Abstract
We study necessary conditions and sufficient conditions for the existence of local-in-time solutions of the Cauchy problem for superlinear fractional parabolic equations. Our conditions are sharp and clarify the relationship between the solvability of the Cauchy problem and the strength of the singularities of the initial measure.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
We consider the Cauchy problem for a superlinear fractional parabolic equation
where \(\mu \) is a nonnegative Radon measure in \({{ \mathbb {R}}}^N\). Throughout the paper, we assume that \(N\ge 1\), \(0<\theta \le 2\), and \(F:[0,\infty )\rightarrow [0,\infty )\) is (at least) continuous.
In general, the existence of local-in-time nonnegative solutions of problem (P) depends crucially on the delicate interplay between the strength of the singularities of the initial measure \(\mu \) and the behavior of \(F(\tau )\) as \(\tau \rightarrow \infty \). In this paper, for a large class of nonlinearities F, we obtain new necessary conditions and new sufficient conditions for the local solvability of problem (P). The prototypical example we have in mind is
As a consequence of our more general results, we are then able to derive sharp results for classes of nonlinearities which include these prototypes as special cases, and quantify this interplay more precisely via ‘optimal singularities.’
Throughout this paper, we use the following notations. For \(T>0\), we set \(Q_T:={{\mathbb {R}}}^N\times (0,T)\) and let \(B(x,\sigma )\) denote the Euclidean ball in \({{ \mathbb {R}}}^N\) center x, radius \(\sigma \). We use for the average value of f over B with respect to the Lebesgue measure dx. The set of nonnegative Lebesgue measurable functions in \({\mathbb R}^N\) is denoted by \({{\mathcal {L}}}_0\), while \({{\mathcal {M}}}\) denotes the set of nonnegative Radon measures in \({\mathbb R}^N\). For \(\mu \in {\mathcal L}_0\), we abuse terminology somewhat by speaking of ‘measure \(\mu \)’ defined via \({\text {d}}\mu =\mu (x){\text {d}}x\).
1.1 Background
The solvability of the Cauchy problem for superlinear parabolic equations has been studied in many papers since the pioneering work by Fujita [14]. The literature is now very extensive, and we refer to the comprehensive monograph [35]. We also mention the following works, some of which are directly related to this paper, others with a different emphasis (higher-order equations, systems, nonlinear boundary conditions): superlinear parabolic equations [2, 6, 7, 14, 29,30,31, 33, 36, 38,39,40,41]; linear heat equation with nonlinear boundary conditions [10, 15, 20, 27, 28]; superlinear parabolic equations with a potential [1, 3, 9, 22, 23, 39]; superlinear parabolic systems [11,12,13, 26, 34]; superlinear fractional parabolic equations [18, 19, 21, 32, 37]; superlinear higher-order parabolic equations [8, 16, 17, 24, 25].
In [19], the second and third authors of this paper considered problem (P) in the special case of the power law nonlinearity \(F(u)=u^p\) with \(p>1\):
There, as here, the exponent \(p_\theta :=1+\theta /N\) plays a critical role. They proved the following necessary conditions for the local existence (cases (i) and (ii)).
-
(i)
Let \(\mu \in {{\mathcal {M}}}\). If problem (1.1) possesses a nonnegative solution in \(Q_T\) for some \(T>0\), then there exists \(C_1=C_1(N,\theta ,p)>0\) such that
$$\begin{aligned} \sup _{x\in {{ \mathbb {R}}}^N}\mu (B(x,\sigma ))\le C_1\sigma ^{N-\frac{\theta }{p-1}}, \quad 0<\sigma \le T^{\frac{1}{\theta }}. \end{aligned}$$(1.2)In the case where \(1<p<p_\theta \), the function \((0,\infty )\ni \sigma \mapsto \sigma ^{N-\theta /(p-1)}\) is decreasing so that relation (1.2) is equivalent to
$$\begin{aligned} \sup _{x\in {{ \mathbb {R}}}^N}\mu \big (B\big (x,T^{\frac{1}{\theta }}\big )\big )\le C_1 T^{\frac{N}{\theta }-\frac{1}{p-1}}. \end{aligned}$$In the case where \(p=p_\theta \), there exists \(C_2=C_2(N,\theta )>0\) such that
$$\begin{aligned} \sup _{x\in {{ \mathbb {R}}}^N}\mu (B(x,\sigma )) \le C_2\left[ \log \biggr (e+\frac{T^{\frac{1}{\theta }}}{\sigma }\biggr )\right] ^{-\frac{N}{\theta }}, \quad 0<\sigma \le T^{\frac{1}{\theta }}. \end{aligned}$$(See [2] for the Laplacian case \(\theta =2\).)
Condition (i) implies the following nonexistence result.
-
(ii)
Let \(p\ge p_\theta \). There exists \(\gamma =\gamma (N,\theta ,p)>0\) such that if \(\mu \in {{\mathcal {L}}}_0\) satisfies
$$\begin{aligned} \begin{array}{ll} \mu (x)\ge \gamma |x|^{-N}\displaystyle {\biggr [\log \left( e+\frac{1}{|x|}\right) \biggr ]^{-\frac{N}{\theta }-1}} &{}\quad \hbox {if}\quad \displaystyle {p=p_\theta },\\ \mu (x)\ge \gamma |x|^{-\frac{\theta }{p-1}} &{}\quad \hbox {if}\quad \displaystyle {p>p_\theta }, \end{array} \end{aligned}$$for almost all (a.a.) x in a neighborhood of the origin, then problem (1.1) possesses no local-in-time nonnegative solutions.
Regarding sufficiency, in [19] they obtained results (iii) and (iv) below.
-
(iii)
Let \(\mu \in {{\mathcal {M}}}\) and \(1<p<p_\theta \). There exists \(c=c(N,\theta ,p)>0\) such that if
$$\begin{aligned} \sup _{x\in {{ \mathbb {R}}}^N}\mu \big (B\big (x,T^{\frac{1}{\theta }}\big )\big )\le c T^{\frac{N}{\theta }-\frac{1}{p-1}} \end{aligned}$$for some \(T>0\), then problem (1.1) possesses a nonnegative solution in \(Q_T\).
-
(iv)
Let \(\mu \in {{\mathcal {L}}}_0\) and \(p\ge p_\theta \). There exists \(\varepsilon =\varepsilon (N,\theta ,p) >0\) such that if
$$\begin{aligned} \begin{array}{ll} 0\le \mu (x)\le \varepsilon |x|^{-N}\displaystyle {\biggr [\log \left( e+\frac{1}{|x|}\right) \biggr ]^{-\frac{N}{\theta }-1}}+K &{}\quad \hbox {if}\quad \displaystyle {p=p_\theta },\\ 0\le \mu (x)\le \varepsilon |x|^{-\frac{\theta }{p-1}}+K &{}\quad \hbox {if}\quad \displaystyle {p>p_\theta }, \end{array} \end{aligned}$$for a.a. \(x\in {{\mathbb {R}}}^N\) for some \(K>0\), then problem (1.1) possesses a local-in-time nonnegative solution.
For \(\mu \in {{\mathcal {L}}}_0\), the results in (ii) and (iv) demonstrate that the ‘strength’ of the singularity at the origin of the functions
is the critical threshold for the local solvability of problem (1.1). We term such a singularity in the initial data an optimal singularity for the solvability for problem (1.1). Of course, by translation invariance the singularity could be located at any point of \({{ \mathbb {R}}}^N\).
Subsequently, the results of [19] were extended to some related parabolic problems with a power law nonlinearity (see [20,21,22,23,24,25]). However, one cannot apply the arguments in these papers to problem (P) with a general nonlinearity F since they depend heavily upon the homogeneous structure of the power law nonlinearity.
1.2 The main result
In this paper, we improve the arguments in [19] to obtain necessary conditions and sufficient conditions for the existence of local-in-time solutions of problem (P) for a significantly larger class of nonlinearities F and determine the optimal singularities of the initial data for the solvability of problem (P).
Let \(f_1\) and \(f_2\) be real-valued functions defined in an interval \([L,\infty )\), where \(L\in {{ \mathbb {R}}}\). We write \(f_1(t)\preceq f_2(t)\) as \(t\rightarrow \infty \) if there exists \(C>0\) such that \(f_1(t)\le Cf_2(t)\) for all large enough \(t\in [L,\infty )\). We define \(\succeq \) in the obvious way, namely \(f_2(t)\succeq f_1(t)\) as \(t\rightarrow \infty \) if and only if \(f_1(t)\preceq f_2(t)\) as \(t\rightarrow \infty \). We write \(f_1(t)\asymp f_2(t)\) as \(t\rightarrow \infty \) whenever \(f_1(t)\preceq f_2(t)\) and \(f_1(t)\succeq f_2(t)\) as \(t\rightarrow \infty \), i.e., there exists \(C>0\) such that \(C^{-1}f_2(t)\le f_1(t)\le Cf_2(t)\) for large enough \(t\in [L,\infty )\).
