JAM Journal of Applied Mathematics 1687-0042 1110-757X Hindawi Publishing Corporation 589483 10.1155/2013/589483 589483 Research Article Exact Number of Positive Solutions for a Class of Two-Point Boundary Value Problems Niu Yanmin Yan Baoqiang Liu Yansheng 1 School of Mathematical Sciences Shandong Normal University Jinan 250014 China sdnu.edu.cn 2013 24 11 2013 2013 25 09 2013 04 11 2013 2013 Copyright © 2013 Yanmin Niu and Baoqiang Yan. 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.

This paper considers the existence of positive solutions of the boundary value problems v+λ(vr+vp-vq)=0 and v(-1)=v(1)=0, where p>r>q>-1 and λ is a positive parameter. Using a time-map approach, we obtain the exact number of positive solutions in different cases.

1. Introduction and Main Results

The study of multiplicity results to boundary value problem (1)v′′+λf(v)=0,t(a,b),v(a)=v(b)=0, where λ>0 is a positive parameter, is very interesting because of its applications. As we know, when f(v)=vp(x)+vq(x), the boundary value problem (2)-v′′(x)=μvp(x)+vq(x),axb,v(x)>0,x(a,b),v(a)=v(b)=0, where 0<q<1<p and k0 are fixed given numbers and μ>0 is a parameter, comes from the elliptic problem (3)-Δu=λuq+up,xΩ,u>0,xΩ,u=0,xΩ, with 0<q<1<p, which was raised by Ambrosetti et al. in .

Under different assumptions on f, there are many results for the above problems and elliptic equations (see ).

In [9, 10], Liu considered the case of f(v)=vp+(1/λ)vq and f(v)=vp+vq+kv and gave the exact number of solutions and many interesting properties of the solutions.

Cheng  investigated the following two-point boundary value problem: (4)-y′′=λ(yp-yq),t(-1,1),y(-1)=y(1)=0, where λ>0 is a positive parameter and p>q>-1 and got the exact number of positive solutions.

Now, in this paper we consider the more general case (5)-v′′=λ(vr+vp-vq),t(-1,1),v(-1)=v(1)=0, where λ>0 is a positive parameter, p>r>q>-1, and (1/(r+1))+(1/(p+1))-(1/(q+1))<0.

Define β, where β satisfies (6)βrr+1+βpp+1-βqq+1=0.

For p>r>q and -1<q<1, let λ1 be given by(7)λ1=r+12β1-rvv×(01dttq+1[(1-tr-q)+((r+1)/(p+1))βp-r(1-tp-q)])2.The main results of this paper are as follows.

Theorem 1.

If p>r>q1, (5) has exactly one positive solution for any λ>0.

Theorem 2.

If p>r1>q>-1, (5) has exactly one positive solution for λ(0,λ1] and none for λ>λ1.

Theorem 3.

If p>1>r>q0, (5) has exactly one positive solution for λ(0,λ1] and none for λ>λ1.

Theorem 4.

If (1/3)p>r>q0, (5) has exactly one positive solution for λ(λ1,+) and none for λ<λ1.

Theorem 5.

Assume that 1>p>r>0>q>-1. Define (8)θ(r,p,q)=011-r[tq+1(1-m1βr-qtr-q-m2βp-qtp-q)]1/2dt-(r-q)011-tq+1[tq+1(1-m1βr-qtr-q-m2βp-qtp-q)]3/2dt-(p-r)(q+1)p+1βp-q×011-tp+1[tq+1(1-m1βr-qtr-q-m2βp-qtp-q)]3/2dt, where m1=(q+1)/(r+1) and m2=(q+1)/(p+1). One has the following.

If  θ(r,p,q)0, (5) has exactly one positive solution for λ(λ1,+) and none for λ(0,λ1).

If θ(r,p,q)<0, there exists λ0(0,λ1) such that (5) has exactly two positive solutions for λ(λ0,λ1], exactly one for λ(λ1,+) or λ=λ0, and none for λ(0,λ0).

