This paper is concerned with boundary value problems for a fourth-order nonlinear difference equation. Via variational methods and critical point theory, sufficient conditions are obtained for the existence of at least two nontrivial solutions, the existence of n distinct pairs of nontrivial solutions, and nonexistence of solutions. Some examples are provided to show the effectiveness of the main results.
National Natural Science Foundation of China115011941. Introduction
Throughout this paper, we denote by N, Z, R the sets of all natural numbers, integers, and real numbers, respectively. Let the symbol ∗ denote the transpose of a vector. For any integers c and d with c≤d, [c,d]Z is defined by the discrete interval {c,c+1,…,d}.
Now, we are concerned with the existence and nonexistence solutions to the fourth-order nonlinear difference equation(1)Δ2pt-2Δ2xt-2+qtxt=ft,xt,t∈1,TZ,satisfying the boundary value conditions(2)Δkx-1=ΔkxT-1,k=0,1,2,3,where 1≤T∈Z, Δ is the forward difference operator defined by Δx(t)=x(t+1)-x(t), Δkx(t)=Δ(Δk-1x(t))(2≤k≤4), Δ0x(t)=x(t), p(t)∈C[-1,T]Z,R with p(-1)=p(T-1), p(0)=p(T), q(t)∈C[1,T]Z,R, f(t,x)∈C([1,T]Z×R,R).
As usual, a solution of (1), (2), in other words, a function x:[-1,T+2]Z→R, satisfies both (1) and (2).
We may think of boundary value problem (BVP) (1), (2) as being a discrete analogue of the following fourth-order nonlinear differential equation:(3)psx′′s′′-qsxs=ft,xs,t∈0,1,with boundary value conditions(4)xk0=xk1,k=0,1,2,3.(3) includes the following differential equation:(5)x4s=fs,xs,s∈R,which is used to describe the bending of an elastic beam [1]. Equations similar in structure to (3) have been studied by many researchers using a variety of methods; see, for example, [2–12].
It is well-known that the study of nonlinear difference equations [13–29] has long been an important one as a result of the fact that they arise in numerical solutions of both ordinary and partial differential equations as well as in applications to different areas of applied mathematics and physics.
Domshlak and Matakaev [17] in 2001 investigated the oscillation properties of the delay difference equation(6)xn+1-xn+bnxn-k=0,bn>0,n≥1,for k=2 and k=3 near the 2-periodic critical states with respect to its oscillation properties. By making use of “the Sturmian comparison method: discrete version”, they obtained some conditions for the existence and for the nonexistence of eventually positive solution.
Using the critical point theory, Yang [27] studied the following higher order nonlinear difference equation:(7)∑i=0nrixk-i+xk+i=fk,xk+M,…,xk,…,xk-M,n∈N,k∈Z1,T,with boundary value conditions(8)x1-m=x2-m=⋯=x0=0,xT+1=xT+2=⋯=xT+m=0.Some sufficient conditions for the existence of the solution to the boundary value problem (7), (8) are obtained.
In 2010, He, Yang and Yang [18] considered the following second-order three-point discrete boundary value problem:(9)Δ2xt-1+ft,xt=0,t∈1,nZ,x0=0,xn+1=αxm.By using the topological degree theory and the fixed point index theory, they provided sufficient conditions for the existence of sign-changing solutions, positive solutions, and negative solutions.
Investigating the high order difference equation(10)Δnrt-nϕcΔnxt-n=-1nft,xt+1,xk,xt-1,t∈Z,Leng [22] established some new criteria for the existence and multiplicity of periodic and subharmonic solutions of (10) based on the linking theorem in combination with variational technique.
Leszczyński [23] considered the difference equation with both advance and retardation,(11)Δ2γt-1ϕptΔ2xt-2=ft,xt+1,xt,xt-1,t∈Z1,k,with boundary value conditions(12)Δx-1=Δx0=0,xk+1=xk+2=0.They applied the direct method of the calculus of variations and the mountain pass technique to prove the existence of at least one and at least two solutions. Nonexistence of nontrivial solutions is also undertaken.
In this paper, we shall study the boundary value problem for a fourth-order nonlinear difference equation (1), (2). Via variational methods and critical point theory, sufficient conditions are obtained for the existence of at least two nontrivial solutions, the existence of n distinct pairs of nontrivial solutions, and nonexistence of solutions. The motivation for the present work stems from the recent papers [3, 5].
