The Scott correction in Dirac-Fock theory (Q778832)

Knowledge about the ground state energy and the distribution of the involved particles in large Coulomb systems, such as atoms or molecules, is of fundamental interest in atomic physics and quantum chemistry. Recently, there has been interest in establishing the asymptotic expansion of the ground state energy for relativistically described heavy atoms, when the nuclear charge \(Z\) and the number of electrons \(N\) tend to infinity, and the rescaled nuclear charge \(\kappa=\alpha Z\) (where \(\alpha\) denotes Sommerfeld's fine structure constant) is kept fixed. In the present article, the authors describe neutral atoms (\(N=Z\)) in reduced Dirac-Fock theory which can be seen as the relativistic analog of reduced Hartree-Fock theory where the exchange term is neglected, see also [\textit{M. J. Esteban} and \textit{E. Séré}, Ann. Henri Poincaré 2, No. 5, 941--961 (2001; Zbl 1092.81520)]. Their main result is the proof of the Scott correction, i.e., the derivation of the first two terms in the asymptotic expansion of the infimum of the Dirac-Fock functional; more precisely, they recover the relativistic Scott correction for the many-particle Furry operator obtained in [\textit{M. Handrek} and \textit{H. Siedentop}, Commun. Math. Phys. 339, No. 2, 589--617 (2015; Zbl 1327.81303)]. For \(0\leq\kappa<1\), introduce the self-adjoint Coulomb-Dirac operator in the sense of \textit{G. Nenciu} [Commun. Math. Phys. 48, 235--247 (1976; Zbl 0349.47014)], \[ D_\kappa := D_0 - \frac{\kappa}{|x|}\otimes\mathbf{1}_{\mathbb{C}^4} \equiv -i\boldsymbol{\alpha}\cdot\nabla + \beta - \frac{\kappa}{|x|}\otimes\mathbf{1}_{\mathbb{C}^4} \quad \text{in}\ L^2(\mathbb{R}^3:\mathbb{C}^4)\,, \] and, for \(0\leq\rho\in L^1(\mathbb{R}^3)\), the perturbed mean field Dirac operator \[ D_{\kappa,\rho} := D_\kappa + \rho\ast\frac{1}{|x|}\otimes\mathbf{1}_{\mathbb{C}^4}\quad \text{in}\ L^2(\mathbb{R}^3:\mathbb{C}^4)\,. \] For \(\kappa=0\), let \[ \nu_0 := \inf_{\rho\geq0,\,\sqrt\rho\in H^{1/2}(\mathbb{R}^3),\,\int_{\mathbb{R}^3}\rho=1}\nu_0(\rho)\,, \quad \text{where}\ \nu_0(\rho) := \min\{\nu>0:\,0\in\mathrm{spec}(D_{0,\nu\rho})\}\,. \] In [\textit{M. J. Esteban}, \textit{M. Lewin}, and \textit{É. Séré}, ``Dirac-Coulomb operators with general charge distribution. II. The lowest eigenvalue'', Preprint, \url{arXiv:2003.04051}] it is shown that \(\nu_0\in[2/(2/\pi+\pi/2),1]\) and conjectured that \(\nu_0=1\). Next, let \(\gamma\) be a one-particle density matrix in \(L^2(\mathbb{R}^3:\mathbb{C}^4)\) satisfying \(0\leq\gamma\leq1\) and \(\mathrm{tr}\,\gamma=N\). Then the reduced Dirac-Fock functional reads \[ \mathcal{E}_{\kappa,\alpha}^{\mathrm{DF}}[\gamma] := \mathrm{tr}(D_\kappa-1)\gamma + \frac{\alpha}{2}\iint_{\mathbb{R}^3\times\mathbb{R}^3} \frac{\rho_\gamma(x)\rho_\gamma(y)}{|x-y|}\,\mathrm{d}x\,\mathrm{d}y\,, \quad \rho_\gamma(x) := \mathrm{tr}_{\mathbb{C}^4}\,\gamma(x,x)\,. \] Although being unbounded from below, \(\mathcal{E}_{\kappa,\alpha}^{\mathrm{DF}}\) has critical points \(\gamma\) that satisfy the self-consistency equation \([\gamma,D_{\kappa,\alpha\rho_\gamma}] = 0\). The projection onto the positive spectral subspace of \(D_{\kappa,\alpha\rho_\gamma}\) is denoted by \(P_{\kappa,\gamma}^+ := \mathbf{1}_{[0,\infty)}(D_{\kappa,\alpha\rho_\gamma})\). In particular, if \(\mathcal{E}_{\kappa,\alpha}^{\mathrm{DF}}[\gamma]\) is restricted to states \(0\leq\gamma\leq1\) satisfying \((1-P_{\kappa,\gamma}^+)\gamma=0\) and \(\alpha\,\mathrm{tr}\,\gamma\leq\nu\) with \(0\leq\nu<\nu_0\), then the resulting functional is bounded from below for \(\kappa\in[0,1)\) and sufficiently small \(\alpha\) (Lemma 8). The main result (Theorem 9) concerns the asymptotic expansion of \[ E^{\mathrm{DF}}(\kappa,\alpha,N) := \inf_{0\leq\gamma\leq P_{\kappa,\gamma}^+,\,\mathrm{tr}\,\gamma\leq N}\mathcal{E}_{\kappa,\alpha}^{\mathrm{DF}}[\gamma] \] for neutral atoms, i.e., \(\kappa=\nu\). For \(0<\kappa<\nu_0\), they prove \[ E^{\mathrm{DF}}(\kappa,\alpha,N) = c_{\mathrm{TF}} \alpha^2 N^{7/3}+ c_{\mathrm S}(\kappa) + \mathcal{O}(\alpha^2 N^{47/24}) \quad \text{as}\ N\to\infty \quad \text{with}\ \alpha N\to\kappa\,. \] Here, \(c_{\mathrm{TF}}\) denotes the hydrogenic (non-relativistic) Thomas-Fermi energy and \[ c_{\mathrm S}(\kappa) = \frac{\kappa^2}{2} + \left[\mathrm{tr}\,[P_\kappa^+\left(D_\kappa-1\right)P_\kappa^+]_- - \mathrm{tr}\,\left(-\frac{1}{2}\Delta-\frac{\kappa}{|x|}\right)_-\right] \] (with the ``Furry projection'' \(P_\kappa^+=\mathbf{1}_{[0,\infty)}(D_\kappa)\)) equals Scott's original (non-relativistic) coefficient plus the anticipated relativistic lowering, given by the spectral shift between the non-relativistic hydrogen and the Furry operator. Their proof readily generalizes to ions, as soon as the Scott correction for the many-particle Furry operator applies to ions; the corresponding result in the non-relativistic setup is proved in [\textit{V. Bach}, Rep. Math. Phys. 28, No. 2, 213--248 (1989; Zbl 0732.58042)]. Moreover, since the dropped exchange term is (in these units) usually \(\mathcal{O}(N^{-1/3})\), it is expected that this result continues to hold for the (full) Dirac-Fock functional. For the corresponding result in the non-relativistic setup (where also the third term in the expansion is derived), see [\textit{V. Bach}, Commun. Math. Phys. 147, No. 3, 527--548 (1992; Zbl 0771.46038)], [\textit{V. Bach}, Commun. Math. Phys. 155, No. 2, 295--310 (1993; Zbl 0802.47061)], and [\textit{G. M. Graf} and \textit{J. P. Solovej}, Rev. Math. Phys. 6, No. 5a, 977--997 (1994; Zbl 0843.47041)]. The main idea of the proof is to replace the self-consistent projection \(P_{\kappa,\gamma}^+\) by the (fixed) Furry projection \(P_\kappa^+\), to estimate the resulting energy difference (which is essentially contained in Propositions 13 (lower bound on \(E^{\mathrm{DF}}(\kappa,\alpha,N)\) in terms of the Coulomb-Dirac operator plus direct term in the Furry picture) and 14 (existence of \(\gamma_N\) with \(P_{\kappa,\gamma_N}^-\gamma_N=0\) such that \(\mathcal{E}_{\kappa,\alpha}^{\mathrm{DF}}[\gamma_N]\) in the Dirac-Fock picture is comparable to \(\mathcal{E}_{\kappa,\alpha}^{\mathrm{DF}}[d_N]\) for given \(d_N\) with \(P_\kappa^-d_N=0\) in the Furry picture)), and to apply the results of [\textit{M. Handrek} and \textit{H. Siedentop}, Commun. Math. Phys. 339, No. 2, 589--617 (2015; Zbl 1327.81303)]. While the proof of Proposition 13 essentially relies on weighted Schatten space estimates on \(\mathbf{1}_{(-\infty,0]}(D_\kappa)-\mathbf{1}_{(-\infty,0]}(D_\kappa+V)\) for \(V\in L^6(\mathbb{R}^3)\) (Proposition 12), one of the main ingredients in the proof of Proposition 14 is their Theorem 4 which asserts that there is a constant \(c_{\kappa,\nu}>0\) such that \(c_{\kappa,\nu}|D_0| \leq |D_{\kappa,\rho}|\), whenever \(\kappa\in[0,1)\), \(\nu\in[0,\nu_0)\), \(\sqrt\rho\in H^{1/2}(\mathbb{R}^3)\), and \(\int_{\mathbb{R}^3}\rho=\nu\).