We consider nonlinearities which are asymptotic to the prototypical example (F), in this sense:
-
(F1)
F is locally Lipschitz continuous in \([0,\infty )\);
-
(F2)
\(F(\tau )\asymp \tau ^p[\log \tau ]^q\) as \(\tau \rightarrow \infty \) for some \(p>1\) and \(q\in {{ \mathbb {R}}}\).
Theorem 1.1
Assume conditions (F1) and (F2).
-
(i)
Let \(\mu \in {{\mathcal {M}}}\) and either
$$\begin{aligned} {\mathrm{(i)}}\quad 1<p<p_\theta \qquad \hbox {or}\qquad {\mathrm{(ii)}}\quad p=p_\theta \hbox { and } q<-1 . \end{aligned}$$Problem (P) possesses a local-in-time solution if and only if \({\sup _{z\in {{ \mathbb {R}}}^N}\mu (B(z,1))<\infty }\).
-
(ii)
Suppose \(\mu \in {{\mathcal {L}}}_0\).
-
(1)
Let \(p=p_\theta \) and \(q=-1\). There exists \(\gamma _1>0\) such that if
$$\begin{aligned} \mu (x)\ge \gamma _1|x|^{-N}|\log |x||^{-1}[\log |\log |x||]^{-\frac{N}{\theta }-1} \end{aligned}$$(1.3)in a neighborhood of \(x=0\), then problem (P) possesses no local-in-time solutions. On the other hand, for any \(R\in (0,1)\), there exists \(\varepsilon _1>0\) such that if
$$\begin{aligned} 0\le \mu (x)\le \varepsilon _1|x|^{-N}|\log |x||^{-1}[\log |\log |x||]^{-\frac{N}{\theta }-1} \chi _{B(0,R)}(x)+K_1, \quad x\in {{\mathbb {R}}}^N,\nonumber \\ \end{aligned}$$(1.4)for some \(K_1>0\), then problem (P) possesses a local-in-time solution.
-
(2)
Let \(p=p_\theta \) and \(q>-1\). There exists \(\gamma _2>0\) such that if
$$\begin{aligned} \mu (x)\ge \gamma _2|x|^{-N}|\log |x||^{-\frac{N(q+1)}{\theta }-1} \end{aligned}$$(1.5)in a neighborhood of \(x=0\), then problem (P) possesses no local-in-time solutions. On the other hand, for any \(R\in (0,1)\), there exists \(\varepsilon _2>0\) such that if
$$\begin{aligned} 0\le \mu (x)\le \varepsilon _2|x|^{-N}|\log |x||^{-\frac{N(q+1)}{\theta }-1}\chi _{B(0,R)}(x)+K_2, \quad x\in {{ \mathbb {R}}}^N, \end{aligned}$$(1.6)for some \(K_2>0\), then problem (P) possesses a local-in-time solution.
-
(3)
Let \(p>p_\theta \). There exists \(\gamma _3>0\) such that if
$$\begin{aligned} \mu (x)\ge \gamma _3|x|^{-\frac{\theta }{p-1}}|\log |x||^{-\frac{q}{p-1}} \end{aligned}$$(1.7)in a neighborhood of \(x=0\), then problem (P) possesses no local-in-time solutions. On the other hand, for any \(R\in (0,1)\), there exists \(\varepsilon _3>0\) such that if
$$\begin{aligned} 0\le \mu (x)\le \varepsilon _3|x|^{-\frac{\theta }{p-1}}|\log |x||^{-\frac{q}{p-1}}\chi _{B(0,R)}(x)+K_3, \quad x\in {{ \mathbb {R}}}^N, \end{aligned}$$(1.8)for some \(K_3>0\), then problem (P) possesses a local-in-time solution.
-
(1)
While Theorem 1.1 provides sharp results on the identification of optimal singularities for the solvability of problem (P), we point out that we have obtained several other interesting and powerful results in this paper regarding necessary conditions and sufficient conditions for existence under very general conditions on F. We mention, in particular, Theorems 3.1, 4.1, 4.2, and 4.3.
Subject to mild assumptions on F (essentially that of majorizing a convex function with suitable monotonicity properties), we follow the strategy in [19] and obtain necessary conditions for the existence in Theorem 3.1. However, the iteration step in [19] to obtain the estimate for the optimal singularity relies on the homogeneity of the pure power law nonlinearity considered there. For the class of nonlinearities satisfying (F1)–(F2), we combine the arguments in [19] with the method introduced in [32], to obtain a sharper necessary condition in Corollary 3.1. Conversely, in order to derive sharp sufficient conditions we require delicate arguments for F satisfying (F1)–(F2). Indeed, the arguments are separated into three cases: (i) \(1<p<p_\theta \) (see Theorem 4.1), \(p>p_\theta \) (see Theorem 4.3), and (iii) \(p=p_\theta \) (see Theorem 4.2). The arguments in case (i) are somewhat standard but the other cases involve certain intricacies, in particular, for the critical case \(p=p_\theta \).
The rest of this paper is organized as follows. In Sect. 2, we recall some properties of the fundamental solution \(\Gamma _\theta \) and prove some preliminary lemmas. In Sect. 3, we obtain necessary conditions for the existence of local-in-time solutions of problem (P). In Sect. 4, we prove several theorems on sufficient conditions for the existence of local-in-time solutions of problem (P). In Sect. 4.4, we also provide a necessary and sufficient condition on the nonlinearity F for which problem (P) is solvable for the case of initial data a Dirac measure (Corollary 4.4). Finally, in Sect. 5 we complete the proof of our main result, Theorem 1.1, and outline some analogous results for nonlinearities which are asymptotic to further log-refinements of the cases above (see Remark 5.1).
2 Preliminaries
In this section, we prove some important technical lemmas, modifying the arguments in [19] for the more general nonlinearities considered here. We make precise our notion of solution used throughout this paper, which implicitly considers nonnegative functions only. The word ‘solvability’ for problem (P) is always used with respect to this solution concept. In all that follows, we will use C to denote generic positive constants which depend only on N, \(\theta \), and F and point out that C may take different values within a calculation. We begin by recalling some properties of the kernel for the fractional Laplacian.
Let \(\Gamma _\theta =\Gamma _\theta (x,t)\) be the fundamental solution of
The function \(\Gamma _\theta \) satisfies
for all \(x\in {{ \mathbb {R}}}^N\) and \(t>0\) and has the following properties:
for all \(x,y\in {{ \mathbb {R}}}^N\) and \(0<s<t\) (see for example [4, 5, 37]). Furthermore, we have the following smoothing estimate for the semigroup associated with \(\Gamma _\theta \) (see [19, Lemma 2.1]).
Lemma 2.1
For any \(\mu \in {{\mathcal {M}}}\), set
Then, there exists \(C=C(N,\theta )>0\) such that
Remark 2.1
(i) \(S(t)\mu \) is possibly infinite everywhere in \({ \mathbb {R}}^N\); (ii) if \(\mu \in {{\mathcal {M}}}\) is such that
for some \(r>0\), then for any \(R\ge r\) there exists \(C\ge 1\) such that
See for example [27, Lemma 2.1] or [12, Lemma 2.4].
We now make precise our solution concepts for problem (P).
Definition 2.1
Let \(T>0\) and u be a nonnegative, measurable, finite almost everywhere function in \(Q_T\). Let F be a nonnegative and continuous function in \([0,\infty )\).
-
(i)
We say that u satisfies
$$\begin{aligned} \partial _t u+(-\Delta )^{\frac{\theta }{2}}u=F(u) \end{aligned}$$(2.5)in \(Q_T\) if, for a.a. \(\tau \in (0,T)\), u satisfies
$$\begin{aligned} \qquad u(x,t)=\int _{{{ \mathbb {R}}}^N}\Gamma _\theta (x-y,t-\tau )u(y,\tau )\,{\textrm{d}}y+\int _\tau ^t\int _{{{ \mathbb {R}}}^N}\Gamma _\theta (x-y,t-s)F(u(y,s))\,{\textrm{d}}y\,{\textrm{d}}s \end{aligned}$$for a.a. \((x,t)\in {{ \mathbb {R}}}^N\times (\tau ,T)\).
-
(ii)
Let \(\mu \in {{\mathcal {M}}}\). We say that u is a solution of problem (P) in \(Q_T\) if u satisfies
$$\begin{aligned} u(x,t)=\int _{{{ \mathbb {R}}}^N}\Gamma _\theta (x-y,t)\,{\textrm{d}}\mu (y)+\int _0^t\int _{{{ \mathbb {R}}}^N}\Gamma _\theta (x-y,t-s)F(u(y,s))\,{\textrm{d}}y\,{\textrm{d}}s \nonumber \\ \end{aligned}$$(2.6)for a.a. \((x,t)\in Q_T\). If u satisfies (2.6) with ‘\(=\)’ replaced by ‘\(\ge \),’ then u is said to be a supersolution of problem (P) in \(Q_T\).
Next, we recall a lemma on the existence of solutions of problem (P) in the presence of a supersolution (see [19, Lemma 2.2]).
