Abstract
We prove that the 2-body operator \(\gamma _{2}^{\Psi }\) of a fermionic N-particle state \(\Psi \) obeys \(\Vert \gamma _{2}^{\Psi }\Vert _{\textrm{HS}}\le \sqrt{5}N\), which complements the bound of Yang (Rev Mod Phys 34:694, 1962) that \(\Vert \gamma _{2}^{\Psi }\Vert _{\textrm{op}}\le N\). This estimate furthermore resolves a conjecture of Carlen–Lieb–Reuvers (Commun Math Phys 344:655–671, 2016) concerning the entropy of the normalized 2-body operator. We also prove that the Hilbert–Schmidt norm of the truncated 2-body operator \(\gamma _{2}^{\Psi ,T}\) obeys the inequality \(\Vert \gamma _{2}^{\Psi ,T}\Vert _{\textrm{HS}}\le \sqrt{5N\,\textrm{tr}\,(\gamma _{1}^{\Psi }(1-\gamma _{1}^{\Psi }))}\).
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Let \(\left({\mathfrak {h}},\left\langle \cdot ,\cdot \right\rangle \right)\) be a (separable) Hilbert space and consider the space of fermionic N-particle states, \(\bigwedge ^{N}{\mathfrak {h}}\). Given a normalized state \(\Psi \in \bigwedge ^{N}{\mathfrak {h}}\), one defines the 1- and 2-body operators associated to \(\Psi \), \(\gamma _{1}^{\Psi }:{\mathfrak {h}}\rightarrow {\mathfrak {h}}\) and \(\gamma _{2}^{\Psi }:{\mathfrak {h}}\otimes {\mathfrak {h}}\rightarrow {\mathfrak {h}}\otimes {\mathfrak {h}}\), by
and
for any \(\varphi _{1},\varphi _{2},\psi _{1},\psi _{2}\in {\mathfrak {h}}\). Here \(c^{*}\left(\cdot \right)\) and \(c\left(\cdot \right)\) denote the fermionic creation and annihilation operators, which obey the canonical anticommutation relations (CAR)
We recall some well-known properties of \(\gamma _{1}^{\Psi }\) and \(\gamma _{2}^{\Psi }\): Firstly, they are non-negative. This is obvious in the case of \(\gamma _{1}^{\Psi }\), and for \(\gamma _{2}^{\Psi }\) this follows by noting that if \(\left(u_{k}\right)_{k=1}^{\infty }\) is an orthonormal basis for \({\mathfrak {h}}\), then a general tensor \(\Phi \in {\mathfrak {h}}\otimes {\mathfrak {h}}\) can be written as
for \(\Phi _{k,l}=\left\langle \Phi ,u_{k}\otimes u_{l}\right\rangle \), from which it readily follows that the inner product \(\left\langle \Phi ,\gamma _{2}^{\Psi }\Phi \right\rangle \) is given by
where \(c_{k}^{*}=c^{*}\left(u_{k}\right)\) and \(c_{k}=c\left(u_{k}\right)\) denotes the creation and annihilation operators associated to the basis \(\left(u_{k}\right)_{k=1}^{\infty }\).
Secondly, they are trace-class, with
which follows from the identity \(\sum _{k=1}^{\infty }c_{k}^{*}c_{k}=M\,\textrm{id}_{\bigwedge ^{M}{\mathfrak {h}}}\), which holds on \(\bigwedge ^{M}{\mathfrak {h}}\) for any \(M\in {\mathbb {N}}\).
This combined with their non-negativity implies the trivial estimates \(\left\| \gamma _{1}^{\Psi }\right\| _{\textrm{op}}\le N\) and \(\left\| \gamma _{2}^{\Psi }\right\| _{\textrm{op}}\le N\left(N-1\right)\). In the bosonic case these are in fact optimal, but this is not so for the fermionic case: For \(\gamma _{1}^{\Psi }\) one may note that
where we simply added the non-negative quantity \(\left\| c^{*}\left(\varphi \right)\Psi \right\| ^{2}\) and applied the CAR. This shows that \(\left\| \gamma _{1}^{\Psi }\right\| _{\textrm{op}}\le 1\), which is usually described as a consequence of the Fermi exclusion principle, which informally speaking prohibits more than one fermion from occupying any given one-particle state.
In terms of the operators \(c_{k}\) this bound can equivalently be phrased as
which we can also use to improve our bound on \(\left\| \gamma _{2}^{\Psi }\right\| _{\textrm{op}}\) as follows: By the triangle and Cauchy–Schwarz inequalities, we can apply this to equation (1.5) for
which implies that \(\left\| \gamma _{2}^{\Psi }\right\| _{\textrm{op}}\le N\) - a bound which Yang proved was optimalFootnote 1 in the classic paper [7].
