Abstract
We consider Dirichlet heat kernel \(p_a^{(\mu )}(t,x,y)\) for the Bessel differential operator \(L^{(\mu )}=\frac{d^2}{dx^2}+\frac{2\mu +1}{2x}\), \(\mu \in \mathbf {R}\), in half-line \((a,\infty )\), \(a>0\), and provide its asymptotic expansions for \(xy/t\rightarrow \infty \).
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Let \(\mu \in \mathbf {R}\) and \(a>0\). We consider Dirichlet heat kernel \(p_a^{(\mu )}(t,x,y)\) in half-line \((a,\infty )\) (with respect to the reference measure \(m^{(\mu )}(dy)=m^{(\mu )}(y)dy=y^{2\mu +1}dy\)) for the Bessel differential operator
This function is a fundamental solution of the heat equation \(2\partial _t=L^{(\mu )}\) with Dirichlet boundary condition at the point a. Our main objective was to derive the precise asymptotic expansion of \(p_a^{(\mu )}(t,x,y)\) for \(xy/t\rightarrow \infty \). The main result of the paper can be stated as follows.
Theorem 1
For \(\mu \in \mathbf {R}, \ a>0\) we have
where \(x,y>a, \ t>0\) and estimates of the error term are given by
while \(xy/t\rightarrow \infty \), with the constant \(t_0(\mu )>0\) defined as follows
Here \(f{\mathop {\lesssim }\limits ^{\mu }} g\) means that there exists a constant \(C=C(\mu )>0\) such that \(f\le Cg\) for all indicated arguments.
It is worth emphasizing, that in some régime of the space parameters we are able to give more precise (long-time) estimates of the error term. In particular, when x, y are bounded away from the boundary, i.e. level \(a>0\), we can write explicitly the first n (for given \(n\in \mathbf {N}\)) terms in the expansion of \(E_{\mu }(t,x,y)\). Moreover, even when one of the space variables is located near the boundary, we can give one term more in an expansion of the error term in comparison with (2). However, for this purpose, additional assumption on the quantity \((x-a)(y-a)/t\) is made.
Theorem 2
Let \(\mu \in \mathbf {R}, \ a>0, \ n\in \mathbf {N}\) and \(t_0(\mu )\) is defined as in (3). We have
and estimates of the error term for \(xy/t\rightarrow \infty \) are given by
-
(i)
$$\begin{aligned} \left| E_\mu ^{(n)}(t/a^2,x/a,y/a)-\sum _{k=0}^{n-1} c_k^{(\mu )}\left( -\frac{t}{xy}\right) ^{k}\right| {\mathop {\lesssim }\limits ^{\mu }} \left( \frac{t}{xy}\right) ^{n}, \end{aligned}$$
whenever \(t\ge t_0(\mu )\) and \(x,y>2a\). The coefficients \(c_k^{(\mu )}\) are expressed explicitly as
$$\begin{aligned} c_{0}^{(\mu )}=0, \qquad c_k^{(\mu )}=\frac{(4\mu ^2-1^2)\ldots (4\mu ^2-(2k-1)^2)}{8^k k!}, \quad k\in \mathbf {N}, \end{aligned}$$(4) -
(ii)
$$\begin{aligned} \left| E_\mu ^{(2)}(t/a^2,x/a,y/a)-\frac{1-4\mu ^2}{8}\frac{t}{xy}\right| {\mathop {\lesssim }\limits ^{\mu }} \left( \frac{t}{xy}\right) ^2, \end{aligned}$$
whenever \( t\ge t_0(\mu ),\ a<x<2a<y\) and \(\frac{(x-a)(y-a)}{t}\ge 1\).
In order to understand these results better, notice that the case \(a<x,y<2a, \ t\ge t_0(\mu )\) is excluded from our consideration. Indeed, it holds then \(xy/t<4a^2/t\le 4a^2/t_0(\mu )\), which is a contradiction with the condition \(xy/t\rightarrow \infty \). Moreover, since \(p_a^{(\mu )}(t,x,y)\) is a symmetric function of space variables x, y, i.e.
one can assume, without loss of the generality, that \(x<y\). In consequence, it is enough to consider (for \(t\ge t_0(\mu )\)) two cases: when \(a<x<2a<y\) and when \(x,y>2a\). Note that one-term, long-time expansion of \(E_{\mu }(t,x,y)\) was derived just in one situation, namely when \(a<x<2a<y\) and \(0<\frac{(x-a)(y-a)}{t}<1\), while in the other cases we are able to provide longer expansions of the error term.