Lemma 2.2
Let F be an increasing, nonnegative continuous function in \([0,\infty )\). Let \(\mu \in {{\mathcal {M}}}\) and \(0<T\le \infty \). If there exists a supersolution v of problem (P) in \(Q_T\), then there exists a solution u of problem (P) in \(Q_T\) such that \(0\le u(x,t)\le v(x,t)\) in \(Q_T\).
Combining Lemma 2.2 and parabolic regularity theory, we have:
Lemma 2.3
Let \(\mu \in {{\mathcal {M}}}\) be such that \(\displaystyle {\sup _{z\in {{ \mathbb {R}}}^N}}\mu (B(z,1))<\infty \). Suppose
-
(i)
\(F_1\) is nonnegative and locally Lipschitz continuous in \([0,\infty )\);
-
(ii)
\(F_2\) is an increasing and continuous function in \([0,\infty )\) such that \(F_1(\tau )\le F_2(\tau )\) for all \(\tau \in [0,\infty )\).
If there exists a supersolution v of (P) in \(Q_T\) with F replaced by \(F_2\) such that for all \(\tau \in (0,T)\)
then there exists a solution u of (P) in \(Q_T\) with F replaced by \(F_1\), with u satisfying \(0\le u(x,t)\le v(x,t)\) in \(Q_T\).
Proof
For any m, \(n\in \mathbb {N}\) set
It follows from Lemma 2.1 that \(S(n^{-1})\mu \in L^\infty ({{ \mathbb {R}}}^N)\). Also, since \(\mu _n=S(n^{-1})S(n^{-1})\mu \), we have that \(\mu _n\in BC({{ \mathbb {R}}}^N)\). For each m, \(n\in \mathbb {N}\) define the sequence \(\{u_{m,n,k}\}_{k=0}^\infty \) by
By (2.4) and Definition 2.1 (ii), we have
for \(x\in {{ \mathbb {R}}}^N\) and \(t\in [0,T-2n^{-1})\). Since \(F_1(\tau )\le F_2(\tau )\) for \(\tau \in [0,\infty )\), by induction, we obtain
for all \(x\in {{ \mathbb {R}}}^N\), \(t\in [0,T-2n^{-1})\), and \(k\ge 0\). Here, we used the assumption that \(F_2\) is increasing. Since \(F_{1,m}\) is globally Lipschitz in \([0,\infty )\), we may apply the standard theory of evolution equations to see that the pointwise limit
exists in \({{ \mathbb {R}}}^N\times [0,\infty )\) and satisfies
for all \(x\in {{ \mathbb {R}}}^N\) and \(t>0\). Furthermore, by (2.8) we see that
for all \(x\in {{ \mathbb {R}}}^N\) and \(t\in [0,T-2n^{-1})\). Then, by (2.7), for any \(\tau \in (0,T-2n^{-1})\) we have
Applying the standard parabolic regularity theory to integral equation (2.9), we find \(\alpha \in (0,1)\) such that
for any compact set \(K\subset Q_T\). By the Ascoli–Arzelá theorem and the diagonal argument we obtain a subsequence \(\{u_{m,n'}\}\) of \(\{u_{m,n}\}\) and a function \(u_m\in C(Q_T)\) such that
Since \(F_{1,m}\) is bounded and continuous in \((0,\infty )\), by (2.9), we have
in \(Q_T\). Furthermore, by (2.10) and (2.12), we see that
for all \(x\in {{ \mathbb {R}}}^N\) and \(t\in [0,T)\).
Similarly to (2.11), using (2.14), instead of (2.10), we have
for any compact set \(K\subset Q_T\). By the Ascoli–Arzelá theorem and the diagonal argument we obtain a subsequence \(\{u_{m'}\}\) of \(\{u_m\}\) and a function \(u\in C(Q_T)\) such that
Since \(F_{1,m}(\tau )\le F_1(\tau )\le F_2(\tau )\) for \(\tau \in (0,\infty )\), by (2.14) we see that
for a.a. \((x,t)\in Q_T\). Then, by (2.13) and (2.15) we apply Lebesgue’s dominated convergence theorem to see that u is a solution of problem (P) in \(Q_T\) with F replaced by \(F_1\) and \(0\le u(x,t)\le v(x,t)\) in \(Q_T\). Thus, Lemma 2.3 follows. \(\square \)
Next, we provide two lemmas on the relationship between the initial measure and the initial trace for problem (P).
Lemma 2.4
Let F be a nonnegative continuous function in \([0,\infty )\).
-
(i)
Let u satisfy (2.5) in \( Q_T\) for some \(T>0\). Then,
$$\begin{aligned} \underset{0<t<T-\varepsilon }{{\textrm{ess}\; \textrm{sup}}}\,\int _{B(0,R)}u(y,t)\,{\textrm{d}}y<\infty \end{aligned}$$for all \(R>0\) and \(0<\varepsilon <T\). Furthermore, there exists a unique \(\nu \in {{\mathcal {M}}}\) as an initial trace of the solution u; that is,
$$\begin{aligned} \underset{t\rightarrow +0}{{\textrm{ess}\; \textrm{lim}}} \int _{{{ \mathbb {R}}}^N}u(y,t)\eta (y)\,{\textrm{d}}y=\int _{{{ \mathbb {R}}}^N}\eta (y)\,{\textrm{d}}\nu (y) \end{aligned}$$for all \(\eta \in C_0({{ \mathbb {R}}}^N)\).
-
(ii)
Let u be a solution of problem (P) in \(Q_T\) for some \(T>0\). Then, assertion (i) holds with \(\nu \) replaced by \(\mu \).
The proof of Lemma 2.4 is the same as in [19, Lemma 2.3]. Furthermore, by assertion (i) we can apply the same argument as in the proof of [19, Theorem 1.2] to obtain the following lemma.
Lemma 2.5
Let F be a nonnegative continuous function in \([0,\infty )\) and \(T>0\). Let u satisfy (2.5) in \(Q_T\). Let \(\mu \in {{\mathcal {M}}}\) be the unique initial trace of u guaranteed by Lemma 2.4. If \(\sup _{z\in {{ \mathbb {R}}}^N}\mu (B(z,1))<\infty \), then u is a solution of problem (P) in \(Q_T\).
In the rest of this section, we prepare preliminary lemmas.
Lemma 2.6
Let \(a>0\) and b, \(c\in {{\mathbb {R}}}\). Set
Then, there exists \(L\in (e,\infty )\) such that \(\varphi '>0\) in \((L,\infty )\) and the inverse function \(\varphi ^{-1}:(\varphi (L),\infty )\rightarrow (L,\infty )\) exists. Furthermore,
as \(\tau \rightarrow \infty \).
Proof
Since \(a>0\), we can find \(L\in (e,\infty )\) such that
for all \(\tau \in (L,\infty )\). Since \(\varphi (\tau )\rightarrow \infty \) as \(\tau \rightarrow \infty \), it follows that \(\varphi ^{-1}:(\varphi (L),\infty )\rightarrow (L,\infty )\) exists and satisfies \(\varphi ^{-1}(\tau )\rightarrow \infty \) as \(\tau \rightarrow \infty \). Now,
as \(\tau \rightarrow \infty \), so that
as \(\tau \rightarrow \infty \). Then, by (2.17) we have
as \(\tau \rightarrow \infty \). Hence,
as \(\tau \rightarrow \infty \), from which (2.16) follows and completes the proof of Lemma 2.6. \(\square \)
Lemma 2.7
Let \(a>0\), \(b\ge 0\), and \(c\in { \mathbb {R}}\).
-
(i)
There exists \(C_1>0\) such that
$$\begin{aligned} \int _A^B\tau ^{a-b-1}(\log \tau )^c\,{\textrm{d}}\tau \ge C_1 A^aB^{-b}(\log A)^c\log \frac{B}{A} \end{aligned}$$for all A, \(B\in [2,\infty )\) with \(A\le B\).
-
(ii)
There exists \(C_2\in [1,\infty )\) such that
$$\begin{aligned}{} & {} C_2^{-1}\tau ^a[\log (e+\tau )]^c\le \int _{\tau /2}^\tau s^{a-1}[\log (e+s)]^c\,{\textrm{d}}s \\{} & {} \le \int _0^\tau s^{a-1}[\log (e+s)]^c\,{\textrm{d}}s \le C_2\tau ^a[\log (e+\tau )]^c \end{aligned}$$
for all \(\tau \in [0,\infty )\).
Proof
We first prove assertion (i). Thanks to \(a>0\), by Lemma 2.6 we find \(R_1\in [2,\infty )\) such that the function \((R_1,\infty )\ni \tau \mapsto \tau ^a(\log \tau )^c\) is increasing. Then, we have
for all A, \(B\in [2,\infty )\) with \(A\le B\). We notice that
We observe that
for all A, \(B\in [2,\infty )\) with \(A\le B\). Then, assertion (i) follows.
