Abstract
The existence condition \(H<1/d\) for first-order derivative of self-intersection local time for \(d\ge 3\) dimensional fractional Brownian motion was obtained in Yu (J Theoret Probab 34(4):1749–1774, 2021). In this paper, we establish a limit theorem under the nonexistence critical condition \(H=1/d\).
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Consider a d-dimensional fractional Brownian motion (fBm) with Hurst parameter \(H\in (0,1)\), which is a d-dimensional centered Gaussian process \(B^H=\{B_t^H, ~t\ge 0\}\) with component processes being independent copies of a 1-dimensional centered Gaussian process \(B^{H,i}\), \(i=1,2,\ldots ,d\) and the covariance function given by
Note that \(B_t^{\frac{1}{2}}\) is a classical standard Brownian motion. Let \(D=\{(r,s): 0<r<s<t\}\). The self-intersection local time (SLT) of fBm was first investigated in Rosen [13] and formally defined as
where \(B^H\) is a 2-dimensional fBm and \(\delta \) is the Dirac delta function. It was further investigated in Hu [4], Hu and Nualart [6]. In particular, Hu and Nualart [6] showed its existence whenever \(Hd<1\). Moreover, \(\alpha _t(y)\) is Hölder continuous in time of any order strictly less than \(1-H\) which can be derived from Xiao [15].
The derivative of self-intersection local time (DSLT) for fBm was first considered in the works by Yan et al. [16, 17], where the ideas were borrowed form Rosen [14]. The DSLT for fBm has two versions. One is extended by the Tanaka formula (see in Jung and Markowsky [9]):
The other is from the occupation-time formula (see Jung and Markowsky [10]):
Motivated by the first-order DSLT for fBm in Jung and Markowsky [10] and the k-th-order derivative of intersection local time (ILT) for fBm in Guo et al. [3], we will consider the following k-th-order DSLT for fBm in this paper,
where \(k=(k_1,\ldots ,k_d)\) is a multi-index with all \(k_i\) being nonnegative integers and \(|k|=k_1+k_2+\cdots +k_d\), \(\delta \) is the Dirac delta function of d variables and \(\delta ^{(k)}(y)=\frac{\partial ^k}{\partial _{y_1}^{k_1}\ldots \partial _{y_d}^{k_d}}\delta (y)\) is the k-th-order partial derivative of \(\delta \).
Set
where \(\langle p,x\rangle =\sum _{j=1}^dp_jx_j\) and \(|p|^2=\sum _{j=1}^dp_j^2\).
Since the Dirac delta function \(\delta \) can be approximated by \(f_\varepsilon (x)\), we approximate \(\delta ^{(k)}\) and \(\widehat{\alpha }_t^{(k)}(y)\) by
and
respectively.
If \(\widehat{\alpha }^{(k)}_{t,\varepsilon }(y)\) converges to a random variable in \(L^p\) as \(\varepsilon \rightarrow 0\), we denote the limit by \(\widehat{\alpha }_t^{(k)}(y)\) and call it the k-th DSLT of \(B^H\).
Recently, Yu [18] studied the existence and Hölder continuity conditions of \(\widehat{\alpha }_t^{(k)}(y)\) and related limit theorem in critical case. We recall the existence condition for \(\widehat{\alpha }_t^{(k)}(y)\) in \(L^2\) as follows.
Theorem 1.1
[18] For \(0<H<1\) and \(\widehat{\alpha }^{(k)}_{t,\varepsilon }(y)\) defined in (1.1), let \(\#:=\#\{k_i ~is ~odd, ~i=1, 2, \ldots d\}\) denotes the odd number of \(k_i\), for \(i=1, 2, \ldots , d\). If \(H<\min \{\frac{2}{2|k|+d},\frac{1}{|k|+d-\#}, \frac{1}{d}\}\) for \(|k|=\sum _{j=1}^dk_j\), then \(\widehat{\alpha }^{(k)}_{t}(0)\) exists in \(L^2\).
Note that, if \(|k|=1\), the existence condition of \(\widehat{\alpha }^{(k)}_{t}(0)\) is \(H<1/d\), and \(Hd=1\) is the critical condition of \(\widehat{\alpha }_t^{(k)}(y)\) for any \(d\ge 2\). When \(Hd=1\) for \(d=2\), Markowsky [11] proved the limit theorem for \(|k|=1\).
