Asymptotic behavior of solutions to a planar Hartree equation with isolated singularities Thanks: ∗ Minbo Yang was partially supported by the National Key Research and Development Program of China (No. 2022YFA1005700), National Natural Science Foundation of China (12471114) and Natural Science Foundation of Zhejiang Province (LZ22A010001). Thanks: *Corresponding author. Tao Feng, Minbo Yang, Xianmei Zhou* Abstract . In this paper we investigate the isolated singularities of the Hartree type equation − Δ ​ u ​ ( x ) = ( 1 | x | α ∗ e u ) ​ e u ⁡ ( x ) in ​ B 1 ∖ { 0 } , -\Delta u(x)=\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\quad\text{in }B_{1}\setminus\{0\}, where α > 0 \alpha>0 , 1 | x | α ∗ e u ≜ ∫ B 1 ∖ { 0 } e u ​ ( y ) | x − y | α ​ 𝑑 y \displaystyle\frac{1}{|x|^{\alpha}}*e^{u}\triangleq\int_{B_{1}\setminus\{0\}}\frac{e^{u}(y)}{|x-y|^{\alpha}}dy , and the punctured ball B 1 ∖ { 0 } ⊂ ℝ 2 B_{1}\setminus\{0\}\subset\mathbb{R}^{2} . Under the finite total curvature condition, by establishing a representation formula for singular solutions, we obtain the asymptotic behavior of the solutions near the origin. We also extend this asymptotic behavior results to the case with a general non-negative coefficient K ⁡ ( x ) K(x) , and to the higher-order Hartree-type equations in any dimension n ≥ 3 n\geq 3 . 2020 Mathematics Subject Classification 35A21; 35B40; 35J91 1. Introduction and main results In the present paper, we are interested in the following planar Hartree type equation with an isolated singularity at the origin: (1.1) -\Delta u(x)=\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\quad\text{in }B_{1}\setminus\{0\}\subset\mathbb{R}^{2}, where \alpha>0 , B_{1}\setminus\{0\} is the punctured ball in \mathbb{R}^{2} and the convolution \displaystyle\frac{1}{|x|^{\alpha}}*e^{u}\triangleq\int_{B_{1}\setminus\{0\}}\frac{e^{u}(y)}{|x-y|^{\alpha}}dy . We mainly consider the singular solutions u\in C^{2}(B_{1}\setminus\left\{0\right\}) for nonlocal equation ( 1.1 ) under the finite total curvature condition (1.2) \int_{B_{1}\setminus\{0\}}\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\mathrm{d}x<+\infty. Elliptic equations involving the exponential nonlinearity and/or the isolated singularity at the origin have been widely studied in recent decades, since it plays an essential role in various geometric and physical problems. Let (M,g) be a complete Riemannian manifold and K be a given function on M . The Gaussian curvature problem asks whether there exists a metric g_{1} , conformal to g on M , whose Gaussian curvature is \frac{K}{2} ? Writing g_{1}=e^{u}g and specializing to the case M=\mathbb{R}^{2} with g the standard Euclidean metric and K\equiv 1 , the problem above is equivalent to solving the following classical elliptic equation (called the prescribed scalar curvature equation) (1.3) -\Delta u=e^{u}\quad\text{in }\mathbb{R}^{2}. Using the moving planes method, all the C^{2} smooth solutions of ( 1.3 ) satisfying the finite total curvature condition \int_{\mathbb{R}^{2}}e^{u(x)}\mathrm{d}x<\infty were classified by Chen-Li [ 9 , 10 ] . There are also some classification results for global smooth solutions of the higher-order Q-curvature equation (1.4) (-\Delta)^{\frac{n}{2}}u=e^{u}\quad\text{in }\mathbb{R}^{n}\ (n\geq 3). For even dimensions n , under the assumption \int_{\mathbb{R}^{n}}e^{u(x)}\mathrm{d}x<\infty and the decay condition u(x)=o(|x|^{2}) as |x|\to+\infty , Lin [ 36 ] , Wei and Xu [ 45 ] established a complete classification of all solutions u\in C^{n}(\mathbb{R}^{n}) of ( 1.4 ) for n=4 and n=2m (m\geq 2) , respectively. For the general integer n , Chang and Yang [ 5 ] classified the C^{n} smooth solutions of equation ( 1.4 ) with the decay condition u(x)=\ln\frac{2}{1+|x|^{2}}+w\left(\zeta(x)\right) as |x|\to+\infty , where w is a smooth function defined on the unit n -sphere \mathbb{S}^{n} . Moreover, Lin [ 36 ] , Xu [ 46 ] , Martinazzi [ 38 ] and Hyder [ 29 ] obtained the asymptotic behavior of solutions to ( 1.4 ) at infinity. It is worth noting that by using the Kelvin transform, the asymptotic behavior of solutions to ( 1.3 ) at infinity directly shows the asymptotic behavior near the singularity 0 for the corresponding singular problem -\Delta u=|x|^{-4}e^{u}\quad\text{in }\mathbb{R}^{2}\setminus\left\{0\right\}. Consequently, it is necessary for us to focus on the asymptotic behavior of singular elliptic solutions with isolated singularities. Over the past few decades, several significant results on the asymptotic behavior of singular solutions with isolated singularities have received significant attention in the literature. Note that if \alpha=0 , the equation ( 1.1 ) is reduced to the conformal Gaussian curvature equation (1.5) -\Delta u=e^{u}\quad\text{in }B_{1}\setminus\left\{0\right\}\subset\mathbb{R}^{2} with the finite total curvature condition (1.6) \int_{B_{1}\setminus\left\{0\right\}}e^{u(x)}\mathrm{d}x<+\infty. Equation ( 1.5 ) arises in the prescribed Gaussian curvature problem on a Riemannian surface with conical singularities [ 11 ] . The earliest investigation for local singular solutions u\in C^{2}(B_{1}\setminus\left\{0\right\}) of equation ( 1.5 ) was conducted by Chou and Wan in [ 13 ] . They showed that there exists a constant b>-2 such that (1.7) u(x)=b\ln|x|+\mathcal{O}(1)\quad\text{as }|x|\to 0 by applying complex-analytic techniques. As mentioned in [ 13 ] , the integrability