Applied Mathematics E-Notes, 11(2011), 197–207 c Available free at mirror sites of http://www.math.nthu.edu.tw/ amen/ ISSN 1607-2510 Positive Solutions For Singular Boundary Value Problems Involving p-Laplacian Operators Changxiu Songy, Xuejun Gaoz Received 6 September 2010 Abstract The existence of positive solutions for singular boundary value problems involving p-Laplacian operators are investigated. By applying the …xed point theorem of cone expansion and compression of norm type, su¢ cient conditions are established for the existence of positive solutions. 1 Introduction In this paper, we study the following singular boundary value problem (BVP) involving p-Laplacian operators 8 ( (x0 ))0 + a(t)f (xt ; y t ) = 0; 0 < t < 1; > > < p 0 0 ( p (y )) + b(t)g(xt ; y t ) = 0; 0 < t < 1; (1) x(t) = '(t); 1 t 1 + ; x0 (0) = 0; > > : 0 y(t) = (t); 1 t 1 + ; y (0) = 0; where xt = x(t + ); 2 [0; ]; 0 < 1; p ( ) is the p-Laplacian operator; '; : [1; 1 + ] ! [0; +1) are continuous, and '(1) = (1) = 0: For p-Laplacian equations, many results have been obtained, for example see papers [1-4]. But most of them are concerned with ordinary di¤erential equations. Recently, the study of BVP of functional di¤erential equations [5-6] is of signi…cance since they arise and have applications in variational problems of control theory and in other areas of applied mathematics. In this paper, by constructing an integral equation which is equivalent to BVP (1), we study the existence of positive solutions of nonlinear singular BVP of the form (1). Let C = C([0; ]; R) be a Banach space with a norm jj!jjC = sup0 j!( )j and C + = f! 2 C : !( ) 0; 2 [0; ]g: We assume the following: Mathematics Subject Classi…cations: 34B15 of Applied Mathematics, Guangdong University of Technology, Guangzhou 510006, P. R. China z School of Applied Mathematics, Guangdong University of Technology, Guangzhou 510006, P. R. China y School 197 198 Positive Solutions for Singular Boundary Problems (H1 ) f; g : C + C + ! (0; +1) are continuous; (H2 ) a; b : (0; 1) ! [0; +1) are continuous, and 0< Z Z a(t) E 0< Z 0 1 1 a(t)dt < +1; 0 < 0 Z q( Z b(t) E s a(r)dr)ds < +1; 0 < 0 Z 1 0 In this paper, we may choose a Z Z 1 b(t)dt < +1; 0 Z q( s b(r)dr)ds < +1: 0 2 (0; minf 14 ; 1 4 g) by (H2 ) such that 1 a(t)dt > 0 and Z 1 b(t)dt > 0: De…ne C = f! 2 C + : 0 < jj!jjC !