Theorem 1.2
[11] \(\widehat{\alpha }^{(k)}_{t,\varepsilon }(y)\) is defined in (1.1) with \(y=0\). Suppose that \(H=\frac{1}{2}\), \(d=2\) and \(|k|=1\), then as \(\varepsilon \rightarrow 0\),
In this paper, we will consider the case of \(Hd=1\) for any \(d\ge 3\) and \(|k|=1\), and prove a limit theorem for \(\widehat{\alpha }^{(k)}_{t,\varepsilon }(0)\). Without loss of generality, we assume that \(k_1=1,k_2=0,\ldots ,k_d=0\) for the multi-index \(k=(k_1,\ldots ,k_d)\), and for the convenience of writing, we will abbreviate \(\widehat{\alpha }^{(1,0,\ldots ,0)}_{t,\varepsilon }(0)\) as \(\widehat{\alpha }^{(1)}_{t,\varepsilon }(0)\) in the subsequent content of this paper without causing confusion.
Theorem 1.3
\(\widehat{\alpha }^{(k)}_{t,\varepsilon }(y)\) is defined in (1.1) with \(y=0\). Suppose that \(Hd=1\) for any \(d\ge 3\) and \(|k|=1\). Then, as \(\varepsilon \rightarrow 0\), we have
where \(\sigma ^2=\frac{2Ht^{3-4H}}{(2\pi )^d(1-2H)^2}\).
When \(|k|=1\), under the condition \(H>1/d\), the behavior of \(\widehat{\alpha }^{(1)}_{t,\varepsilon }(0)\) as \(\varepsilon \rightarrow 0\) is also of interest. One would expect a central limit theorem to exist, but this remains unproved. Nevertheless, we venture the following conjecture
(1) If \(H=\frac{2}{d+2}>\frac{1}{d}\) and \(d\ge 3\), \((\log \frac{1}{\varepsilon })^{\gamma _1(H)}\widehat{\alpha }^{(1)}_{t,\varepsilon }(0)\) converges in distribution to a normal law for some \(\gamma _1(H)<0\);
(2) If \(H>\frac{1}{2}\ge \frac{2}{d+2}\) and \(d\ge 2\), \(\varepsilon ^{\gamma _2(H)}\widehat{\alpha }^{(1)}_{t,\varepsilon }(0)\) converges in distribution to a normal law for some \(\gamma _2(H)>0\);
(3) If \(\frac{2}{d+2}<H<\frac{1}{2}\) and \(d\ge 3\), \(\varepsilon ^{\gamma _3(H)}(\log \frac{1}{\varepsilon })^{\gamma _4(H)}\widehat{\alpha }^{(1)}_{t,\varepsilon }(0)\) converges in distribution to a normal law for some \(\gamma _3(H)>0\) and \(\gamma _4(H)<0\).
The paper has the following structure. We state some preliminary lemmas in Sect. 2. Section 3 is to prove the main result. Throughout this paper, if not mentioned otherwise, the letter C, with or without a subscript, denotes a generic positive finite constant and may change from line to line.
2 Preliminaries
In this section, we present two basic lemmas, which will be used in Sect. 3. The first lemma gives the bounds on the quantity of \(\lambda \rho -\mu ^2\), which could be obtained from the Appendix B in [9] or Lemma 3.1 in [4]. In fact, \(\lambda , \rho \) and \(\mu \) represent the three quantities of the covariance matrix of the increment of fBm, and the bound estimation of \(\lambda \rho -\mu ^2\) is beneficial for the subsequent calculation of the convergence of multiple integrals, which will bring a lot of convenience to the proof in Sect. 3.
