Abstract
We re-examine the combined semi-classical and mean-field limit in the N-body fermionic Schrödinger equation with pure state initial data using the Husimi measure framework. The Husimi measure equation involves three residue types: kinetic, semiclassical, and mean-field. The main result of this paper is to provide better estimates for the kinetic and mean-field residue than those in Chen et al. (J Stat Phys 182(2):1–41, http://arxiv.org/abs/1910.09892v4, 2021). Especially, the estimate for the mean-field residue is shown to be smaller than the semiclassical residue by a mixed-norm estimate of the two-particle reduced density matrix factorization. Our analysis also updates the oscillation estimate parts in the residual term estimates appeared in Chen et al. (J Stat Phys 182(2):1–41, http://arxiv.org/abs/1910.09892v4, 2021).
Similar content being viewed by others
1 Introduction
In this article, we consider the following N-particle mean-field Schrödinger equation
where \(\{\textrm{e}_j\}_{j=1}^N\) is a family of orthonormal basis in \(L^2(\mathbb {R}^3)\) and \(\Delta _{x_j}\) is the Laplacian on j-th particle. The initial data in (1.1) is in the form of a Slater determinant, which stays in the antisymmetric subspace \(L^2_a(\mathbb {R}^{3N})\) of \(L^2(\mathbb {R}^3)\) with \(\Vert \psi _{N,0}\Vert _2=1\), where
In the above formulation, \(S_N\) is the symmetric group.
Note that the number of the terms for interaction is of order \(N^2\). Hence, with the mean-field constant 1/N in front of the interaction, we could think that the size of interaction energy is of order N. Since we are interested in the regime where the size of kinetic energy is similar to the size of interaction energy, we have from Tomas–Fermi theory that \( \hbar ^2 N^{5/3} = N. \) This gives \( \hbar = N^{-1/3}. \) For more details, we refer to [5, 6, 8].
As it is difficult to solve the Schrödinger equation in (1.1) numerically when the number of particle N is large, we aim to derive its corresponding effective evolution equation. In fact, we consider the k-particle reduced density matrix where its corresponding integral kernel is given by
where \(1\leqslant k \leqslant N\). Moreover, we denote the expectation of the one-particle observable as follows,
The one-particle reduce density matrix of the initial data given in (1.1) is \( \omega _N = \sum _{j=1}^N |\textrm{e}_j\rangle \langle \textrm{e}_j|, \) where its corresponding integral kernel is \(\omega _N(x;y) = \sum _{j=1}^N \overline{\textrm{e}_j(y)} \textrm{e}_j(x)\).
1.1 Short Review of Mean-Field Limit, \(N\rightarrow \infty \)
It is well known that the Hartree–Fock equation
is used to approximate the Schrödinger equation in the mean-field limit. Here we use the conventional notation \([A,B]:=AB-BA\) for commutator of operators.
The mean field limit result for fixed \(\hbar \) has been given in [17] for short time. Under the scaling \(\hbar = N^{-1/3}\), the rates of convergence in the trace norm and the Hilbert–Schmidt norm are obtained for arbitrary given time in [8] when the initial data is an approximation of the Slater determinant. Later on, the case with mixed state initial data has been considered in [7, 10]. Furthermore, for Coulomb and Riesz potentials, the rate of convergence is obtained in [40, 41]. We refer more references on this topic to [6, 19, 37,38,39] and the references therein.
1.2 Short Review of Semi-classical Limit, \(\hbar \rightarrow 0\)
The Vlasov equation can be obtained via semi-classical limit of the Hartree or the Hartree–Fock equations. It has been first investigated in [32] by using Wigner measure for smooth potentials. Recently, the rates of convergence in the trace norm as well as the Hilbert–Schmidt norm has been studied in [11] with regularity assumptions on the mixed state initial data and a class of regular potentials. The k-particle Wigner measure used in the [11, 32] reads
Some of the recent developments in the semi-classical limit are the following: One can find results for the inverse power law potential in [43], for the rate of convergence in the Schatten norm in [30], for the Coulomb potential and mixed states in [42], and for the convergence in the Wasserstein distance in [28, 29]. Relativistic fermionic system has been studied in [16]. Further analyses of the semi-classical limit can be found in [1,2,3, 6, 20, 34].
1.3 Combined Mean-Field and Semi-classical Limits
Narnhofer and Sewell, and Spohn independently derived Vlasov equation (1.8) from the N-body Schrödinger equation (1.1) with \(\hbar = N^{-1/3}\), in [36, 46]. Without assuming \(\hbar = N^{-1/3}\), a rate of convergence was obtained in [25] in a weak formulation. The rate of convergence of the combined limits was studied in [22,23,24] by using the Wasserstein (pseudo-)distance. Under a generalized Husimi measure framework, the authors in [13] obtained the convergence for regular potentials. Recently, the combined limit for the singular potential case with regular mixed state initial data was obtained in [14].
It is well-known that the Wigner measure in (1.4) is not a (proper) probability measure, as there might be some point having negative sign. (We refer, e.g., [12, 26, 27, 33, 45] for further references on Wigner measure.) It has been shown that the Husimi measure, the convolution of the Wigner measure with a Gaussian function, is a nonnegative probability measure [15, 18, 48]. In particular, from [18, p.21], given a specific Gaussian coherent state, the relation between the Husimi measure and Wigner measure is given by the following convolution: for any \(1 \leqslant k \leqslant N\),
where
The Wigner transform of Hartree (or Hartree–Fock) equation shares the structure of Vlasov equation, see [9, Eq. (6.15)] for example.
