--- title: 'Non-Gaussian Hardy Equation: Critical Insights' url: https://www.emergentmind.com/topics/non-gaussian-hardy-equation type: topic --- # Non-Gaussian Hardy Equation: Critical Insights Searching arXiv for recent and foundational papers on the non-Gaussian Hardy equation and related Hardy–Hénon / Hardy–Poincaré formulations. The non-Gaussian Hardy equation denotes, in current arXiv usage, a family of Hardy-type PDEs and inequalities in which the singular, weighted, or diffusive structure is not Gaussian. Representative formulations include the higher-order Hardy–Hénon equation \[ (-\Delta)^m u = |x|^\sigma u^p \quad \text{in }\mathbb{R}^n, \] the second-order Hardy–Hénon equation \[ -\Delta u = |x|^\sigma u^p \quad \text{in }\mathbb{R}^n, \] the Hardy–Poincaré framework with polynomially weighted measures \(d\mu_\alpha(x)=(1+|x|^2)^\alpha dx\), and the time-fractional Hardy-type equation \[ \partial_t^\alpha(u-u_0)+\Psi_\beta(-i\nabla)u=|x|^{-\gamma}|u|^{p-1}u \quad \text{in }(0,\infty)\times\mathbb{R}^d. \] In these settings, the non-Gaussian character comes from algebraic weights such as \(|x|^\sigma\), polynomially decaying measures, Riesz kernels, or stable Lévy generators with non-Gaussian heat kernels rather than Gaussian profiles [2007.09652] [2201.08159] [1111.1516] [2507.11743]. ## 1. Terminology and principal formulations In the cited literature, the phrase is not restricted to a single canonical equation. It is used for Hardy-type problems in which the Hardy singularity interacts with a non-Gaussian ambient structure. For the Hardy–Hénon equation, the relevant weight is the power law \(|x|^\sigma\): \(\sigma>0\) is Hénon-type, \(\sigma=0\) is Lane–Emden-type, and \(\sigma<0\) is Hardy-type. In the Hardy–Poincaré setting, the non-Gaussian feature is the family of weights \(h_\alpha(x)=(1+|x|^2)^\alpha\), which interpolate between the pure Hardy and Gaussian Poincaré regimes after rescaling. In the time-fractional setting, the non-Gaussian feature is the stable-type pseudo-differential operator \(\Psi_\beta(-i\nabla)\), whose fundamental solution has heavy polynomial tails and anisotropic scaling rather than Gaussian decay [2007.09652] [1111.1516] [2507.11743]. | Formulation | Non-Gaussian feature | Reference | |---|---|---| | \( (-\Delta)^m u = |x|^\sigma u^p \) | power-law Hardy–Hénon weight | [2007.09652] | | \( -\Delta u = |x|^\sigma u^p \) | algebraic weight, full parameter classification | [2201.08159] | | Hardy–Poincaré inequalities | measure \( (1+|x|^2)^\alpha dx \) | [1111.1516] | | \( \partial_t^\alpha(u-u_0)+\Psi_\beta(-i\nabla)u=|x|^{-\gamma}|u|^{p-1}u \) | stable Lévy generator and fractional time | [2507.11743] | A recurrent misconception is that a Hardy equation must place the singular factor on the linear side as a term like \(|x|^{-2}u\). The Hardy–Hénon literature shows that the Hardy weight may instead appear in the nonlinear source \(|x|^\sigma u^p\), while retaining the same analytical role of controlling singularity at the origin and modifying scaling [2007.09652]. This suggests that “Hardy” is best understood structurally rather than syntactically. ## 2. Higher-order Hardy–Hénon equations For