1.1 Main results
In this paper we present two results on \(\gamma _{2}^{\Psi }\). The first result concerns \(\left\| \gamma _{2}^{\Psi }\right\| _{\textrm{HS}}\): Given the identity \(\left\| \gamma _{2}^{\Psi }\right\| _{\textrm{tr}}=\textrm{tr}\left(\gamma _{2}^{\Psi }\right)=N\left(N-1\right)\) and the general inequality \(\left\| \gamma _{2}^{\Psi }\right\| _{\textrm{op}}\le N\), one can deduce that the Hilbert–Schmidt norm of \(\gamma _{2}^{\Psi }\) must obey
and since \(\left\| \gamma _{2}^{\Psi }\right\| _{\textrm{tr}}=N\left(N-1\right)\) is an identity and \(\left\| \gamma _{2}^{\Psi }\right\| _{\textrm{op}}\le N\) is optimal, one might suppose that this is at least nearly optimal. Below we will however prove that this is far from the case, as the following holds:
Theorem 1
For any \(N\in {\mathbb {N}}\) and normalized \(\Psi \in \bigwedge ^{N}{\mathfrak {h}}\) it holds that
This result is interesting in two respects: Firstly, it is of the same order (with respect to N) as the optimal bound \(\left\| \gamma _{2}^{\Psi }\right\| _{\textrm{op}}\le N\), even though \(\left\| \gamma _{2}^{\Psi }\right\| _{\textrm{tr}}=N\left(N-1\right)\). This informally implies that although it is possible for \(\gamma _{2}^{\Psi }\) to have eigenvalues of order N, it can not have “too many” large eigenvalues.
Secondly, it is easy to compute that for a Slater state \(\Psi =u_{1}\wedge \cdots \wedge u_{N}\), \(\left\| \gamma _{2}^{\Psi }\right\| _{\textrm{HS}}=\sqrt{2N\left(N-1\right)}\), which is also \(O\left(N\right)\). This suggests that an \(O\left(N\right)\) behavior of the Hilbert–Schmidt norm of fermionic 2-body operators might be a general feature.
This bound also resolves a conjecture of Carlen–Lieb–Reuvers concerning the entropy of the trace-normalized 2-body operator \({\overline{\gamma }}_{2}=\frac{1}{N\left(N-1\right)}\gamma _{2}^{\Psi }\). In [4] they conjectured (Conjecture 2.6) that the entropy of this should always obey a bound of the form
As noted in [4], Jensen’s inequality for the convex function \(x\mapsto -\log \left(x\right)\) implies that
for any density operator \({\overline{\gamma }}\), so as a consequence of Theorem 1 we find that
for any \(N\ge 2\), which is to say \(S\left({\overline{\gamma }}_{2}\right)\ge 2\log \left(N\right)-\log \left(5\right)+o\left(1\right)\) as \(N\rightarrow \infty \).
1.1.1 Truncated 2-Body Operators
Our second result concerns the truncated 2-body operator: It is well-known that \(\Psi \) is a Slater state if and only if \(\gamma _{1}^{\Psi }\) is a projection, i.e. \(\left(\gamma _{1}^{\Psi }\right)^{2}=\gamma _{1}^{\Psi }\), and that in this case the 2-body operator can be expressed in terms of \(\gamma _{1}^{\Psi }\) as
where \(\textrm{Ex}:{\mathfrak {h}}\otimes {\mathfrak {h}}\rightarrow {\mathfrak {h}}\otimes {\mathfrak {h}}\) acts according to \(\textrm{Ex}\left(\varphi \otimes \psi \right)=\psi \otimes \varphi \).
For a general state \(\Psi \) one would expect that if \(\Psi \) is “nearly Slater”, in an appropriate sense, then the same expression should be approximately valid for \(\gamma _{2}^{\Psi }\) - this leads to the Hartree-Fock functional as an approximation to the energy of a fermionic system, for instance.
