the traveling wave solutions, we also have the following reflection invariance resulta direct verification is easy and can be found in2.

(6)

Lemma 2.3cf. proof of2, Theorem 3. Given anyδ∈Z\ {0}andτ ∈Z withτ, δ 1.If {vti }is a traveling wave solution of 2.2with velocity−δ/τ,then{wit}{vt−i}is also a traveling wave solution of 2.2with velocityδ/τ.

Let−δ ∈ Z andΔ,Υ, τ ∈ Z,where τ, δ 1.Suppose that v {vit} is aΔ,Υ- periodic traveling wave solution of2.2with velocity −δ/τ. Then it is easy to check that w {wti } {vt−i}is also temporal Δ-periodic and spatialΥ-periodic. From this fact and Lemma 2.3, when we want to consider theΔ,Υ-periodic traveling wave solutions of2.2 with velocity−δ/τ, it is sufficient to consider theΔ,Υ-periodic traveling wave solutions of 2.2with velocityδ/τ. In conclusion, from now on, we may restrict our attention to the case where

τ ∈Z, δN with τ, δ 1. 2.16

3. Basic Facts

Some additional basic facts are needed. Let us state these as follows. First, letAξbe a circulant matrix defined by

A2

2 −2

−2 2

,

Aξ

⎢⎢

⎢⎢

⎢⎢

⎢⎢

⎢⎢

⎢⎣

2 −1 0 −1

−1 2 −1 0

· · ·

· · ·

0 −1 2 −1

−1 0 −1 2

⎥⎥

⎥⎥

⎥⎥

⎥⎥

⎥⎥

⎥⎦

ξ×ξ

, ξ≥3.

3.1

Second, we set

λi,ξ4sin2

ξ

, i∈Z, ξ∈Z, 3.2

ui,ξm 1

ξcos2miπ/ξ sin2miπ/ξ, m, i∈Z; ξ∈Z. 3.3

It is knownsee, e.g.,3that for anyξ≥2,the eigenvalues ofAξ areλ1,ξ, . . . , λξ,ξand the eigenvector corresponding toλi,ξis

ui,ξ

ui,ξ1 , . . . , ui,ξξ

fori∈ {1, . . . , ξ}, 3.4

(7)

and that u1,ξ, u2,ξ, . . . , uξ,ξ are orthonormal. It is also clear that u0,ξ uξ,ξ, λ0,ξ λξ,ξ, λi,ξλξ−i,ξ, and

uξ−i,ξm 1

ξcos2miπ/ξ−sin2miπ/ξ ∀m, i∈Z. 3.5

Therefore,λ0,ξ, . . . , λξ/2,ξare all distinct eigenvalues ofAξwith corresponding eigenspaces span{uξ},span{u1, uξ−1}, . . . ,span{uξ/2, uξ−ξ/2},respectively.

Given any finite sequencev {v1, v2, . . . , vξ}or vectorv

v1, v2, . . . , vξ

, where ξ≥1, itsperiodicextension is the sequencev{vi}i∈Zdefined by

vivimodξ, i∈Z. 3.6

Suppose that Υ,Δ ∈ Z and τ, δ satisfy 2.16. When we want to know whether a double sequence is aΔ,Υ-periodic traveling wave solution of2.2with velocity−δ/τ,the following two results will be useful.

Lemma 3.1. Letξ, η∈Zwithξ2 and letui,ξbe defined by3.4.

iSupposeξ ≥ 4.Letj, k ∈ {1, . . . ,ξ/2}withj /kanda, b, c, d ∈ Rsuch thatauj,ξ buξ−j,ξandcuk,ξduξ−k,ξare both nonzero vectors. Thenηis a period of the extension of the vectorauj,ξbuξ−j,ξcuk,ξduξ−k,ξif and only ifηj/ξ∈Zandηk/ξ∈Z. iiSuppose ξ ≥ 3.Let j ∈ {1, . . . ,ξ/2} and a, b, c ∈ Rsuch that buξ−j,ξcuj,ξ is a

nonzero vector. Thenauξ,ξbuξ−j,ξcuj,ξisξ-periodic if and only ifj, ξ 1.

iiiSupposeξ2.Leta, b∈Rsuch thatb /0.Thenau2,2bu1,2is 2-periodic.