2. The Proofs of Theorems <xref ref-type="statement" rid="thm1.1">1</xref>–<xref ref-type="statement" rid="thm1.4">4</xref>

We assume throughout this section that p>r>q>-1 and (9)1r+1+1p+1-1q+1<0. Denote that E=(β,+), where β is given by (6). For aβ and t(0,1), let (10)P(a,t)=ata(zr+zp)dz,Q(a,t)=atazqdz. Define a function F:E(0,+) as (11)F(a)=0a[2va(zr+zp-zq)dz]-1/2dvfor  aE. It is clear that 0β(zr+zp-zq)dz=0. Now we claim that ar+ap-aq>0 for aβ and 0a[2va(zr+zp-zq)dz]-1/2dv< if and only if aE.

Let g1(z)=zp and g2(z)=zr-zq; then f(z)=g1(z)+g2(z). Immediately, we get that g2(x)=rzr-1-qzq-1=zq-1(rzr-q-q). In order to judge the sign of g2(x), we just judge the sign of T(z)=rzr-q-q. Since T(0)=-q, T(1)=r-q, and T(z)=r(r-q)zr-q-1, we can get the following two results.

For r>q>0, the function g2(z) has a stable point z0=(q/r)1/(r-q). When z(0,z0), we have g2(z)<0, and when z(z0,+), we have g2(z)>0.

For r>0>q, we have g2(z)>0 on z(0,+). Combining g1(x)=pzp-1 and g1′′(x)=p(p-1)zp-2 with the above two results, we obtain the monotony of f(z).

For p1>r>q>0, the function f(z) has a stable point z* on (0,1). When z(0,z*), we have f(z)<0 and when z(0,z*), we have f(z)>0.

For 1>p>r>q>0, we have that f(z)>0 on (0,+).

For p>r>0>q or p>0>r>q, we have that f(z)>0 on (0,+).

Then from 0β(zr+zp-zq)dz=0, we infer that ar+ap-aq>0 for aβ.

We consider the integration 0a[2va(zr+zp-zq)dz]-1/2dv. It is clear that v=a is a flaw point. Since 0a[2va(zr+zp-zq)dz]-1/2dv<0a[2va(zp-zq)dz]-1/2dv on E, we consider the integration 0a[2va(zp-zq)dz]-1/2dv.

Using Lagrange theorem, we obtain that (12)0a[2va(zp-zq)dz]-1/2dv=0a[2(ap+1p+1-vp+1p+1)-2(aq+1q+1-vq+1q+1)]-1/2dv=0a{2[v+θ1(a-v)]p(a-v)-2[v+θ2(a-v)]q(a-v)}-1/2dv=0a[2(a-v)]-1/2{[v+θ1(a-v)]p-[v+θ2(a-v)]q}-1/2dv<, where θ1,θ2(0,1) are constants.

The following Lemma 6 is listed to show that to study the number of positive solutions of (5) is equivalent to study the shape of the map F(a) on E. Lemmas 79 show some properties of F(a) on E.

Lemma 6.

Let u(a,t) be the unique solution of the problem (13)0u(t)a,  u(t)a[2va(zr+zp-zq)dz]-1/2dv=F(a)t,t[0,1], where aE. One has the following.

If λ>0 and v is a positive solution of (5), v(0)E, F(v(0))=λ, and v(t)=u(v(0),|t|) for t[-1,1].

If aE and F(a)=λ, v(t)=u(v(0),|t|) is a positive solution of (5) with v(0)=a.

Proof.

(1) Assume that λ>0 and v is a positive solution of (5). Let τ[-1,1] satisfy v(τ)=maxt[-1,1]v(t)=a. It follows from v(τ)=0 that (14)[v(s)]2=τs2v(η)v(η)dη=2λv(s)a(zr+zp-zq)dz,s(-1,1). This implies that aβ and v(s)0 if v(s)<a. And combine v(τ)=-λ(ar+ap-aq)<0 to obtain (15)v(s)>0for  s(-1,τ),v(s)<0for  s(τ,1). Then, we have that (16)v(s)[2v(s)a(zr+zp-zq)dz]-1/2=λ,s(-1,τ),-v(s)[2v(s)a(zr+zp-zq)dz]-1/2=λ,s(τ,1). It follows that (17)v(t)a[2va(zr+zp-zq)dz]-1/2dv=(τ-t)λ,t[-1,τ],(18)v(t)a[2va(zr+zp-zq)dz]-1/2dv=(t-τ)λ,t[τ,1],(19)(1-τ)λ=0a[2va(zr+zp-zq)dz]-1/2dv=(τ+1)λ. From (19) and aβ we have that τ=0. With (17) and (18) we obtain the result (1) of this theorem.