Lemma 2.1
Let
and
Case (i) Suppose that \(D_1=\{(r,r',s,s')\in [0,t]^4 ~|~ r<r'<s<s'\}\), let \(r'-r=a\), \(s-r'=b\), \(s'-s=c\). Then, there exists a constant \(K_1\) such that
and
Case (ii) Suppose that \(D_2=\{(r,r',s,s')\in [0,t]^4 ~|~ r<r'<s'<s\}\), let \(r'-r=a\), \(s'-r'=b\), \(s-s'=c\). Then, there exists a constant \(K_2\) such that
and
Case (iii) Suppose that \(D_3=\{(r,r',s,s')\in [0,t]^4 ~|~ r<s<r'<s'\}\), let \(s-r=a\), \(r'-s=b\), \(s'-r'=c\). Then, there exists a constant \(K_3\) such that
and
The second lemma shows the Wiener chaos expansion of \(\widehat{\alpha }^{(k)}_{t,\varepsilon }(0)\) with \(|k|=1\). Before that, we need to explain some notations. We will denote by \(\mathcal {H}\) the Hilbert space obtained by taking the completion of the space of step functions endowed with the inner product
where \(B^{H,1}\) is a 1-dimensional fBm. The mapping \(\mathbbm {1}_{[0,t]}\rightarrow B_t^{H,1}\) can be extended to a linear isometry between \(\mathcal {H}\) and a Gaussian subspace \(L^2(\Omega , \mathcal {F}, \mathbb {P})\). For any integer \(q\in \mathbb {N}\), we denote by \(\mathcal {H}^{\otimes q}\) and \(\mathcal {H}^{\odot q}\) the q-th tensor product of \(\mathcal {H}\), and the q-th symmetric tensor product of \(\mathcal {H}\), respectively.
Similarly, for d-dimensional fBm \(B^H=(B^{H,1},\ldots ,B^{H,d})\), we can define corresponding Hilbert space \(\mathcal {H}^d\) and tensor product spaces \((\mathcal {H}^d)^{\otimes q}\) and \((\mathcal {H}^d)^{\odot q}\). If \(h=(h^1,\ldots , h^d)\in \mathcal {H}^d\), we set \(B^H(h)=\sum _{j=1}^dB^{H,j}(h^j)\). Then \(h\mapsto B^H(h)\) is a linear isometry between \(\mathcal {H}^d\) and the Gaussian subspace of \(L^2(\Omega _1\times \cdots \times \Omega _d, \mathcal {F}_1\times \cdots \times \mathcal {F}_d, \mathbb {P}_1\times \cdots \times \mathbb {P}_d)\) generated by \(B^H\). The q-th Wiener chaos of \(L^2(\Omega _1\times \cdots \times \Omega _d, \mathcal {F}_1\times \cdots \times \mathcal {F}_d, \mathbb {P}_1\times \cdots \times \mathbb {P}_d)\), denoted by \(\mathfrak {H}_q\), is the closed subspace of \(L^2(\Omega _1\times \cdots \times \Omega _d, \mathcal {F}_1\times \cdots \times \mathcal {F}_d, \mathbb {P}_1\times \cdots \times \mathbb {P}_d)\) generated by the variables
where \(H_q\) is the q-th Hermite polynomial, defined by
For \(q\in \mathbb {N}\), \(q\ge 1\) and \(h\in \mathcal {H}^d\) of the form \(h=(h^1,\ldots ,h^d)\) with \(\Vert h^j\Vert _{\mathcal {H}}=1\), we can write
For such h, we define the mapping
where \(q_j(i_1,\ldots ,i_q)\) denotes the number of indices in \((i_1,\ldots ,i_q)\) equal to j. The range of \(I_q\) is contained in \(\mathfrak {H}_q\). This mapping provides a linear isometry between \((\mathcal {H}^d)^{\odot q}\) (equipped with the norm \(\sqrt{q!}\Vert \cdot \Vert _{(\mathcal {H}^d)^{\otimes q}}\)) and \(\mathfrak {H}_q\) (equipped with the \(L^2\)-norm). Here multiple stochastic integral \(I_n\) is the d-dimensional version see in Jaramillo and Nualart [8] (or in Flandoli and Tudor [2]).
It also holds that \(I_n(f)=I_n(\widetilde{f})\), where \(\widetilde{f}\) denotes the symmetrization of f. We recall that any square integrable random variable F which is measurable with respect to the \(\sigma \)-algebra generated by \(B^H\) can be expanded into an orthogonal sum of multiple stochastic integrals
where \(f_n\in (\mathcal {H}^d)^{\odot n}\) are (uniquely determined) symmetric functions and \(I_0(f)=\mathbb {E}(F)\).