Studying the behaviour of heat kernels (for second-order differential operators) is of primary importance in many areas of the mathematical analysis, especially in the theory of partial differential equations, harmonic analysis, as well as in the potential theory. However, even for Laplacian, the problem of providing precise description of the behaviour of the corresponding Dirichlet heat kernels is very difficult. Exact formulas, in terms of elementary functions, of these heat kernels for Laplace operator are available only in a few special cases (e.g. half-line, \(\mathbf {R}^d\) or half-space). As a consequence, providing estimates of the heat kernels is one of the most important approach to describe behaviour of these objects. Some classical results involving short-time estimates of the Dirichlet heat kernel for Laplacian were derived by Varadhan in [19, 20]. It is also worth mentioning that estimates of heat kernels for Laplace operator in bounded \(C^{1,1}\) domains were proved by Davies in [6] (upper bound) and Zhang in [21] (lower bound)—see also [22] for the case of domains with bounded complements. These results are only qualitatively sharp, which means that exponential terms in upper and lower bound are different. Notice that estimates of the Dirichlet heat kernels for Laplacian in a ball, with the same exponents in upper and lower bounds (so-called sharp estimates), were obtained very recently by Małecki and Serafin in [13]. For more information (including more general second-order differential operators and domains e.g. manifolds) in this topic we refer the Reader to the monographs [7, 9] and references therein.
From the probabilistic point of view, \(p_a^{(\mu )}(t,x,y)\) is a transition probability density of the Bessel process with index \(\mu \in \mathbf {R}\) starting from \(x>a\) and killed upon leaving a half-line \((a,\infty ), \ a>0\). In this context, results in this article are a continuation of the research related to the first hitting times at a given level by the Bessel process. Namely, sharp estimates of the density of these hitting times were obtained recently in [4, 10]. The next natural step was to describe the transition probability density \(p_a^{(\mu )}(t,x,y)\). T. Takemura in [16] derived the integral representations involving highly oscillating functions. Then, in papers [2, 3], sharp estimates of \(p_a^{(\mu )}(t,x,y)\) in a full range of the variables \(x,y>a\) and \(t>0\) were obtained. In this context, providing asymptotic expansion of the considered Bessel heat kernels is a natural improvement of these results. It is worth noting that the analogous expansions for the density of the first hitting time of Bessel processes were derived recently in [15, 17].
To compare the main results of this paper with the known ones for the heat kernels of the Bessel operator in half-line, let us invoke the estimates from [2, 3] related to the case \(xy/t\rightarrow \infty \). In particular, it was shown that for \(\mu \in \mathbf {R}\) we have
whenever \(x,y>a, \ t>0\) and \(xy\ge t\). Here \(f\wedge g\) denotes the minimum of f and g. In fact, it should be stressed that the estimates of \(p_a^{(\mu )}(t,x,y)\) given in [2, 3] are valid for the whole range of the parameters \(x,y>a,\ t>0\) and \(\mu \in \mathbf {R}\). However, our results are more precise (but on limited range of parameters) and the estimate (6) for xy / t large enough is a consequence of those given in Theorem 1. To make it clearer, let us point out the main differences between above mentioned results. At first, notice that the leading term given in (6), i.e.,
is of a correct asymptotic order, but it is not of the exact asymptotic form. Secondly, the function
describes the behaviour of the heat kernel near the boundary (the level \(a>0\)) more precise than the expression \(1\wedge \frac{(x-a)(y-a)}{t}\) from (6). This difference is the most apparent when \((x-a)(y-a)/t\) is small due to the expansion of \(z\mapsto 1-e^{-z}\) near zero. Finally, our third remark is that, in opposite to (6), Theorem 1 contains estimates of the error term. This accuracy strongly depends (in the case \(xy/t\rightarrow \infty \)) on the régime of the time parameter [see, (2)]. As it was already mentioned, under stronger assumptions on the space-variables x, y, even more precise bounds can be obtained—see Theorem 2.
The paper is organized as follows. In Preliminaries basic notation and known results on modified Bessel functions as well as on Bessel processes are introduced. The next two sections are devoted to the proofs of Theorems 1–2. In particular, Sect. 3 contains short-time estimates of \(p_a^{(\mu )}(t,x,y)\) given in Theorem 1 (see also Proposition 1), while the long-time ones are provided in series of the Propositions 2–5 in Sect. 4. Finally, in Appendix, some useful lemmas are collected.
2 Preliminaries
2.1 Notation
The symbol \(f{\mathop {\lesssim }\limits ^{\mu }} g\) means that there exists a constant \(C=C(\mu )>0\), depending on \(\mu \), such that \(f\le Cg\) for all indicated arguments. Moreover, if the conditions \(f{\mathop {\lesssim }\limits ^{\mu }}g\) and \(g{\mathop {\lesssim }\limits ^{\mu }} f\) are simultaneously satisfied, then we write \(f{\mathop {\approx }\limits ^{\mu }}g\). The notation \(f=O_{\mu }(g)\) denotes \(|f|{\mathop {\lesssim }\limits ^{\mu }} g\). However, in proofs we omit writing \(\mu \) over the signs \(\lesssim \) and \(\approx \). Similarly, the dependence of the constants on the parameters is not indicated in proofs throughout this article. To simplify the notation, we introduce the function
where \(t>0\) and the coefficients A, B are given by
Moreover, the following function (see Theorem 1)
where \(x,y>1, \ t>0\) and \(\mu \ge 0\), will be used in the paper.