the higher-order equation \[ (-\Delta)^m u = |x|^\sigma u^p \quad \text{in }\mathbb{R}^n, \] the basic parameters are the polyharmonic order \(m\in\mathbb N\), the dimension \(n\in\mathbb N\), the weight exponent \(\sigma\in\mathbb R\), and the nonlinearity exponent \(p>1\). A central quantity is \[ \delta := \frac{2m+\sigma}{p-1}, \] which measures the singular behavior of the formal power profile \(u(x)\sim |x|^{-\delta}\). The sharp necessary condition for the existence of nonnegative nontrivial distributional solutions is \[ n-2m-\frac{2m+\sigma}{p-1}>0. \] If this inequality fails, then no nonnegative nontrivial distributional solution exists. When \(n>2m\), this threshold can be rewritten as \[ p>p_c(m,\sigma):=\frac{n+2m+2\sigma}{n-2m}. \] For \(\sigma>-2m\), the condition is sharp for distributional solutions: if \(p>p_c(m,\sigma)\), the obstruction disappears and a distributional solution of power type can be built [2007.09652]. For classical solutions, the same exponent separates the subcritical and critical/supercritical regimes. If \(n\ge 2m\), \(\sigma>-2m\), and \[ 11\). By contrast, if \(n>2m\), \(\sigma>-2m\), and \[ p\ge \frac{n+2m+2\sigma}{n-2m}, \] then positive radially symmetric classical solutions exist [2007.09652]. The exponent \[ p_c(m,\sigma)=\frac{n+2m+2\sigma}{n-2m} \] is the higher-order weighted Hardy–Hénon critical exponent. Its origin is scaling: under \(u_\lambda(x)=\lambda^\alpha u(\lambda x)\), invariance forces \[ \alpha=\frac{2m+\sigma}{p-1}=\delta. \] The weight \(|x|^\sigma\) changes the balance between the operator and the nonlinearity. For \(\sigma>-2m\), the theory is largely complete; \(\sigma=-2m\) is a Hardy borderline, and the data specifically note that Mitidieri–Pohozaev show even punctured supersolutions fail to exist there [2007.09652]. ## 3. Distributional formulation, integral representation, and super polyharmonicity The higher-order theory distinguishes classical, punctured, and distributional solutions. For \(\sigma\ge0\), a classical solution belongs to \(C^{2m}(\mathbb R^n)\); for \(\sigma<0\), it belongs to \(C^{2m}(\mathbb R^n\setminus\{0\})\cap C(\mathbb R^n)\). A punctured solution is \(C^{2m}\) on \(\mathbb R^n\setminus\{0\}\). A distributional solution satisfies \[ u\in L^1_{\mathrm{loc}}(\mathbb R^n), \qquad |x|^\sigma u^p\in L^1_{\mathrm{loc}}(\mathbb R^n), \] and \[ \int_{\mathbb R^n}u(-\Delta)^m\varphi\,dx = \int_{\mathbb R^n}|x|^\sigma u^p\varphi\,dx \quad \forall \varphi\in C_c^\infty(\mathbb R^n). \] Throughout that theory, “solution” means nonnegative and nontrivial [2007.09652]. A decisive structural result is that, for \(n>2m\) and \(\sigma>-2m\), every distributional solution satisfies a Riesz-potential integral equation. With \[ G(x)=C(2m)|x|^{2m-n}, \] one has \[ u(x)=C(2m)\int_{\mathbb R^n}\frac{|y|^\sigma u^p(y)}{|x-y|^{n-2m}}\,dy \quad \text{for a.e. }x\in\mathbb R^n. \] The proof uses a ring condition at infinity, a representation theorem of Caristi–D’Ambrosio–Mitidieri, and testing against radial kernels [2007.09652]. The same paper proves weak and strong super polyharmonicity. The weak form states that for \(i=1,\dots,m-1\) and every nonnegative \(\varphi\in