condition ( 1.6 ) is necessary for u being asymptotically radial. Using the asymptotic estimate ( 1.7 ), Guo, Wan and Yang [ 27 ] later derived an asymptotic expansion up to arbitrary orders of u near the origin. There are also some results for the higher-order conformal Q-curvature equation (1.8) (-\Delta)^{\frac{n}{2}}u=e^{u}\quad\text{in }B_{1}\setminus\left\{0\right\}\subset\mathbb{R}^{n} with the important condition ( 1.6 ), where n\geq 3 is an integer. When n=4 , under an additional decay condition at the origin (1.9) u(x)=o(|x|^{-2})\quad\text{as}\ |x|\rightarrow 0, Guo and Liu [ 25 ] established the asymptotic behavior of local singular solutions u\in C^{4}(B_{1}\setminus\left\{0\right\}) of ( 1.8 ) near the origin via ODE methods, refined subsequently to a complete asymptotic expansion in [ 26 ] by using spherical averaging and ODE analysis. For the general case n\geq 3 , under a mild decay assumption near the origin weaker than ( 1.9 ), that is \int_{B_{r}\setminus\left\{0\right\}}|u(x)|\mathrm{d}x=o\left(r^{n-2}\right)\quad\text{as}\ r\to 0, Yang and Yang [ 47 ] developed a PDE method to study the local singular solutions of ( 1.8 ). Specifically, for odd integer n\geq 3 , they consider the equation ( 1.8 ) in the distributional sense, while for even dimensions n , the solution u belongs to C^{n}(B_{1}\setminus\left\{0\right\}) . Based on the characterization of isolated singularities for the Poisson equation, the authors in [ 47 ] derived a representation formula for the local singular solutions of ( 1.8 ), and subsequently established its asymptotic behavior. They also extended the asymptotic behavior results for both ( 1.5 ) and ( 1.8 ) to the general isolated singularity problem (1.10) (-\Delta)^{\frac{n}{2}}u=K(x)e^{u}\quad\text{in }B_{1}\setminus\left\{0\right\}\subset\mathbb{R}^{n}, where K(x)\in L^{\infty}(B_{1}) is non-negative. It is well-known that for n\geq 3 , the higher-dimensional analogue of equations ( 1.3 ) and ( 1.4 ) is the following higher-order Yamabe equation (1.11) \left(-\Delta\right)^{\frac{m}{2}}u=u^{\frac{n+m}{n-m}}\quad\text{in }B_{1}\setminus\left\{0\right\}\subset\mathbb{R}^{n}, where m\in(0,n) is an integer. This equation is related to the problem of finding a conformal Riemannian metric on a conformally flat n -dimensional manifold with a prescribed curvature. When m=2 , Caffarelli, Gidas and Spruck [ 4 ] used a “measure theoretic” variation of the Alexandrov reflection technique to obtain the asymptotic symmetry of local singular positive solutions u\in C^{2}(B_{1}\setminus\left\{0\right\}) of second order Yamabe equation. Later, Li [ 33 ] established the same result by employing a simplified moving plane method. Subsequent works refined the asymptotics using spectral properties of the linearized Fowler operator, see [ 31 ] . When m\geq 2 is an even integer, Jin and Xiong [ 30 ] recently studied the higher-order Yamabe equation the asymptotic radial symmetry of singular positive solutions of ( 1.11 ), by establishing the representation formula of solutions of ( 1.11 ). Moreover, for the radial symmetry and classification of global singular positive solutions u\in C^{m}\left(\mathbb{R}^{n}\setminus\left\{0\right\}\right) to equation ( 1.11 ) in \mathbb{R}^{n}\setminus\left\{0\right\} , we refer readers to the works of Gidas, Ni and Nirenberg [ 19 ] , Carffarelli, Gidas and Spruck [ 4 ] , Schoen [ 43 ] , Chen and Li [ 9 ] , Chen, Li and Ou [ 12 ] , Lin [ 36 ] , Wei and Xu [ 45 ] , Li [ 34 ] as well as Du and Yang [ 15 ] . Regarding results for the higher-order Yamabe equation with more general singular sets, see [ 15 , 28 , 6 , 7 , 49 ] . Additionally, constructions of Delaunay-type solutions of higher-order Yamabe equation are provided in [ 17 , 24 , 2 ] . For the Yamabe equation with a general prescribed curvature function K(x) , we refer to [ 7 , 8 , 37 , 44 , 49 , 16 ] . The analysis of isolated singularities in critical Hartree-type equations is driven by motivations from both physical models and fundamental analytic inequalities, see [ 35 , 40 , 41 ] . A prototypical example is the following critical Hartree equation (1.12) -\Delta u=(\mathcal{R}_{\alpha}\ast u^{\frac{2n-\alpha}{n-2}})u^{\frac{n+2-\alpha}{n-2}}\quad\text{in }\Omega\subset\mathbb{R}^{n}, where \mathcal{R}_{\alpha}(x)\ast u^{\frac{2n-\alpha}{n-2}}:=\int_{\Omega}\frac{u^{\frac{2n-\alpha}{n-2}}(y)}{|x-y|^{\alpha}}dy with n\geqslant 3 and \alpha\in(0,n) . If \Omega=\mathbb{R}^{n} , a complete classification of the global classical positive solutions to equation ( 1.12 ) has been established in the works of Miao et al. [ 39 ] , Du and Yang [ 14 ] , Gao et al. [ 18 ] , and Le [ 32 ] , with related results also in [ 22 ] . Recently, Andrade, Feng, Piccione, and Yang [ 1 ] studied the singular solutions to ( 1.12 ), by establishing the integral representation of positive solutions of ( 1.12 ). On the one hand, when \Omega=\mathbb{R}^{n}\setminus\left\{0\right\} , the authors in [ 1 ] obtained the radial symmetry of global positive solutions u\in C^{2}\left(\mathbb{R}^{n}\setminus\left\{0\right\}\right)\cap L^{\frac{2n-\alpha}{n+2}}(\mathbb{R}^{n}) of ( 1.12 ). On the other hand, when \Omega=B_{r}\setminus\left\{0\right\} with 00 , and the finite total curvature condition ( 1.2 ) holds. Inspired by [ 47 ] , we establish a representation formula for local singular solutions of ( 1.1 ) and subsequently derive their asymptotic behavior at the origin. Our