2.2 Modified Bessel function
The modified Bessel functions of the first and second kind are denoted by \(I_{\mu }(z)\) and \(K_{\mu }(z)\), respectively, and defined as follows
for \(\mu \not \in \mathbf {N}\); otherwise we take a limit when \(\mu \rightarrow n\in \mathbb {N}\) of the last expression. Their asymptotic behaviour at infinity can be described as follows (see 9.7.1 and 9.7.2 in [1])
where the coefficients \(c_k^{(\mu )}\) are defined in (4). On the other hand, the asymptotic behaviour of \(I_{\mu }(z)\) at zero is given by (see 9.6.7 in [1])
Finally, we recall the following two-sided bounds for the ratio of the modified Bessel function of the first kind, provided by Laforgia in [11]
Here the upper bound holds for \(\mu >-1/2\) , while the lower estimate is true for \(\mu \ge 1/2\). Moreover, the inequality \(I_{\mu }(x)\le I_{\mu }(y)\) is valid whenever \(\mu \ge 0\) and \(y\ge x>0\) (see formula 8.445 in [8]).
2.3 Bessel processes
We work on the canonical path space \(C([0,\infty ),\mathbf {R})\), equipped with the filtration \(\{\mathcal {F}_t\}_{t\ge 0}\) generated by trajectories up to time t. Then, for a given trajectory \(R\in C([0,\infty ),\mathbf {R})\), the first hitting time \(T_a\) at given level \(a>0\) is defined as
Moreover, we define the Bessel process\(BES^{(\mu )}(x)\), with index \(\mu \in \mathbf {R}\) and starting point \(x>0\), as a one-dimensional diffusion on \([0,\infty )\) with infinitesimal generator
where the killing condition at zero for \(\mu <0\) is imposed. The probability law and the corresponding expected value of the Bessel process \(BES^{(\mu )}(x)\) are denoted by \(\mathbf {P}_x^{(\mu )}\) and \(\mathbf {E}^{(\mu )}_x\), respectively. Notice that the laws of Bessel processes with different indices are absolutely continuous and their Radon–Nikodym derivative is given by
where the equality holds \(\mathbf {P}^{(\nu )}_x\)—a.s. on \(\{T_0>t\}\).
The density of \(T_a\) for the Bessel process \(BES^{(\mu )}(x)\) is denoted by
For \(a=1\) we write simply \(q_x^{(\mu )}(s)\), where \(x>1\) and \(s>0\). Sharp estimates of this function were obtained in [4], i.e. for \(x>1\) and \(s>0\) we have
and
However, some more precise estimates of \(q_x^{(\mu )}(s)\) are also available. In particular, the following asymptotic expansion (see Theorem B in [18]) will be used in our article
This result can be strengthened by giving more precise estimates of the error term for \(0\le \mu <1/2\) (see Lemma 4 in [4]), namely we have
where \(0\le E_{\mu }(s,x)\le \frac{1-4\mu ^2}{8}\frac{s}{x}\) for \(x>1, \ s>0\). Some improvement is available also in the case \(\mu \ge 1/2\), i.e. Lemma 1 in [2] states that
where \(\mu \ge 1/2\) and \(x>1, \ s>0\).
Recall now that \(m^{(\mu )}(dy)=m^{(\mu )}(y)dy=y^{2\mu +1}dy\). The crucial fact is that the considered Bessel heat kernel (for \(a=0\)) is the transition probability density \(p^{(\mu )}(t,x,y)\) of the Bessel process \(BES^{(\mu )}(x)\)
Moreover, this function can be expressed in terms of the modified Bessel function of the first kind in the following way
for \(x,y> 0\) and \(t>0\). In particular, using (8), one can write
where \(x,y>0, \ t>0\) and \(xy/t\rightarrow \infty \). On the other hand, in the case \(a>0\) the considered Bessel heat kernel is the transition probability density of the Bessel process starting from \(x>a\) and killed upon leaving a half-line \((a,\infty )\)
We are able to represent the considered heat kernel \(p_a^{(\mu )}(t,x,y)\) in terms of the above mentioned densities \(p^{(\mu )}(t,x,y)\) and \(q_x^{(\mu )}(s)\), namely it holds
In fact, the last formula is a starting point for deriving estimates of the Bessel heat kernels. Notice that putting \(\mu =-\nu \ge 0\) in (12) gives
which shows us that it is enough to consider indices \(\mu \ge 0\). Moreover, we can focus only on the case \(a=1\) due to the following scaling property
Finally, let us recall the well-known results in the case \(\mu =1/2\) (see formulas 2.13, 2.16 and 2.17 in [2])
where \(x,y>a>0\) and \(t>0\). Observe that the function \(p_1^{(1/2)}(t,x,y)\) appears in the expansion of \(p_1^{(\mu )}(t,x,y)\) given in the main results of the paper (see Theorems 1–2). Therefore, we will often refer to the formula (25), especially in proofs of propositions in the next two sections.