Proof. To seei, we need to consider five mutually exclusive and exhaustive cases:aj, k∈ {1, . . . ,ξ/2−1};b ξ is odd, j ∈ {1, . . . ,ξ/2−1} andk ξ−1/2;c ξ is odd, k ∈ {1, . . . ,ξ/2−1}andj ξ−1/2;dξis even,j ∈ {1, . . . ,ξ/2−1}andk ξ/2;eξis even,k∈ {1, . . . ,ξ/2−1}andjξ/2.

Suppose that caseaholds. Take

uauj,ξbuξ−j,ξcuk,ξduξ−k,ξ, 3.7

wherea, b, c, d ∈Rsuch thatauj,ξbuξ−j,ξ andcuk,ξduξ−k,ξare both nonzero vectors.

Letu{ui}i∈Zbe the extension ofu,so thatuiuimodξfori∈Z.Then it is clear that for any i∈Z,

uiauj,ξi buξ−j,ξi cuk,ξi duξ−k,ξi 1

ξ

abcos2ijπ

ξ absin2ijπ

ξ cdcos2ikπ

ξ cdsin2ikπ ξ

.

3.8

(8)

By direct computation, we also have

u 1

ξcos2ηjπ ξ

abcos2ijπ

ξ absin2ijπ ξ

1

ξsin2ηjπ ξ

a−bcos2ijπ

ξ −absin2ijπ ξ

1

ξcos2ηkπ ξ

cdcos2ikπ

ξ cdsin2ikπ ξ

1

ξsin2ηkπ ξ

c−dcos2ikπ

ξ −cdsin2ikπ ξ

.

3.9

By3.8and3.9, we see thatηis a period ofu, that is,uiu 0 for alli∈Z,if, and only if, given anyi∈Z,

0 1 ξ

cos2ηjπ ξ −1

abcos2ijπ

ξ a−bsin2ijπ ξ

1 ξ

cos2ηkπ ξ −1

cdcos2ikπ

ξ cdsin2ikπ ξ

1

ξsin2ηjπ ξ

a−bcos2ijπ

ξ −absin2ijπ ξ

1

ξsin2ηkπ ξ

c−dcos2ikπ

ξ −cdsin2ikπ ξ

.

3.10

By3.3and3.5, we may rewrite3.10as

0

cos2ηjπ

ξ −1

auj,ξi buξ−j,ξi

sin2ηjπ ξ

−buj,ξi auξ−j,ξi

cos2ηkπ

ξ −1

cuk,ξi duξ−k,ξi

sin2ηkπ ξ

−cuk,ξi duξ−k,ξi .

3.11

By3.3again, we haveui,ξ ui,ξm for eachi, m∈Z.Hence we see thatηis a period ofuif, and only if,

0, . . . ,0

cos2ηjπ

ξ −1

auj,ξbuξ−j,ξ

sin2ηjπ ξ

−buj,ξauξ−j,ξ

×

cos2ηkπ

ξ −1

cuk,ξduξ−k,ξ

sin2ηkπ ξ

−cuk,ξduξ−k,ξ .

3.12

(9)

Note thatj ∈ {1, . . . ,ξ/2−1}implies thatuj,ξanduξ−j,ξ are distinct and hence they are linearly independent. Thus, the fact thatauj,ξbuξ−j,ξis not a zero vector implies|a||b|/0.

Similarly, we also have|c||d|/0.Then it is easy to check that auj,ξ buξ−j,ξ,−buj,ξ auξ−j,ξ, cuk,ξduξ−k,ξand−cuk,ξduξ−k,ξare linear independent. Hence we have thatη is a period ofuif and only if

cos2ηjπ

ξ −1cos2ηkπ

ξ −1sin2ηjπ

ξ sin2ηkπ

ξ 0. 3.13

In other words,ηis a period ofuif, and only if,ηj/ξ∈Zandηk/ξ∈Z.