The proof process of Wiener chaos expansion also requires the knowledge of Malliavin derivative \(\mathbb {D}\) with respect to fBm \(B^H\). Denote by \(C_b^{\infty }(\mathbb {R}^n)\) the space of bounded smooth functions on \(\mathbb {R}^n\). Consider the space of random variables
The Malliavin derivative of \(F\in \mathcal {S}\), denoted by \(\mathbb {D}F\), is given by
By iteration, we can define the n-th derivatives \(D^n\) for every \(n\ge 2\), which is an element of \(L^2(\Omega ,(\mathcal {H}^d)^{\otimes n})\). For example, we write for the smooth function f,
where \(h_j=\mathbbm {1}_{[0,t]}e_j, j=1,2,\ldots ,d\) and
Similarly,
where \(\mathbbm {1}_{[0,t]}e_{i_j}\in \mathcal {H}^d\), \(\bigotimes _{j=1}^n(\mathbbm {1}_{[0,t]}e_{i_j})\in (\mathcal {H}^d)^{\otimes n}, i_j\in \{1,2,\ldots ,d\}, j=1,2,\ldots ,n\).
More detailed introductions to Malliavin derivative and multiple stochastic integral can be found in Nualart [12], Hu [5] and the references therein.
Lemma 2.2
Let \(\widehat{\alpha }^{(k)}_{t,\varepsilon }(y)\) be defined in (1.1), then we have the Wiener chaos expansion for \(k=(1,0\ldots ,0)\),
(i) If \(d=2\), \(f_{2q-1,\varepsilon }\) is the element of \((\mathcal {H}^2)^{\otimes (2q-1)}\) given by
where \(\beta _q=\frac{(-1)^q}{2\pi (2q-1)!}\sum _{q_1+q_2=q, q_1\ge 1}\frac{(2q-1)!(2q_1)!}{(2q_1-1)!(q_1)!(q_2)!2^q}\) and \(\mathbbm {1}_{[r,s]}e_{i_j}\in \mathcal {H}^2, i_j\in \{1,2\}, j=1,2,\ldots ,2q-1\), (\(e_{i_j}\) defined in (2.2)).
(ii) If \(d\ge 3\), \(f_{2q-1,\varepsilon }\in (\mathcal {H}^d)^{\otimes (2q-1)}\)
where \(\beta _{q,d}=\frac{(-1)^q}{(2q-1)!(2\pi )^{d/2}}\sum _{q_1+\cdots +q_d=q, q_1\ge 1}\frac{(2q-1)!(2q_1)!}{(2q_1-1)!(q_1)!\cdots (q_d)!2^q}\) and \(\mathbbm {1}_{[r,s]}e_{i_j}\in \mathcal {H}^d, i_j\in \{1,2,\ldots ,d\}, j=1,2,\ldots ,2q-1\).
Proof
The proof adopts a method similar to Lemma 7 in Hu and Nualart [6] (or the Appendix A in Das and Markowsky [1]).
(i) For the case \(d=2\), by Stroock’s formula,
where \(f_{2q-1,\varepsilon }\in (\mathcal {H}^2)^{\otimes (2q-1)}\) and
with \(x_j\in [0,t]\) for all \(j=1,2,\ldots ,n\).
Let \(i_j\in \{1,2\}\) for all \(j=1,2,\ldots ,n\). Then by (2.3), we can compute the expectation
where \((\mathbbm {1}_{[r,s]}e_{i_j})(x_j)\in \mathcal {H}^2, i_j\in \{1,2\}, j=1,2,\ldots ,n\), (\(e_{i_j}\) defined in (2.2)) and
with the independent identical distribution standard Gaussian random variables \(X_{i_n}\) and
Then, for \(n=2q-1=2(q_1+q_2)-1\) with the number of \(i_k=1\) is \(2q_1-1\) and the number of \(i_k=2\) is \(2q_2\), the summation
where \(\#\{i_k=x\}\) denotes the number of \(i_k=x\). This gives
Thus, we have
where \(\beta _q=\frac{(-1)^q}{2\pi (2q-1)!}\sum _{q_1+q_2=q, q_1\ge 1}\frac{(2q-1)!(2q_1)!}{(2q_1-1)!(q_1)!(q_2)!2^q}\).
(ii) Similarly, we can prove the case of \(d\ge 3\).
where \(f_{2q-1,\varepsilon }\in (\mathcal {H}^d)^{\otimes (2q-1)}\) and
with \((\mathbbm {1}_{[r,s]}e_{i_j})(x_j)\in \mathcal {H}^d, i_j\in \{1,2,\ldots ,d\}, j=1,2,\ldots ,n\).