3 Short-time estimates
As it was already mentioned, it is enough to find estimates of \(p_a^{(\mu )}(t,x,y)\) only for \(a=1\) and \(1<x<y\) [see (22) and (5)]. Moreover, due to (21), one can assume that \(\mu \ge 0\). We begin with the part of Theorem 1 related to the case \(0<t<t_0(\mu )\).
Proposition 1
For \(\mu \in \mathbf {R}\) we have
where \(x,y>1\) and \(0<t<t_0(\mu )\).
Proof
Let \(A\subset (1,\infty )\) be a Borel set. We apply the absolute continuity of the laws of Bessel processes (12) to write for \(\mu \ge \nu \ge 0\)
Notice that it holds \(0\le \int _0^t \frac{ds}{R^2(s)}\le \int _0^t ds=t\) on the set \(\{t<T_1\}\), which implies
Since \(\mathbf {E}^{(\mu )}_x[t<T_1,R_t\in A]=\int \limits _A p_1^{(\mu )}(t,x,y)m^{(\mu )}(y)dy\), the set A is given arbitrary and \(m^{(\mu )}(y)=y^{2\mu +1}\), we get
where \(x,y>1\) and \(t>0\). Putting \(\mu \ge \nu =1/2\) and estimating the above-given exponent from below by \(1-\frac{\mu ^2-1/4}{2}t\), we obtain
while for \(\mu =1/2\ge \nu \ge 0\) it holds
Existence of the constant \(C_1^{(\nu )}>0\) is justified by the behaviour of \(t\mapsto \exp \left( \frac{\mu ^2-\nu ^2}{2}t\right) \) for \(0<t<1\). To end the proof, it remains to apply the explicit formula for \(p_1^{(1/2)}(t,x,y)\), given in (25), to the inequalities (26) and (27). \(\square \)
4 Long-time estimates
In this section we provide estimates of the Bessel heat kernel \(p_a^{(\mu )}(t,x,y)\) for large times t. As previously, we will assume that \(a=1, \ 1<x<y\) and \(\mu \ge 0\). Moreover, as it was mentioned in Introduction, we need to investigate only two cases: \(x,y>2\) and \(1<x<2<y\). We start with the first one. In fact, using below given Proposition 2, the part (i) of Theorem 2 is also proved.
Proposition 2
Let \(\mu \ge 0\) be fixed. Then we have for \(x,y>2\) and \(t>0\)
whenever \(xy/t\rightarrow \infty \).
Proof
Since \(p_1^{(\mu )}(t,x,y)\le p^{(\mu )}(t,x,y)\), it is enough to focus on the proof of an appropriate lower bound for the considered heat kernel. We begin with the case \(x\ge t\). Notice that, due to the asymptotic behaviour of the modified Bessel function at zero (10) and at infinity (8), we have for \(w>0\)
Combining this inequality (for \(w=y/(t-s)\) and \(\alpha =-1/2\)) with the following bound [see (17) and (16)]
valid for \(x>1\) and \(s>0\), we can write
The last line is a consequence of the exact formula for \(\int _0^t f_{A,B,t}(s)ds\) given in Lemma 1 and estimates of \(\int _0^t sf_{A,B,t}(s)ds\) from Lemma 2. Both of these lemmas can be found in Appendix. Since \(xy/t\rightarrow \infty \) we can use (19) to justify the existence of \(C_1>0\) such that
where the inequality \(y>t\) was used (since \(y>x\ge t\)). Moreover, because of \(x,y>2\) the last exponent can be bounded from above by \(\exp {(-xy/(4t))}\). Hence, the Hunt formula (20) gives us finally
Therefore, it remains to deal with the situation when \(x<t\). To achieve this, we split integral defining the expression \(r_1(t, x, y)\) as follows
and then we estimate separately each of these integrals, starting with the part \(\mathcal {J}_1^{(\mu )}(t,x,y)\). Let \(0<s<x\). Taking \(\alpha =-1/2, \ w=\frac{y}{t-s}\) in (28) and using the following inequality [see (13) and (14)]
we obtain
The last estimate is a consequence of the exact formula for \(\int _0^t f_{a,b}(s)ds\) given in Lemma 1 from Appendix. Since \(xy/t\rightarrow \infty \) we can use (19) to justify the following inequalities
where the second one is a consequence of the assumption \(x,y>2\). To deal with the part \(\mathcal {J}_2^{(\mu )}(t,x,y)\) we use (28) with \(w=\frac{y}{t-s}, \ \alpha =\mu \) and the following upper bound [see again (13) and (14)]
Namely, we infer that
Moreover, making substitution \(w=\frac{1}{s}-\frac{1}{t}\), one can rewrite this inequality as follows