(2) Since u(a,0)=a>0 and u(a,t) is a positive solution of the boundary value problem (20)u+[F(a)]2(ur+up-uq)=0,t(0,1),u(0)=0,u(1)=0, we have that v(t)=u(a,|t|) is a positive solution of (5).

Lemma 7.

F is differentiable on (β,), and (21)F(a)=14201H0(a,t)dt,a>β,(22)F(a)=14201H1(a,t)dt,a>β, where (23)H0(a,t)=4a[P(a,t)-Q(a,t)]-1/2,H1(a,t)=2[P(a,t)-Q(a,t)]-3/2×[1-rr+1ar+1(1-tr+1)+1-pp+1ap+1(1-tp+1)-1-qq+1aq+1(1-tq+1)].

Proof.

Equation (21) can be obtained by (11), immediately. From (24)P(a,t)a=a×[1-rr+1ar+1(1-tr+1)+1-pp+1ap+1(1-tp+1)]=ar(1-tr+1)+ap(1-tp+1),Q(a,t)a=a[1q+1aq+1(1-tq+1)]=aq(1-tq+1), we have that (25)H0(a,t)a=2[P(a,t)-Q(a,t)]-3/2×[1-rr+1ar+1(1-tr+1)+1-pp+1ap+1(1-tp+1)-1-qq+1aq+1(1-tq+1)]=H1(a,t). It follows from (21) that F is differentiable on (β,) and (22) is true.

Lemma 8.

Consider the following:(26)limaβF(a)={+,q1,λ1,q<1, where λ1 is given by (7).

Proof.

From (27)limaββadv2va(zr+zp-zq)dz=0 and the Lebesgue theorem, we have that (28)limaβF(a)=limaβ0βdv2va(zr+zp-zq)dz=0βdv2vβ(zr+zp-zq)dz. On the other hand, from 0β(zr+zp-zq)dz=0, we can obtain that (29)0βdv2vβ(zr+zp-zq)dz=0βdv20v(zq-zr-zp)dz=01(βdt)×(2[(βt)q+1q+1-(βt)r+1r+1-(βt)p+1p+1])-1=01(βdt)×(2[βr+1r+1tq+1+βp+1p+1tq+100000000000-(βt)r+1r+1-(βt)p+1p+1])-1/2=r+12β(1-r)/2×01dttq+1[(1-tr-q)+((r+1)/(p+1))βp-r(1-tp-q)]. This completes the proof of Lemma 8.

Lemma 9.

Consider the following:(30)lima+F(a)={0,p>1,r1orp>1,r<1,π22,p=1,r=1,+,p<1,r<1.

Proof.

From (31)lima+Q(a,t)P(a,t)=lima+aq+1(1-tq+1)/(q+1)ar+1(1-tr+1)/(r+1)+ap+1(1-tp+1)/(p+1)=lima+((1-tq+1)(r+1)(p+1))000000000×((q+1)(p+1)ar-q(1-tr+1)0000000000000+(q+1)(r+1)ap-q(1-tp+1))-1=0, we have that(32)lima+aP(a,t)-Q(a,t)=lima+aP(a,t)=lima+1(1/(r+1))ar-1(1-tr+1)+(1/(p+1))ap-1(1-tp+1).From Lemma 7 and the Lebesgue theorem, we can obtain the results of this lemma.

In the following section, we give the proofs of Theorems 15. For convenience, we denote that (33)I(a,t)=1-rr+1ar+1(1-tr+1)+1-pp+1ap+1(1-tp+1)-1-qq+1aq+1(1-tq+1). Hence from Lemma 7, we have that (34)F(a)=142012(P-Q)-3/2I(a,t)dt,for  a>β.

Proof of Theorem <xref ref-type="statement" rid="thm1.1">1</xref>.

From p>r>q1, we obtain that (35)I(a,t)<1-rr+1ar+1(1-tr+1)+1-pp+1ap+1(1-tp+1)-1-qq+1ap+1(1-tp+1)=1-rr+1ar+1(1-tr+1)+2(q-p)(p+1)(q+1)ap+1×(1-tp+1)<0, for a>β, t(0,1). Thus F(a)<0 for a>β. By Lemmas 6, 8, and 9, we have the results of this theorem.

Proof of Theorem <xref ref-type="statement" rid="thm1.2">2</xref>.