Next, we prove assertion (ii). Let \(\varepsilon \in (0,a)\). By Lemma 2.6 we find \(R_2>0\) such that the function \((R_2,\infty )\ni \tau \mapsto (e+\tau )^\varepsilon (\log (e+\tau ))^c\) is increasing. Then, we have
for all \(\tau \in [R_2,\infty )\). On the other hand,
for all \(\tau \in (0,R_2)\). These imply that
for all \(\tau \in [0,\infty )\). On the other hand, since
we have
for \(\tau >0\). This yields
for all \(\tau \in [0,\infty )\). By (2.18) and (2.19), we have assertion (ii). The proof is complete. \(\square \)
Lemma 2.8
Let \(p>1\), \(d\in [1,p)\), \(q\in {{\mathbb {R}}}\), and \(R\ge 0\). Define a function f in \([0,\infty )\) by
Then,
-
(i)
the function \((0,\infty )\ni \tau \mapsto \tau ^{-d}f(\tau )\) is increasing;
-
(ii)
f is convex in \([0,\infty )\);
-
(iii)
\(f(\tau )\asymp \tau ^p(\log \tau )^q\) as \(\tau \rightarrow \infty \).
Proof
By the definition of f, we easily obtain property (i). Since
for \(\tau \in (R,\infty )\), we observe that \(f'\) is increasing in \([0,\infty )\), so that property (ii) holds.
We prove property (iii). Since \(d\in (1,p)\), by Lemma 2.7 (ii) we have
for all \(\tau > R\) and
for all \(\tau > 4R\). By (2.20) and (2.21), we obtain assertion (iii). Thus, Lemma 2.8 follows. \(\square \)
3 Necessary conditions for solvability
In this section, we establish necessary conditions for the solvability of problem (P). We begin in Theorem 3.1 by imposing only weak constraints on the nonlinearity F, before specializing to the case where F satisfies (F1) and (F2) in Corollary 3.1.
Theorem 3.1
Let F be a continuous function in \([0,\infty )\). Assume that there exists a convex function f in \([0,\infty )\) with the following properties:
-
(f1)
\(F(\tau )\ge f(\tau )\ge 0\) in \([0,\infty )\);
-
(f2)
the function \((0,\infty )\ni \tau \mapsto \tau ^{-d}f(\tau )\) is increasing for some \(d>1\).
Let u satisfy (2.5) in \(Q_T\) for some \(T>0\) and let \(\mu \) be the initial trace of u. Then, there exists \(\gamma =\gamma (N,\theta ,f)\ge 1\) such that
for all \(z\in {{ \mathbb {R}}}^N\) and \(\sigma \in (0,T^{\frac{1}{\theta }})\), where \(m_\sigma (z):=\mu (B(z,\sigma ))\).
Proof
It follows from Definition 2.1 (i) and property (f1) that, for a.a. \(\tau \in (0,T)\),
for a.a. \(x\in {{ \mathbb {R}}}^N\) and a.a. \(t\in (\tau ,T)\). This implies that
for a.a. \(x\in {{ \mathbb {R}}}^N\) and a.a. \(t\in (0,T/2)\).
Let \(0<\rho <(T/2)^{\frac{1}{\theta }}\). It follows from Definition 2.1 (i), property (f1), and (2.4) that
for all \(z\in {{ \mathbb {R}}}^N\) and a.a. \(t\in (0,T)\). On the other hand, by (2.1) we have
for all \(z\in {{ \mathbb {R}}}^N\) and \(t\ge \rho ^\theta \), where \(\sigma :=2^{\frac{1}{\theta }}\rho \in (0,T^{\frac{1}{\theta }})\). Furthermore, by (2.2) and (2.3) we see that
for all y, \(z\in {{ \mathbb {R}}}^N\) and \(0<s<t\). Combining (3.2), (3.3), (3.4), and (3.5), we find \(C_*\ge 1\) such that
for a.a. \(z\in {{ \mathbb {R}}}^N\) and a.a. \(t\in (\rho ^\theta ,T/2)\). Thanks to the convexity of f, by (2.2) we may apply Jensen’s inequality to obtain
for a.a. \(z\in {{ \mathbb {R}}}^N\) and a.a. \(t\in (\rho ^\theta ,T/2)\).
Since f is convex in \([0,\infty )\), it is Lipschitz continuous in any compact subinterval of \([0,\infty )\). We may then let \(\zeta \) denote the unique local solution of the integral equation
Hence, \(\zeta \) is the unique local solution of
By (3.6), applying the standard theory for ordinary differential equations to (3.7), we see that the solution \(\zeta \) exists in \([\rho ^\theta ,T/2)\) and satisfies
It follows from (3.8) and property (f2) that
for all \(t\in [\rho ^\theta ,T/2)\). Then, we have
Recalling (3.8) and setting \(\eta :=C_*^{-1}\mu (B(z,\sigma )) s^{-\frac{N}{\theta }}\), we take large enough \(C_*\) if necessary so that
for all \(z\in {\mathbb {R}}^N\) and \(\sigma \in (0,T^{\frac{1}{\theta }})\), where \(\gamma =2^{-\frac{N}{\theta }}C_*\). Here, we used the relation \(\sigma =2^{\frac{1}{\theta }}\rho \in (0,T^{\frac{1}{\theta }})\). Thus, inequality (3.1) holds, and the proof of Theorem 3.1 is complete. \(\square \)
Corollary 3.1
Assume conditions (F1) and (F2). Let u satisfy
in \(Q_T\) for some \(T>0\). Then, there exists a unique \(\nu \in {{\mathcal {M}}}\) as the initial trace of u. Furthermore,
-
(i)
u is a solution of problem (P) in \(Q_T\) with \(\mu =\nu \);
-
(ii)
there exists \(C=C(N,\theta ,F)>0\) such that
$$\begin{aligned} \sup _{z\in {{ \mathbb {R}}}^N}\mu (B(z,\sigma ))\le \left\{ \begin{array}{ll} C\sigma ^{N-\frac{\theta }{p-1}}|\log \sigma |^{-\frac{q}{p-1}} &{}\quad \text {if}\quad p\ne p_\theta ,\\ C|\log \sigma |^{-\frac{N(q+1)}{\theta }} &{}\quad \text {if}\quad p=p_\theta ,\,\,\,q\ne -1,\\ C[\log |\log \sigma |]^{-\frac{N}{\theta }} &{}\quad \text {if}\quad p=p_\theta ,\,\,\,q=-1, \end{array} \right. \end{aligned}$$for all small enough \(\sigma >0\).
Proof
The existence and uniqueness of the initial trace of u follows from Lemma 2.4. Let \(d\in (1,p)\), \(R>0\), and \(\kappa >0\). Set
By Lemma 2.8 (i) and (ii), we see that f is convex in \((0,\infty )\) and (f2) in Theorem 3.1 holds. Furthermore, thanks to Lemma 2.8 (iii), taking small enough \(\kappa >0\) and large enough \(R>0\), by (F2) we can ensure that \(F(\tau )\ge f(\tau )\) in \([0,\infty )\) and consequently (f1) in Theorem 3.1 also holds. In particular, we find \(L\in (R,\infty )\) such that
By Theorem 3.1, we also find \(\gamma \ge 1\) such that
for all \(z\in \mathbb {R}^N\) and \(\sigma \in (0,T^{\frac{1}{\theta }})\).
We show that
For then by Remark 2.1 (ii), we have
and assertion (i) will follow from Lemma 2.5.
Suppose that \(\sigma \in (0,T^{\frac{1}{\theta }})\) but (3.11) does not hold. Then, there exists a sequence \(\{z_n\}\subset \mathbb {R}^N\) such that \(m_\sigma (z_n)\rightarrow \infty \) as \(n\rightarrow \infty \). Consequently,
for all n large enough. By (3.9), (3.10), (3.13), and Lemma 2.7 (i) (with \(a=p-1\), \(b=\theta /N\), and \(c=q\)), we obtain
Hence,
Letting \(n\rightarrow \infty \) in (3.14) yields a contradiction and thus (3.11) holds.
We now prove assertion (ii). Consider first the case where \(p\not =p_\theta \). We show that there exist \(C>0\) and \(\sigma _*>0\) such that
for all \(z\in { \mathbb {R}}^N\) and \(\sigma \in (0,\sigma _*)\). Suppose, for contradiction, that there exist sequences \(\{z_n\}\subset \mathbb {R}^N\) and \(\{\sigma _n\}\subset (0,\infty )\) such that
Set \(M_n:=m_{\sigma _n}(z_n)\). It follows from (3.12) that
for all n large enough. By (3.16), we necessarily have
so that for n large enough,
Similar to the proof of part (i), it follows from (3.9), (3.10), and (3.18) that
Applying Lemma 2.7 (i) (with \(a=p-1\), \(b=\theta /N\), and \(c=q\)), we obtain (for n large enough)
where \(\tau _n:=\sigma _n^{-N}M_n\). For n large enough, and rescaling with \(s_n=C\tau _n\) in (3.19), we can apply Lemma 2.6 (with \(a=p-1>0\), \(b=q\), and \(c=0\)) to obtain (after rescaling back to \(\tau _n\))
Consequently, for such n,
contradicting (3.16). Thus, (3.15) holds, as required.