( ); 2 [0; ]g and E = ft 2 [0; 1] : 0 t+ 1; 0 g = [0; 1 ]: Note that for t 2 [ ; 1 ] E; we have xt0 = y0t = 0: DEFINITION 1. A function (x; y) 2 C 1 [0; 1] C 1 [0; 1] is called a positive solution of BVP (1) if it satis…es the following: 1. (x; y) satis…es BVP(1); 2. x(t) > 0; y(t) > 0; t 2 (0; 1); and 3. ( p (x0 ); p (y 0 )) is absolutely continuous on [0; 1]: Suppose (x(t); y(t)) is a solution of BVP (1). Then 8 R1 Rs r r t 1; > q [ 0 a(r)f (x ; y )dr]ds; 0 t > < x(t) = '(t); 1 t 1+ ; R1 Rs (2) r r > t 1; > q [ 0 b(r)g(x ; y )dr]ds; 0 t : y(t) = (t); 1 t 1+ : Suppose that (x0 (t); y0 (t)) is the solution of 8 0; 0 > > < x0 (t) = '(t); 1 0; 0 > > : y0 (t) = (t); 1 BVP (1) with f t t t t 0; g 0: Then 1; 1+ ; 1; 1+ : (3) If (x(t); y(t)) is the solution of BVP (1) and (u(t); v(t)) = (x(t) x0 (t); y(t) y0 (t)), noting that (u(t); v(t)) = (x(t); y(t)) for 0 t 1; we have from (2) that 8 R1 Rs r r r r t 1; > q [ 0 a(r)f (u + x0 ; v + y0 )dr]ds; 0 t > u(t) = < 0; 1 t 1+ ; R1 Rs (4) r r r r > ; v + y )dr]ds; 0 t 1; [ b(r)g(u + x > q 0 0 t 0 : v(t) = 0; 1 t 1+ : Let K be a cone in Banach space X = C[0; 1 + ] K = f(u; v) 2 X : u(t) 0; v(t) 0; u(t) + v(t) C[0; 1 + ] de…ned by g(t)jj(u; v)jj; t 2 [0; 1]g; C. X. Song and X. J. Gao 199 where jj(u; v)jj = jjujj + jjvjj; jjujj = sup ju(t)j; jjvjj = sup jv(t)j; and t2[0;b] t2[0;b] 1 t; 0 0; 1 g(t) = t t 1; 1+ : De…ne A(u; v)(t) = R1 Rs q[ 0 a(r)f (ur + xr0 ; v r + y0r )dr]ds; 0 1 t t 1; 1+ ; R1 Rs q[ 0 b(r)g(ur + xr0 ; v r + y0r )dr]ds; 0 1 t t 1; 1+ ; t 0; B(u; v)(t) = t 0; and (u; v)(t) = (A(u; v)(t); B(u; v)(t)); 0 t 1+ : (5) Under assumptions (H1 ) and (H2 ), BVP (1) has a solution if and only if has a …xed point (u; v), that is, (u; v) = (u; v): The following lemma will play an important role in the proof of our results and can be found in the book [7]. LEMMA 1. Assume that X is a Banach space and K X is T a cone in X; 1 , 2 are open subsets of X, and 0 2 1 : Furthermore, let : K ( 2 n 1 ) ! K be 2 a completely continuous operator satisfying one of the following conditions: (i) jj (x)jj (ii) jj (x)jj jjxjj; 8 x 2 K T @ 1 ; jj (x)jj T jjxjj; 8 x 2 K @ 2 ; jj (x)jj Then there is a …xed point of LEMMA 2. The map in K T ( 2 n jjxjj; 8 x 2 K T @ 2; T jjxjj; 8 x 2 K @ 1 : 1 ). : X ! X in (5) is completely continuous and The proof of Lemma 2 can be found in [5-6]. 2 Main Results In the sequel, we let f0 := f0 := ! 1 ;! 2 2C f1 := f1 := lim (jj! 1 jjC +jj! 2 jjC )!0 f (! 1 ; ! 2 ) (jj! 1 jjC + jj! 2 jjC )p ; f (! 1 ; ! 2 ) ;(jj! 1 jjC +jj! 2 jjC )!0 (jj! 1 jjC + jj! 2 jjC )p lim f (! 1 ; ! 2 ) (jj! 1 jjC +jj! 2 jjC )!1 (jj! 1 jjC + jj! 2 jjC )p ! 1 ;! 2 2C 1 lim 1 ; ; f (! 1 ; ! 2 ) ;(jj! 1 jjC +jj! 2 jjC )!1 (jj! 1 jjC + jj! 2 jjC )p lim 1 1 ; (K) K: 200 Positive Solutions for Singular Boundary Problems g0 := g0 := lim (jj! 1 jjC +jj! 2 jjC )!0 1 g(! 1 ; ! 2 ) (jj! 1 jjC +jj! 2 jjC )!1 (jj! 1 jjC + jj! 2 jjC )p lim and ! 1 ;! 2 2C ; f (! 1 ; ! 