Therefore, estimating \((1+tw)^{2\mu }\le (t/x)^{2\mu }\) for \(w<\frac{1}{x}-\frac{1}{t}\) and extending the range of integration to \((0,\infty )\), provide us
Here the second line is a consequence of (43) and (44) (see Appendix), since \(\frac{(x-1)(y-1)}{t}\ge \frac{xy}{2t}\rightarrow \infty \) for \(x,y>2\) and \(xy/t\rightarrow \infty \). On the other hand, the last one follows from asymptotic behaviour of \(p^{(\mu )}(t,x,y)\) for \(xy/t\rightarrow \infty \) [see (19)]. Since we assumed \(xy/t\rightarrow \infty \) and \(2<x<y\), we get
Summarizing, in view of (29) and (30), the usage of the Hunt formula (20) immediately ends the proof. \(\square \)
Proof of the part (i) of Theorem 2
Let us fix \(n\in \mathbf {N}\). Due to the asymptotic behaviour of \(p^{(\mu )}(t,x,y)\) for \(xy/t\rightarrow \infty \) [see (19)] we can rewrite the result obtained in Proposition 2 as follows
where \(x,y>2,\ t\ge 4\) and \(xy/t\rightarrow \infty \). Notice that for \(x,y>2\) we have
Thus we can replace the term \(\exp {\left( -\frac{(x-y)^2}{2t}\right) }\) in (31) by the expression
Moreover, since the exponential growth dominates the power one, the product of the expressions in square brackets in (31) can be simply replaced by the first of these. Therefore, in this way we have proved the part (i) of Theorem 2. \(\square \)
The proof of Theorem 1 in the case \(1<x<2<y, \ t\ge t_0(\mu )\) will be provided separately in the situations when \(\frac{(x-1)(y-1)}{t}\) is bounded away from zero or bounded away from infinity. To simplify the notation in the proof of the next proposition, we introduce the following functions
where \(\mu \ge 0, \ x,y>1, \ t>0\). Below given estimates of \(\mathcal {I}_1^{(\mu )}(t,x,y)\) and \(\mathcal {I}_1^{(\mu )}(t,x,y)\) are crucial to prove the estimate (2) under the additional assumptions \(\frac{(x-1)(y-1)}{t}\ge 1, \ 1<x<2<y, \ t\ge t_0(\mu )\) (see also the part (ii) of Theorem 2).
Proposition 3
Let \(\mu \ge 0\). We have
for \(xy/t\rightarrow \infty \), where \(1<x<2<y, \ t\ge t_0(\mu )\) and \(\frac{(x-1)(y-1)}{t}\ge 1\). It holds also
whenever \(y>x, \ y>2\) and \(xy/t\rightarrow \infty \).
Proof
At first, we deal with \(\mathcal {I}_1^{(\mu )}(t,x,y)\). Since \(y/(t-s)\ge y/t\rightarrow \infty \) for \(1<x<2\) and \(xy/t\rightarrow \infty \), asymptotic behaviour of \(I_{\mu }(z)\) at infinity (8) (for \(n=2\)) enable us to write
or equivalently
where we have also used \(((t-s)/y)^2\lesssim (t/(xy))^2\) for \(1<x<2\). Hence, using exact formula for \(q_{x}^{(1/2)}(s)\) (see (23)) and \(f_{a,b,t}(s)\) [see (7)], we obtain
where
Applying Lemma 1 from Appendix we get
Moreover, due to (47) one can estimate
where the last inequality is a consequence of \(\frac{(x-1)(y-1)}{t}\ge 1\). On the other hand, we apply Lemma 2 from Appendix to calculate \(\mathcal {I}_{1,2}^{(\mu )}\) as follows
Since \(1<x<2<y\) and \(xy/t\rightarrow \infty \) we get \(\frac{x+y-2}{\sqrt{t}}\ge \frac{y-1}{\sqrt{t}}\rightarrow \infty \), which enable us to use (46) and write simply
Here the second line follows from \(y>2>x\) and the estimate (47). The last one is valid, because of the assumptions \(1<x<2\) and \(t\ge t_0(\mu )\). Collecting together (32) and (33) we have
and it is exactly what we needed. To deal with \(\mathcal {I}_2^{(\mu )}(t,x,y)\) we apply (8) (for \(n=1\)) together with exact formula for \(q_{x}^{(1/2)}(s)\) [see (23)] to arrive at
In view of Lemma 2 from Appendix (and the estimate given therein) we can write
where the last estimate follows from assumption \(1<x<2<y\) and inequality (47) given in Appendix. This ends the proof. \(\square \)
Now we are in position to deal with the part (ii) of Theorem 2, i.e. when \(1<x<2<y, \ t\ge t_0(\mu )\) and \(\frac{(x-1)(y-1)}{t}\ge 1\). We follow the notation given in the previous proposition.