Now consider the case when \(p=p_\theta \) and \(q\not =-1\). We show that there exist \(C>0\) and \(\sigma _*>0\) such that
for all \(z\in { \mathbb {R}}^N\) and \(\sigma \in (0,\sigma _*)\). Suppose, for contradiction, that there exist sequences \(\{z_n\}\subset \mathbb {R}^N\) and \(\{\sigma _n\}\subset (0,\infty )\) such that
Set \(M_n:=m_{\sigma _n}(z_n)\). Since \(\sigma ^{-N/2}\ge |\log \sigma |^{\frac{N(q+1)}{\theta }}\) for all \(\sigma >0\) small enough, by (3.21), we necessarily have
so that for n large enough,
Similar to the proof of part (i), it follows from (3.9), (3.10), and (3.22) that
Now set \(c_q:=1/2\) if \(q\ge 0\) and \(c_q:=1\) if \(q<0\). Then, by (3.23) we have
so that
Setting \(\tau _n:=\gamma ^{-1}\sigma _n^{-Nc_q}M_n\), (3.24) can be written as
Applying Lemma 2.6 to (3.25) (with \(a=\theta /N\), \(b=q\), and \(c=0\)) then yields
for all n large enough. Consequently for such n,
contradicting (3.21). Thus, (3.20) holds, as required.
Finally, consider the case when \(p=p_\theta \) and \(q=-1\). We show that there exist \(C>0\) and \(\sigma _*>0\) such that
for all \(z\in { \mathbb {R}}^N\) and \(\sigma \in (0,\sigma _*)\). Suppose, for contradiction, that there exist sequences \(\{z_n\}\subset \mathbb {R}^N\) and \(\{\sigma _n\}\subset (0,\infty )\) such that
Set \(M_n:=m_{\sigma _n}(z_n)\). Since \(\sigma ^{-N}\ge \left( \log |\log \sigma |\right) ^{\frac{N}{\theta }}\) for all \(\sigma >0\) small enough, by (3.27), we necessarily have
Then, combining (3.17), we find \(L'>0\) such that
for all n large enough. Once again, by (3.9), (3.10), and (3.28) we have
for all n large enough, where \(\tau _{n}:=\gamma ^{-1}\sigma _n^{-N}M_n\). By (3.22) and Lemma 2.6 (with \(a=\theta /N\), \(b=0\), and \(c=1\)), we have
so that
contradicting (3.27). Hence, (3.26) holds, as required. The proof of Corollary 3.1 is complete. \(\square \)
4 Sufficient conditions for solvability
In this section, we establish sufficient conditions for the existence of a supersolution, and consequently of a local-in-time solution of problem (P), for three general classes of nonlinearity F (see Theorems 4.1, 4.2, and 4.3). As corollaries, we obtain the corresponding results when specializing to nonlinearities satisfying (F1) and (F2) (Corollaries 4.1, 4.2, and 4.3). Indeed, for F satisfying (F1) and (F2) the classification of initial data for which problem (P) is locally solvable separates naturally into the following three cases:
-
(A):
\( \hbox {either}\quad {\mathrm{(i)}}\quad 1<p<p_\theta \hbox { and } q\in {{\mathbb {R}}}\quad \hbox {or}\quad {\mathrm{(ii)}}\quad { p=p_\theta \hbox { and } q<-1}; \)
-
(B):
\(p=p_\theta \) and \(q\ge -1\);
-
(C):
\(p>p_\theta \).
4.1 Sufficiency: case (A)
We begin with nonlinearities F which generalize case (A).
Theorem 4.1
Let F be a nonnegative continuous function in \([0,\infty )\) and assume the following conditions:
-
(A1)
there exists \(R\ge 0\) such that the function \((R,\infty )\ni \tau \mapsto \tau ^{-1}F(\tau )\) is increasing;
-
(A2)
\(\displaystyle {\int _1^\infty \tau ^{-p_\theta -1}F(\tau )\,{\textrm{d}}\tau <\infty }\).
If \(\mu \in {{\mathcal {M}}}\) satisfies
then problem (P) possesses a solution u in \(Q_T\) for some \(T>0\), with u satisfying
in \(Q_T\) for some \(C>0\).
Proof
Let \(T\in (0,1)\) be chosen later. Set \(w(x,t):=R+2[S(t)\mu ](x)\). It follows from Lemma 2.1 and (4.1) that
for \(0<t\le T\) and small enough T, where \(M:=C{\sup _{x\in {{ \mathbb {R}}}^N}}\mu (B(x,1))+1<\infty \). Then, by (A1) and (4.2) we have
Noting that
then by (A2) and (4.3) we obtain
for small enough T. This means that w is a supersolution in \(Q_T\) and the desired result follows from Lemma 2.2 and (4.2). \(\square \)
Corollary 4.1
Assume conditions (F1) and (F2) with
If \(\mu \in {{\mathcal {M}}}\) satisfies
then problem (P) possesses a solution u in \(Q_T\) for some \(T>0\), with u satisfying
in \(Q_T\) for some \(R>0\) and \(C>0\).
Proof
Set
It follows from Lemma 2.8 (iii) (with \(d=1\) and \(R=0\)) that \(g(\tau )\asymp \tau ^p[\log \tau ]^q\) as \(\tau \rightarrow \infty \). Hence, since either \(1<p<p_\theta \), or \(p=p_\theta \) and \(q<-1\), we have
Let \(\kappa >0\) and \(L>0\). Set
Clearly, \(f(\tau )\asymp g(\tau )\asymp \tau ^p[\log \tau ]^q\) as \(\tau \rightarrow \infty \) and so by (F1)–(F2) we may choose \(\kappa \) and L large enough such that
Now,
for all \(\tau \) large enough (\(\tau >R=R(\kappa , L)\)). Hence, f satisfies hypothesis (A1) of Theorem 4.1. Furthermore, by (4.4) and (4.5), f also satisfies hypothesis (A2) of Theorem 4.1.
Hence, by Theorem 4.1, there exists \(T>0\) and a solution v in \(Q_T\) of problem (P) with F replaced by f, with v satisfying
in \(Q_T\) for some \(C>0\). This together with Lemma 2.3 implies that problem (P) possesses a solution u in \(Q_T\) such that
in \(Q_T\). Thus, Corollary 4.1 follows. \(\square \)
4.2 Sufficiency: case (B)
We consider nonlinearities F which generalize case (B).
Theorem 4.2
Let \(\mu \in {{\mathcal {L}}}_0\) and let F be an increasing, nonnegative continuous function in \([0,\infty )\). Assume that there exist \(R>0\), \(\alpha >0\), and positive functions \(G\in C([R,\infty ))\) and \(H\in C^1([R,\infty ))\) satisfying the following conditions (B1)–(B5):
-
(B1)
\(\tau ^{-p_\theta }F(\tau )\asymp G(\tau )\) as \(\tau \rightarrow \infty \);
-
(B2)
(i) for any \(a\ge 1\) and \(b>0\), \(G(a\tau ^b)\asymp G(\tau )\) as \(\tau \rightarrow \infty \). Furthermore, (ii) \({\lim _{\tau \rightarrow \infty }}\tau ^{-\delta }G(\tau )=0\) for all \(\delta >0\);
-
(B3)
(i) \(H'(\tau )\asymp \tau ^{-1}G(\tau )>0\) and (ii) \(G(\tau H(\tau )^{-1})\asymp G(\tau )\) as \(\tau \rightarrow \infty \). Furthermore, (iii) \({\lim _{\tau \rightarrow \infty }}H(\tau )=\infty \) and (iv) \({\lim _{\tau \rightarrow \infty }}\tau ^{-\delta }H(\tau )=0\) for all \(\delta >0\);
-
(B4)
there exists a strictly increasing and convex function \(\Phi _\alpha \) in \([R,\infty )\) such that
$$\begin{aligned} \Phi _\alpha ^{-1}(\tau )=\tau H(\tau )^{-\alpha } \end{aligned}$$for all \(\tau \in [\Phi _\alpha (R),\infty )\);
-
(B5)
there exists \(\eta \in (0,\theta /N)\) such that the function \(P:(R,\infty )\ni \tau \mapsto \tau ^\eta H(\tau )^{-\alpha } G(\tau )\) is increasing.
Then, there exists \(\varepsilon >0\) such that if \(\mu \) satisfies
![](http://media.springernature.com/lw395/springer-static/image/art%3A10.1007%2Fs00028-022-00853-z/MediaObjects/28_2022_853_Equ65_HTML.png)
for all small enough \(\sigma >0\), then problem (P) possesses a solution u in \(Q_T\) for some \(T>0\), with u satisfying
in \(Q_T\) for some \(C>0\).
We prepare a preliminary lemma.
Lemma 4.1
Let \(R>0\) and \(\alpha >0\). Let G and H be positive functions in \([R,\infty )\) such that \(G\in C([R,\infty ))\) and \(H\in C^1([R,\infty ))\). Assume also that conditions (B2)-(i), (B3)-(i), (iii), (iv), and (B4) in Theorem 4.2 hold. Then, for any \(a>0\), \(b>0\), and \(c\in {{\mathbb {R}}}\),
as \(\tau \rightarrow \infty \).