2 ) (jj! 1 jjC + jj! 2 jjC )p lim ! 1 ;! 2 2C ;(jj! 1 jjC +jj! 2 jjC )!0 g1 := g1 := f (! 1 ; ! 2 ) (jj! 1 jjC + jj! 2 jjC )p 1 1 ; ; g(! 1 ; ! 2 ) ;(jj! 1 jjC +jj! 2 jjC )!1 (jj! 1 jjC + jj! 2 jjC )p lim 1 : THEOREM 1. Assume (H1 ) and (H2 ) hold. Then BVP (1) has at least one positive solution if one of the following conditions is satis…ed: (H3 ) f0 = 0; f1 = +1; g0 = 0; '(t) = (t) 0; t 2 [1; 1 + ]; or (H4 ) f0 = +1; f1 = 0; g1 = 0: PROOF. Suppose that (H3 ) is satis…ed. By (t) 0; '(t) 0; t 2 [1; 1 + ]; we know xt0 = y0t = 0; t 2 [0; 1 + ]: Since f0 = 0; for " > 0 (we choose " satisfying R1 Rs " 0 q [ 0 a(r)dr]ds 12 ), there is a 1 > 0 such that f (! 1 ; ! 2 ) ("(jj! 1 jjC + jj! 2 jjC ))p 1 ;0 jj! 1 jjC + jj! 2 jjC 1: De…ne 1 For (u; v) 2 @ 1 = f(u; v) 2 X : jj(u; v)jj < \ K; we deduce that jjur jjC + jjv r jjC jjA(u; v)jj = Z 1 0 Z Z [ q 1 for r 2 [0; 1] and thus s a(r)f (ur ; v r )dr]ds 0 1 0 Z [ q s 0 a(r)("(jjur jj + jjv r jj))p "(jjujj + jjvjj) Z 1 (jjujj + jjvjj): 2 1 2 (jjujj Similarly, we have B(u; v) 1 g: 0 1 Z [ q 1 dr]ds s a(r)dr]ds 0 + jjvjj): This implies jj (u; v)jj = jjA(u; v)jj + jjB(u; v)jj jj(u; v)jj; (u; v) 2 @ 1 \ K: On the other hand, since f1 = +1; for M > 0 we can choose M satisfying R1 Rs a(r)dr]ds 1, there exists a 2 > 1 such that M q[ f (! 1 ; ! 2 ) (M (jj! 1 jjC + jj! 2 jjC ))p 1 ; ! 1 ; ! 2 2 C ; jj! 1 jjC + jj! 2 jjC De…ne 2 = f(u; v) 2 X : jj(u; v)jj < 2 g: 2: C. X. Song and X. J. Gao For (u; v) 2 @ 2 201 \ K; we deduce jjur jjC jjujj g(t)jjujj u(t); t 2 [ ; 1 ]; r 2 [0; 1]; jjv r jjC jjvjj g(t)jjvjj v(t); t 2 [ ; 1 ]; r 2 [0; 1]; which implies that ur ; v r 2 C for r 2 [ ; 1 jjur jjC jjujj = Thus, for (u; v) 2 @ 2 jjA(u; v)jj = ] and jjv r jjC 2; jjvjj = 2; r 2 [ ;1 ]: \ K; we have Z Z q[ 1 0 Z s a(r)f (ur ; v r )dr]ds 0 1 Z [ q s a(r)(M (jjur jjC + jjv r jjC ))p M (jjujj + jjvjj) jjujj + jjvjj = jj(u; v)jj: Z Z [ q 1 1 dr]ds s a(r)dr]ds That is, jj (u; v)jj jj(u; v)jj; (u; v) 2 @ 2 \ K: According to the …rst part of Lemma 1, it follows that has a …xed point(u; v) 2 T K ( 2 n 1 ): Now, suppose that (H4 ) is satis…ed. Since f0 = +1; for M > 0 (choose M satisfying R1 Rs M a(r)dr]ds 1), there exists a 1 > 0 such that q[ (M (jj! 1 jjC + jj! 2 jjC ))p f (! 1 ; ! 2 ) 1 ; ! 1 ; ! 2 2 C jj! 1 jjC + jj! 2 jjC 1: De…ne 1 For (u; v) 2 @ 1 = f(u; v) 2 X : jj(u; v)jj < 1 g: \ K; we deduce jjur jjC jjujj g(t)jjujj u(t); t 2 [ ; 1 ]; r 2 [0; 1]; jjv r jjC jjvjj g(t)jjvjj v(t); t 2 [ ; 1 ]; r 2 [0; 1]; which implies that ur ; v r 2 C for r 2 [ ; 1 jjur jjC jjujj = 1; jjv r jjC ] and jjvjj = 1; r 2 [ ;1 ]: (6) 202 Positive Solutions for Singular Boundary Problems ]; we have xr0 = y0r = 0: Thus, for (u; v) 2 @ For r 2 [ :1 jjA(u; v)jj = Z Z q[ 1 0 Z a(r)f (ur ; v r )dr]ds Z [ q s a(r)(M (jjur jjC + jjv r jjC ))p M (jjujj + jjvjj) jjujj + jjvjj = jj(u; v)jj; which implies jj (u; v)jj Z Z [ q 1 jj(u; v)jj; 8(u; v) 2 @ 1 ("(jj! 1 jjC + jj! 2 jjC ))p Choose a positive constant 2 > 1+maxff q 1 2 (! 