A way to make this precise is to consider the truncated 2-body operator \(\gamma _{2}^{\Psi ,T}:{\mathfrak {h}}\otimes {\mathfrak {h}}\rightarrow {\mathfrak {h}}\otimes {\mathfrak {h}}\), defined by
and establish that this can be controlled in terms of \(\textrm{tr}\left(\gamma _{1}^{\Psi }\left(1-\gamma _{1}^{\Psi }\right)\right)=\textrm{tr}\left(\gamma _{1}^{\Psi }-\left(\gamma _{1}^{\Psi }\right)^{2}\right)\), which can be viewed as a measure of the “Slaterness” of the state \(\Psi \). A result of this form was first derived by Bach in [1], where he proved the following (see also Bach’s paper [2] and Graf–Solovej’s paper [6] for related generalizations):
Theorem
(Bach [1]). Let \(X:{\mathfrak {h}}\rightarrow {\mathfrak {h}}\) be an orthogonal projection. Then for any normalized \(\Psi \in \bigwedge ^{N}{\mathfrak {h}}\) it holds that
This can be seen as a sort of lower bound for \(\gamma _{2}^{\Psi ,T}\). Below we will complement this result by proving the following Hilbert-Schmidt bound for \(\gamma _{2}^{\Psi ,T}\):
Theorem 2
For any \(N\in {\mathbb {N}}\) and \(\Psi \in \bigwedge ^{N}{\mathfrak {h}}\) it holds that
We remark that at worst \(\textrm{tr}\left(\gamma _{1}^{\Psi }\left(1-\gamma _{1}^{\Psi }\right)\right)\le \textrm{tr}\left(\gamma _{1}^{\Psi }\right)=N\), in which case this bound essentially reduces to that of Theorem 1 (up to the quantity \(\left\| \left(1-\textrm{Ex}\right)\gamma _{1}^{\Psi }\otimes \gamma _{1}^{\Psi }\right\| _{\textrm{HS}}\)), but in view of that theorem this bound is non-trivial for any \(\Psi \) with \(\textrm{tr}\left(\gamma _{1}^{\Psi }\left(1-\gamma _{1}^{\Psi }\right)\right)\ll N\).
2 Proof of the Theorems
To prove Theorem 1 we first observe that by the characterization \(\left\| T\right\| _{\textrm{HS}}=\sup _{A}\left| \textrm{tr}\left(AT\right)\right| \), where the supremum is over all Hilbert–Schmidt operators A with \(\left\| A\right\| _{\textrm{HS}}=1\), it suffices to estimate a quantity of the form \(\left| \textrm{tr}\left(A\gamma _{2}^{\Psi }\right)\right| \). To do this we begin by writing
where \(A_{k,l,m,n}=\left\langle \left(u_{k}\otimes u_{l}\right),A\left(u_{m}\otimes u_{n}\right)\right\rangle \). The Cauchy–Schwarz inequality thus allows us to estimate
where we recognized that \(\sum _{n=1}^{\infty }\left\| c_{n}\Psi \right\| ^{2}=\left\langle \Psi ,{\mathcal {N}}\Psi \right\rangle =N\), and used that as in the proof of the inequality \(\gamma _{1}^{\Psi }\le 1\), we are free to add the quantity \(\left\| \left(\sum _{k,l,m=1}^{\infty }\overline{A_{k,l,m,n}}c_{m}^{*}c_{l}c_{k}\right)^{*}\Psi \right\| ^{2}\) to obtain the anticommutator of \(\sum _{k,l,m=1}^{\infty }\overline{A_{k,l,m,n}}c_{m}^{*}c_{l}c_{k}\) with its adjoint.
The reason for this last step is an argument which was recently used in the study of the correlation energy of Fermi gases in [5], although it has since been learned this was first used by Bell in [3]: Although the CAR does not imply a useful anticommutator identity for sums of an even number of creation and annihilation operators, it does for sums of an odd number of such operators.
In the present case we note the following identity:
Proposition 3
Let \(A_{k,l,m,n}\in {\mathbb {C}}\) for \(k,l,m,n\in {\mathbb {N}}\) obey \(A_{l,k,m,n}=-A_{k,l,m,n}\). Then it holds for any \(n\in {\mathbb {N}}\) that \(T_{n}=\left\{ \left(\sum _{k,l,m=1}^{\infty }\overline{A_{k,l,m,n}}c_{m}^{*}c_{l}c_{k}\right)^{*},\sum _{k,l,m=1}^{\infty }\overline{A_{k,l,m,n}}c_{m}^{*}c_{l}c_{k}\right\} \) is given by
Proof
By applying the CAR we find that for any \(k,l,m,k',l',m'\in {\mathbb {N}}\)
and by standard commutator identities
so
Consequently \(T_{n}\) is given by
The claim now follows by renaming indices and using the assumed antisymmetry of \(A_{k,l,m,n}\). \(\square \)
With this identity we now obtain the central estimate of this paper:
Proposition 4
For any Hilbert–Schmidt operator \(A:{\mathfrak {h}}\otimes {\mathfrak {h}}\rightarrow {\mathfrak {h}}\otimes {\mathfrak {h}}\) and \(N\in {\mathbb {N}}\) it holds that with respect to \(\bigwedge ^{N}{\mathfrak {h}}\)
Proof
We begin by noting that as
and \(\sum _{k,l,m,n}\left| A_{k,l,m,n}-A_{l,k,m,n}\right| ^{2}\le 4\sum _{k,l,m,n}\left| A_{k,l,m,n}\right| ^{2}\), we can assume without loss of generality that \(A_{l,k,m,n}=-A_{k,l,m,n}\).