Proof of the part (ii) of Theorem 2
It is enough to use the Hunt formula (20) and estimate each component separately. We start with the subtrahend \(r_1^{(\mu )}(t,x,y)\). Namely, using the asymptotic expansion (15) of \(q_x^{(\mu )}(s)\), one can write
Hence, Proposition 3 implies that
The next step is to provide an appropriate expansion for the expression \(p^{(\mu )}(t,x,y)\). Since \(xy/t\rightarrow \infty \), applying the formula (19) for \(n=2\) gives us
where the last equality follows from (48). Then, applying (35) and (34) to the Hunt formula [see (20)], we arrive at
The proof of the part (ii) of Theorem 2 is complete. \(\square \)
Finally, we deal with the last case, namely \(1<x<2<y, \ t\ge t_0(\mu )\) and \(\frac{(x-1)(y-1)}{t}<1\). In this régime of the variables we provide an upper bound for \(p_a^{(\mu )}(t,x,y)\), where \(0\le \mu <1/2\) (see Proposition 4), as well as a lower bound for \(\mu \ge 1/2\) (see Proposition 5). Moreover, the proof of Theorem 1 is complete in this case, since we have for \(0\le \mu <1/2\) [see, (26)]
while for \(\mu \ge 1/2\) it holds [see (27)]
Proposition 4
Let \(0\le \mu \le 1/2\). There exists a constant \(C_1^{(\mu )}>0\) such that
whenever \(xy/t\rightarrow \infty \) and \(1<x<2, \ y>2, t>t_0(\mu )=1,\ \frac{(x-1)(y-1)}{t}\le 1\).
Proof
Notice that for \(0<s<t\) we have
Combining these with
[see (16)] we get
where the last equality follows from Lemma 1 in Appendix. In view of the bound given in (11) we have
which immediately yields [see also definition of \(p^{(\mu )}(t,x,y)\) given in (18)]
The Hunt formula (20) gives us therefore
Notice that the expression in square bracket can be decomposed as the sum of
and
We show that the second one, divided by \(\mathcal {A}_1\), can be bounded from above by t / (xy) (up to some multiplicative positive constant). To achieve this let us denote
and observe that for \(\frac{(x-1)(y-1)}{t}\le 1\) we have (for some \(1<x_0<x<2\))
Since \(g'(x)=-g(x)\frac{\sqrt{y^2+1}-(y-1)}{t}\left[ 1+\frac{t}{x}\frac{\mu +1/2}{\sqrt{y^2+1}-(y-1)}\right] \) we have that \(g'(x)\approx -g(x)\) for \(1<x<2<y, \ t>1\), which implies
Hence, for some \(C_1>0\), we have
Finally, notice that due to (19) we can estimate \(p^{(\mu )}(t,x,y)\) for \(xy/t\rightarrow \infty \) as follows
where \(C_2>0\) is a constant. Thus, the proof is complete with the constant \(C_1^{(\mu )}:=C_1+C_2>0\). \(\square \)
Proposition 5
Let \(\mu >1/2\). There exists a constant \(C_2^{(\mu )}>0\) such that
whenever \(xy/t\rightarrow \infty , \ 1<x<2, \ y>2, t>t_0(\mu )=8/(4\mu ^2-1)\) and \(\frac{(x-1)(y-1)}{t}\le 1\).