Note that
Then, for \(n=2q-1=2(q_1+\cdots +q_d)-1\) with the number of \(i_k=1\) is \(2q_1-1\) and the number of \(i_k=\ell \) is \(2q_{\ell }\), the summation
This gives
Thus,
where \(\beta _{q,d}=\frac{(-1)^q}{(2q-1)!(2\pi )^{d/2}}\sum _{q_1+\cdots +q_d=q, q_1\ge 1}\frac{(2q-1)!(2q_1)!}{(2q_1-1)!(q_1)!\cdots (q_d)!2^q}\). \(\square \)
Lemma 2.3
If \(Hd=1\), as \(\varepsilon \rightarrow 0\), we have
(i)
and
(ii)
Proof
For (i), by L’Hôspital’s rule, we have
where we use the condition \(Hd=1\) in the second equality.
For (ii), take the variable transformation \(x=y\varepsilon ^{\frac{1}{2H}}\),
where we use L’Hôspital’s rule and the condition \(Hd=1\) in the second equality.
\(\square \)
3 Proof of Theorem 1.3
In this section, the proof of Theorem 1.3 is taken into account, we will consider the case of \(Hd=1\) for any \(d\ge 3\) and \(|k|=1\). By Lemma 2.2, \(\widehat{\alpha }^{(1)}_{t,\varepsilon }(0)\) has the following chaos decomposition
where
with \(D=\{(r,s): 0<r<s<t\}\), where
For \(q=1\),
where \(\mathcal {H}^d\) is the Hilbert space obtained by taking the completion of the step functions (see in Sect. 2).
For \(q>1\), we have to describe the terms \(\langle f_{2q-1,\varepsilon ,s_1,r_1},f_{2q-1,\varepsilon ,s_2,r_2}\rangle _{(\mathcal {H}^d)^{\otimes (2q-1)}}\), where \((\mathcal {H}^d)^{\otimes (2q-1)}\) is the \((2q-1)\)-th tensor product of \(\mathcal {H}^d\). For every \(x, ~u_1, ~u_2>0\), we define
For \(j=1,2,\ldots ,2q-1\), \(i_j\in \{1,2,\ldots ,d\}\),
Then, we have
where
Note that equations (3.2)–(3.4) here can degenerate into the case \(d=1\) of the equations (2.18)–(2.19) in Jaramillo and Nualart [7].
Before completing the proof of the main result, we give some useful lemmas below.
Lemma 3.1
where \(\sigma ^2=\frac{2Ht^{3-4H}}{(2\pi )^d(1-2H)^2}\).
Proof
Let \((X,Y)\in \mathbb {R}\times \mathbb {R}\) be a jointly Gaussian vector with mean zero and covariance \(A=(A_{i,j})_{i,j=1,2}\), let \(f_A\) is the density of (X, Y) and \(f_{1,\varepsilon }(x)=\frac{1}{\sqrt{2\pi \varepsilon }}e^{-\frac{x^2}{2\varepsilon }}, x\in \mathbb {R}\) be a 1-dimensional density function. Then,
where \(\widetilde{A}:=(\varepsilon ^{-1}I+A^{-1})^{-1}\), \(f_{\widetilde{A}}\) denotes the density of a Gaussian vector with mean zero and covariance \(\widetilde{A}=(\widetilde{A}_{i,j})_{i,j=1,2}\).
Similarly, let \(\Sigma =(\Sigma _{i,j})_{i,j=1,2}\) be the covariance matrix of \((B^{H,1}_s-B^{H,1}_r,B^{H,1}_{s'}-B^{H,1}_{r'})\), and \(\Sigma ^{d-1}\) be the covariance matrix of \((\widetilde{B}^{H}_s-\widetilde{B}^{H}_r,\widetilde{B}^{H}_{s'}-\widetilde{B}^{H}_{r'})\) (\(\widetilde{B}^H\) denotes the \((d-1)\)-dimensional fBm). The notations \(f_{\Sigma }\) and \(f_{\Sigma ^{d-1}}\) represent their density functions, respectively. It is easy to find that \(\Sigma ^{d-1}\) is a block diagonal matrix, and that the dimension of it is \(2(d-1)\times 2(d-1)\). Then,
Thus, for any Gaussian vector \((X, Y)\in \mathbb {R}^d\times \mathbb {R}^d\) and k-th (\(k=(1,0,\ldots ,0)\)) order derivative, we have
where \(\widetilde{x}=(x_2,\ldots ,x_d), \widetilde{y}=(y_2,\ldots ,y_d)\).
