84 M. Franca0. Furthermore assume that we are in the Hypotheses of Proposition 1 B, then we alsohave that lim d→D R 1 (d) = ∞. Therefore the Dirichlet problem in the ball of radiusR > 0 for equation (3) admits at least one solution for any R > 0.THEOREM 21. Assume that Hyp. F1, F2 and F4 are satisfied, then (3) admitsuncountably many S.G.S.REMARK 13. Assume that f is as in (23), that the functions k i (r) are uniformlypositive and bounded for any r ≥ 0 and p < q M < p ∗ . Then hypotheses F1 and F2are satisfied. Moreover if −k i (r) and k j (r) are decreasing for any r > 0, 1 ≤ i ≤ Nand N < j ≤ M, hypothesis F3 is satisfied; finally if q M > p ∗ F4 holds.The proof of Theorem 20 can be found in [17] and follows, with some minorchanges, the scheme introduced in [25] and then used in [19], [22] and [16].When Hyp. F1 and F2 are satisfied the initial value problem (3), with u(0) =d > 0, u ′ (0) = 0 admits at least a solution. Moreover such a solution, denoted byu(d, r) is unique for any d ≥ A and u ′ (r) ≤ 0 for r small. All these solutions can becontinued in J(d) = (0, R d ) = {r > 0|u ′ (r) < 0 < u(r)}, where R d can also beinfinite. This was proved for the spatial independent problem in [19] and then adaptedwith some trivial changes to the spatial dependent problem in [17]. Since u(d, r) ispositive and decreasing for r < R d , the limit lim r→Rd u(d, r) exists and is nonnegative,so we can define the following set:I := {d ≥ A | limr→R du ′ (d, r) < 0}Using an energy analysis we can prove that A ̸∈ I , when F1, F2 and F3 are satisfied.Moreover using a continuity argument on the auxiliary system (5), we prove that I isopen in [A,∞) whenever F1 and F2 hold, see [17]. The difficult part of the proof is toshow that I ̸= ∅. In fact we show that, if F1 and F2 are satisfied there is D ̸∈ I suchthat (D,∞) ⊂ I .For this purpose we introduce a dynamical system of the form (5), using (4)with l = q, where q is the parameter defined in [12]. Then we show that for r smalland u large we can approximate our system with an autonomous subcritical system oftype (8). Through a careful analysis of the phase portrait we are able to construct abarrier set E τ ⊂ {(x, y) ∈ R 2 | y ≤ 0 ≤ x}. Then, using Wazewski’s principle, weshow that there are M > 0 and δ > 0 such that, for any τ < −M, there is an unstableset ˜W u (τ) ⊂ E τ , which intersects the y negative semi-axis in a compact connectedset, say ζ(τ). It follows that the trajectories x τ q (Qu (τ), t) ∈ E τ for any t < 0 and thatlim t→−∞ x τ q (Qu (τ), t) = O. So they correspond to regular solutions u(d(τ), r), thatare positive and decreasing for r ≤ exp(τ) and they become null with nonzero slope atr = exp(τ), so they are crossing solutions.It follows that if F1, F2 and F3 hold there is D > A, such that (D,∞) ∈ I , butD ̸∈ I . Then we show that u(D, r) is a monotone decreasing G.S. using a continuityargument on (5), and Theorem 20 and Corollary 5 follow.To prove the existence of S.G.S. we have to consider system (7) and to construct
Radial solutions for p-Laplace equation 85a stable set ˜W s (τ) through Proposition 2. We choose τ < −M, so that ˜W u (τ) crossesthe y negative semi-axis, in view of Corollary 5. We denote by B(τ) the bounded setenclosed by ˜W u (τ) and the y negative semi-axis. We choose one of the uncountablymany points in ˜W s (τ) ∩ B(τ), say Q s , and we follow backwards in t the trajectoriesx τ q (Qs , t) where Q s ∈ ˜W s (τ). We show that x τ q (Qs , t) is forced to stay in B(t + τ)for any t < 0 and we conclude that x τ q (Qs , t) is uniformly positive as t → −∞.Then, from Proposition 3 we deduce that the corresponding solutions v(r) of (3) aremonotone decreasing S.G.S.7. Appendix: reduction of div(g(|x|)∇u|∇u| p−2 ) + f (u,|x|) = 0 and natural dimensionIn this subsection we want to show how we can pass from the analysis of radial solutionsof an equation of the following class(24) div(g(|x|)∇u|∇u| p−2 ) + ¯ f (u,|x|) = 0to the analysis of solutions of an equation of the form (3). Here again x ∈ R n andg(|x|) ≥ 0 for |x| ≥ 0.We repeat the argument developed in Appendix B of [17]. In fact we exploithere an idea already used in [33] and [20], and we follow quite closely the latter paper,in which the concept of natural dimension is introduced. First of all observe that aradial solutions u(r) of (24) satisfy the following ODE:(25) (r n−1 g(r)u ′ |u ′ | p−2 ) ′ + r n−1 ¯ f (u, r) = 0.Set a(r) = r n−1 g(r) and assume that one of the Hypotheses below is satisfiedH1 a −1/(p−1) ∈ L 1 [1,∞]\ L 1 [0, 1]H2 a −1/(p−1) ∈ L 1 [0, 1]\L 1 [1,∞)Then we make the following change of variables borrowed from [20]. Let N > p be aconstant and assume that Hyp. H1 is satisfied; we define s(r) =(∫ ∞ra(τ) −1/(p−1) dτ ) −p+1N−p. Obviously s : R + 0→ R+ 0, s(0) = 0, s(∞) = ∞ ands(r) is a diffeomorphism of R + 0into itself with inverse r = r(s) for s ≥ 0. If u(r) is asolution of (25), v(s) = u(r(s)) is a solution of the following transformed equation(26) (s N−1 v s |v s | p−2 ) s + s N−1 h(s) f (v, s) = 0,where f (v, s) = ¯ f (v, r(s)) andh(s) =( ) ( )N − pp p/(p−1)g(r(s)) 1/p r(s) n−1p − 1s N−1 .If we replace Hyp. H1 by Hyp. H2 we can define s(r) as follows s(r) =(∫ r0 a(τ)−1/(p−1) dτ ) p−1N−pand obtain again (26) from (25), with the same expression for