Proof
Using (11) to estimate ratios \(I_{\mu }\left( \frac{y}{t-s}\right) /I_{\mu }\left( \frac{y}{t}\right) \) and \(I_{\mu }\left( \frac{y}{t}\right) /I_{\mu }\left( \frac{xy}{t}\right) \) from above, we get
Combining this together with the inequality \(q_x^{(\mu )}(s)\le x^{1/2-\mu }q_x^{(1/2)}(s)\), valid for \(s>0\) and \(\mu \ge 1/2\) [see (17)], we obtain
with the function \(f_{A,B,t}(s)\) defined in (7). We split the last integral into three parts
and estimate them separately. To simplify the notation let us denote
To deal with \(\mathcal {K}_1\) it is sufficient to use Lemma 1 from Appendix. Indeed, we have
Now we focus our attention on the expression \(\mathcal {K}_2\). Observe that for \(s\in (0,t/2)\) we have \(t/(t-s))^{\mu +1/2}-1\approx s/(t-s)\). Hence, making substitution \(w=\frac{1}{s}-\frac{1}{t}\) in the integral \(\mathcal {K}_2\), we can write
where we have substituted \(r=b/(t^2w)\). Since \(b/t\approx y^2/t\rightarrow \infty \) for \(y>2>x\) and \(xy/t\rightarrow \infty \), the last integral behaves like a constant. It implies
To estimate \(\mathcal {K}_3\) from above, let us observe that it holds \((t/(t-s))^{\mu +1/2}-1\approx (t/(t-s))^{\mu +1/2}\), whenever \(t/2<s<t\). Hence, putting once again \(w=\frac{1}{s}-\frac{1}{t}\) we get
Note that since \(a/t\lesssim 1\) we have \(e^{-aw}\approx 1\). Thus, the substitution \(r=b/(t^2w)\) leads to
The last integral is the upper incomplete Gamma function \(\Gamma (\mu ,b/t)\), which behaves like \((b/t)^{\mu -1}e^{-b/t}\) since \(b/t\approx y^2/t\rightarrow \infty \) (see 8.357 in [8]). Therefore, we have
and consequently
due to \(\frac{(x-y)^2}{2t}-\frac{(y-1)^2}{2t}=\frac{-(x-1)(2y-x-1)}{2t}<0\). Combining together (36), (37) and (38) we obtain
for some constants \(C_1, C_2>0\). Hence, from the Hunt formula (20), we obtain
Let us decompose the last expression as a sum \(\mathcal {B}_1+\mathcal {B}_2-C_2\mathcal {B}_3\), where
Observe that the term \(\mathcal {B}_1\) contribute to the leading part for the expansion of \(p_1^{(\mu )}(t,x,y)\). Hence we start with estimating \(\mathcal {B}_2\). One can see that it holds
since \(\frac{(x-1)(y-1)}{t}<1\) and \(1<x<2<y\). Thus, there exists constant \(C_3>0\) such that
On the other hand, we can estimate \(\mathcal {B}_3\) as follows
with the same justification as for \(\mathcal {B}_2.\) This estimate implies that for some \(C_4, C_5>0\) we have
Collecting together (39) and (40) we arrive at
On the other hand, because of (19) for \(n=1\), there exists \(C_6>0\) such that
where \(xy/t\rightarrow \infty \). Putting the last inequality to (41) we get the desired result with the constant \(C_2^{(\mu )}:=C_3+C_5+C_6>0\). \(\square \)
References
Abramowitz, M., Stegun, I.A.: Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables, 9th edn. Dover, New York (1972)
Bogus, K., Małecki, J.: Sharp estimates of transition probability density for Bessel process in half-line. Potential Anal. 43, 1–22 (2015)
Bogus, K., Małecki, J.: Heat kernel estimates for the Bessel differential operator in half-line. Math. Nachr. 289(17–18), 2097–2107 (2016)
Byczkowski, T., Małecki, J., Ryznar, M.: Hitting times of Bessel processes. Potential Anal 38, 753–786 (2013)
Byczkowski, T., Ryznar, M.: Hitting distribution of geometric Brownian motion. Studia Math. 173, 19–38 (2006)
Davies, E.B.: The equivalence of certain heat kernel and Green function bounds. J. Funct. Anal. 71, 88–103 (1987)
Davies, E.B.: Heat kernels and spectral theory (Cambridge Tracts in Mathematics), vol. 92, p. 208. Cambridge Univ. Press, Cambridge (1990)
Gradshteyn, I.S., Ryzhik, I.M.: Table of integrals, series, and products, 7th edn, p. 1171. Academic Press, California (2007)
Gyrya, P., Saloff-Coste, L.: Neumann and Dirichlet heat kernels in inner uniform domains, viii+144p., Société Mathématique de France 336, Paris (2011)
Hamana, Y., Matsumoto, H.: The probability distributions of the first hitting times of Bessel processes. Trans. Am. Math. Soc. 365, 5237–5257 (2013)
Laforgia, A.: Bounds for modified Bessel functions. J. Comput. Appl. Math. 34, 263–267 (1991)
Małecki, J., Serafin, G., Żórawik, T.: Fourier-Bessel heat kernel estimates. J. Math. Anal. Appl. 439(1), 91–102 (2016)
Małecki, J., Serafin, G.: Dirichlet heat kernel for the Laplacian in a ball. Potential Anal. (2018). https://doi.org/10.1007/s11118-018-9750-2