Throughout the whole paper, we suppose that there exists a function G(t,x) such that(13)Gt,x=∫0xft,vdv,for any (t,x)∈[1,T]Z×R.
The remainder of this article is organized as follows. In Section 2, we shall give some preliminary lemmas and establish the variational structure of BVP (1), (2). In Section 3, we shall give sufficient conditions to the existence and nonexistence solutions. In Section 4, we shall complete the proofs of the main results. Some examples illustrating our main results are given in Section 5.
For the basic knowledge of variational methods, the reader is referred to [30–32].
2. Preliminary Lemmas
Assume that X is a real Banach space and I∈C1(X,R) is a continuously Fréchet differentiable functional defined on X. As usual, I is said to satisfy the Palais-Smale condition if any sequence xkk=1∞⊂X for which Ixkk=1∞ is bounded and I′xk→0 as k→∞ possesses a convergent subsequence. Here, the sequence xkk=1∞ is called a Palais-Smale sequence.
Let X be a real Banach space. We denote by the symbol Br the open ball in X about 0 of radius r, ∂Br its boundary, and B-r its closure.
In the present article, we define a vector space X by(16)X≔x:-1,T+2Z→R∣Δkx-1=ΔkxT-1,k=0,1,2,3,and for any x∈X, define(17)x,y≔∑t=1Txtyt,∀x,y∈X,and(18)x≔∑t=1Tx2t1/2,∀x∈X.
Remark 1.
For any x∈X, it is easy to see that(19)x-1=xT-1,x0=xT,x1=xT+1,x2=xT+2.As the case stands, X is isomorphic to RT. In the following and in the sequel, when we write x=x(1),x(2),…,x(T)∈RT, we always imply that x can be extended to a vector in X so that (19) is satisfied.
For any x∈X, let the functional I be denoted by(20)Ix≔12∑t=1Tpt-2Δ2xt-22+12∑t=1Tqtxt2-∑t=1TGt,xt.Then I∈C1(X,R) and(21)∂I∂xt=Δ2pt-2Δ2xt-2+qtxt-ft,xt,t∈1,TZ.
Thus, I′(x)=0 if and only if(22)Δ2pt-2Δ2xt-2+qtxt=ft,xt,t∈1,TZ.
Thereupon a function x∈X is a critical point of the functional I on X if and only if x is a solution of BVP (1), (2).
Let P be the T×T matrix. If T≥5, let(23)P=b1a1p10⋯0pT-1aTa1b2a2p2⋯00pTp1a2b3a3⋯0000p2a3b4⋯000⋯⋯⋯⋯⋯⋯⋯⋯0000⋯aT-3pT-300000⋯bT-2aT-2pT-2pT-1000⋯aT-2bT-1aT-1aTpT00⋯pT-2aT-1bT,where b(k)=p(k)+4p(k-1)+p(k-2),a(k)=-2(p(k-1)+p(k)),k=1,2,…,T.
If T=4, let(24)P=p-1+4p0+p1-2p0+p1p1+p3-2p3+p4-2p0+p1p0+4p1+p2-2p1+p2p2+p4p1+p3-2p1+p2p1+4p2+p3-2p2+p3-2p3+p4p2+p4-2p2+p3p2+4p3+p4.
If T=3, let(25)P=p-1+4p0+p1p2-2p0+p1p1-2p2+p3p2-2p0+p1p0+4p1+p2p3-2p1+p2p1-2p2+p3p3-2p1+p2p1+4p2+p3.
If T=2, let(26)P=p-1+4p0+3p1-2p0-4p1-2p2-2p0-4p1-2p2p0+4p1+3p2.
If T=1, let P=(0).
Then the functional I(x) can be rewritten as(27)Ix=12x∗Px+12∑t=1Tqtxt2-∑t=1TGt,xt.
Lemma 2 (linking theorem [30]).
Let X be a real Banach space, X=X1⊕X2, where X1 is finite dimensional. Assume that I∈C1(X,R) satisfies the Palais-Smale condition and the following:
There are positive constants r and ρ such that I|∂Bρ∩X2≥r.
There is a ν∈∂B1∩X2 and a positive constant c~≥ρ such that I|∂Φ≤0, where Φ=(B-c~∩X1)⊕{sν∣0<s<c~}.