C_c^\infty(\mathbb R^n)\), \[ \int_{\mathbb R^n}u(-\Delta)^{m-i}\varphi\,dx\ge0. \] For classical solutions in the critical and supercritical range, this bootstraps to \[ (-\Delta)^i u>0 \quad \text{in }\mathbb R^n\setminus\{0\}, \qquad i=1,\dots,m-1. \] The super polyharmonic property restores a positivity mechanism that plays the role ordinarily filled by a maximum principle in second-order theory [2007.09652]. The threshold \[ n-2m-\frac{2m+\sigma}{p-1}>0 \] has an additional interpretation: it is exactly the condition under which a punctured or classical solution is automatically a distributional solution. In the Hardy regime \(\sigma<0\), explicit punctured power profiles may satisfy the PDE pointwise while failing to satisfy the distributional formulation when this integrability condition is violated [2007.09652]. ## 4. Exhaustive second-order classification For the second-order Hardy–Hénon equation \[ -\Delta u = |x|^\sigma u^p \quad \text{in }\mathbb R^n, \] the 2022 classification treats arbitrary \(n\ge1\), arbitrary \(p\in\mathbb R\), and arbitrary \(\sigma\in\mathbb R\). For \(n\ge2\), the existence theorem is complete: there exists a nontrivial nonnegative classical solution in \(\mathbb R^n\) if and only if \[ n\ge3,\qquad \sigma>-2,\qquad p\ge \frac{n+2+2\sigma}{n-2}. \] Equivalently, there is no nontrivial nonnegative classical solution in dimension \(n=2\) for any \(p\) and \(\sigma\), and for \(n\ge3\) the supercritical Hardy–Sobolev regime is exactly \[ p\ge p_S(n,\sigma):=\frac{n+2+2\sigma}{n-2},\qquad \sigma>-2. \] This recovers the classical Lane–Emden picture at \(\sigma=0\), the Hénon picture at \(\sigma>0\), and the Hardy regime at \(\sigma<0\) [2201.08159]. The one-dimensional classification is qualitatively different. On the full line \(\mathbb R\), the ODE \[ -u''(x)=|x|^\sigma u(x)^p \] admits a nontrivial nonnegative classical solution if and only if one of the following holds: \[ \sigma<-2 \text{ and } p>-1-\sigma, \qquad\text{or}\qquad -2<\sigma<0 \text{ and } p<-1-\sigma. \] On the half-line \((0,\infty)\), the existence criterion becomes \[ \sigma<-2 \text{ and } p>-1-\sigma, \qquad\text{or}\qquad \sigma>-2 \text{ and } p<-1-\sigma. \] In the one-dimensional existence regimes, there are explicit power-type solutions \[ u_a(x)=c_a|x|^a,\qquad a=\frac{2+\sigma}{1-p},\qquad c_a=\bigl(a(1-a)\bigr)^{1/(p-1)}, \] with \(a\in(0,1)\) precisely in those regimes [2201.08159]. The same paper develops a detailed one-dimensional nonuniqueness theory. In the regime \(\sigma>-2\) and \(p<-1-\sigma\), there exists a one-parameter family \(\{u^\alpha\}_{\alpha>0}\) of positive classical solutions on \((0,\infty)\), all satisfying \[ \lim_{x\searrow0}\frac{u^\alpha(x)}{u_a(x)}=1, \] and strictly ordered by the parameter. A concrete example is the case \(\sigma=1\), \(p=-4\), where \[ u^\alpha(x)=\left(\frac{25}{6}\right)^{1/5}x^{3/5}(1+\alpha x)^{2/5},\qquad \alpha>0, \] all solve \(u''+xu^{-4}=0\) on \(x>0\). In the regime \(\sigma<-2\), \(p>-1-\sigma\), additional asymptotic behaviors occur, and when \(a=1/2\) the paper proves the existence of oscillatory positive solutions for which \(u(x)/u_a(x)\) remains