The other casesb–ecan be proved in similar manners and hence their proofs are skipped.

To proveii, we first setuauξ,ξbuξ−j,ξcuj,ξ.As ini, we also know thatηis a period ofu{ui}i∈Z,whereuiuimodξ,if and only ifηj/ξ∈Z.That is,

η∈Z| ηj ξ ∈Z

η∈Z|η is a period ofu

. 3.14

Supposeξ, j 1.Ifηj/ξ∈Zfor someη∈Z,then we haveη0 modξbecauseξ, j 1.

Hence we have

min

η∈Z|η is a period ofu min

η∈Z| ηj ξ ∈Z

ξ. 3.15

In other words, ifξ, j 1,thenu isξ-periodic. Next, supposeξ, j η1/1; that is, there exists someξ1, j1 >1 such thatξη1ξ1andj η1j1.Note thatj < ξand hence we also have η1 < ξ1.Sinceξ η1ξ1, η1 < ξandη1 > 1,we have 1 < ξ1 < ξ.Takingη ξ1,then we have ηj/ξ∈Z.Henceηis a period ofuandη < ξ.That is,uis notξ-periodic. In conclusion, ifuis ξ-periodic, thenξ, j 1.

The proof of iii is done by recalling that u2,2 1/√

21,1 and u1,2 1/√

2−1,1and checking thatau2,2bu1,2is truly 2-periodic. The proof is complete.

The above can be used, as we will see later, to determine the spatial periods of some special double sequences.

Lemma 3.2. Letui,ξ be defined by3.4. Letξ ≥ 3, j ∈ {0,1, . . . ,ξ/2}andk ∈ {1, . . . ,ξ/2}

withj /k. Let further

uauj,ξbuξ−j,ξcuk,ξduξ−k,ξ, u−auj,ξbuξ−j,ξcuk,ξduξ−k,ξ,

3.16

wherea, b, c, d∈Rsuch thatauj,ξbuξ−j,ξis a nonzero vector. Define v{vit}by

vti

i∈Z

⎧⎨

u, iftis odd,

u, iftis even. 3.17

(10)

iSuppose thatj /0 andcuk,ξduξ−k,ξis a nonzero vector. Then v is spatialξ-periodic if, and only if,ηj/ξ /∈Zorηk/ξ /∈Zfor anyη∈ {1, . . . , ξ−1}withη|ξ.

iiSuppose thatj 0 andcuk,ξi duξ−k,ξi is a nonzero vector. Then v is spatialξ-periodic if, and only if,k, ξ 1.

iiiSuppose thatcuk,ξduξ−k,ξis a zero vector. Then v is spatialξ-periodic if, and only if, j, ξ 1.

Proof. To seei, suppose thatj /0 andcuk,ξduξ−k,ξis a nonzero vector.Note that the fact thatj, k ∈ {1, . . . ,ξ/2}withj /kimpliesξ ≥ 4.ByLemma 3.1i,ηis a period ofuif, and only if,ηj/ξ∈Zandηk/ξ∈Z. ByLemma 3.1iagain,ηj/ξ∈Zandηk/ξ∈Zif, and only if,ηis a period ofu.Hence the least period ofuis the same asuand v is spatialξ-periodic if, and only if,uisξ-periodic. Note thatξis a period ofu. ByLemma 2.1andLemma 3.1i, we haveuisξ-periodic if and only ifηj/ξ /∈Zorηk/ξ /∈Zfor anyη∈ {1, . . . , ξ−1}withη|ξ.

The assertionsiiandiiican be proved in similar manners. The proof is complete.