From p>r1>q>-1, (33), and (34), we have that F(a)<0 for a>β. It follows from Lemmas 6, 8, and 9 that the results of Theorem 2 hold.

Proof of Theorem <xref ref-type="statement" rid="thm1.3">3</xref>.

Conditions p>1>r>q0 and (1/(r+1))+(1/(p+1))-(1/(q+1))<0 imply that rp-1>q(r+p+2)>0. With (33), we obtain that (36)I(a,t)<1-rr+1ap+1(1-tp+1)+1-pp+1ap+1(1-tp+1)-1-qq+1aq+1(1-tq+1)<2(1-rp)(p+1)(r+1)ap+1(1-tp+1)-1-qq+1aq+1×(1-tq+1)<0, for a>β, t(0,1), which means that F(a)<0 for a>β. Thus by Lemmas 6, 8, and 9, we have the results of this theorem.

Proof of Theorem <xref ref-type="statement" rid="thm1.4">4</xref>.

From (1/3)p>r>q0 and (33), we have that (37)I(a,t)>1-rr+1aq+1(1-tq+1)+1-pp+1aq+1(1-tq+1)-1-qq+1aq+1(1-tq+1)>(21-pp+1-1-qq+1)aq+1(1-tq+1)=q(3-p)+(1-3p)(p+1)(q+1)aq+1(1-tq+1)>0, for a>β, t(0,1). Hence F(a)>0 for a>β. By Lemmas 6, 8, and 9, we have the results of this theorem.

3. The Proof of Theorem <xref ref-type="statement" rid="thm1.5">5</xref>

In this section we always assume that 1>p>r>0>q>-1. Denote that (38)S(a,t)=P(a,t)Q(a,t),a>β,t(0,1),(39)h1(s)=(s-1)(s-n),s(-,),(40)h2(s)=-3(s-m)(s-n)+2(s-1)(s-l),s(-,), where P(a,t), Q(a,t), and β are given by (10), (6), and (41)m=m(t)=q+1r+1-(p-r)(q+1)(p+1)(r+1)ap-q1-tp+11-tq+1,a>β,t(0,1),n=n(t)=1-q1-r+(p-r)(q+1)(p+1)(1-r)ap-q1-tp+11-tq+1,a>β,t(0,1),l=l(t)=1-q21-r2+(p2-r2)(q+1)(p+1)(1-r2)ap-q1-tp+11-tq+1,a>β,t(0,1).

Remark 10.

From 1>p>r>0q>-1, (6), and (41), it is obvious that m<1, n>1, and l<n.

Lemma 11.

For a>β and t(0,1), (42)q+1r+1ar-q+q+1p+1ap-q<P(a,t)Q(a,t)<ar-q+ap-q.

Proof.

Let (43)K(t)=arr+1[(r-q)tr+q+1+(q+1)tq-(r+1)tr]+app+1[(p-q)tp+q+1+(q+1)tq-(p+1)tp]. Condition 1>p>r>0>q>-1 implies that (44)K(t)=arr+1[tr-1(r-q)(r+q+1)tr+q+(q+1)qtq-1-(r+1)rtr-1]+app+1[tp-1(p-q)(p+q+1)tp+q+(q+1)qtq-1-(p+1)ptp-1]<arr+1[(r-q)(r+q+1)+(q+1)q-(r+1)r]tr+q+app+1[(p-q)(p+q+1)+(q+1)q-(p+1)p]tp+q=0,t(0,1). With K(1)=0 and (44), we have that K(t)>0 for t(0,1).

It follows that (45)tP(a,t)Q(a,t)=q+1(1-tq+1)2a-qK(t)>0,t(0,1). Combining (46)limt0P(a,t)Q(a,t)=q+1r+1ar-q+q+1p+1ap-q,limt1P(a,t)Q(a,t)=ar-q+ap-q, we have the results of this lemma.

Lemma 12.

For 1>p>r>0>q>-1, (47)limaβF(a)=122(q+1βq+1)1/2θ(r,p,q).

Proof.