bounded but has no limit as \(x\searrow0\) [2201.08159]. Methodologically, the classification combines spherical averaging, monotonicity, integral estimates, and a one-dimensional Kelvin transform \[ v(r)=r\,u(1/r), \] which maps \[ -u''(r)=r^\sigma u(r)^p \] to the same equation with transformed parameter \[ \tilde\sigma=-p-\sigma-3. \] This duality connects distinct Hardy-type regimes and is one reason the classification is exhaustive [2201.08159]. ## 5. Hardy–Poincaré inequalities and the non-Gaussian Hardy equation as an ODE A complementary use of the term appears in the theory of improved Hardy, Gaussian Poincaré, and Hardy–Poincaré inequalities. For \(\alpha\le0\), define \[ h_\alpha(x)=(1+|x|^2)^\alpha,\qquad d\mu_\alpha(x)=h_\alpha(x)\,dx. \] The Hardy–Poincaré inequality takes the form \[ \Lambda_{\alpha,d}\int_{\mathbb R^d}|u-\mu_{\alpha-1}(u)|^2\,d\mu_{\alpha-1} \le \int_{\mathbb R^d}|\nabla u|^2\,d\mu_\alpha, \] with explicit sharp spectral constant \(\Lambda_{\alpha,d}\). This family interpolates between the classical Hardy inequality as \(\alpha\to0^{-}\) and the Gaussian Poincaré inequality as \(\alpha\to-\infty\) after rescaling [1111.1516]. The paper then studies improved inequalities by recursive “expansion of the square.” In the Hardy setting, this yields the Filippas–Tertikas asymptotic expansion with positive remainder terms \(W_k\). In the Gaussian setting, the quadratic form \[ \mathsf G[u] = \int_{\mathbb R^d}|\nabla u|^2\,d\mu +\frac d2\int_{\mathbb R^d}|u|^2\,d\mu -\frac14\int_{\mathbb R^d}|x|^2|u|^2\,d\mu \] admits an improved inequality \[ \mathsf G[u] \ge \frac14(d-2)^2\int_{\mathbb R^d}\frac{u^2}{|x|^2}\Big(\sum_{k=0}^\infty W_k(t)\Big)\,d\mu, \] with asymptotically optimal coefficients. The same recursive strategy is carried into the Hardy–Poincaré family [1111.1516]. In this framework, the “non-Gaussian Hardy equation” is the radial ODE governing the critical potential obtained from expansion of a square. For the Hardy–Poincaré family, the key equation is \[ (1+r^2)r h'(r)+[(d-2)r^2+d]h(r)-r^2h(r)^2=0. \] It is presented as the non-Gaussian analogue of the Gaussian equation \[ r h'(r)+(d-2)h(r)-h(r)^2=0 \] and the Hardy equation \[ r g'(r)-g(r)^2=0. \] The improved Hardy–Poincaré inequalities are then obtained by solving this equation asymptotically and iterating suitable changes of variables, producing weights \(X_k\), \(Y_k\), \(Z_k\), and \(W_k\) with sharp coefficient \(1\) at each order [1111.1516]. The conceptual significance is that a Hardy equation need not be a semilinear PDE at all. It may also be the Euler–Lagrange or ground-state ODE underlying a weighted quadratic inequality. This broadens the scope of the term while preserving its central analytic themes: singularity, scale, optimal remainder terms, and asymptotic sharpness [1111.1516]. ## 6. Fractional time and non-Gaussian stable generators The time-fractional non-Gaussian Hardy-type equation introduces both memory and nonlocal spatial diffusion: \[ \partial_t^\alpha(u-u_0)(t,x)+\Psi_\beta(-i\nabla)u(t,x) = |x|^{-\gamma}|u(t,x)|^{p-1}u(t,x), \] with \[ \alpha\in(0,1),\qquad \beta\in(0,2),\qquad p>1,\qquad \gamma>0. \] The time derivative is