Lemma 3.3. Letξbe even withξ3 and letui,ξbe defined by3.4. Letj, k∈ {1, . . . ,ξ/2}and

uauj,ξbuξ−j,ξcuk,ξduξ−k,ξ, u−auj,ξbuξ−j,ξcuk,ξduξ−k,ξ,

3.18

wherea, b, c, d∈Rsuch thatauj,ξi buξ−j,ξi is a nonzero vector. Let v{vti }be defined by

vti

i∈Z

⎧⎨

u, iftis odd,

u, iftis even. 3.19

iSuppose thatcuk,ξduξ−k,ξis a nonzero vector. Thenvit1 vtiξ/2 for alli∈Zand tNif and only ifjis odd andkis even.

iiSuppose thatcuk,ξduξ−k,ξis a zero vector. Thenvit1vtiξ/2 for alli∈ZandtN if and only ifjis odd.

Proof. To seei, we first suppose thatjis odd andkis even. Note that

vi0 1 ξ

abcos2ijπ

ξ absin2ijπ

ξ cdcos2ikπ

ξ cdsin2ikπ ξ

, 3.20

v1i 1 ξ

−abcos2ijπ

ξ −a−bsin2ijπ

ξ cdcos2ikπ

ξ cdsin2ikπ ξ

. 3.21

(11)

For anys, i∈Z,it is clear that

cos2iξ/2sπ

ξ

⎧⎪

⎪⎪

⎪⎪

⎪⎩

cos2isπ

ξ ifsis even,

−cos2isπ

ξ ifsis odd,

sin2iξ/2sπ

ξ

⎧⎪

⎪⎪

⎪⎪

⎪⎩

sin2isπ

ξ ifsis even,

−sin2isπ

ξ ifsis odd.

3.22

Sincejis odd andkis even, by3.22, it is easy to see thatv1iv0iξ/2for alli∈Z.By the definition of v,we also have

vt2i vit, vti vt ∀i∈Z, t∈N. 3.23

In particular, we havevit1vtiξ/2for alli∈ZandtN.

For the converse, suppose thatj is even orkis odd. We first focus on the case thatj andkare both even. By3.20and3.22, we have

v0iξ/2 1 ξ

abcos2ijπ

ξ absin2ijπ

ξ cdcos2ikπ

ξ cdsin2ikπ ξ

. 3.24 Ifvt1i vtiξ/2for alli∈ZandtN,it is clear thatvi1 viξ/20 for alli∈Z. By3.21and 3.24, we have

2abcos2ijπ

ξ 2a−bsin2ijπ

ξ 0 ∀i∈Z. 3.25

That is, 2auj,ξbuξ−j,ξ 0.This is contrary to our assumption. That is, ifj, kare both even, then we havevit1/vtiξ/2for sometNandi∈Z.By similar arguments, in case wherej, k are both odd or wherej is even andkis odd, we also havevit1/vtiξ/2for sometNand i∈Z.In summary, ifvit1vtiξ/2for alli, t,thenjis odd andkis even.

The assertioniiis proved in a manner similar to that ofi. The proof is complete.

4. Necessary Conditions

Let Υ,Δ ∈ Z,in this section, we want to find the necessary and sufficient conditions for Δ,Υ-periodic traveling wave solutions of2.2with velocity−δ/τ,under the assumptions 2.14,2.15, and2.16.

(12)

We first consider the case wherefx 1 for allx∈R.Then we may rewrite2.2as vit1vitα

vti−1−2vti vti1

κ, i∈Z, t∈N, κ /0. 4.1

Suppose that v {vti }is a Δ,Υ-periodic traveling wave solutions of4.1with velocity

−δ/τ. For anyl, s∈Z,it is clear that Δ

t1

Υ i1

vtsil Δ

t1

Υ i1

vti . 4.2

Then we have

0Δ

t1

Υ i1

vit1vti α

Δ t1

Υ i1

vti−1−2vti vti1 Δ

t1

Υ i1

κ ΔΥκ /0. 4.3

This is a contradiction. In other words,Δ,Υ-periodic traveling wave solutions of4.1with velocity−δ/τdo not exist.