From Lemma 7 we have that (48)22F(a)=01((1-rr+1ar+1(1-tr+1)+1-pp+1ap+1(1-tp+1)0000000000000-1-qq+1aq+1(1-tq+1))000000000×([P(a,t)-Q(a,t)]3/2)-11-pp+1)dt=011-r[P(a,t)-Q(a,t)]1/2dt-r-qq+1aq+1011-tq+1[P(a,t)-Q(a,t)]3/2dt-p-rp+1ap+1011-tp+1[P(a,t)-Q(a,t)]3/2dt. It follows from Lebesgue theorem that (49)limaβ011-r[P(a,t)-Q(a,t)]1/2dt=011-r[P(β,t)-Q(β,t)]1/2dt,limaβr-qq+1aq+1011-tq+1[P(a,t)-Q(a,t)]3/2dt+limaβp-rp+1ap+1011-tp+1[P(a,t)-Q(a,t)]3/2dt=r-qq+1βq+1011-tq+1[P(β,t)-Q(β,t)]3/2dt+p-rp+1βp+1011-tp+1[P(β,t)-Q(β,t)]3/2dt. Finally, 0β(zr+zp-zq)dz=0 implies that (50)P(β,t)-Q(β,t)=0βt(zq-zr-zp)dz=(βt)q+1q+1(1-m1βr-qtr-q-m2βp-qtp-q). Combing (49) and (50), we complete the proof of this lemma.

Lemma 13.

F ( a ) has continuous derivatives up to second order on (β,) as follows: (51)F(a)=1-r2201G(a,t)h1(S(a,t))dt,a>β,(52)F′′(a)=1-r24a201G(a,t)h2(S(a,t))dt,a>β, where G(a,t)=[S(a,t)-1]-5/2[Q(a,t)]-1/2.

Proof.

Equation (51) can be obtained by (53)H1(a,t)=2[P(a,t)-Q(a,t)]-3/2×[p-rp+1(1-r)P(a,t)-(1-q)Q(a,t)-p-rp+1ap+1(1-tp+1)]=2[P(a,t)Q(a,t)-1]-5/2[Q(a,t)]-1/2×(1-r)[P(a,t)Q(a,t)-1-q1-r-(p-r)(q+1)(p+1)(1-r)×ap-q1-tp+11-tq+1][P(a,t)Q(a,t)-1]=2(1-r)G(a,t)h1(S(a,t)) and Lemma 7, immediately. From (24) we have that (54)aH1(a,t)a=-3(P-Q)-5/2×[p-rp+1(r+1)P-(q+1)Q+p-rp+1ap+1(1-tp+1)]×[(1-r)P-(1-q)Q-p-rp+1ap+1×(1-tp+1)p-rp+1]+2(P-Q)-3/2×[p2-r2p+1(1-r2)P-(1-q2)Q-p2-r2p+1ap+1(1-tp+1)]=(PQ-1)-3/2(1-r2)×{-3[P-q+1r+1Q+(p-r)(p+1)(r+1)×ap+1(1-tp+1)q+1r+1]×[P-1-q1-rQ-(p-r)(p+1)(1-r)ap+1×(1-tp+1)1-q1-r]+2(P-Q)×[P-1-q21-r2Q-(p2-r2)(p+1)(1-r2)×ap+1(1-tp+1)1-q21-r2]}=(1-r2)G(a,t)h2(S(a,t)), and so (52) holds.

Lemma 14.

There exist δ(r,p,q)>0 and η(r,p,q)>0 such that (55)F′′(a)+δ(r,p,q)F(a)1+r2a>η(r,p,q)(ar-q+ap-q)-5/2a-(3+q)/2for  a>β.

Proof.

From (41) and 1>p>r>0>q>-1, we have that (56)mins[1,n]h2(s)=min{h2(1),h2(n)}=(n-1)min{3(1-m),2(n-l)}2(n-1)min{(1-m),(n-l)}>2(n-1)min{r-qr+1,(1-q)(r-q)1-r2}=2(n-1)r-qr+1. And since h1((1+n)/2)=-((n-1)2/4)<0, let δ(r,p,q)>0 satisfy (57)mins[1,n][h2(s)+δ(r,p,q)h1(s)]=2(n-1)r-qr+1. Set (58)h(s)=h2(s)+δ(r,p,q)h1(s),s(-,). From h(s)=(δ-1)s2+(3m+3n-2l-2-δ-δn)s-3mn+2l+δn, we have that δ(r,p,q)-1>0 and (59)h(s)mins[1,n]h(s)=2(n-1)r-qr+1,s(-,). On the other hand, from (10), (38), and Lemma 11, we have that (60)0<Q(a,t)<1q+1aq+1,0<q+1aq+10a(zr+zp-zq)dz=q+1r+1ar-q+q+1p+1ap-q-1<S(a,t)-1<S(a,t)<ar-q+ap-q, for a>β, t(0,1).