the Riemann–Liouville derivative \[ \partial_t^\alpha v(t)=\frac{d}{dt}(g_{1-\alpha}*v)(t), \qquad g_\rho(t)=\frac{1}{\Gamma(\rho)}t^{\rho-1}, \] and the spatial operator \(\Psi_\beta(-i\nabla)\) is a pseudo-differential operator of order \(\beta\), identified as the generator of a stable, possibly anisotropic, Lévy process. When \(\beta=2\) and the symbol is isotropic, one recovers the classical Laplacian; for \(\beta\in(0,2)\), the heat kernel is non-Gaussian [2507.11743]. The linear problem is governed by kernels \(Z\) and \(Y\), with \(Z\) represented by subordination in terms of the stable heat kernel \(G\) and a time-fractional kernel \(G_\alpha\). The scaling laws are \[ Z(t,x)=t^{-\frac{\alpha d}{\beta}}\,Z\bigl(1,t^{-\frac{\alpha}{\beta}}x\bigr), \qquad Y(t,x)=t^{-\frac{\alpha d}{\beta}+\alpha-1}\,Y\bigl(1,t^{-\frac{\alpha}{\beta}}x\bigr), \] and the pointwise bounds exhibit algebraic tails together with a logarithmic correction at the critical dimension \(d=\beta\). A mild solution is defined by \[ u(t)=S(t)u_0+\int_0^t R(t-s)\bigl(|\cdot|^{-\gamma}|u|^{p-1}u(s)\bigr)\,ds, \] where \(S(t)v=Z(t,\cdot)*v\) and \(R(t)v=Y(t,\cdot)*v\) [2507.11743]. The critical exponent for this theory is \[ q_c:=\frac{d}{\beta-\gamma}(p-1). \] If \[ 0<\gamma<\min\{\beta,d\} \quad\text{and}\quad q>\max\{p,q_c\}, \] then the equation is locally well posed in \(C([0,T];L_q(\mathbb R^d))\) for sufficiently small \(T\). For global small-data theory, the paper assumes \[ q_c\ge \max\Bigl(1,\frac d\beta\Bigr), \qquad \alpha p>p-1, \qquad \max\{p,q_c\}q\Bigl(1+\frac{\beta}{\alpha(d-\gamma)}\Bigr), \] then there exists positive \(u_0\in L_q(\mathbb R^d)\) such that no local positive mild solution exists, and any fixed point fails to belong to \(L_{1,\mathrm{loc}}(\mathbb R^d)\) for any \(t>0\). A corollary shows the same instantaneous blow-up mechanism when the initial datum dominates the critical Hardy profile \[ |x|^{-d/q_c}\mathbf 1_{B(0,R)}. \] Under additional moment and integrability hypotheses, global solutions satisfy the large-time asymptotic expansion \[ u(t,\cdot)-A Z(t,\cdot)-B Y(t,\cdot)\to0 \quad \text{in }L_q, \] with \[ A=\int_{\mathbb R^d}u_0(y)\,dy, \qquad B=\int_0^\infty\int_{\mathbb R^d}|y|^{-\gamma}|u(s,y)|^{p-1}u(s,y)\,dy\,ds. \] This is a fractional-time, non-Gaussian analogue of classical Hardy and Fujita threshold phenomena [2507.11743]. ## 7. Hardy–Sobolev singular potentials, radial transforms, and structural synthesis A related Euclidean formulation places the Hardy singularity on the linear side: \[ -\Delta u-\lambda \frac{u}{|x|^2}=u^p. \] For radial functions, the map \[ v(r)=r^a u(r^b) \] removes the Hardy singularity and transforms the equation into a semilinear equation without the singular potential, in an effective dimension \(M\) and with modified coefficient \(A(\lambda,p)\). In the critical whole-space case \(p=(N+2)/(N-2)\), this reduces the radial Hardy–Sobolev equation to the standard critical equation \[ -\Delta v = N(N-2)\,v^{\frac{N+2}{N-2}} \quad\text{in }\mathbb R^N, \] so the full family of radial positive solutions follows from the Caffarelli–Gidas–Spruck classification. In the subcritical ball case \(1