Next, we consider the casefx xand focus on the equation vt1ivit α

vti−1−2vti vi1t

κvit, i∈Z, t∈N, κ /0. 4.4

Before dealing with this case, we give some necessary conditions for the existence ofΔ,Υ- periodic traveling wave solutions of4.4with velocity−δ/τ.

Lemma 4.1. Letα, κ ∈ Rwithκ /0 andτ, δsatisfy2.16. Suppose that v {vit}is aΔ,Υ- periodic traveling wave solution of 4.4with velocity −δ/τ,where Δ 1 and Υ > 1.Then the matrixκIΥαAΥis not invertible and

v01 , . . . , vΥ0

is a nonzero vector in kerκIΥαAΥ.

Proof. Let v{vti }be a1,Υ-periodic traveling wave solution of2.2with velocity−δ/τ.

It is clear that

vt1i vti , vitvt ∀i, t. 4.5

Since v satisfies4.4, by4.5, we have κIΥαAΥ

v1t, . . . , vΥt

0, . . . ,0 ∀t∈N. 4.6

This fact implies that

vt1 , . . . , vΥt

is a vector in kerκIΥαAΥ.IfκIΥαAΥis invertible or

v1t, . . . , vtΥ

0, by direct computation, we have vt1 , . . . , vtΥ

0, . . . ,0 ∀t∈N, 4.7

(13)

and hencevti 0 for alltN andi ∈ Z.This is contrary toΔbeing the least among all spatial periods andΔ>1. That is,κIΥαAΥis not invertible and

v01 , . . . , v0Υ

is a nonzero vector in kerκIΥαAΥ.The proof is complete.

Lemma 4.2. Letα, κ ∈ Rwithκ /0 andτ, δsatisfy2.16. Suppose that v {vit}is aΔ,Υ- periodic traveling wave solution of 4.4with velocity−δ/τ, whereΔ > 1 andΥ 1.ThenΔ 2, κ−2,eachv0i /0 and

vti

⎧⎨

v0i , iftis even, i∈Z,

−v0i , iftis odd, i∈Z. 4.8

Proof. From the assumption of v,we have

vivti , vti vti1 ∀i, t. 4.9 Note that v also satisfies4.4. Hence by4.9and computation, we have

vt1i 1κt1v0i , i∈Z, t∈N. 4.10

Ifvj00 for somej,by4.9and4.10, we havev0i 0 for alliandvti 0 for anyi, t.This is contrary toΔbeing the least among all temporal periods andΔ>1.Hence we havevi0/0 for alli.Then it is clear thatvti is divergent ast → ∞if|1κ|>1 andvtiλast → ∞ for alli∈Zif|1κ|<1.This is impossible because v is temporalΔ-periodic andΔ>1. Thus we know that|1κ|1.Sinceκ /0,we know thatκ−2.By4.10, we have

vit

⎧⎨

v0i , if tis even, i∈Z,

−v0i , if tis odd, i∈Z. 4.11

Lemma 4.3. Letα, κ∈Rwithκ /0, τ, δsatisfy2.16andλi,ξare defined by3.2. Suppose that v{vti }is aΔ,Υ-periodic traveling wave solution of 4.4with velocity−δ/τ,whereΔ>1 and Υ>1.Then the following results are true.

iFor anytN,one has

vt11 , . . . , vt1Υ

1κIΥαAΥt1

v01 , . . . , v0Υ

. 4.12

iiThe vector

v10, . . . , v0Υ

is the sum of the vectorsuandw,whereuis an eigenvector of 1κIΥαAΥcorresponding to the eigenvalue−1 andwis either the zero vector or an eigenvector of1κIΥαAΥcorresponding to the eigenvalue 1.

iiiThe matrix1κIΥαAΥhas an eigenvalue−1,that is,1κα4 sin2jπ/Υ −1 for somej ∈ {0,1, . . . ,Υ/2}.

iv Δ 2.