Thus,
with
where \(D_i\) (i=1, 2, 3) defined in Lemma 2.1 and \(\Sigma \) is a covariance matrix with \(\Sigma _{1,1}=\lambda \), \(\Sigma _{2,2}=\rho \), \(\Sigma _{1,2}=\mu \) given in Lemma 2.1.
Next, we will split the proof into three parts to consider \(V_1(\varepsilon )\), \(V_2(\varepsilon )\) and \(V_3(\varepsilon )\), respectively.
For the \(V_1(\varepsilon )\) term, changing the coordinates \((r, r', s, s')\) by \((r, a=r'-r, b=s-r', c=s'-s)\) and integrating the r variable, we get
Applying Lemma 2.1 Case (i), for some \(C>0\), we get
where we use the Young’s inequality in the second to last inequality.
Substituting (3.6) and
into the integrand of \(\widetilde{V_1}(\varepsilon )\),
where we make the change of variable \(c=c\,\varepsilon ^{-\frac{1}{H}}\) in the last inequality.
By L’Hôspital’s rule, we have
where we have used Lemma 2.3 in the last equality.
So, we can obtain
For the \(V_2(\varepsilon )\) term, changing the coordinates \((r, r', s, s')\) by \((r, a=r'-r, b=s'-r', c=s-s')\) and integrating the r variable, then by (3.5), we get
By Lemma 2.1 Case (ii),
Then, we have
Next, we need to estimate this integral over the regions \(\{b\le (a\vee c)\}\) and \(\{b > (a\vee c)\}\) separately, and denote these two integrals by \(\widetilde{V_{2,1}}(\varepsilon )\) and \(\widetilde{V_{2,2}}(\varepsilon )\), respectively. Note that
If \(b\le (a\vee c)\), we choose \(|\mu |\le b^{2H}\). Thus,
where we have used Lemma 2.3 in the last equality and the following fact
If \(b>(a\vee c)\), we choose \(|\mu |\le 2Hb(a\wedge c)^{2H-1}\). Similarly, we have
where we have used the following fact
So, by the above result, we can obtain
For the \(V_3(\varepsilon )\) term.
By changing the coordinates \((r, r', s, s')\) by \((r, a=s-r, b=r'-s, c=s'-r')\), then from (3.2) and Lemma 2.1 Case (iii), we can write
and \(|\varepsilon I+\Sigma |=\varepsilon ^2+\varepsilon (a^{2\,H}+c^{2\,H})+(ac)^{2\,H}-\mu (a+b,a,c)^2\). It is not hard to see that
where we change the coordinates (a, b, c) by \((\varepsilon ^{-\frac{1}{2H}}a, b, \varepsilon ^{-\frac{1}{2H}}c)\) in the last equality. Denote
where \(O_{\varepsilon ,3}=\{[0,t\varepsilon ^{-\frac{1}{2H}}]^2\times [(\log \frac{1}{\varepsilon })^{-1},t]\}\).
We conclude that
Indeed,
where we use
in the first inequality and use \(\mu (a,a,c)\le (a\wedge c)^{2H}\),
in the second inequality.
By the definition of \(\mu (a+b,a,c)\), it is easy to find
and use the Taylor’s theorem for integrand,
Similarly, the denominator of the integrand in \(V_3(\varepsilon )\) can be rewritten as
It is easy to see that
and
Then, by L’Hôspital’s rule, we have
Together (3.7), (3.8), (3.9) and (3.11), we can see
\(\square \)
Lemma 3.2
For \(I_1(f_{1,\varepsilon })\) given in (3.1), then
Proof
From (3.1), we can find
where \(V_i^{(1)}(\varepsilon )=2\int _{D_i}\langle f_{1,\varepsilon ,s_1,r_1},f_{1,\varepsilon ,s_2,r_2}\rangle _{\mathcal {H}^d}dr_1dr_2ds_1ds_2\) for \(i=1, 2, 3\), and \(\langle f_{1,\varepsilon ,s_1,r_1},f_{1,\varepsilon ,s_2,r_2}\rangle _{\mathcal {H}^d}\) was defined in (3.4). Then we have
Combining (3.13) with (3.7) and (3.8), we can see
Thus, we only need to consider \(\Big (\varepsilon ^{-\frac{1}{H}}\log 1/\varepsilon \Big )^{2H-1}V_3^{(1)}(\varepsilon )\) as \(\varepsilon \rightarrow 0\).