Then I possesses a critical value c≥r, where(28)c=infφ∈Πsupx∈ΠIφx,
and Π={φ∈C(Φ-,X)∣φ|∂Ω=id}, where id denotes the identity operator.
Lemma 3 (Clark theorem [30]).
Let X be a real Banach space, I∈C1(X,R), with I being even, bounded from below, and satisfying Palais-Smale condition. Suppose I(0)=0, there is a set Ω⊂X such that Ω is homeomorphic to ST-1 (T-1 dimension unit sphere) by an odd map, and supΩI<0. Then I has at least T distinct pairs of nonzero critical points.
3. Main Results
We now state our main theorems in this paper.
Theorem 4.
Suppose that the function G(t,x)≥0 and the following conditions are satisfied:
There are constants ρ1>0 and c1∈0,1/2(λ_+q_) such that(29)Gt,x≤c1x2,∀t∈1,TZ,x≤ρ1.
There are constants c2∈1/2λ-+q-,+∞ and c3>0 such that(30)Gt,x≥c2x2-c3,∀t∈1,TZ,x∈R.
Here λ_ and λ- are constants which can be referred to (32) and (33).
Then BVP (1), (2) has at least three solutions.
Corollary 5.
Suppose that the function G(t,x)≥0 and the conditions (p), (q), (W), (G1), and (G2) are satisfied. Then BVP (1), (2) has at least two nontrivial solutions.
Theorem 6.
Suppose that the function G(t,x)≥0, (p), (q), (W) and the following conditions are satisfied:
For any (t,x)∈[1,T]Z×R, lim|x|→0G(t,x)/x2=0.
For any (t,x)∈[1,T]Z×R, there are constants c4>0, α>2 and c5>0 such that(31)Gt,x≥c4xα-c5.
Then BVP (1), (2) has at least three solutions.
Corollary 7.
Suppose that the function G(t,x)≥0, (p), (q), (W) and the conditions (G3) and (G4) are satisfied. Then BVP (1), (2) has at least two nontrivial solutions.
Theorem 8.
Suppose that the function G(t,x)≥0, (p), (q), (W), (G1), (G2) and the following condition are satisfied:
f(t,-x)=-f(t,x),∀(t,x)∈[1,T]Z×R.
Then BVP (1), (2) has at least n distinct pairs of nontrivial solutions, where n is the dimension of X2 which can be referred to (34).
Theorem 9.
Suppose that the following conditions are satisfied:
For any t∈[-1,T]Z, p(t)≤0.
For any t∈[1,T]Z, q(t)≤0.
For any x∉0,t∈[1,T]Z, xf(t,x)>0.
Then BVP (1), (2) has no nontrivial solutions.
4. Proofs of the Main Results
In this section, we shall finish proofs of the main results via variational methods.
Proof of Theorem 4.
The matrix P satisfies that P is positive semidefinite. In fact, from (W), it is obvious that 0 is an eigenvalue of P with an eigenvector (1,1,…,1)∗. Define the eigenvalues of P by λ1,λ2,…,λT.
Let X1={(d,d,…,d)∗∈X∣d∈R}. Clearly, X1 is an invariant subspace of X. The space X2 is defined by(34)X=X1⊕X2.
Let xkk∈N⊂X be such that Ixkk∈N is bounded and I′xk→0 as k→∞. Thus, for any k∈N, there exists a positive constant C such that(35)-C≤Ixk≤C.For any xkk∈N⊂X, it comes from (27) and (G2) that(36)-C≤Ixk=12xk∗Pxk+12∑t=1Tqtxkt2-∑t=1TGt,xkt≤12λ-xk2+12q-xk2-∑t=1Tc2xkt2-c3=12λ-+q--c2xk2+c3T.Therefore,(37)c2-12λ-+q-xk2≤C+c3T.From (G2), we have c2>1/2λ-+q-. (37) implies that xkk∈N is a bounded in X. Since the dimension of X is finite, xkk∈N possesses a convergent subsequence. Consequently, the functional I(x) satisfies the Palais-Smale condition. Therefore, it suffices to prove that I(x) satisfies the conditions (I1) and (I2) of Lemma 2.
For any t∈Z[1,T], it follows from (q), (G1), and (G2) that G(t,0)=0 and f(t,0)=0. Hence, x=0 is a trivial solution of BVP (1), (2).