first main result is stated below. Theorem 1.1 . Suppose that u\in C^{2}\left(B_{1}\setminus\{0\}\right) is a solution of ( 1.1 ) and \alpha>0 . If u satisfies ( 1.2 ), then there exists a constant b>-2 such that u(x)=v(x)+h(x)+b\ln|x|\quad\text{for }x\in B_{1}\setminus\{0\}, where h\in C^{\infty}(B_{1}) is a solution of -\Delta h=0 in B_{1} and v is defined by v(x)=\frac{1}{2\pi}\int_{B_{1}\setminus\left\{0\right\}}\ln\frac{5}{|x-y|}\left(\frac{1}{|y|^{\alpha}}*e^{u}\right)e^{u(y)}\mathrm{d}y. Moreover, if \alpha\in\left(0,2\right) and b>\frac{\alpha-4}{2} , then there exists a H \ddot{o} lder continuous function \varphi\in C_{loc}^{\gamma}(B_{1}) with \gamma\in(0,1) such that (1.13) u(x)=\varphi+b\ln|x|\quad\text{near the origin}. More generally, we will consider the following Hartree type equations in higher-order case (1.14) (-\Delta)^{\frac{n}{2}}u(x)=\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\quad\text{in }B_{1}\setminus\{0\}\subset\mathbb{R}^{n}, where n\geq 3 is an integer. Obviously, equation ( 1.14 ) is a polyharmonic equation in even dimensions and a higher-order fractional equation in odd dimensions. Based on the analysis in the proof of Theorem 1.1 , we can again obtain a representation formula for the singular solutions of ( 1.14 ). However, compared with the two dimensional case ( n=2 ), for the higher-order Hartree-type equations in dimensions n\geq 3 , the following additional decay conditions on the solutions of ( 1.14 ) are required to eliminate the higher-order derivative terms: (1.15) \int_{B_{r}\setminus\left\{0\right\}}|u(x)|\mathrm{d}x=o\left(r^{n-2}\right)\quad\text{as }r\to 0. We first consider the even-dimensional case. In this case, a fundamental solution of (-\Delta)^{\frac{n}{2}} is given by \phi(x)=c_{n}\ln\frac{5}{|x|} , where c_{n}>0 is a constant. The second result of this paper is Theorem 1.2 . Suppose that n\geq 4 is an even integer and \alpha>0 . Let u\in C^{n}\left(B_{1}\setminus\{0\}\right) be a solution of ( 1.14 ). If u satisfies ( 1.2 ) and the decay condition ( 1.15 ), then there exists a constant b>-n such that u(x)=v(x)+h(x)+b\ln|x|\quad\text{for }x\in B_{1}\setminus\{0\}, where h\in C^{\infty}(B_{1}) is a solution of (-\Delta)^{\frac{n}{2}}h=0 in B_{1} and v is defined by (1.16) v(x)=c_{n}\int_{B_{1}\setminus\left\{0\right\}}\ln\frac{5}{|x-y|}\left(\frac{1}{|y|^{\alpha}}*e^{u}\right)e^{u(y)}\mathrm{d}y, Moreover, if \alpha\in\left(0,n\right) and b>\frac{\alpha-2n}{2} , then there exists a H \ddot{o} lder continuous function \varphi\in C_{loc}^{\gamma}(B_{1}) with \gamma\in(0,1) such that (1.17) u(x)=\varphi(x)+b\ln|x|\quad\text{near the origin}. Next, we consider the odd-dimensional case. Notice that when n\geq 3 is an odd integer, (-\Delta)^{\frac{n}{2}} is a nonlocal integral operator. Therefore, in this dimension, we study equation ( 1.14 ) in the distributional sense. Definition 1.3 . Let n\geq 3 be an odd integer and \alpha>0 . We say that u\in C^{n}\left(B_{1}\setminus\left\{0\right\}\right) is a solution of ( 1.14 ), if the following conditions hold: (i) The function u is an element of \mathcal{L}_{\frac{n}{2}}(\mathbb{R}^{n}):=\left\{u\in L^{1}_{loc}(\mathbb{R}^{n}):\int_{\mathbb{R}^{n}}\frac{|u(x)|}{1+|x|^{2n}}\mathrm{d}x<\infty\right\} ; (ii) \left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\in L^{1}_{loc}(B_{1}\setminus\left\{0\right\}) ; (iii) For every test function \varphi\in C^{\infty}_{c}(B_{1}\setminus\left\{0\right\}) , the following integral equality holds: \int_{\mathbb{R}^{n}}u(x)\left(-\Delta\right)^{\frac{n}{2}}\varphi(x)\mathrm{d}x=\int_{B_{1}\setminus\left\{0\right\}}\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\varphi(x)\mathrm{d}x. Then we have the following result. Theorem 1.4 . Suppose that n\geq 3 is an odd integer and \alpha>0 . Let u\in\mathcal{L}_{\frac{n}{2}}(\mathbb{R}^{n})\cap C^{n}(B_{1}\setminus\left\{0\right\}) be a solution of ( 1.14 ). If u satisfies ( 1.2 ) and ( 1.15 ), then there exists a constant b>-n such that u(x)=v(x)+h(x)+b\ln|x|\quad\text{for }x\in B_{1}\setminus\{0\}, where v is defined by ( 1.16 ) and h\in\mathcal{L}_{\frac{n}{2}}(\mathbb{R}^{n})\cap C^{\infty}(B_{1}) is a solution of (-\Delta)^{\frac{n}{2}}h=0 in B_{1} . Moreover, if \alpha\in\left(0,n\right) and b>\frac{\alpha-2n}{2} , then there exists a H \ddot{o} lder continuous function \varphi\in C_{loc}^{\gamma}(B_{1}) with \gamma\in(0,1) such that (1.18) u(x)=\varphi(x)+b\ln|x|\quad\text{near the origin}. Remark 1.5 (Asymptotic behavior) . For \alpha\in\left(0,n\right) and b>\frac{\alpha-2n}{2} , we can describe the asymptotic behavior of the solutions to ( 1.1 ) and ( 1.14 ) at the origin. In fact, by ( 1.18 ) and \varphi\in C^{\gamma}_{loc}(B_{1}) for some \gamma\in(0,1) , there exist constants C_{1} and C_{2} such that (1.19) C_{1}+b\ln|x|\leq u(x)\leq C_{2}+b\ln|x|\quad\text{near the origin}. Remark 1.6 (Valid range of b ) . In Theorem 1.1 , 1.2 and 1.4 , for \alpha\in(0,n) and n\geq 2 , the solution exhibits the asymptotic behavior given in ( 1.13 ), ( 1.17 ) and ( 1.18 ) near the origin only when b>\frac{\alpha-2n}{2}. To be more precise, assume that there holds u(x)=\varphi(x)+b\ln|x| near the origin with \varphi\in C^{\gamma}_{loc}(B_{1}) for some 0<\gamma<1 , then ( 1.19 ) also holds. Hence, for sufficiently small r>0 , we have \displaystyle\int_{B_{1}\setminus\{0\}}\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\mathrm{d}x\geq