Proof
We first prove (4.8). Consider the case where \(a\ge 1\) and \(b\ge 1\). By (B3)-(i), we see that H is increasing for large enough \(\tau \). Then, we take large enough \(R'\in (R,\infty )\) so that \(a\tau ^b\ge \tau \ge R'\) for \(\tau \in [R',\infty )\), \((R'/a)^{1/b}\ge R\), and
for all \(\tau \in [R',\infty )\), where \(\xi =(s/a)^{1/b}\). Then, by (B2)-(i) and (B3)-(i), (iii) we take large enough \(R''\in (R',\infty )\) so that
for large enough \(\tau \). Thus, (4.8) holds for \(a\ge 1\) and \(b\ge 1\). In particular, we have
as \(\tau \rightarrow \infty \) for \(a\ge 1\) and \(b\ge 1\). Then, we see that
as \(\tau \rightarrow \infty \) for \(a\ge 1\) and \(b\ge 1\). By (4.11) and (4.12), for any \(a>0\) and \(b>0\), we obtain
as \(\tau \rightarrow \infty \), and (4.8) holds.
Next, we prove (4.9). Let \(\delta >0\) be such that \(b-\delta |c|>0\). By (B3)-(iii), (iv), we see that \(1\le H(\tau )\le \tau ^{\delta }\) for large enough \(\tau \). Since H is increasing for large enough \(\tau \), we have
as \(\tau \rightarrow \infty \). This together with (4.8) implies that \(H(\tau ^b H(\tau )^c)\asymp H(\tau )\) as \(\tau \rightarrow \infty \), that is, (4.9) holds.
Furthermore, we observe from (B4) and (4.9) (with \(b=1\) and \(c=\alpha \)) that
as \(\tau \rightarrow \infty \). Then, we find \(C\ge 1\) such that
for large enough \(\tau \), which together with (B4) implies that
for large enough \(\tau \). Then, by (4.8) we see that
as \(\tau \rightarrow \infty \), yielding (4.10). The proof of Lemma 4.1 is complete. \(\square \)
Proof of Theorem 4.2
Let \(\varepsilon \in (0,1)\) and \(L\in (R,\infty )\) be chosen later. Set
It follows from (4.8) that
for all \(t\in (0,T)\) and small enough T. Furthermore, by (B3)-(iv) we see that \(\rho (t^{\frac{1}{\theta }})\rightarrow \infty \) as \(t\rightarrow 0\). We apply Lemmas 2.1 and 4.1 to obtain
in \(Q_T\) for small enough T. Since
as \(t\rightarrow 0\) (see (B3)-(iii), (iv)), by (B3)-(i), (4.8), (4.9), and (4.14), we have
in \(Q_T\) for small enough T. By (B4) and (4.16), we have
in \(Q_T\) for small enough T. Since \(H^{-\alpha }\) is monotone decreasing for large enough \(\tau \), by (4.15) we have
for all \(t\in (0,T)\) and small enough T. This together with (4.9) implies that
for all \(t\in (0,T)\) and small enough T. By (4.17) and (4.18), we obtain
in \(Q_T\) for small enough T. Then, taking large enough L if necessary, by (B1) and (4.10) we have
in \(Q_T\), where P is as in (B5). Furthermore, by (B5) and (4.19) we obtain
in \(Q_T\) for small enough T. On the other hand, by (B3)-(iv) we see that \(t^{-1}H(t^{-1})^{-1}\rightarrow \infty \) as \(t\rightarrow 0\). Then, by (B2)-(i) and (B3)-(ii) we see that
for all \(t\in (0,T)\) and small enough T. This together with (4.8), (4.9), and (4.21) implies that
in \(Q_T\). Since \(0<\eta <\theta /N\) (see (B5)), by (4.19), (4.20), and (4.22) we obtain
in \(Q_T\). Therefore, we deduce from (B3)-(i) and (4.23) that
for all \(t\in (0,T)\). Similarly, by (4.14) and Lemma 4.1, we have
in \(Q_T\). Therefore, taking small enough \(\varepsilon \in (0,1)\), by (4.24) and (4.25) we obtain
in \(Q_T\), where we have used the fact that
by Jensen’s inequality. Hence, w is a supersolution in \(Q_T\) and Theorem 4.2 now follows from Lemma 2.2 and (4.19). \(\square \)
Corollary 4.2
Let \(\mu \in {{\mathcal {L}}}_0\) and assume conditions (F1) and (F2) hold with \(p=p_\theta \) and \(q\ge -1\). Let \(\alpha >0\) and set
for \(\tau \in (0,\infty )\). Then, there exists \(\varepsilon >0\) such that if \(\mu \) satisfies
![](http://media.springernature.com/lw330/springer-static/image/art%3A10.1007%2Fs00028-022-00853-z/MediaObjects/28_2022_853_Equ84_HTML.png)
for small enough \(\sigma >0\), then problem (P) possesses a solution u in \(Q_T\) for some \(T>0\), with u satisfying
in \(Q_T\) for some \(C>0\).
Proof
Let \(\alpha >1\). Set
Then, for any \(a\ge 1\) and \(b>0\), we have
as \(\tau \rightarrow \infty \). Thus, condition (B2) holds. Let \(\Phi _\alpha \) be the inverse function of \(\Psi _\alpha \), and we show that conditions (B3)–(B5) in Theorem 4.2 hold.
Consider the case of \(q>-1\). Then,
as \(\tau \rightarrow \infty \). We see that
Thus, condition (B3) holds. Furthermore, we observe that \(\Psi _\alpha \) is strictly increasing and concave for large enough \(\tau \), that is, the inverse function \(\Phi _\alpha \) of \(\Psi _\alpha ^{-1}\) exists and it is strictly increasing and convex for large enough \(\tau \). Then, condition (B4) holds. In addition, for any \(\eta >0\), setting
by Lemma 2.6, we see that \(P'(\tau )>0\) for large enough \(\tau \). This implies that condition (B5) also holds. Thus, conditions (B3)–(B5) hold in the case of \(q>-1\).
Consider the case of \(q=-1\). It follows that
as \(\tau \rightarrow \infty \). Similarly to the case of \(q>-1\), we have
Thus, condition (B3) holds. Furthermore, we see that \(\Psi _\alpha \) is strictly increasing and concave for large enough \(\tau \), that is, the inverse function \(\Psi _\alpha ^{-1}\) exists and it is strictly increasing and convex for large enough \(\tau \). Then, condition (B4) holds. In addition, for any \(\eta >0\), setting
by Lemma 2.6 we see that \(P'(\tau )>0\) for large enough \(\tau \). This implies that condition (B5) also holds. Thus, conditions (B3)–(B5) hold in the case of \(q=-1\).
Assume (4.26). By Lemma 2.6 (with \(a=1\), \(b=-\alpha (q+1)\), and \(c=0\) for \(q>-1\) and with \(a=1\), \(b=0\), and \(c=-\alpha \) for \(q=-1\)), we have
as \(\tau \rightarrow \infty \). Since
for large enough \(\tau \), taking large enough \(R>0\) if necessary, we see that
![](http://media.springernature.com/lw494/springer-static/image/art%3A10.1007%2Fs00028-022-00853-z/MediaObjects/28_2022_853_Equ85_HTML.png)
Furthermore, we see that
for \(\tau >0\). Then, by (4.26) and (4.27) we see that
![](http://media.springernature.com/lw486/springer-static/image/art%3A10.1007%2Fs00028-022-00853-z/MediaObjects/28_2022_853_Equ217_HTML.png)
for all small enough \(\sigma >0\).
Let f be as in (4.5). Since \(f(\tau )\asymp \tau ^{p_\theta }(\log \tau )^q\) as \(\tau \rightarrow \infty \), condition (B1) holds with F replaced by f. We deduce from Theorem 4.2 that problem (P) with F replaced by f possesses a solution v in \(Q_T\) for some \(T>0\) such that
for all \((x,t)\in Q_T\). This together with \(f(\tau )\ge F(\tau )\) (by (4.6)) and Lemma 2.3 implies that problem (P) possesses a solution u in \(Q_T\) such that
for all \((x,t)\in Q_T\). Thus, Corollary 4.2 follows. \(\square \)
4.3 Sufficiency: case (C)
In this section, we consider nonlinearities F which generalize case (C).
Theorem 4.3
Let \(\mu \in {{\mathcal {L}}}_0\) and let F be an increasing, nonnegative continuous function in \([0,\infty )\) such that
-
(C1)
there exist \(R\ge 0\) and \(d>1\) such that the function \((R,\infty )\ni \tau \mapsto \tau ^{-d}F(\tau )\in (0,\infty )\) is increasing.
Furthermore, assume that there exists a continuous function G in \([R,\infty )\) satisfying the following conditions:
-
(C2)
there exists \(p\in [d,d+1)\) such that \(G(\tau )\succeq \tau ^{-p}F(\tau )>0\) as \(\tau \rightarrow \infty \);
-
(C3)
for any \(a\ge 1\), \(b>0\), and \(c\in {{ \mathbb {R}}}\), \(G(a\tau ^b G(\tau )^c)\asymp G(\tau )\) as \(\tau \rightarrow \infty \);
-
(C4)
there exists \(\delta \in (0,1)\) such that the function \((R,\infty )\ni \tau \mapsto \tau ^{-\delta }G(\tau )\) is decreasing.