1 ; ! 2 ) : 0 1 1 dr]ds s a(r)dr]ds \ K: On the other hand, since f1 = 0, for 8 " > 0; 9N > f (! 1 ; ! 2 ) \ K; we have s 0 1 2 1 such that ; jj! 1 jjC + jj! 2 jjC > N: such that jj! 1 jjC +jj! 2 jjC Z 1 N +jju0 jj+jjv0 jjg q [ (a(r)+b(r))dr]: 0 De…ne 2 = f(u; v) 2 X : jj(u; v)jj < For (u; v) 2 @ 2 \ K; we have from the facts: x0 (t) 0; t 2 [0; 1 + ]; that for r 2 [0; 1]; jjur + xr0 jjC + jjv r + y0r jjC 2 g: 0; u(t) 0; y0 (t) jjur jjC + jjv r jjC > N; for jjur jjC + jjv r jjC > N; and jjur + xr0 jjC + jjv r + y0r jjC for jjur jjC + jjv r jjC jjur jjC + jjxr0 jjC + jjv r jjC + jjy0r jjC N + jjx0 jj + jjy0 jj; N: Let := maxff q 1 0; v(t) (! 1 ; ! 2 ) : 0 jj! 1 jjC + jj! 2 jjC N + jjx0 jj + jjy0 jjg: C. X. Song and X. J. Gao 203 R1 q[ 0 Thus, for " satisfying 0 < "(1 + jjx0 jj + jjy0 jj) jjA(u; v)jj = Z 1 s a(r)f (ur + xr0 ; v r + y0r )dr]ds 0 0 Z Z [ q a(r)dr] < 21 ; we have 1 0 Z [ q 1 a(r)f (ur + xr0 ; v r + y0r )dr]ds 0 Z = a(r)f (ur + xr0 ; v r + y0r )dr q[ jjur jjC +jjv r jjC >N Z + a(r)f (ur + xr0 ; v r + y0r )dr] 0 jjur jjC +jjv r jjC N maxf"(jjur + xr0 jjC + jjv r + y0r jjC ); g Z maxf"(jju + x0 jj + jjv + y0 jj); g q [ 1 1 maxf (jjujj + jjvjj) + ; 2 2 1 < (jjujj + jjvjj) 2 1 = : 2 2 Similarly, we have B(u; v) 1 2 (jjujj Z q[ Z [ q 1 a(r)dr] 0 1 a(r)dr] 0 1 a(r)dr]g 0 + jjvjj): This implies jj (u; v)jj = jjA(u; v)jj + jjB(u; v)jj jj(u; v)jj; (u; v) 2 @ 2 \ K: According to the second part of Lemma 1, it follows that has a …xed point(u; v) 2 T K ( 2 n 1 ): T Suppose that (u(t); v(t)) is the …xed point of in K ( 2 n 1 ); then (x(t); y(t)) = (u(t) + x0 (t); v(t) + y0 (t)) is a positive solution of BVP (1). This completes the proof. Similarly, we have the next theorem. THEOREM 2. Assume (H1 ) and (H2 ) hold. Then BVP (1) has at least a positive solution if one of the following conditions is satis…ed: (H03 ) f0 = 0; g1 = +1; g0 = 0; '(t) = (t) 0; t 2 [1; 1 + ]; or (H04 ) f1 = 0; g1 = 0; g0 = +1: In what follows, we shall consider the existence of multiple positive solutions for BVP (1). THEOREM 3. Assume (H1 ); (H2 ) hold and the following conditions are satis…ed: (H5 ) f0 = +1; f1 = +1; (H6 ) 9 a p1 > 0 such that for 8 0 jj! 1 jjC + jj! 2 jjC p1 + p0 ; one has and g(! 1 ; ! 2 ) (l1 p1 )p R1 Rs max f (t); '(t)g; l1 = f2 0 q [ 0 a(r)dr]dsg 1 : f (! 1 ; ! 