It then follows from the previous proposition that
and the same argument we applied in equation (1.9) now shows that for any \(\Psi \in \bigwedge ^{N}{\mathfrak {h}}\)
i.e. \(\sum _{m,n=1}^{\infty }\left| \sum _{k,l=1}^{\infty }\overline{A_{k,l,m,n}}c_{l}c_{k}\right| ^{2}\le N\left\| A\right\| _{\textrm{HS}}^{2}\), and as also \(\left\| \sum _{k}\alpha _{k}c_{k}^{*}\right\| _{\textrm{op}}\le \sqrt{\sum _{k}\left| \alpha _{k}\right| ^{2}}\), it similarly holds that
\(\square \)
As noted at the beginning of this section, Theorem 1 immediately follows since we now have from equation (2.2) that
2.1 Proof of Theorem 2
We note the following identity for \(\gamma _{2}^{\Psi ,T}\):
Proposition 5
For any \(\varphi _{1},\varphi _{2},\psi _{1},\psi _{2}\in {\mathfrak {h}}\) it holds that
Proof
By the CAR we have that
which by applying \(\left\langle \Psi ,\left(\cdot \right)\Psi \right\rangle \) yields
Consequently
which implies the claim since \(\left(1\otimes \gamma _{1}^{\Psi }\right)\left(\gamma _{1}^{\Psi }\otimes 1\right)\left(1-\textrm{Ex}\right)=\left(1-\textrm{Ex}\right)\left(\gamma _{1}^{\Psi }\otimes \gamma _{1}^{\Psi }\right)\). \(\square \)
Theorem 2 can now be concluded in the following form:
Proposition 6
For any Hilbert–Schmidt operator \(A:{\mathfrak {h}}\otimes {\mathfrak {h}}\rightarrow {\mathfrak {h}}\otimes {\mathfrak {h}}\), \(N\in {\mathbb {N}}\) and normalized \(\Psi \in \bigwedge ^{N}{\mathfrak {h}}\) it holds that
Proof
By the previous proposition we can write the trace as
so by Cauchy–Schwarz and Proposition 4
Now we note that by the CAR
and
for the claim. \(\square \)
Data Availability
Data sharing not applicable to this article as no datasets were generated or analysed during the current study.
Notes
In fact he proved the stronger statement that \(\left\| \gamma _{2}^{\Psi }\right\| _{\textrm{op}}\le \frac{M-N+2}{M}N\) when \(M=\dim \left({\mathfrak {h}}\right)<\infty \) and M and N are even, and characterized the optimizers in this case.
References
Bach, V.: Error bound for the Hartree–Fock energy of atoms and molecules. Commun. Math. Phys. 147, 527–548 (1992)
Bach, V.: Accuracy of mean field approximations for atoms and molecules. Commun. Math. Phys. 155, 295–310 (1993)
Bell, J.S.: On a conjecture of C. N. Yang. Phys. Lett. 2, 116 (1962)
Carlen, E.A., Lieb, E.H., Reuvers, R.: Entropy and entanglement bounds for reduced density matrices of fermionic states. Commun. Math. Phys. 344, 655–671 (2016)
Christiansen, M.R., Hainzl, C., Nam, P.T.: The Gell–Mann–Brueckner formula for the correlation energy of the electron gas: a rigorous upper bound in the mean-field regime. Commun. Math. Phys. (to appear) (2023)
Graf, G.M., Solovej, J.P.: A correlation estimate with applications to quantum systems with Coulomb interactions. Rev. Math. Phys. 06, 977–997 (1994)
Yang, C.N.: Concept of off-diagonal long-range order and the quantum phases of liquid He and of superconductors. Rev. Mod. Phys. 34, 694 (1962)
Acknowledgements
I am grateful to Phan Thành Nam for informing me of the application of Theorem 1 to the conjecture of Carlen–Lieb–Reuvers, and for providing me with valuable feedback during the writing of this paper. This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)—Project-ID 470903074—TRR 352.
Funding
Open Access funding enabled and organized by Projekt DEAL.
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.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Christiansen, M.R. Hilbert–Schmidt Estimates for Fermionic 2-Body Operators. Commun. Math. Phys. 405, 18 (2024). https://doi.org/10.1007/s00220-023-04914-3
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00220-023-04914-3