Let \(\alpha >1\). Then, there exists \(\varepsilon >0\) such that if \(\mu \) satisfies
![](http://media.springernature.com/lw342/springer-static/image/art%3A10.1007%2Fs00028-022-00853-z/MediaObjects/28_2022_853_Equ86_HTML.png)
for small enough \(\sigma >0\), then problem (P) possesses a solution u in \(Q_T\) for some \(T>0\), with u satisfying
in \(Q_T\) for some \(C>0\).
Proof
Let \(\varepsilon \in (0,1)\) be chosen later, and assume (4.28). Without loss of generality, we may assume that \(\alpha \in (1,d)\). Indeed, if \(\alpha \ge d\) and (4.28) holds, then for any \(\alpha '\in (1,d)\) we can write \(\mu ^{\alpha }=(\mu ^{\alpha '})^\frac{\alpha }{\alpha '}\) and apply Jensen’s inequality to give
![](http://media.springernature.com/lw552/springer-static/image/art%3A10.1007%2Fs00028-022-00853-z/MediaObjects/28_2022_853_Equ218_HTML.png)
for small enough \(\sigma >0\). Consequently, (4.28) also holds for \(\alpha '\in (1,d)\).
Set
It follows from (C3), Lemma 2.1, and (4.28) that
in \(Q_T\) for small enough T. On the other hand, by (C1) and (C2) we see that
since \(p<d+1\). These imply that
in \(Q_T\). Since \(1<\alpha <d\), by (C1)–(C3), (4.30), and (4.31) we obtain
in \(Q_T\) for small enough T. Similarly, by (4.31) we have
in \(Q_T\). On the other hand, by (C4) we see that
for all \(t\in (0,T)\) and small enough T. Therefore, taking small enough \(\varepsilon \), by Jensen’s inequalities, (4.32), (4.33), and (4.34) we obtain
in \(Q_T\). Hence, w is a supersolution in \(Q_T\) and Theorem 4.3 now follows from Lemma 2.2, (4.29), and (4.30). \(\square \)
Corollary 4.3
Let \(\mu \in {{\mathcal {L}}}_0\) and assume conditions (F1) and (F2) hold. For any \(\alpha >1\), there exists \(\varepsilon >0\) such that if \(\mu \) satisfies
![](http://media.springernature.com/lw335/springer-static/image/art%3A10.1007%2Fs00028-022-00853-z/MediaObjects/28_2022_853_Equ93_HTML.png)
for all small enough \(\sigma >0\), then problem (P) possesses a solution u in \(Q_T\) for some \(T>0\), with u satisfying
in \(Q_T\) for some \(R, C>0\).
Proof
Let \(d\in (1,p)\) with \(d>p-1\). Let \(\kappa \), \(L>0\), and set
for \(\tau \in (0,\infty )\). It follows from Lemma 2.8 (iii) that
as \(\tau \rightarrow \infty \). We take large enough \(\kappa \) and L so that \(F(\tau )\le f(\tau )\) in \([0,\infty )\). On the other hand, since
for large enough \(\tau \), condition (C1) in Theorem 4.3 holds in \((R,\infty )\) with F replaced by f for some \(R>0\).
Taking large enough R if necessary and setting \(G(\tau ):=(\log \tau )^q\) for \(\tau \in (R,\infty )\), we see that the function \((R,\infty )\ni \tau \mapsto \tau ^{-\frac{1}{2}}G(\tau )\) is decreasing (i.e., \(\delta =1/2\) in (C4)). By (4.36) and (4.37), we find \(C>0\) such that
for all \(\tau \in (R,\infty )\). Then, conditions (C2)–(C4) in Theorem 4.3 hold with F replaced by f. Therefore, by Theorem 4.3 there exists \(\varepsilon >0\) such that if \(\mu \) satisfies (4.35), then problem (P) with F replaced by f possesses a solution v in \(Q_T\) for some \(T>0\) such that
in \(Q_T\), for some \(C>0\). This together with Lemma 2.3 implies that problem (P) possesses a solution u in \(Q_T\) such that
in \(Q_T\). Thus, Corollary 4.3 follows. \(\square \)
4.4 A special case: Dirac measure as initial data
Here, we provide a necessary and sufficient condition on the nonlinearity F for the solvability of problem (P) in the special case when \(\mu =\delta _y\), the Dirac measure in \({{ \mathbb {R}}}^N\) based at point y. This problem was considered in [7] for the opposite sign pure power law case \(F(u)=-u^p\), i.e., dissipative F.
Corollary 4.4
Suppose F satisfies
-
(D1)
F is nonnegative and locally Lipschitz continuous in \([0,\infty )\);
-
(D2)
there exist \(R>0\) and \(d>1\) such that
-
(i)
the function \((R,\infty )\ni \tau \mapsto \tau ^{-d}F(\tau )\in (0,\infty )\) is increasing;
-
(ii)
F is convex in \((R,\infty )\).
-
(i)
Let \(y\in \mathbb {R}^N\). Then, problem (P) possesses a local-in-time solution with \(\mu =\delta _y\) if and only if
Proof
Assume that problem (P) possesses a solution with \(\mu =\delta _y\) in \(Q_T\) for some \(T>0\). Set
Then, by (D2)-(i) we see that f is increasing and \(F\ge f\) in \([0,\infty )\). Applying Theorem 3.1 with \(z=y\), so that \(m_\sigma (z)=\delta _y(B(y,\sigma ))\equiv 1\), we find \(\gamma \ge 1\) such that
Letting \(\sigma \rightarrow 0\), we have
This together with (4.39) implies (4.38).
Conversely, under condition (4.38), we apply Theorem 4.1 to obtain a local-in-time solution of problem (P) with \(\mu =\delta _y\). Thus, Corollary 4.4 follows. \(\square \)
We mention that the integral condition (4.38) also appears in [30, Theorem 5.1] as a necessary and sufficient condition for existence with \(L^1\) initial data. See also the informal argument preceding the proof of Theorem 4.1 of that work, where a Dirac delta function is considered as initial data.
5 Proof of the main theorem
Proof of Theorem 1.1
Assertion (i) is proved by Corollary 3.1 (ii), Remark 2.1, and Corollary 4.1.
We now prove the nonexistence parts of statements (1) and (2) in assertion (ii). Suppose first that (1.3) holds and there exists a local solution of problem (P). Then,
for small enough \(\sigma >0\). For large enough \(\gamma _1\), we then obtain a contradiction to Corollary 3.1. Hence, no local solution can exist for such \(\gamma _1\). Now suppose that (1.5) holds. Then, there exists \(C_2>0\) such that
for small enough \(\sigma >0\). Again, we can obtain a contradiction to Corollary 3.1 for large enough \(\gamma _2\) and deduce that problem (P) possesses no local-in-time solution for such \(\gamma _2\).
Next, we prove the existence parts of statements (1) and (2) in assertion (ii). Assume therefore that either (1.4) with \(\varepsilon _1\in (0,1)\) or (1.6) with \(\varepsilon _2\in (0,1)\) hold. Let \(\alpha >0\) and set
If (1.4) holds, then
This implies that
![](http://media.springernature.com/lw539/springer-static/image/art%3A10.1007%2Fs00028-022-00853-z/MediaObjects/28_2022_853_Equ219_HTML.png)
for small enough \(\sigma >0\).
Similarly, if (1.6) holds, then
This implies that
![](http://media.springernature.com/lw530/springer-static/image/art%3A10.1007%2Fs00028-022-00853-z/MediaObjects/28_2022_853_Equ220_HTML.png)
for small enough \(\sigma >0\). Therefore, by Corollary 4.2 we see that, if \(\varepsilon _1>0\) (respectively \(\varepsilon _2>0\)) is small enough, then problem (P) possesses a local-in-time solution. Thus, statements (1) and (2) in assertion (ii) follow.
Finally, we prove statement (3) in assertion (ii). Assume that (1.7) holds. Then,
for small enough \(\sigma >0\). This together with Corollary 3.1 implies that problem (P) possesses no local-in-time solution for large enough \(\gamma _3\). Conversely, suppose that (1.8) holds. Since \(p>p_\theta \), we find \(\alpha >1\) such that \(\alpha \theta /(p-1)<N\). Then, we have
![](http://media.springernature.com/lw439/springer-static/image/art%3A10.1007%2Fs00028-022-00853-z/MediaObjects/28_2022_853_Equ221_HTML.png)
for small enough \(\sigma >0\). By Corollary 4.3, we see that, if \(\varepsilon _3>0\) is small enough, then problem (P) possesses a local-in-time solution. Thus, statement (3) in assertion (ii) follows. The proof is complete. \(\square \)
Remark 5.1
The arguments in the proof of Theorem 1.1 are readily adapted to further log-refinements. For example, suppose that (F2) is replaced by
- (F2\(^\prime \)):
-
\(F(\tau )\asymp \tau ^p[\log \tau ]^q[\log (\log \tau )]^r\) as \(\tau \rightarrow \infty \) for some \(p>1\) and q, \(r\in { \mathbb {R}}\).