C\int_{B_{r}}\left(\int_{B_{r}}\frac{|y|^{b}}{r^{\alpha}}\mathrm{d}y\right)|x|^{b}\mathrm{d}x, which implies the finite total curvature condition ( 1.2 ) fails whenever b\leq\frac{\alpha-2n}{2} . Therefore, it must have b>\frac{\alpha-2n}{2} . Remark 1.7 (General coefficient K(x) ) . Replace equations ( 1.1 ) and ( 1.14 ) by equations (1.20) -\Delta u(x)=K(x)\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\quad\text{in }B_{1}\setminus\{0\}\subset\mathbb{R}^{2} and (1.21) (-\Delta)^{\frac{n}{2}}u(x)=K(x)\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\quad\text{in }B_{1}\setminus\{0\}\subset\mathbb{R}^{n},\ n\geq 3 respectively, where the only difference is the presence of an additional non-negative coefficient K(x)\in L^{\infty}(B_{1}) . Then, Theorem 1.1 , 1.2 , 1.4 and Remark 1.6 remain valid in this setting. In these cases, the expression of v becomes (1.22) v(x)=c_{n}\int_{B_{1}\setminus\left\{0\right\}}\ln\frac{5}{|x-y|}K(y)\left(\frac{1}{|y|^{\alpha}}*e^{u}\right)e^{u(y)}\mathrm{d}y, where c_{n}>0 is a constant and c_{2}=\frac{1}{2\pi} . It is worth noting that when K(x)\equiv 1 , equations ( 1.20 ) and ( 1.21 ) reduce to the original equations ( 1.1 ) and ( 1.14 ) respectively. Therefore, in the proofs below, we shall work directly with equations ( 1.20 ) and ( 1.21 ). Inspired by [ 47 ] , a key step in our approach is to obtain a representation formula for the solutions of differential equations ( 1.20 ) and ( 1.21 ). This method not only avoids the failure of ODE-based techniques caused by the convolution term, but also bypasses the breakdown of the maximum principle induced by the nonlocality of the equation. This idea plays a significant role in analyzing the properties of solutions to higher-order equations with isolated singularities, as showed in [ 15 , 30 ] . This paper is organized as follows. In Section 2, we firstly deduce a representation formula for singular solutions of equations ( 1.20 ) and ( 1.21 ) in even dimensions n\geq 3 , and then prove Theorem 1.1 and Theorem 1.2 , by using the elliptic regularity theory. In Section 3, we generalize the asymptotic behavior at the origin for the solutions to ( 1.21 ) in odd dimensions n\geq 3 , and prove Theorem 1.4 . In this paper, c , C will be used to denote different constants. 2. Asymptotic behavior at isolated singularities in even dimension case In this section, we focus on the singular solutions of equation ( 1.21 ) with even dimensions n\geq 2 for simplicity. First, under the finite total curvature assumption ( 1.2 ), we derive a crucial representation formula for solutions of ( 1.21 ) expressed in term of v . Subsequently, for n=2 and for even n\geq 4 , we verify the local Hölder continuity of v under necessary conditions, which allows us to obtain the asymptotic behavior of the solution at the origin. 2.1. Representation formula Motivated by [ 47 ] and based on the characterization of isolated singularities for the Poisson equation, we now present the representation formula for solutions of ( 1.21 ) in the case of even dimension n\geq 2 . Proposition 2.1 (Representation Formula) . Let n\geq 2 be a positive even integer and \alpha>0 . Suppose that u\in C^{n}\left(B_{1}\setminus\{0\}\right) is a solution of ( 1.21 ), where K\in L^{\infty}(B_{1}) is non-negative and u satisfies ( 1.2 ), then there exists constants a_{\beta}\in\mathbb{R} with |\beta|\leq n-1 , such that (2.1) u(x)=v(x)+h(x)+\sum_{|\beta|\leq n-1}a_{\beta}D^{\beta}\phi(x)\quad\text{for }x\in B_{1}\setminus\{0\}, where \phi(x)=c_{n}\ln\frac{5}{|x|} is a fundamental solution of (-\Delta)^{\frac{n}{2}} in \mathbb{R}^{n} , h\in C^{\infty}(B_{1}) is a solution of (-\Delta)^{\frac{n}{2}}h=0 in B_{1} and v is given by ( 1.22 ). Proof. Define u^{+}(x):=\max\left\{u(x),0\right\} . It is easy to verify that u^{+}(x)\leq e^{{u}(x)} . Thus, we have \displaystyle\int_{B_{1}\setminus\{0\}}K(x)\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\mathrm{d}x \displaystyle\geq C\int_{B_{1}\setminus\{0\}}\left(\int_{B_{1}\setminus\{0\}}\frac{u^{+}(y)}{|x-y|^{\alpha}}\mathrm{d}y\right)u^{+}(x)\mathrm{d}x \displaystyle\geq\frac{C}{2^{\alpha}}\left(\int_{B_{1}\setminus\{0\}}u^{+}(x)\mathrm{d}x\right)^{2}. This, together with ( 1.2 ), implies that \int_{B_{1}\setminus\{0\}}u^{+}(x)\mathrm{d}x<+\infty. Applying the results in [ 47 , Theorem 1.4] with s=0 , we conclude the proof of this proposition. ∎ 2.2. Two-dimensional case For the case n=2 , a fundamental solution of -\Delta in \mathbb{R}^{2} is given by \phi(x)=\frac{1}{2\pi}\ln\frac{5}{|x|} . We will give the following specific representation for solutions of equation ( 1.20 ). Lemma 2.2 . Let \alpha>0 . Suppose that u\in C^{2}\left(B_{1}\setminus\{0\}\right) is a solution of ( 1.20 ) satisfying condition ( 1.2 ), and K\in L^{\infty}(B_{1}) is non-negative. Then u decomposes as u(x)=v(x)+h(x)+a_{0}\phi(x)\quad\text{for }x\in B_{1}\setminus\{0\}, where v is given by ( 1.22 ), h\in C^{\infty}(B_{1}) is a harmonic function in B_{1} satisfying -\Delta h=0 , and the constant a_{0} satisfies the constraint a_{0}<4\pi . Proof. Using proposition 2.1 with n=2 , we obtain that (2.2) u(x)=v(x)+h(x)+\sum_{|\beta|\leq 1}a_{\beta}D^{\beta}\phi(x)\quad\text{for }x\in B_{1}\setminus\{0\}, where v is given by ( 1.22 ). We will prove that a_{\beta}=0 for |\beta|=1 . By ( 1.2 ) and ( 2.2 ), we obtain that (2.3) \int_{B_{\frac{1}{2}}\setminus\{0\}}\left(\int_{B_{\frac{1}{2}}\setminus\{0\}}\frac{e^{v(y)}e^{h(y)}e^{\sum_{|\beta|\leq 