From the proof of (36), we have that I(x) is bounded from above in X. Denote(38)c0=supx∈XIx.As a consequence, on the one hand, there is a sequence xi in X such that(39)c0=limi→∞Ixi.By (27), on the other hand, the functional I(x) satisfies(40)Ix≤12λ-+q--c2x2+c3T,∀x∈X.(40) means that lim‖x‖→+∞I(x)=-∞ which implies that xi is bounded. As a result, xi has a convergent subsequence in X denoted by xik. Let(41)x0=limk→+∞xik.By the reason of the continuity of I(x) in x, it is easy to see that I(x0)=c0. That is, x0∈X is a critical point of I(x).
For any x∈X2,‖x‖≤ρ1, via (27) and (G1), the functional I(x) satisfies(42)Ix=12x∗Px+12∑t=1Tqtxt2-∑t=1TGt,xt≥12λ_x2+12q_x2-c1∑t=1kxt2≥12λ_+q_-c1x2.Take(43)r=12λ_+q_-c1ρ12.Then(44)Ix≥r,∀x∈X2∩∂Bρ1.In other words, there exist two positive constants r and ρ1 such that I|∂Bρ1∩X2≥r. Hence, I(x) satisfies (I1) of Lemma 2.
For any x∈X1, Px=0. Combining with (q), we have(45)Ix=12x∗Px+12∑t=1Tqtxt2-∑t=1TGt,xt≤-∑t=1TGt,xt≤0.Therefore, x0∉X1 and the critical point x0 of I(x) corresponding to the critical value c0 is a nontrivial solution of BVP (1), (2).
Next, we shall prove (I2) of Lemma 2.
Take y∈∂B1∩X2; for any z∈X1 and ξ∈R, let x=ξy+z. By (F2), we have(46)Ix=12ξy+z∗Pξy+z+12∑t=1Tqtξyt+zt2-∑t=1TGt,ξyt+zt≤12ξy∗Pξy+12∑t=1Tqtξyt+zt2-∑n=1kc2ξyt+zt2-c3≤12λ-ξ2+12q-∑t=1Tξyt+zt2-c2∑t=1Tξyt+zt2+c3T=12λ-+q--c2ξ2+12q--c2z2+c3T≤12q--c2z2+c3T.Thereby, there exists a constant c~>ρ1 such that(47)Ix≤0,∀x∈∂Φ,where Φ=(B-c~∩X1)⊕{sν∣0<s<c~}. According to Lemma 2, I(x) has a critical value c≥r>0, where(48)c=infφ∈Πsupx∈ΠIφx,and Π={φ∈C(Φ-,X)∣φ|∂Ω=id}.
Similar to the proof of Theorem 1.1 in [22], we can prove that BVP (1), (2) has at least three solutions. For simplicity, its proof is omitted.
Remark 10.
From the course of the proof of Theorem 4, the conclusion of Corollary 5 is evidently correct.
Remark 11.
The techniques of the proof of Theorem 6 are just the same as those carried out in the proof of Theorem 4. We do not repeat them here.
Remark 12.
According to Theorem 6, it is easy to see that the conclusion of Corollary 7 is true.
Proof of Theorem 8.
For any t∈[1,T]Z, by the continuity of f(t,x) in x, I(x) can be viewed as a continuously differentiable functional defined on X. It comes from (q) and (G1) that I(0)=0. Owing to the condition (ψ), I(x) is even. From the process of proof of Theorem 4, I(x) is bounded from below and satisfies the Palais-Smale condition. Next, in the light of Clark Theorem, we shall find a set Ω and an odd map such that Ω is homeomorphic to ST-1 by an odd map.
Set(49)Ω=∂Bρ1∩X2.It is obvious that Ω is homeomorphic to ST-1 by an odd map. (44) implies that supΩ(-I(x))<0. As a result of Clark Theorem, I(x) has at least T distinct pairs of nonzero critical points. As a consequence, BVP (1), (2) has at least T distinct pairs of nontrivial solutions. The desired result is obtained.
Proof of Theorem 9.