2 ) where p0 = 1 t 1+ (l1 p1 )p 1 1 ; 204 Positive Solutions for Singular Boundary Problems Then BVP (1) has at least two positive solutions. PROOF. By (H5 ), there exists a 1 : 0 < 1 < p1 such that (M (jj! 1 jjC + jj! 2 jjC ))p 1 ; jj! 1 jjC + jj! 2 jjC R1 R1 where M satis…es M 1 a(r)dr]ds 1: De…ne q[ f (! 1 ; ! 2 ) = f(u; v) 2 X : jj(u; v)jj < 1 For (u; v) 2 @ 1 !1 ; !2 2 C ; 1 g: \ K; similar to (6) one has ur ; v r 2 C and jjur jjC + jjv r jjC 1 1; (jjujj + jjvjj) = 1; r 2 [ ;1 ]: Hence, we obtain an analogous inequality: jj (u; v)jj = jjA(u; v)jj + jjB(u; v)jj Similarly, there exists a 3 1 \ K: > p1 such that (M (jj! 1 jjC + jj! 2 jjC ))p f (! 1 ; ! 2 ) jj(u; v)jj; (u; v) 2 @ 1 ; jj! 1 jjC + jj! 2 jjC 3; !1 ; !2 2 C ; M is chosen as above. De…ne 3 For (u; v) 2 @ 3 = f(u; v) 2 X : jj(u; v)jj < 3 g: \ K; one has ur ; v r 2 C and jjur jjC + jjv r jjC (jjujj + jjvjj) = 3; r 2 [ ;1 ]: Furthermore, we have jj (u; v)jj = jjA(u; v)jj + jjB(u; v)jj By (H6 ), let one has 2 = p1 , de…ne 2 jjA(u; v)jj = = f(u; v) 2 X : jj(u; v)jj < Z 1 0 l 1 p1 Z q[ Z 0 = = = Similarly, we have B(u; v) jj(u; v)jj; (u; v) 2 @ 0 1 2 g: 3 \ K: For (u; v) 2 @ s a(r)f (ur + xr0 ; v r + y0r )dr]ds Z [ q s a(r)dr]ds 0 1 p1 2 1 2 2 1 jj(u; v)jj: 2 1 2 (jjujj + jjvjj): This implies jj (u; v)jj = jjA(u; v)jj + jjB(u; v)jj jj(u; v)jj; (u; v) 2 @ 2 \ K: 2 \ K; C. X. Song and X. J. Gao 205 According to Lemma 1, it follows that has two …xed points, that is to say, BVP (1) has at least two positive solutions. This completes the proof. From above, the following theorems are obvious. THEOREM 4. Assume (H1 ) and (H2 ) hold and the following conditions are satis…ed: (H7 ) f0 = 0; f1 = 0; g0 = 0; g1 = 0; (t) = '(t) 0: (H8 ) 9 a p2 > 0 such that for 8 p2 jj! 1 jjC + jj! 2 jjC p2 ; one has f (! 1 ; ! 2 ) (l2 p2 )p 1 ; nR o 1 Rs 1 where l2 = 0 : q [ 0 a(r)dr]dt Then BVP (1) has at least two positive solutions. THEOREM 5. Assume (H1 ) and (H2 ) hold and the following conditions are satis…ed: (H05 ) g0 = +1; g1 = +1; (H6 ) 9 a p1 > 0 such that for 8 0 jj! 1 jjC + jj! 2 jjC p1 + p0 ; one has and g(! 1 ; ! 2 ) (l1 p1 )p n R o Rs 1 max f (t); '(t)g and l1 = 2 0 q [ 0 a(r)dr]ds f (! 1 ; ! 2 ) where p0 = (l1 p1 )p 1 1 ; 1 1 t 1+ : Then BVP (1) has at least two positive solutions. THEOREM 6. Assume (H1 ) and (H2 ) hold and the following conditions are satis…ed: (H7 ) f0 = 0; f1 = 0; g0 = 0; g1 = 0; (t) = '(t) 0: (H09 ) 9 a p2 > 0 such that for 8 p2 jj! 1 jjC + jj! 2 jjC p2 ; one has g(! 1 ; ! 2 ) (l2 p2 )p nR o 1 Rs 1 where l2 = 0 : q [ 0 b(r)dr]dt Then BVP (1) has at least two positive solutions. 