1}a_{\beta}D^{\beta}\phi(y)}}{|x-y|^{\alpha}}\mathrm{d}y\right)e^{v(x)}e^{h(x)}e^{\sum_{|\beta|\leq 1}a_{\beta}D^{\beta}\phi(x)}\mathrm{d}x<+\infty. Since h\in C^{\infty}(B_{1}(0)) , we get that C_{1}0 . By the nonnegativity of K , then e^{v}\geq 1 in B_{1/2}\setminus\{0\} . For x,y\in B_{1/2} , we have \frac{1}{|x-y|^{\alpha}}\geq 1 . Hence, from ( 2.3 ), we yields \int_{B_{\frac{1}{2}}\setminus\{0\}}e^{\sum_{|\beta|\leq 1}a_{\beta}D^{\beta}\phi(x)}\mathrm{d}x<+\infty. Note that D^{(i,j)}\phi represents the derivative of \phi , taken i times with respect to x_{1} and j times with respect to x_{2} . Thus, based on the expression of the fundamental solution of -\Delta in \mathbb{R}^{2} , we obtain that (2.4) \int_{B_{\frac{1}{2}}\setminus\{0\}}e^{a_{0}\frac{1}{2\pi}\ln\frac{5}{|x|}+a_{(1,0)}\left(-\frac{x_{1}}{2\pi|x|^{2}}\right)+a_{(0,1)}\left(-\frac{x_{2}}{2\pi|x|^{2}}\right)}\mathrm{d}x<+\infty. Consider the polar coordinates \begin{cases}x_{1}=r\cos\theta,\\ x_{2}=r\sin\theta,\end{cases}\quad 0\leq r<1,\ -\pi\leq\theta\leq\pi. Now, we show that if a_{(0,1)}\leq 0 , then a_{(1,0)}=0 . Let D_{1}=\left\{(x_{1},x_{2})\in B_{1/2}\setminus\{0\}\bigg|0\leq r<\frac{1}{2},0\leq\theta\leq\frac{\pi}{4}\right\}. By ( 2.4 ), we get that \int_{0}^{\frac{1}{2}}\int_{0}^{\frac{\pi}{4}}\left(\frac{5}{r}\right)^{\frac{1}{2\pi}a_{0}}e^{\frac{1}{2\pi}\left[a_{(1,0)}\left(-\frac{\cos\theta}{r}\right)+a_{(0,1)}\left(-\frac{\sin\theta}{r}\right)\right]}r\mathrm{d}\theta\mathrm{d}r<\infty. Since a_{(0,1)}\leq 0 and 0<\theta<\frac{\pi}{4} , we obtain that \int_{0}^{\frac{1}{2}}\int_{0}^{\frac{\pi}{4}}\left(\frac{5}{r}\right)^{\frac{1}{2\pi}a_{0}}e^{\frac{1}{2\pi}\left[a_{(1,0)}\left(-\frac{\cos\theta}{r}\right)\right]}r\mathrm{d}\theta\mathrm{d}r<\infty. Using \cos\theta\geq\frac{\sqrt{2}}{2} in D_{1} and the following equality \int_{0}^{\frac{1}{2}}r^{a}e^{\frac{b}{r}}\mathrm{d}r=+\infty\quad\forall a\in\mathbb{R},b>0, we have a_{(1,0)}\geq 0 . On the other hand, we set D_{2}=\left\{(x_{1},x_{2})\in B_{1/2}\setminus\{0\}\bigg|0\leq r<\frac{1}{2},\frac{3\pi}{4}\leq\theta\leq\pi\right\}. Then we can also get a_{(1,0)}\leq 0 . Hence, if a_{(0,1)}\leq 0 holds true, then a_{(1,0)}=0 holds as well. Similarly, if a_{(0,1)}\geq 0 , we can also deduce that a_{(1,0)}=0 , by analyzing the integral in D_{3}=\left\{(x_{1},x_{2})\in B_{1/2}\setminus\{0\}\bigg|0\leq r<\frac{1}{2},-\frac{\pi}{4}\leq\theta\leq 0\right\} and D_{4}=\left\{(x_{1},x_{2})\in B_{1/2}\setminus\{0\}\bigg|0\leq r<\frac{1}{2},-\pi\leq\theta\leq-\frac{3\pi}{4}\right\}. Thus, we obtain a_{(1,0)}=0 . Moreover, we can also get a_{(0,1)}=0 . Combined with ( 2.4 ), we have \int_{B_{\frac{1}{2}}\setminus\{0\}}\left(\frac{5}{\sqrt{x_{1}^{2}+x_{2}^{2}}}\right)^{\frac{1}{2\pi}a_{0}}\mathrm{d}x_{1}\mathrm{d}x_{2}<\infty, which yields that a_{0}<4\pi . ∎ Next, we proceed to prove Theorem 1.1 . We establish the local Hölder continuity of v , from which the asymptotic behavior of u at the origin follows. Proof of Theorem 1.1 . Since a_{0}<4\pi and a_{0}\phi(x)=\frac{1}{2\pi}a_{0}\left(\ln 5-\ln|x|\right) , by Lemma 2.2 , there exists a constant b:=-\frac{1}{2\pi}a_{0}>-2 , such that u satisfies (2.5) u(x)=v(x)+h(x)+b\ln|x|\quad\text{for }x\in B_{1}\setminus\{0\}, where v is given by ( 1.22 ) and h\in C^{\infty}(B_{1}) is a harmonic function in B_{1} . Here h has been redefined by absorbing the constant -b\ln 5 . We will prove that if \alpha\in\left(0,2\right) and b>\frac{\alpha-4}{2} , then v\in C_{loc}^{\gamma}(B_{1}) for some 0<\gamma<1 . By ( 2.5 ), we know that v is a non-negative solution of \displaystyle-\Delta v(x) \displaystyle=K(x)\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)} \displaystyle=K(x)\left(\int_{B_{1}\setminus\{0\}}\frac{e^{v(y)}e^{h(y)}|y|^{b}}{|x-y|^{\alpha}}\mathrm{d}y\right)e^{v(x)}e^{h(x)}|x|^{b}\quad\text{in }B_{1} in the sense of distributions. According to K\in L^{\infty}(B_{1}) and ( 1.2 ), we have \int_{B_{1}\setminus\{0\}}K(x)\left(\int_{B_{1}\setminus\{0\}}\frac{e^{v(y)}e^{h(y)}|y|^{b}}{|x-y|^{\alpha}}\mathrm{d}y\right)e^{v(x)}e^{h(x)}|x|^{b}\mathrm{d}x<\infty. We claim that K(x)\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\in L^{p_{0}}_{loc}(B_{1}) for some p_{0}>1 . In fact, we fix an arbitrary r\in(0,1) . For convenience, we define f(x):=e^{u(x)}=|x|^{b}e^{h(x)}e^{v(x)} and g_{r}(x):=\int_{B_{r}\setminus\{0\}}\frac{f(y)}{|x-y|^{\alpha}}\mathrm{d}y . We divide K(x)\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)} into two parts: \displaystyle K(x)\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}= \displaystyle K(x)g_{r}(x)f(x)+K(x)\left(\int_{B_{r_{1}}\setminus B_{r}}\frac{e^{u(y)}}{|x-y|^{\alpha}}\mathrm{d}y\right)f(x) \displaystyle+K(x)\left(\int_{B_{1}\setminus B_{r_{1}}}\frac{e^{u(y)}}{|x-y|^{\alpha}}\mathrm{d}y\right)f(x) \displaystyle:= \displaystyle\mathcal{I}_{1}(x)+\mathcal{I}_{2}(x)+\mathcal{I}_{3}(x)\quad\text{in }B_{r}, where we choose r_{1} satisfying r0 . Since b>-2 , there exists p_{f}>1 , such that bp_{f}>-2 , whence |x|^{b}\in L^{p_{f}}(B_{r}) . We consider two cases: (i) b\geq 0 ; (ii) -21 such that |x|^{b}\in L^{p_{f}}(B_{r}) . Therefore, f\in L^{q}(B_{r}) for any q>1 . Since 0<\alpha<2 , by the Hardy–Littlewood–Sobolev inequality, we easily get that there exists some q_{1}>1 such that g_{r}\in L^{q_{1}}(B_{r}) , and hence \mathcal{I}_{1}(x)\in L^{q_{1}}(B_{r}) . For case (ii), by the Hölder inequality, we can choose some