Then, we can show that problem (P) possesses a local-in-time solution if and only if
in the cases when (i) \(p<p_\theta \), (ii) \(p=p_\theta \) and \(q<-1\), and (iii) \(p=p_\theta \), \(q=-1\), and \(r<-1\). In the other cases, we divide condition (F2\('\)) into four cases:
-
(1)
\(p=p_\theta \) and \(q=r=-1\).
-
(2)
\(p=p_\theta \), \(q=-1\), and \(r>-1\).
-
(3)
\(p=p_\theta \), \(q>-1\), and \(r\in { \mathbb {R}}\).
-
(4)
\(p>p_\theta \) and q, \(r\in { \mathbb {R}}\),
and we can identify the optimal singularities of the initial data for solvability of problem (P). Since inclusion of the proofs here would make the paper unduly long, we leave the details to the interested reader.
Data availability
Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.
References
P. Baras and R. Kersner, Local and global solvability of a class of semilinear parabolic equations, J. Differential Equations 68 (1987), 238–252,
P. Baras and M. Pierre, Critère d’existence de solutions positives pour des équations semi-linéaires non monotones, Ann. Inst. H. Poincaré Anal. Non Linéaire 2 (1985), 185–212.
B. Ben Slimene, S. Tayachi, and F. B. Weissler, Well-posedness, global existence and large time behavior for Hardy-Hénon parabolic equations, Nonlinear Anal. 152 (2017), 116–148.
K. Bogdan and T. Jakubowski, Estimates of heat kernel of fractional Laplacian perturbed by gradient operators, Comm. Math. Phys. 271 (2007), 179–198.
L. Brandolese and G. Karch, Far field asymptotics of solutions to convection equation with anomalous diffusion, J. Evol. Equ. (2008), 307–326.
H. Brezis and T. Cazenave, A nonlinear heat equation with singular initial data, J. Anal. Math., 68 (1996), 277–304.
H. Brézis and A. Friedman, Nonlinear parabolic equations involving measures as initial conditions, J. Math. Pures Appl. 62 (1983), 73–97.
G. Caristi and E. Mitidieri, Existence and nonexistence of global solutions of higher-order parabolic problems with slow decay initial data, J. Math. Anal. Appl. 279 (2003), 710–722.
N. Chikami, Composition estimates and well-posedness for Hardy–Hénon parabolic equations in Besov spaces, J. Elliptic Parabol. Equ. 5 (2019), 215–250.
K. Deng, M. Fila, and H. A. Levine, On critical exponents for a system of heat equations coupled in the boundary conditions, Acta Math. Univ. Comenian. (N.S.) 63 (1994), 169–192.
M. Escobedo and M. A. Herrero, Boundedness and blow up for a semilinear reaction–diffusion system, J. Differential Equations 89 (1991), 176–202.
Y. Fujishima and K. Ishige, Initial traces and solvability of Cauchy problem to a semilinear parabolic system, to appear in J. Math. Soc. Japan (arXiv:1910.06546).
Y. Fujishima and K. Ishige, Optimal singularities of initial functions for solvability of a semilinear parabolic system, to appear in J. Math. Soc. Japan (arXiv:2012.05479).
H. Fujita, On the blowing up of solutions of the Cauchy problem for\(u_{t}=\Delta u+u^{1+\alpha }\), J. Fac. Sci. Univ. Tokyo Sect. I 13 (1966), 109–124.
V. A. Galaktionov and H. A. Levine, On critical Fujita exponents for heat equations with nonlinear flux conditions on the boundary, Israel J. Math. 94 (1996), 125–146.
V. A. Galaktionov and S. I. Pohozaev, Existence and blow-up for higher-order semilinear parabolic equations: majorizing order-preserving operators, Indiana Univ. Math. J. 51 (2002), 1321–1338.
F. Gazzola and H.-C. Grunau, Global solutions for superlinear parabolic equations involving the biharmonic operator for initial data with optimal slow decay, Calc. Var. Partial Differential Equations 30 (2007), 389–415.
M. Guedda and M. Kirane., A note on nonexistence of global solutions to a nonlinear integral equation, Bull. Belg. Math. Soc. Simon Stevin, 6 (1999), 491–497.
K. Hisa and K. Ishige, Existence of solutions for a fractional semilinear parabolic equation with singular initial data, Nonlinear Anal. 175 (2018), 108–132.
K. Hisa and K. Ishige, Solvability of the heat equation with a nonlinear boundary condition, SIAM J. Math. Anal. 51 (2019), 565–594.
K. Hisa, K. Ishige, and J. Takahashi, Existence of solutions for an inhomogeneous fractional semilinear heat equation, Nonlinear Anal. 199 (2020), 111920, 28.
K. Hisa and M. Sierżęga, Existence and nonexistence of solutions to the Hardy parabolic equation, (arXiv:2102.04079).
K. Hisa and J. Takahashi, Optimal singularities of initial data for solvability of the Hardy parabolic equation, J. Differential Equations 296 (2021), 822–848.
K. Ishige, T. Kawakami, and S. Okabe, Existence of solutions for a higher-order semilinear parabolic equation with singular initial data, Ann. Inst. H. Poincaré Anal. Non Linéaire 37 (2020), 1185–1209.
K. Ishige, T. Kawakami, and S. Okabe, Existence of solutions to nonlinear parabolic equations via majorant integral kernel, preprint (arXiv:2101.06581).
K. Ishige, T. Kawakami, and M. Sierżȩga, Supersolutions for a class of nonlinear parabolic systems, J. Differential Equations 260 (2016), 6084–6107.
K. Ishige and R. Sato, Heat equation with a nonlinear boundary condition and uniformly local\(L^r\)spaces, Discrete Contin. Dyn. Syst. 36 (2016), 2627–2652.
K. Ishige and R. Sato, Heat equation with a nonlinear boundary condition and growing initial data, Differential Integral Equations 30 (2017), 481–504.
H. Kozono and M. Yamazaki, Semilinear heat equations and the Navier–Stokes equation with distributions in new function spaces as initial data, Comm. Partial Differential Equations 19 (1994), 959–1014.
R. Laister, J. C. Robinson, M. Sierżęga, A. Vidal-López, A complete characterisation of local existence for semilinear heat equations in Lebesgue spaces, Ann. Inst. H. Poincaré Anal. Non Linéaire 33 (2016), 1519–1538.
R. Laister, M. Sierżęga, Well-posedness of semilinear heat equations in\(L^1\), Ann. Inst. H. Poincaré Anal. Non Linéaire, 37 (2020), 709–725.
R. Laister, M. Sierżęga, A blow-up dichotomy for semilinear fractional heat equations, Math. Ann. 381 (2021), 75–90.
Y. Miyamoto, A doubly critical semilinear heat equation in the\(L^1\)space, J. Evol. Equ. 21 (2021), 151–166.
P. Quittner and P. Souplet, Admissible\(L_p\)norms for local existence and for continuation in semilinear parabolic systems are not the same, Proc. Roy. Soc. Edinburgh Sect. A 131 (2001), 1435–1456.
P. Quittner and P. Souplet, Superlinear parabolic problems, Birkhäuser Advanced Texts: Basler Lehrbücher. [Birkhäuser Advanced Texts: Basel Textbooks], 2019.
J. C. Robinson and M. Sierżęga, Supersolutions for a class of semilinear heat equations, Rev. Mat. Complut. 26 (2013), 341–360.
S. Sugitani, On nonexistence of global solutions for some nonlinear integral equations, Osaka Math. J. 12 (1975), 45–51.
J. Takahashi, Solvability of a semilinear parabolic equation with measures as initial data, Geometric properties for parabolic and elliptic PDE’s, Springer Proc. Math. Stat., vol. 176, Springer, [Cham], 2016, pp. 257–276.
X. Wang, On the Cauchy problem for reaction–diffusion equations, Trans. Amer. Math. Soc. 337 (1993), 549–590.
F. B. Weissler, Local existence and nonexistence for semilinear parabolic equations in\(L^{p}\), Indiana Univ. Math. J. 29 (1980), 79–102.
F. B. Weissler, Existence and nonexistence of global solutions for a semilinear heat equation, Israel J. Math. 38 (1981), 29–40.
Acknowledgements
YF was supported partially by JSPS KAKENHI Grant Number 19K14569. KH and KI were supported in part by JSPS KAKENHI Grant Number JP19H05599. RL was supported partially by a Daiwa Anglo-Japanese Foundation grant (Ref.: 4646/13713).
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Fujishima, Y., Hisa, K., Ishige, K. et al. Solvability of superlinear fractional parabolic equations. J. Evol. Equ. 23, 4 (2023). https://doi.org/10.1007/s00028-022-00853-z
Accepted:
Published:
DOI: https://doi.org/10.1007/s00028-022-00853-z