Revuz, D., Yor, M.: Continuous martingales and Brownian motion, 3rd edn, p. 542. Springer, New York (1999)
Serafin, G.: Exit times densities of Bessel process. Proc. Am. Math. Soc. 145, 3165–3178 (2017)
Takemura, T.: Elementary solutions of Bessel processes with boundary conditions. Ann. Rep. Grad. Sch. Hum. Sci. Nara Women Univ. 23, 265–278 (2008)
Uchiyama, K.: Asymptotics of the densities of the first passage time distributions for Bessel diffusions. Trans. Am. Math. Soc. 367(4), 2719–2742 (2015)
Uchiyama, K.: Density of space-time distribution of Brownian first hitting of a disc and a ball. Potential Anal. 44, 497–541 (2016)
Varadhan, S.R.S.: On the behavior of the fundamental solution of the heat equation with variable coefficients. Comm. Pure Appl. Math. 20(2), 431–455 (1967)
Varadhan, S.R.S.: Diffusion processes in a small time interval. Comm. Pure Appl. Math. 20(4), 659–685 (1967)
Zhang, Q.S.: The boundary behavior of heat kernels of Dirichlet Laplacians. J. Differ. Equ. 182, 416–430 (2002)
Zhang, Q.S.: The global behavior of heat kernels in exterior domains. J. Funct. Anal. 200(1), 160–176 (2003)
Acknowledgements
The author is very grateful to Jacek Małecki and Grzegorz Serafin for critical remarks and comments which improve the presentation of the paper.
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.
The project was funded by the National Science Centre (Poland) Grant ETIUDA No. 2016/20/T/ST1/00257.
Appendix
Appendix
In this section we collect three lemmas, which play crucial rôle in the proofs of the main results of the paper. To make the paper more self-contained we recall the following formula involving the modified Bessel function of the second kind \(K_{\mu }(z)\) (see 3.471. 9 in [8])
which is valid for \(c,d>0\) and \(\mu \in \mathbf {R}\). Moreover, since \(K_{\mu }(z)\approx z^{-1/2}e^{-z}\) at infinity (9), we obtain for \(1\lesssim cd\) that
On the other hand, if we put \(\mu =1/2\) in (43), then we have
The last equality will be used to prove Lemma 1 given below. Namely, we derive an explicit formula for the integral \(\int _0^t f_{A,B,t}(s)ds\) defined in (7).
Lemma 1
Let \(t>0\). Then we have
Proof
Making the substitution \(w=\frac{1}{s}-\frac{1}{t}\), the last integral takes a form
Using formula (45) we obtain thesis. \(\square \)
Next Lemma corresponds to \(\int _0^t s f_{A,B,t}(s)ds\). We derive exact formula for this integral in terms of the error function.
Lemma 2
Let \(c,d>0\) and \(t>0\). We have
where \({{\,\mathrm{erf}\,}}(z)=\frac{2}{\sqrt{\pi }}\int _z^{\infty }\exp {(-u^2)}du, \ z>0,\) denotes the error function. Moreover if \((\sqrt{c}+\sqrt{d})/\sqrt{t}\rightarrow \infty \) then we get
Proof
Let \(C>0\) be given arbitrary. Integrating the equality from Lemma 1 with respect to c on interval \((C,\infty )\), we have
Since the integrand on the left-hand side (LHS) is positive, we can use the Fubini–Tonelli theorem to change the order of integration as follows
On the other hand, making substitution \(w=\frac{\sqrt{c}+\sqrt{d}}{\sqrt{t}}\) in the right-hand side (RHS) we have
Comparing LHS with RHS give us thesis of the first part of Lemma 2. The second part follows from the asymptotic behaviour of the error function at infinity, namely we have \(1-{{\,\mathrm{erf}\,}}(z)\approx e^{-z^2}/z\) as \(z\rightarrow \infty \). \(\square \)
The last lemma contains some estimates of exponents, which play important rôle especially in the proof of Proposition 4.
Lemma 3
Fix \(C>0\) and assume that \(x,y>1\) and \(t>0\). Then it holds that
while for \(\frac{(x-1)(y-1)}{t}\ge C\) we have
Proof
Since
and the right-hand side behaves like positive constant for \(\frac{(x-1)(y-1)}{t}\ge C>0\), the proof of (48) is complete. To prove (47) it is enough to write
and simply estimate denominator from below by using the inequality \(e^z-1\ge z\), which is true for \(z>0\). \(\square \)
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Cite this article
Bogus, K. Asymptotic behaviour of the Bessel heat kernels. Math. Z. 293, 1601–1621 (2019). https://doi.org/10.1007/s00209-019-02339-0
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00209-019-02339-0
Keywords
- Bessel differential operator
- Heat kernel
- Asymptotic expansion
- Half-line
- Bessel process
- Transition probability density