p_{1} satisfying 11 satisfying (2.6) \frac{1}{t}=\frac{1}{p_{1}}+\frac{\alpha}{2}-1, such that (2.7) \|g_{r}\|_{L^{t}(B_{r})}\leq C\|f\|_{L^{p_{1}}(B_{r})} for some constant C>0 . Next, we consider two cases here: (a) -\frac{2}{b}\geq\frac{2}{2-\alpha} , i.e. \alpha-2\leq b<0 , (b) -\frac{2}{b}<\frac{2}{2-\alpha} , i.e. -21 satisfying \frac{1}{s}+\frac{1}{t}=\frac{1}{q_{2}} , so that \int_{B_{r}\setminus\{0\}}\left|f(x)g_{r}(x)\right|^{q_{2}}\mathrm{d}x\leq\int_{B_{r}\setminus\{0\}}\left|f(x)\right|^{q_{2}}\left|g_{r}(x)\right|^{q_{2}}\mathrm{d}x\leq\|f^{q_{2}}\|_{\frac{s}{q_{2}}}\|(g_{r})^{q_{2}}\|_{\frac{t}{q_{2}}}<\infty, where we have used the Hölder inequality. Consequently, it follows that \mathcal{I}_{1}(x)\in L^{q_{2}}(B_{r}). For case (b), we have 1\frac{\alpha-4}{2} , we obtain that \frac{\alpha-4}{2}1 satisfying \frac{1}{s}+\frac{1}{t}=\frac{1}{q_{3}} such that \mathcal{I}_{1}(x)\in L^{q_{3}}(B_{r}). For \mathcal{I}_{2} , by the regularity of u in \bar{B}_{r_{1}}\setminus\left\{0\right\} , we know that there exists a constant M>0 such that u1 such that \mathcal{I}_{2}\in L^{q_{4}}(B_{r}) . For \mathcal{I}_{3} , by the assumption ( 1.2 ) and \alpha\in(0,n) , we have \left(\int_{B_{1}\setminus\{0\}}e^{u(x)}\mathrm{d}x\right)^{2}\leq 2^{\alpha}\int_{B_{1}\setminus\{0\}}\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\mathrm{d}x<\infty, which implies that e^{u}\in L^{1}(B_{1}\setminus\left\{0\right\}) . For x\in B_{r} and y\in B_{1}\setminus B_{r_{1}} , we observe that |x-y|\geq|y|-|x|\geq r_{1}-r>0. Therefore, we get \displaystyle\mathcal{I}_{3}(x) \displaystyle\leq C\left(\int_{B_{1}\setminus B_{r_{1}}}\frac{e^{u(y)}}{|x-y|^{\alpha}}\mathrm{d}y\right)f(x) \displaystyle\leq C\left(r_{1}-r\right)^{-\alpha}\left(\int_{B_{1}\setminus B_{r_{1}}}e^{u(y)}\mathrm{d}y\right)f(x) \displaystyle\leq Cf(x)\quad\text{for }x\in B_{r}. By an argument analogous to that for \mathcal{I}_{1} , we have \mathcal{I}_{3}\in L^{q_{4}}(B_{r}) . To summarize, for \alpha\in\left(0,2\right) and b>\frac{\alpha-4}{2} , we can select p_{0}=\min\left\{q_{1},q_{2},q_{3},q_{4}\right\} , then K(x)\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\in L^{p_{0}}(B_{r}) . Since r\in(0,1) is arbitrary, we deduce that K(x)\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\in L^{p_{0}}_{loc}(B_{1})\quad\text{for some }p_{0}>1. An application of the standard W^{2,p} elliptic regularity theory (see [ 20 ] ) directly yields the improved regularity v\in W^{2,p_{0}}_{loc}(B_{1}) . It then follows from the Sobolev embedding theorem that v is locally Hölder continuous, i.e., v\in C^{\gamma}_{loc}(B_{1}) for some \gamma\in(0,1) . ∎ 2.3. Even-dimensional case ( n\geq 4 ) In this subsection, following a similar detailed analysis in two-dimensional case, we study the Hartree-type equation ( 1.21 ) in even dimensions n\geq 4 . The corresponding polyharmonic equations involve higher-order singularities, which introduce substantial new challenges. Thus, we need an additional decay condition ( 1.15 ) near the origin. Under the finite total curvature assumption ( 1.2 ) and condition ( 1.15 ), we also derive a representation formula for solutions to ( 1.21 ) in the even-dimensional case, and further investigate the asymptotic behavior of these solutions near the origin. If n\geq 4 is even, a fundamental solution of \left(-\Delta\right)^{\frac{n}{2}} is given by \phi(x)=c_{n}\ln\frac{5}{|x|} , with c_{n}>0 being a constant. Then we have Lemma 2.3 . Let n\geq 4 be an even integer and \alpha>0 . Suppose that u\in C^{n}\left(B_{1}\setminus\{0\}\right) is a solution of ( 1.21 ) satisfying condition ( 1.2 ) and ( 1.15 ), and K\in L^{\infty}(B_{1}) is non-negative. Then u decomposes as (2.8) u(x)=v(x)+h(x)+a_{0}\phi(x)\quad\text{for }x\in B_{1}\setminus\{0\}, where v is defined by ( 1.22 ). The function h\in C^{\infty}(B_{1}) is a polyharmonic function in B_{1} satisfying (-\Delta)^{\frac{n}{2}}h=0 , and the constant a_{0} satisfies the constraint a_{0}<\frac{n}{c_{n}} . Proof. If u solves ( 1.21 ) and K\in L^{\infty}(B_{1}) , Proposition 2.1 with s=0 gives the decomposition: (2.9) u(x)=v(x)+h(x)+\sum_{|\beta|\leq n-1}a_{\beta}D^{\beta}\phi(x)\quad\text{for }x\in B_{1}\setminus\{0\}, where v is given by ( 1.22 ) and h\in C^{\infty}(B_{1}) is a solution of (-\Delta)^{\frac{n}{2}}h=0 in B_{1} . Next we will prove that \sum_{1\leq|\beta|\leq n-1}a_{\beta}D^{\beta}\phi(x)\equiv 0 . We can split the proof in two steps. Step 1. We show that all coefficients corresponding to derivatives of order 2\leq|\beta|\leq n-1 must be zero. By the Fubini Theorem and the condition ( 1.2 ), we obtain \displaystyle\int_{B_{r}}v(x)\mathrm{d}x \displaystyle\leq C\int_{B_{r}}\left[\int_{B_{1}}\ln\frac{5}{|x-y|}K(y)\left(\frac{1}{|y|^{\alpha}}\ast e^{u}\right)e^{u(y)}\mathrm{d}y\right]\mathrm{d}x \displaystyle\leq C\int_{B_{1}}\left(\int_{B_{r}}\ln\frac{5}{|x-y|}\mathrm{d}x\right)K(y)\left(\frac{1}{|y|^{\alpha}}\ast e^{u}\right)e^{u(y)}\mathrm{d}y \displaystyle\leq Cr^{n}\ln\frac{1}{r}\int_{B_{1}}K(y)\left(\frac{1}{|y|^{\alpha}}\ast e^{u}\right)e^{u(y)}\mathrm{d}y \displaystyle\leq Cr^{n}\ln\frac{1}{r},\qquad\text{for }00 . Let d_{0}=\frac{1}{2}P_{n-1}(\theta_{0}) . Therefore, by continuity, there exists a geodesic ball neighborhood U_{0}\subset\mathbb{S}^{n-1} of \theta_{0} such that \left|P_{n-1}\right|\geq d_{0}>0 for all \theta\in U_{0} . Using polar coordinates x=\rho\theta with \theta\in\mathbb{S}^{n-1} , the volume