(14)

Proof. To seei, note that the assumption on v implies

vi vti , vti vt ∀i, t. 4.13

Since v is a solution, by4.13, we know that

vt11 , . . . , vt1Υ

1κIΥαAΥ

vt1 , . . . , vtΥ

, tN. 4.14

By direct computation, we have

vt11 , . . . , vΥt1

1κIΥαAΥt1

v01 , . . . , vΥ0

, tN. 4.15

Foriiandiii, by takingt1 Δin4.12, it is clear from4.13that

v01 , . . . , vΥ0

v1Δ, . . . , vΔΥ

1κIΥαAΥΔ

v01 , . . . , v0Υ

. 4.16

Thus

v01 , . . . , vΥ0

is an eigenvector of1κIΥαAΥΔcorresponding to the eigenvalue 1.This implies that the matrix1κIΥαAΥmust have eigenvalue 1 or−1,and

v10, . . . , v0Υ

uw, 4.17

whereuis either the zero vector or an eigenvector of1κIΥαAΥcorresponding to the eigenvalue−1 andwis either a zero vector or an eigenvector of1κIΥαAΥcorresponding to the eigenvalue 1. Suppose that uis the zero vector, or, −1 is not an eigenvalue of1 κIΥαAΥ.Then 1 must be an eigenvalue of1κIΥαAΥandwmust be an eigenvector corresponding to the eigenvalue 1; otherwise,

v10, . . . , v0Υ

0 and this is impossible.

Thus, 1 is a temporal period of v.This is contrary toΔbeing the least among all periods and Δ>1.In conclusion,1κIΥαAΥhas eigenvalue−1 and

v10, . . . , v0Υ

uw,where uis an eigenvector of1κIΥαAΥ corresponding to the eigenvalue−1,andw is either a zero vector or an eigenvector of1κIΥαAΥ corresponding to the eigenvalue 1.Since 1καλ0,Υ, . . . ,1καλΥ/2,Υare all distinct eigenvalues of1κIΥαAΥ,there exists somej∈ {0,1, . . . ,Υ/2}such that1καλj,Υ −1.

To seeiv, recall the result inii. We have

v10, . . . , v0Υ

1κIΥαAΥ2

v01 , . . . , vΥ0

. 4.18

It is also clear that

vt21 , . . . , vt2Υ

1κIΥαAΥ2

vt1 , . . . , vtΥ

, tN. 4.19

(15)

That is, 2 is a temporal period of v.By the definition ofΔandΔ > 1,we haveΔ 2.The proof is complete.

Next, we consider one result about the relation betweenδandΥunder the assumption that doubly-periodic traveling wave solutions of4.4exist.

Lemma 4.4. Letα, κ ∈ Rwithκ /0 andτ, δsatisfy2.16. Suppose that v {vit}is aΔ,Υ- periodic traveling wave solution of 4.4with velocity−δ/τ,whereΔ 2 andΥ≥1.

iIfτis even, thenδT1Υfor some odd integerT1andΥis odd.

iiIfτis odd, thenΥis even andδT1Υ/2 for some odd integerT1. Proof. By the assumption on v,we have

vt2i vti , vitvt ∀i, t. 4.20

Since v is a traveling wave, we also know that

vi vt ∀i, t. 4.21

To seei, suppose thatτis even. Then from4.20and4.21, we have

vti vit2· · ·vivt ∀i, t. 4.22

That is,δis also a spatial period of v.ByLemma 2.1and the definition ofΥ, it is easy to see thatδ0 mod Υ.Sinceδ0 mod Υ, τis even andτ, δ 1,we haveδT1Υfor some odd integerT1andΥis odd.