By (3.1), (3.4) and (3.12) we have
Note that
and
Similar to (3.9), we get
where
According to (3.10) and (3.11), we can find that
where we use \(\beta _{1,d}^2=\frac{1}{(2\pi )^d}\) in the second equality.
Thus,
\(\square \)
Proof of Theorem 1.3
we can see
Since \(I_1(f_{1,\varepsilon })\) is Gaussian, we have, as \(\varepsilon \rightarrow 0\),
Thus,
as \(\varepsilon \rightarrow 0\). This completes the proof. \(\square \)
Data Availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.
References
Das, K., Markowsky, G.: Existence, renormalization, and regularity properties of higher order derivatives of self-intersection local time of fractional Brownian motion. Stoch. Anal. Appl. 40(1), 133–157 (2022)
Flandoli, F., Tudor, C.A.: Brownian and fractional Brownian stochastic currents via Malliavin calculus. J. Funct. Anal. 258(1), 279–306 (2010)
Guo, J., Hu, Y., Xiao, Y.: Higher-order derivative of intersection local time for two independent fractional Brownian motions. J. Theoret. Probab. 32(3), 1190–1201 (2019)
Hu, Y.: Self-intersection local time of fractional Brownian motions-via chaos expansion. J. Math. Kyoto Univ. 41(2), 233–250 (2001)
Hu, Y.: Analysis on Gaussian Spaces. World Scientific Publishing Co. Pte. Ltd., Hackensack (2017)
Hu, Y., Nualart, D.: Renormalized self-intersection local time for fractional Brownian motion. Ann. Probab. 33(3), 948–983 (2005)
Jaramillo, A., Nualart, D.: Asymptotic properties of the derivative of self-intersection local time of fractional Brownian motion. Stoch. Process. Appl. 127(2), 669–700 (2017)
Jaramillo, A., Nualart, D.: Functional limit theorem for the self-intersection local time of the fractional Brownian motion. Ann. Inst. H. Poincaré Probab. Statist. 55(1), 480–527 (2019)
Jung, P., Markowsky, G.: On the Tanaka formula for the derivative of self-intersection local time of fractional Brownian motion. Stoch. Process. Appl. 124(11), 3846–3868 (2014)
Jung, P., Markowsky, G.: Hölder continuity and occupation-time formulas for fBm self-intersection local time and its derivative. J. Theoret. Probab. 28(1), 299–312 (2015)
Markowsky, G.: Renormalization and convergence in law for the derivative of intersection local time in \(R^2\). Stoch. Process. Appl. 118(9), 1552–1585 (2008)
Nualart, D.: The Malliavin Calculus and Related Topics, 2nd edn. Springer, Secaucus (2006)
Rosen, J.: The intersection local time of fractional Brownian motion in the plane. J. Multivar. Anal. 23(1), 37–46 (1987)
Rosen, J.: Derivatives of self-intersection local times. Séminaire de Probabilités XXXVIII 263–281 (2005)
Xiao, Y.: Hölder conditions for the local times and the Hausdorff measure of the level sets of Gaussian random fields. Probab. Theory Related Fields. 109(1), 129–157 (1997)
Yan, L., Yang, X., Lu, Y.: \(p\)-Variation of an integral functional driven by fractional Brownian motion. Statist. Probab. Lett. 78(9), 1148–1157 (2008)
Yan, L., Liu, J., Yang, X.: Integration with respect to fractional local time with Hurst index \(1/2<H<1\). Potential Anal. 30(2), 115–138 (2009)
Yu, Q.: Higher-order derivative of self-intersection local time for fractional Brownian motion. J. Theoret. Probab. 34(4), 1749–1774 (2021)
Acknowledgements
The authors are grateful to the anonymous referees and editors for their insightful and valuable comments.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Q. Yu is partially supported by the Fundamental Research Funds for the Central Universities (NS2022072), National Natural Science Foundation of China (12201294) and Natural Science Foundation of Jiangsu Province, China (BK20220865). X. Yu is supported by National Natural Science Foundation of China (12071493).
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Yu, Q., Yu, X. Limit Theorem for Self-intersection Local Time Derivative of Multidimensional Fractional Brownian Motion. J Theor Probab 37, 2054–2075 (2024). https://doi.org/10.1007/s10959-023-01300-6
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10959-023-01300-6