On the contrary, we suppose that BVP (1), (2) has a nontrivial solution. Therefore, I(x) has a nonzero critical point x~. Since(50)∂I∂x~t=Δ2ptΔ2x~t-2+qtx~t-ft,x~t,t∈1,TZ,we have(51)∑t=1Tft,x~tx~t=∑t=1TΔ2ptΔ2x~t-2+qtx~tx~t=∑t=1TptΔ2x~t2+∑t=1Tqtx~t2.On the one hand, it follows from (p′) and (q′) that(52)∑t=1Tft,x~tx~t≤0.On the other hand, by (ϕ), we have(53)∑t=1Tft,x~tx~t>0,which is a contradiction with (52).
5. Some Examples
In this section, we shall provide three examples to illustrate our main results.
Example 1.
For t∈[1,3]Z, assume that(54)Δ2t-22Δ2xt-2-t3xt=15xt4,satisfying the boundary value conditions(55)x-1=x2,Δx-1=Δx2,Δ2x-1=Δ2x2,Δ3x-1=Δ3x2.We have(56)pt=t2,qt=-t3,ft,ut=15xt4,t∈1,3Z,with(57)p-1=4,p0=9,and(58)Gt,ut=xt5,t∈1,3Z.Besides,(59)P=41-16-25-1617-1-25-126,and the eigenvalues of P are λ1=0,λ2=21, and λ3=63. It is easy to verify that all the suppositions of Theorem 4 are satisfied and then BVP (54), (55) has at least three solutions.
Example 2.
For t∈[1,4]Z, assume that(60)Δ2t-2Δ2xt-2-t6xt=xt3,satisfying the boundary value conditions(61)x-1=x3,Δx-1=Δx3,Δ2x-1=Δ2x3,Δ3x-1=Δ3x3.We have(62)pt=t,qt=-t6,ft,ut=xt3,t∈1,4Z,with(63)p-1=3,p0=4,and(64)Gt,ut=14xt4,t∈1,4Z.Besides,(65)P=20-104-14-1010-664-612-10-146-1018,and the eigenvalues of P are λ1=0,λ2≈6.8154,λ3≈11.1773, and λ4≈42.0074. It is easy to verify that all the suppositions of Theorem 8 are satisfied and then BVP (60), (61) has at least 3 distinct pairs of nontrivial solutions.
Example 3.
For t∈[1,5]Z, assume that(66)-Δ2t-2Δ2xt-2-t6xt=xt5,satisfying the boundary value conditions(67)x-1=x4,Δx-1=Δx4,Δ2x-1=Δ2x4,Δ3x-1=Δ3x4.We have(68)pt=-t,qt=-t6,ft,ut=xt5,t∈1,5Z,with(69)p-1=-4,p0=-5,and(70)Gt,ut=16xt6,t∈1,5Z.It is easy to verify that all the suppositions of Theorem 9 are satisfied and then BVP (66), (67) has no nontrivial solutions.
Data Availability
No data were used to support this study.
Conflicts of Interest
The authors declare that there are no conflicts of interest regarding the publication of this paper.
Authors’ Contributions
All authors contributed equally and significantly in writing this article. All authors read and approved the final manuscript.
Acknowledgments
This project is supported by the National Natural Science Foundation of China (No. 11501194). This work was carried out while visiting Central South University. The author Haiping Shi wishes to thank Professor Xianhua Tang for his invitation.