Forii, suppose thatτ is odd.Then from4.20and4.21, we have

vt1i · · ·vi vt ∀i, t. 4.23

By4.20and4.23, we know that

vti vt2i vt1vti2δ ∀i, t. 4.24

That is, 2δis also a spatial period of v.ByLemma 2.1and the definition ofΥ, it is easy to see that 2δ0 mod Υ.Ifδ0 mod Υ.From4.23, we have

vit1vt· · ·vit ∀i, t. 4.25

Then 1 is a temporal period of v and this is contrary to Δ 2. Thusδ /0 mod Υ. Since 2δ 0 mod Υ,the fact thatΥ is odd impliesδ 0 mod Υ.This leads to a contradiction.

So we must have that Υ is even and δ 0 modΥ/2.Note that δ 0 modΥ/2and δ /0 mod ΥimpliesδT1Υ/2 for some odd integerT1.The proof is complete.

(16)

5. Existence Criteria

Now we turn to our main problem. First of all, letα, κ∈Rwithκ /0 andτ, δsatisfy2.16. If Υ,Δ∈ZwithΔ>1 and if4.4has aΔ,Υ-periodic traveling wave solution of4.4with velocity−δ/τ,by Lemmas4.2and4.3,Δmust be 2.For this reason, we just need to consider five mutually exclusive and exhaustive cases:i Υ Δ 1;ii Δ 1 andΥ>1;iii Δ 2 andΥ 1;iv Δ 2 andΥ 2 andv Δ 2 andΥ≥3.

The conditionΥ Δ 1 is easy to handle.

Theorem 5.1. Let α, κ ∈ R with κ /0 and τ, δ satisfy 2.16. Then the unique 1,1-periodic traveling wave solution of 4.4is{vti 0}.

Proof. If v{vti }is a1,1-periodic traveling wave solution of4.4, thenvitcfor alli∈Z andtN,wherec∈R.Substituting{vit c}into4.4, we havec0.Conversely, it is clear that{vit 0}is a1,1-periodic traveling wave solution.

Theorem 5.2. Letα, κ∈Rwithκ /0 andτ, δsatisfy2.16. Letλi,ξandui,ξbe defined by3.2 and3.4, respectively. Then the following results hold.

iFor any Δ 1 and any Υ ≥ 2,4.4has a 1,Υ-periodic traveling wave solutions of 2.2with velocity −δ/τ if, and only if,δ 0 mod Υ, and καλj,Υ 0 for some j ∈ {1, . . . ,Υ/2}withj,Υ 1.

iiEvery1,Υ-periodic traveling wave solution v{vti }is of the form

vti

vi0

i∈Zu ∀t∈N, 5.1

whereu aujbuΥ−j for somea, b ∈ Rsuch thataujbuΥ−j is a nonzero vector, and the converse is true.

Proof. Fori, let v{vit}be a1,Υ-periodic traveling wave solution of2.2with velocity

−δ/τ.From the assumption on v,we have {vt1i }i∈Z {vti }i∈Zfor alltN andΥis the least spatial period. Hence given anytN,it is easy to see that the extension {vit}i∈Z of vt1 , . . . , vΥtisΥ-periodic. Note that we also have

vt1i vti , vitvt ∀i, t. 5.2

Since v is a traveling wave, from5.2, we know that

vtvi vitτ−1· · ·vti ∀i, t. 5.3

Therefore, given anytN, δis a period of{vit}i∈Z. ByLemma 2.1, we haveδ0 mod Υ.

By Lemma 4.1, we also know that κIΥαAΥ is not invertible and v01 , . . . , v0Υ is a nonzero vector in kerκIΥαAΥ. Note that καλ0,Υ, . . . , καλΥ/2,Υ are all