element becomes \mathrm{d}x=\rho^{n-1}\mathrm{d}\rho\mathrm{d}\theta. By homogeneity, for \theta\in\mathbb{S}^{n-1} and \rho>0 , we have P_{n-1}(x)=\rho^{n-1}P_{n-1}(\theta). Now, define the region V_{r}=[0,r]\times U_{0} . Then, \int_{V_{r}}\left|\sum_{|\beta|=n-1}a_{\beta}D^{\beta}\phi(x)\right|\mathrm{d}x=\int_{U_{0}}\left|P_{n-1}(\theta)\right|\mathrm{d}\theta\int_{0}^{r}\mathrm{d}\rho\geq d_{1}r for some constant d_{1}>0 . On the other hand, a direct computation gives \int_{B_{r}}\left|\sum_{|\beta|\leq n-2}a_{\beta}D^{\beta}\phi(x)\right|\mathrm{d}x\leq C\int_{B_{r}}|x|^{-n+2}\mathrm{d}x\leq Cr^{2}. Thus, combining these two estimates, we have \displaystyle\int_{V_{r}}\left|\sum_{|\beta|\leq n-1}a_{\beta}D^{\beta}\phi(x)\right|\mathrm{d}x \displaystyle\geq\int_{V_{r}}\left|\sum_{|\beta|=n-1}a_{\beta}D^{\beta}\phi(x)\right|\mathrm{d}x-\int_{V_{r}}\left|\sum_{|\beta|\leq n-2}a_{\beta}D^{\beta}\phi(x)\right|\mathrm{d}x \displaystyle\geq\int_{V_{r}}\left|\sum_{|\beta|=n-1}a_{\beta}D^{\beta}\phi(x)\right|\mathrm{d}x-\int_{B_{r}}\left|\sum_{|\beta|\leq n-2}a_{\beta}D^{\beta}\phi(x)\right|\mathrm{d}x \displaystyle\geq d_{1}r-Cr^{2} \displaystyle\geq\frac{d_{1}}{2}r\quad\text{for small }r>0. This contradicts ( 2.10 ). Hence, we conclude that P_{n-1}\equiv 0 , which implies \sum_{|\beta|=n-1}a_{\beta}D^{\beta}\phi(x)=0. By iterating this procedure, we conclude that \sum_{|\beta|=j}a_{\beta}D^{\beta}\phi(x)=0, for every index 2\leq j\leq n-2 . Step 2. We show that \sum_{|\beta|=1}a_{\beta}D^{\beta}\phi(x)=0 and a_{0}<\frac{n}{c_{n}} . By ( 1.2 ), ( 2.9 ) and the results in step 1, we have (2.12) \int_{B_{\frac{1}{2}}\setminus\{0\}}\left(\int_{B_{\frac{1}{2}}\setminus\{0\}}\frac{e^{v(y)}e^{h(y)}e^{\sum_{|\beta|\leq 1}a_{\beta}D^{\beta}\phi(y)}}{|x-y|^{\alpha}}\mathrm{d}y\right)e^{v(x)}e^{h(x)}e^{\sum_{|\beta|\leq 1}a_{\beta}D^{\beta}\phi(x)}\mathrm{d}x<+\infty. To exploit this, we first note that h\in C^{\infty}(B_{1}(0)) implies that C_{1}0 . By the nonnegativity of K , e^{v}\geq 1 in B_{1/2}\setminus\{0\} . For x,y\in B_{1/2} , the kernel satisfies \frac{1}{|x-y|^{\alpha}}\geq 1 . Combining these bounds with ( 2.12 ), we obtain the finiteness of the integral involving the remaining exponential term: \int_{B_{\frac{1}{2}}\setminus\{0\}}e^{\sum_{|\beta|\leq 1}a_{\beta}D^{\beta}\phi(x)}\mathrm{d}x<+\infty. Namely, (2.13) \int_{B_{\frac{1}{2}}\setminus\{0\}}\left(\frac{5}{|x|}\right)^{c_{n}a_{0}}e^{-c_{n}\sum_{i=1}^{n}a_{i}\frac{x_{i}}{|x|^{2}}}\mathrm{d}x<+\infty, where a_{i}=a(0,\cdots,1,\cdots,0) . An argument parallel to the one used in proving Theorem 1.1 , we have a_{i}=0 for every i=1,2,\cdots,n , and hence \sum_{|\beta|=1}a_{\beta}D^{\beta}\phi(x)=0 . From ( 2.13 ), it follows that \int_{B_{\frac{1}{2}}\setminus\{0\}}\left(\frac{5}{|x|}\right)^{c_{n}a_{0}}\mathrm{d}x<+\infty, which implies c_{n}a_{0}0 . Then for any p>1 , we have e^{pv}\in L^{1}(B_{1}). Remark 2.5 . Lemma 2.4 does not require F to be finite everywhere. Any non-negative F\in L^{1}(B_{1}) is admissible, even if F has isolated singularities. Since F is integrable, the potential v(x)=\int_{B_{1}}K(x,y)F(y)\mathrm{d}y is well defined almost everywhere in B_{1} , which is sufficient to get e^{pv}\in L^{1}(B_{1}) . Proof. Since F\geq 0 and F\in L^{1}(B_{1}) , for any \varepsilon>0 , we may choose \lambda>0 such that the truncation F_{1}=F\chi_{\left\{F<\lambda\right\}}\qquad\text{and}\qquad F_{2}=F\chi_{\left\{F\geq\lambda\right\}} satisfying F_{1}\in L^{\infty}(B_{1}) and \|F_{2}\|_{L^{1}(B_{1})}<\varepsilon , where cut-off function \chi_{\Omega} is defined by \chi_{\Omega}=\begin{cases}\displaystyle 1,&\quad\text{in }\Omega,\\[6.0pt] 0,&\quad\text{in }\Omega^{c}.\end{cases} Therefore, for any arbitrarily fixed p>1 , there exists \lambda_{p}>0 such that F_{1}\in L^{\infty}(B_{1}) and \|F_{2}\|_{L^{1}(B_{1})}<\frac{n}{2pC_{1}} . Then the function v admits a decomposition v=v_{1}+v_{2} , in which v_{i}(x)=\int_{B_{1}}K(x,y)F_{i}(y)\mathrm{d}y\qquad i=1,2. By the Hölder inequality, we have \displaystyle\|v_{1}\|_{L^{\infty}(B_{1})} \displaystyle\leq\|K(x,y)\|_{L^{1}(B_{1})}\|F_{1}\|_{L^{\infty}(B_{1})} \displaystyle\leq C_{1}\|F_{1}\|_{L^{\infty}(B_{1})}\int_{B_{2}}\left(1+\left|\ln|y|\right|\right)\mathrm{d}y \displaystyle<\infty, which implies that e^{pv_{1}}\in L^{1}(B_{1}) . On the other hand, using ( 2.14 ) and Jensen’s inequality, we have \displaystyle\int_{B_{1}}e^{p|v_{2}(x)|}\mathrm{d}x \displaystyle\leq\int_{B_{1}}e^{pC_{1}\int_{B_{1}}\left(1+\left|\ln|x-y|\right|\right)F_{2}(y)\mathrm{d}y}\mathrm{d}x \displaystyle\leq C\int_{B_{1}}e^{pC_{1}\|F_{2}\|_{1}\int_{B_{1}}\left(\left|\ln|x-y|\right|\frac{F_{2}(y)}{\|F_{2}\|_{1}}\right)\mathrm{d}y}\mathrm{d}x \displaystyle\leq C\int_{B_{1}}\int_{B_{1}}\left(e^{\left|\ln|x-y|\right|}\right)^{\frac{n}{2}}\frac{F_{2}(y)}{\|F_{2}\|_{1}}\mathrm{d}y\mathrm{d}x \displaystyle\leq C\int_{B_{1}}\left(\frac{1}{|y|}\right)^{\frac{n}{2}}\mathrm{d}y+C\int_{B_{2}\setminus B_{1}}|y|^{\frac{n}{2}}\mathrm{d}y \displaystyle<\infty. Thus, we conclude that for any p>1 , e^{pv}\in L^{1}(B_{1}) . This completes the proof. ∎ Now we begin to prove Theorem 1.2 . Proof of Theorem 1.2 . We define b:=-c_{n}a_{0}>-n. Since a_{0}\phi(x)=-b\ln 5+b\ln|x| , by Lemma 2.3 , we have (2.15) u(x)=v(x)+h(x)+b\ln|x|. Here v is given by ( 1.22 ) and h\in C^{\infty}(B_{1}) is a solution of (-\Delta)^{\frac{n}{2}}h=0 in B_{1} , where we have absorbed the constant -b\ln 5 into the harmonic