GuptaC. P.Existence and uniqueness theorems for the bending of an elastic beam equation198826428930410.1080/00036818808839715MR922976Zbl0611.34015ChenP.TianC.Infinitely many solutions for Schrödinger-Maxwell equations with indefinite sign subquadratic potentials201422649250210.1016/j.amc.2013.10.069MR3144327Zbl1358.35027ChenP.HeX.TangX. H.Infinitely many solutions for a class of fractional Hamiltonian systems via critical point theory20163951005101910.1002/mma.3537MR3474137Zbl1336.340122-s2.0-84961147306GuoC.AgarwalR. P.WangC.O’ReganD.The existence of homoclinic orbits for a class of first-order superquadratic Hamiltonian systems20146183102MR3288213GuoC.O’ReganD.XuY.AgarwalR. P.Existence of homoclinic orbits for a class of first-order differential difference equations20153551077109410.1016/S0252-9602(15)30041-2MR3374045GuoC. J.O’ReganD.XuY. T.AgarwalR. P.Existence and multiplicity of homoclinic orbits of a second-order differential difference equation via variational methods201241115GuoC.O’ReganD.XuY.AgarwalR. P.Existence of subharmonic solutions and homoclinic orbits for a class of even higher order differential equations20119071169118310.1080/00036811.2010.524154MR2811969HeT.ChenC.HuangY.HouC.Infinitely many sign-changing solutions for p-Laplacian Neumann problems with indefinite weight201539737910.1016/j.aml.2014.08.011MR3269221LiYanyanLongYuhuaPeriodic solutions of a class of fourth-order superlinear differential equations20122012846986310.1155/2012/469863MR2984033TangX.ChenS.Ground state solutions of Nehari-Pohozaev type for Schrödinger-POIsson problems with general potentials20173794973500210.3934/dcds.2017214MR3661829Zbl1371.35051TangX. H.ChenS.Ground state solutions of Nehari-Pohozaev type for Kirchhoff-type problems with general potentials201756411013410.1007/s00526-017-1214-9MR3669779Zbl1376.35056ZhouH.YangL.AgarwalP.Solvability for fractional p-Laplacian differential equations with multipoint boundary conditions at resonance on infinite interval2017531-2517610.1007/s12190-015-0957-8MR3595988Zbl1365.340242-s2.0-84946866450AgarwalR. P.WongP. J. Y.1997404Dordrecht, The NetherlandsKluwer Academic PublishersMathematics and Its Applications10.1007/978-94-015-8899-7MR1447437Zbl0878.39001AlSharawiZ.CushingJ. M.ElaydiS.2014New York, NY, USASpringerZbl1297.39001ChenP.Existence of homoclinic orbits in discrete Hamiltonian systems without Palais-Smale condition201319111781179410.1080/10236198.2013.777716MR3173517Zbl1287.390042-s2.0-84885403313ChenP.TangX. H.Infinitely many homoclinic solutions for the second-order discrete p-Laplacian systems2013202193212MR3082759DomshlakY.MatakaevA.The third and fourth order linear difference equations near periodic critical states: sharp results in oscillation theory20017455157510.1080/10236190108808287MR1922589Zbl1009.39013HeT.YangW.YangF.Sign-changing solutions for discrete second-order three-point boundary value problems201020101470538710.1155/2010/705387MR2600329HeT.YangW.ZhaoH.LeiY.Positive periodic solutions for higher order functional difference systems depending on a parameter201310279180610.1007/s00009-012-0221-5MR3045680Zbl1279.390072-s2.0-84876683107HuangM.ZhouZ.Ground state solutions of the periodic discrete coupled nonlinear Schrödinger equations20153881682169510.1002/mma.3181MR3343582Zbl1325.39003HuangMeihuaZhouZhanOn the existence of ground state solutions of the periodic discrete coupled nonlinear schrödinger lattice20132013840436910.1155/2013/404369MR3108948LengJ.Periodic and subharmonic solutions for 2nth-order ϕc-Laplacian difference equations containing both advance and retardation201627490291310.1016/j.indag.2016.05.002MR3542944LeszczyńskiM.Fourth-order discrete anisotropic boundary-value problems20152015238114LinG.ZhouZ.Homoclinic solutions of discrete ϕ-Laplacian equations with mixed nonlinearities20181751723174710.3934/cpaa.2018082MR3809126LiuX.ZhouT.ShiH.Existence of homoclinic orbits for a higher order difference system20171041842185310.22436/jnsa.010.04.44MR3639748LongY.ZengB.Multiple and sign-changing solutions for discrete Robin boundary value problem with parameter dependence2017151549155710.1515/math-2017-0129MR3748192Zbl1382.390042-s2.0-85041088871YangL.Boundary value problems of a higher order nonlinear difference equation201779493102MR3756508ZhouZ.MaD.Multiplicity results of breathers for the discrete nonlinear Schrödinger equations with unbounded potentials201558478179010.1007/s11425-014-4883-2MR3319311Zbl1328.39011ZhouZ.YuJ.ChenY.Homoclinic solutions in periodic difference equations with saturable nonlinearity201154183932-s2.0-7865148313510.1007/s11425-010-4101-9Zbl1239.39010MawhinJ.WillemM.198974New York, NY. USASpringer10.1007/978-1-4757-2061-7MR982267Zbl0676.58017RabinowitzP. H.198665Providence, RI, USAAmerican Mathematical Society10.1090/cbms/065MR845785Zbl0609.58002StruweM.2008344thBerlin, GermanySpringerModern Surveys in MathematicsMR2431434Zbl1284.49004