(17)

distinct eigenvalues of κIΥαAΥ with corresponding eigenspaces span{uΥ,Υ}, . . . , span{uΥ/2,Υ, uΥ−Υ/2,Υ},respectively. Sinceκ−αλ0,Υκ /0 andκIΥ−αAΥis not invertible, we haveκ−αλj,Υ0 for somej ∈ {1, . . . ,Υ/2}.Hence kerκIΥ−αAΥ span{uj,Υ, uΥ−j,Υ} and it is clear that

v01 , . . . , vΥ0

auj,ΥbuΥ−j,Υ, 5.4

wherea, b ∈ Rsuch that auj,ΥbuΥ−j,Υis a nonzero vector. If Υ 2,we see thatj must be 1 sincej ∈ {1, . . . ,Υ/2}.It is clear thatj,Υ 1,2 1.SupposeΥ≥3 and recall that the extension{vi0}i∈Z ofv01 , . . . , v0Υ isΥ-periodic. ByLemma 3.1ii,the extension{v0i } ofv10, . . . , v0Υ isΥ-periodic if and only ifj,Υ 1.

Conversely, supposeδ 0 modΥ; there exists somej ∈ {1, . . . ,Υ/2}such thatκαλj,Υ 0 andj,Υ 1 whenΥ≥ 3. Let v {vit}satisfy5.1. By the definition of v,it is clear that v is temporal 1-periodic andΥis a spatial period of v. SupposeΥ 2 and then we have thatu abu1.The fact thatuis not a zero vector impliesab /0.ByLemma 3.1iii, we have thatuis 2-periodic. By5.1, it is clear that v is spatial 2-periodic. SupposeΥ ≥ 3.

Sincej,Υ 1,byLemma 3.1ii, we haveuisΥ-periodic. By5.1again, it is also clear that v is spatialΥ-periodic. In conclusion, we have that v is spatialΥ-periodic, that is,

vt vit ∀t∈N, i∈Z. 5.5

Sinceu ∈ kerκIΥαAΥ,from the definition of v,it is easy to check that v is a solution of 4.4. Finally, sinceδ0 mod Υ,by5.1and5.5, we know that

vi · · ·vit1vti vt· · ·vt; 5.6 that is, v is traveling wave with velocity−δ/τ.

To seeii, note that from the second part of the proof ini, it is easy to see that any v {vti }satisfying5.1is a1,Υ-periodic traveling wave solution of4.4with velocity−δ/τ.

Also, by the first part of the proof ini, the converse is also true. The proof is complete.

We remark that any 1,Υ-periodic traveling wave solution v {vti } of 4.4 is a standing wave since this v is also a traveling wave with velocity 0, that is,vt1i vitfor all i∈ZandtN.

Theorem 5.3. LetΥ 1,Δ 2, α, κ∈Rwithκ /0 andτ, δsatisfy2.16.Then

i 4.4has aΔ,Υ-periodic traveling wave solution with velocity−δ/τif, and only if,κ

−2 andτis even;

iifurthermore, every such solution v{vit}is of the form

vit

⎧⎨

c, iftis even, i∈Z,

−c, iftis odd, i∈Z. 5.7

wherec /0,and the converse is true.

参照

関連したドキュメント

Key words: spreading speed, convolution model, integro-differential equation, discrete monostable equation, nonlocal evolution equation, Fisher-Kolmogorov equation.. ∗ Department

By using the Fourier transform, Green’s function and the weighted energy method, the authors in [24, 25] showed the global stability of critical traveling waves, which depends on

In this work, we have applied Feng’s first-integral method to the two-component generalization of the reduced Ostrovsky equation, and found some new traveling wave solutions,

In view of the existence of traveling wavefronts for both the nonlocal monos- table equation (1.1) and the bistable non-local delayed diffusion equation [20], it is then expected

A monotone iteration scheme for traveling waves based on ordered upper and lower solutions is derived for a class of nonlocal dispersal system with delay.. Such system can be used

For certain singular limits of phase field models, u is approximately constant near the phase interface, and an asymptotic analysis can be conducted to de- termine the wave profile

7, Fan subequation method 8, projective Riccati equation method 9, differential transform method 10, direct algebraic method 11, first integral method 12, Hirota’s bilinear method

If D ( ρ ) ≥ 0, the existence of solutions to the initial-value problem for (1.1) is more or less classical [24]; however, the fine structure of traveling waves reveals a variety