function h . Therefore, we know that v is a nonnegative solution of \left(-\Delta\right)^{\frac{n}{2}}v=K(x)\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\quad\text{in }B_{1} in the sense of distributions. Using K\in L^{\infty}(B_{1}) and assumption ( 1.2 ), we obtain that \int_{B_{1}\setminus\{0\}}K(x)\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\mathrm{d}x<+\infty. Since c_{n}\ln\frac{5}{|x-y|} satisfies ( 2.14 ), then by Lemma 2.4 and Remark 2.5 , we have e^{v}\in L^{p}(B_{1}) for any p>1 . Next, we will prove that v\in C^{\gamma}_{loc}(B_{1}) for some 0<\gamma<1 . By ( 1.21 ) and ( 2.15 ), we know that v is a non-negative solution of \displaystyle(-\Delta)^{\frac{n}{2}}v(x) \displaystyle=K(x)\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)} \displaystyle=K(x)\left(\int_{B_{1}\setminus\{0\}}\frac{e^{v(y)}e^{h(y)}|y|^{b}}{|x-y|^{\alpha}}\mathrm{d}y\right)e^{v(x)}e^{h(x)}|x|^{b}\quad\text{in }B_{1} in the sense of distributions. Proceeding in a manner similar to the proof of Theorem 1.1 , one finds that, provides \alpha\in(0,n) and b>\frac{\alpha-2n}{2} , it follows that K(x)\left(\int_{B_{1}\setminus\{0\}}\frac{e^{v(y)}e^{h(y)}|y|^{b}}{|x-y|^{\alpha}}\mathrm{d}y\right)e^{v(x)}e^{h(x)}|x|^{b}\in L^{p_{0}}(B_{1}) for some p_{0}>1 . A subsequent application of the standard W^{n,p} elliptic regularity theory yields v\in W^{n,p_{0}}_{loc}(B_{1}) . It then follows from the Sobolev embedding theorem that v\in C^{\gamma}_{loc}(B_{1}) for some 0<\gamma<1 . Thus, by ( 2.15 ), we have u(x)=\varphi(x)+b\ln|x|\quad\text{in }B_{1}\setminus\left\{0\right\}, in which \varphi:=v+h\in C^{\gamma}_{loc}(B_{1}) for some 0<\gamma<1 . We finish the proof. ∎ 3. Asymptotic behavior at isolated singularities in odd dimension case In this section, we study the isolated singularity problem for the Hartree-type equation ( 1.21 ) with odd dimensions n\geq 3 . In this case, the operator (-\Delta)^{\frac{n}{2}} is a nonlocal operator, and we mainly consider the distributional solutions defined in the space u\in\mathcal{L}_{\frac{n}{2}}(\mathbb{R}^{n}) . Under the finite total curvature condition ( 1.2 ) and the decay condition ( 1.15 ), we also show that the asymptotic behavior near the origin has the same form as in the even-dimensional case. The following lemma provide the corresponding representation formula for the solutions to ( 1.21 ) in odd dimensions. Lemma 3.1 . Let n\geq 3 be an odd integer and \alpha>0 . Suppose that u\in\mathcal{L}_{\frac{n}{2}}(\mathbb{R}^{n})\cap C^{n}\left(B_{1}\setminus\left\{0\right\}\right) is a solution of ( 1.21 ) satisfying condition ( 1.2 ) and ( 1.15 ), and K\in L^{\infty}(B_{1}) is non-negative. Then u decomposes as u(x)=v(x)+h(x)+b\ln|x|\quad\text{for }x\in B_{1}\setminus\left\{0\right\}, where v is defined by ( 1.22 ), h\in C^{\infty}(B_{1}) is a polyharmonic function in B_{1} satisfying (-\Delta)^{\frac{n}{2}}h=0 , and the constant b satisfies b>-n . Proof. Let v be defined as in ( 1.22 ). Standard potential estimates imply v\in\mathcal{L}_{\frac{n}{2}}(\mathbb{R}^{n}) , and (-\Delta)^{\frac{n}{2}}v(x)=K(x)\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)}\qquad\text{in }\mathcal{D}^{\prime}(B_{1}). Set w=u-v . Then (-\Delta)^{\frac{n}{2}}w=0\qquad\text{in }B_{1}\setminus\{0\}. Hence, for any \psi\in C_{c}^{\infty}(B_{1}\setminus\{0\}) , we have \langle(-\Delta)^{\frac{n}{2}}w,\psi\rangle=0, which implies that \mathrm{supp}\big((-\Delta)^{\frac{n}{2}}w\big)\subset\{0\}. By the classical structure theorem of Schwartz, we know that (-\Delta)^{\frac{n}{2}}w=\sum_{|\beta|\leq M}a_{\beta}D^{\beta}\delta_{0},\qquad a_{\beta}\in\mathbb{R}. Using (-\Delta)^{\frac{n}{2}}\phi=\delta_{0} , we have w(x)=h(x)+\sum_{|\beta|\leq M}a_{\beta}D^{\beta}\phi(x), where h satisfies (-\Delta)^{\frac{n}{2}}h=0 in B_{1} and \phi(x)=c_{n}\ln\frac{5}{|x|} is a fundamental solution of (-\Delta)^{\frac{n}{2}} . By the elliptic regularity, we have h\in C^{\infty}(B_{1}) . Since u,v,h\in L^{1}_{\mathrm{loc}}(B_{1}) and D^{\beta}\phi(x)\sim|x|^{-|\beta|} , the local integrability near 0 requires |\beta|-n and h has been redefined by absorbing a constant into the harmonic part. ∎ Now we discuss the regularity of v and prove Theorem 1.4 . Proof of Theorem 1.4 . Since c_{n}\ln\frac{5}{|x-y|} satisfies ( 2.14 ), then by Lemma 2.4 and Remark 2.5 , we have e^{pv}\in L^{1}(B_{1}) for any p>1 . By ( 1.21 ) and the expression of u , we know that v is a non-negative solution of \displaystyle(-\Delta)^{\frac{n}{2}}v(x) \displaystyle=K(x)\left(\frac{1}{|x|^{\alpha}}*e^{u}\right)e^{u(x)} \displaystyle=K(x)\left(\int_{B_{1}\setminus\{0\}}\frac{e^{v(y)}e^{h(y)}|y|^{b}}{|x-y|^{\alpha}}\mathrm{d}y\right)e^{v(x)}e^{h(x)}|x|^{b}\quad\text{in }B_{1} in the sense of distributions. An argument analogous to the proof of Theorem 1.1 shows that, provided \alpha\in(0,n) and b>\frac{\alpha-2n}{2} , we have K(x)\left(\int_{B_{1}\setminus\{0\}}\frac{e^{v(y)}e^{h(y)}|y|^{b}}{|x-y|^{\alpha}}\mathrm{d}y\right)e^{v(x)}e^{h(x)}|x|^{b}\in L^{p_{0}}(B_{1}) for some p_{0}>1 . By the expression of v and the classical potential estimate (see, e.g., Gilbarg-Trudinger [ 20 ] ), we obtain that v\in C^{\gamma}_{loc}(B_{1}) for some 0<\gamma<1 . Consequently, we get u(x)=\varphi(x)+b\ln|x|\quad\text{in }B_{1}\setminus\left\{0\right\}, in which \varphi:=v+h\in C^{\gamma}_{loc}(B_{1}) for some 0<\gamma<1 . ∎ Declaration : The authors declared that they have no conflict of interest.