This means that (61)G(a,t)=[S(a,t)-1]-5/2[Q(a,t)]-1/2>(ar-q+ap-q)-5/2a-(1+q)/21+q, for a>β, t(0,1). It follows from Lemma 13 and (58) that (62)F′′(a)+δ(r,p,q)F(a)1+r2a=1-r24a201G(a,t)h(S(a,t))dt. Now, from (59)–(62) we have (55), where (63)η(r,p,q)=(1-r)(r-q)1+q2201[n(t)-1]dt.

Lemma 15.

If θ(r,p,q)0, then F(a)>0 for a>β. If θ(r,p,q)<0, then there exists a*>β, such that F(a)<0 for a(β,a*) and F(a)>0 for a>a*.

Proof.

It follows from Lemma 14 that if (64)F(a)=0,then  F′′(a)>0, and then (65)F(a)  has at most one zero in  (β,). By (40), (42), and (51) we can obtain that (66)F(a)>0for  P(a,t)Q(a,t)>n(t). With (10) and (41), (66) implies that (67)F(a)>0for  q+1r+1ar+(q+1p+1-p-r1-r)ap>1-q1-raq. If θ(r,p,q)0, then from Lemma 12 and (64)–(67) we have the results of this lemma, immediately. If θ(r,p,q)=0, then by Lemmas 12 and 14 we have that F′′(a)>0 for a near β, and so F(a)>0 for a near β. Thus, it follows from (64)–(67) that F(a)>0 for a>β.

Proof of Theorem <xref ref-type="statement" rid="thm1.5">5</xref>.

From Lemmas 6, 8, 9, and 15, we can obtain the results of Theorem 5.

Ambrosetti A. Brezis H. Cerami G. Combined effects of concave and convex nonlinearities in some elliptic problems Journal of Functional Analysis 1994 122 2 519 543 10.1006/jfan.1994.1078 MR1276168 ZBL0805.35028 Korman P. Li Y. Ouyang T. Exact multiplicity results for boundary value problems with nonlinearities generalising cubic Proceedings of the Royal Society of Edinburgh A 1996 126 3 599 616 10.1017/S0308210500022927 MR1396280 ZBL0855.34022 Korman P. Ouyang T. Exact multiplicity results for two classes of boundary value problems Differential and Integral Equations 1993 6 6 1507 1517 MR1235208 ZBL0780.34013 Korman P. Ouyang T. Multiplicity results for two classes of boundary-value problems SIAM Journal on Mathematical Analysis 1995 26 1 180 189 10.1137/S0036141092231872 MR1311887 ZBL0824.34028 Laetsch T. The number of solutions of a nonlinear two point boundary value problem Indiana University Mathematics Journal 1970 20 1 13 MR0269922 10.1512/iumj.1971.20.20001 Smoller J. Wasserman A. Global bifurcation of steady-state solutions Journal of Differential Equations 1981 39 2 269 290 10.1016/0022-0396(81)90077-2 MR607786 ZBL0425.34028 Wang S.-H. Long D.-M. An exact multiplicity theorem involving concave-convex nonlinearities and its application to stationary solutions of a singular diffusion problem Nonlinear Analysis. Theory, Methods & Applications A 2001 44 469 486 10.1016/S0362-546X(99)00272-2 MR1822096 ZBL0992.34014 Zhang L. Q. Uniqueness of positive solutions of Δu+u+up=0 in a ball Communications in Partial Differential Equations 1992 17 7-8 1141 1164 10.1080/03605309208820880 MR1179281 ZBL0782.35025 Liu Z. Exact number of solutions of a class of two-point boundary value problems involving concave and convex nonlinearities Nonlinear Analysis. Theory, Methods & Applications A 2001 46 181 197 10.1016/S0362-546X(99)00454-X MR1849789 ZBL1003.34020 Liu Z. Zhang X. A class of two-point boundary value problems Journal of Mathematical Analysis and Applications 2001 254 2 599 617 10.1006/jmaa.2000.7258 MR1805527 ZBL1003.34020 Cheng J. Exact number of positive solutions for a class of semipositone problems Journal of Mathematical Analysis and Applications 2003 280 2 197 208 10.1016/S0022-247X(02)00539-5 MR1976624 ZBL1028.34027