3 1 ; Examples We have several examples. EXAMPLE 1. Consider BVP 8 1 1 ( (x0 ))0 4 (t + 31 ) + y 3 (t + 31 ) = 0; 0 < t < 1; > > < p 0 01 1 ( p (y )) 3 (t + 31 ) + y 4 (t + 13 ) = 0; 0 < t < 1; > x(t) = 0; 1 t 1 + 31 ; x0 (0) = 0; > : y(t) = 0; 1 t 1 + 31 ; y 0 (0) = 0: (7) Here, a(t) = b(t) = 1; xt = x(t + ) x(t + 31 ); y t = y(t + ) y(t + 31 ); = 13 ; p = 98 ; and 1 1 1 1 1 1 1 1 f (! 1 ; ! 2 ) = ! 14 ( ) + ! 23 ( ); g(! 1 ; ! 2 ) = ! 13 ( ) + ! 24 ( ): 3 3 3 3 206 Positive Solutions for Singular Boundary Problems As jj! 1 jjC + jj! 2 jjC ! 0, we have 1 f (! 1 ; ! 1 ) (jj! 1 jjC + jj! 2 jjC )p 1 1 ! 14 ( 31 ) + ! 23 ( 13 ) = 1 (jj! 1 jjC + jj! 2 jjC ) 8 1 1 jj! 1 jjC4 + jj! 2 jjC3 1 (jj! 1 jjC + jj! 2 jjC ) 8 1 (jj! 1 jjC + jj! 2 jjC ) 8 ! 0 and 1 g(! 1 ; ! 1 ) (jj! 1 jjC + jj! 2 jjC )p 1 1 ! 13 ( 31 ) + ! 24 ( 13 ) = 1 (jj! 1 jjC + jj! 2 jjC ) 8 1 1 jj! 1 jjC3 + jj! 2 jjC4 1 (jj! 1 jjC + jj! 2 jjC ) 8 1 (jj! 1 jjC + jj! 2 jjC ) 8 ! 0; that is to say f0 = 0 and g0 = 0 hold. On the other hand, suppose ! 1 ; ! 2 2 C : Then ! 1 ( ) As ! 1 ; ! 2 2 C ; jj! 1 jjC + jj! 2 jjC ! 1; we get 1 f (! 1 ; ! 1 ) (jj! 1 jjC + jj! 2 jjC )p 1 = jj! 1 jjC ; ! 2 ( ) jj! 2 jjC : 1 ! 14 ( 13 ) + ! 23 ( 13 ) 1 (jj! 1 jjC + jj! 2 jjC ) 8 1 1 ( jj! 1 jjC ) 4 + ( jj! 2 jjC ) 3 1 1 3 (jj! 1 jjC + jj! 2 jjC ) 8 1 (jj! 1 jjC + jj! 2 jjC ) 8 ! +1; which means that f0 = +1 holds. According to Theorem 1, it follows that BVP (7) has at least one positive solution. Acknowledgment. Supported by grants (no. 10871052) and (no. 109010600) from NNSF of China, and by grant (no.10151009001000032) from NSF of Guangdong. References [1] D. Q. Jiang and W. J. Gao, Upper and lower solution method and a singular boundary value problems for the one-dimensional singular p-Laplacian, J. Math. Anal. Appl., 252(2)(2000), 631–648. C. X. Song and X. J. Gao 207 [2] F. H. Wong, Existence of positive solutions for m-Laplacian BVPs, Appl. Math. Letters, 12(1999), 11–17. [3] R. Manasevich and F. Zanolin, Time-mapping and multiplicitiy of solutions for the one-dimensional p-Laplacian, Nonlinear Analysis, 21(4)(1993), 269–291. [4] X. M. He and W. G. Ge, Existence of positive solutions for the one-dimensional p-Laplacian equations, Acta Mathematica Sinica 46(4)(2003), 805–810. [5] P. X. Weng and D. Q. Jing, Existence of positive solutions for boundary value problem of second-order FDE, Computers and Mathematics with Applications, 37(1999), 1–9. [6] C. Song and P. Weng, Existence of positive solutions for p-Laplacian singular boundary value problems of functional di¤erential equation, Appl. Math. J. Chinese Univ. Ser. B, 20(1)(2005), 51–58. [7] M. A. Krasnosel’skii, Positive Solutions of Operator Equations, Noordhoof, Groningen, 1964.