In this paper, following the ideas in [13], we study the equation for Husimi measure. Instead of using the classical definition of Husimi measure in (1.5), we consider the following generalized k-particle Husimi measure, which is given for example in [18]: For any \(p,q \in \mathbb {R}^3\) and \(\psi _{N,t} \in L^2_a(\mathbb {R}^{3N})\), the k-particle Husimi measure is given by
Here \(a^*(f^\hbar _{q,p})\) and \(a(f^\hbar _{q,p})\) are standard creation- and annihilation-operator respectivelyFootnote 1 with respect to the coherent state \(f^\hbar _{q,p}\) given by
where f is any given real-valued function satisfying \(\Vert f\Vert _2 =1\).
Remark 1.1
As stated in [18], the k-particle Husimi measure \(m^{(k)}_{N,t}\) describes how many fermions are within the k-semi-classical boxes of length \(\sqrt{\hbar }\) centered at the phase-spaces \((q_1,p_1), \dots , (q_k,p_k)\).
Remark 1.2
If \(f (x)= \pi ^{-3/4} e^{-|x|^2/2}\), [15] shows that the k-particle Husimi measure \(m_{N,t}^{(k)}\) coincides with the \(\mathfrak {m}^{(k)}_{N,t}\) in (1.5).
1.4 Main Result
Let \(\psi _{N,t}\) be the solution to the Schrödinger equation in (1.1) and denote the one-particle Husimi measure of it by \(m_{N,t}:= m_{N,t}^{(1)}\). From [13, Proposition 2.1], we obtain the following identity:
where
Remark 1.3
The two-particle reduced density matrix in (1.7) is given in (1.2).
Our aim is therefore to obtain the convergence from \(m_{N,t}\), in weak sense, to the solution of Vlasov equation \(m_t\) as follows:
where \(\varrho _{t} = \int \textrm{d}p\, m_t(q,p)\) and the initial data \(m_0 (q,p)\) is the one-particle Husimi measure of \(\psi _{N,0}\) given in (1.1).
The following assumptions are needed in this paper.
Assumption H1
-
1.
(Interaction potential) \(V\in L^1(\mathbb {R}^3)\) and \(V(-x) = V(x)\). Furthermore it holds \(\int \textrm{d}{p} (1+|p|^2) |\widehat{V}(p)| < \infty \), where \(\widehat{V}\) is its Fourier transform of V.
-
2.
(Coherent state) \(f \in H^1(\mathbb {R}^3)\cap L^\infty (\mathbb {R}^3)\) satisfies \(\Vert f\Vert _2 = 1\), and has compact support in \(B_{R_1}\) for a given \(R_1>0\).
-
3.
(Initial data) \(m_{N}\) converges weakly to \(m_0\) in \(L^1(\mathbb {R}^3)\). Furthermore, it satisfies
$$\begin{aligned} \int \textrm{d}q\textrm{d}p\, (|p|^2 + |q|) m_{N}(q,p) <\infty \end{aligned}$$(1.9)uniformly for all N.
-
4.
(Initial data) \(\omega _N\), the one-particle density matrix of \(\psi _{N,0}\) satisfies
$$\begin{aligned} \begin{aligned} \sup _{p\in \mathbb {R}^{3}}\frac{1}{1+|p|}\Vert [e^{\textrm{i}p\cdot x}, \omega _{N}]\Vert _{\text {Tr}}&\leqslant CN\hbar ,\\ \Vert [\hbar \nabla , \omega _{N}]\Vert _{\text {Tr}}&\leqslant CN\hbar , \end{aligned} \end{aligned}$$(1.10)where \(\Vert \cdot \Vert _{\text {Tr}}\) is the trace norm.
Remark 1.4
The assumptions in (1.10) can be explained by the nature of the semi-classical structure. More details can be found in [8] where mean field limit has been studied.
With the assumptions presented above, our final goal is to obtain the following theorem:
Theorem 1.1
Let \(m_{N,t}\) be the one-particle Husimi measure defined in (1.6) with \(\psi _{N,t}\) the solution of Schrödinger equation (1.1), and suppose the aforementioned assumptions hold and \(m_{t}\) is the solution of Vlasov equation in (1.8). Then, for any given \(T>0\), \((m_{N,t})_{N\in \mathbb {N}}\) converges to \(m_t\) weakly (*) in \(L^p ((0,T)\times \mathbb {R}^3 \times \mathbb {R}^3)\) for arbitrary \(1\leqslant p \leqslant \infty \).
Remark 1.5
As a consequence, [47, Theorem 6.9] implies, for \(t\geqslant 0\), that \(W_1 (m_{N,t}, m_t ) \rightarrow 0\) as \( N \rightarrow \infty \), where \(W_1\) is 1-Weisserstein distance.
For convenience, we use the Fock space formalism, which will be briefly introduced in Appendix A. By using Husimi transform given in (1.7), the Schrödinger equation for \(\psi _{N,t}\) can be rewritten into the Vlasov type equation for \(m_{N,t}\) with residual terms. More precisely, from the computations in [13], we have
where \(\varrho _{N,t}(q):= \int \textrm{d}{p} m_{N,t}(q,p)\), the kinetic residue \(\widetilde{\mathcal {R}}\), the semi-classical residue \(\mathcal {R}_\textrm{s}\) and the mean-field residue \(\mathcal {R}_\textrm{m}\) are given by
Remark 1.6
The three terms—kinetic residue \(\widetilde{\mathcal {R}}\), semiclassical residue \(\mathcal {R}_\textrm{s}\), and mean-field residue \(\mathcal {R}_\textrm{m}\)—arise due to the following reasons: The kinetic residue \(\widetilde{\mathcal {R}}\) is bounded by the kinetic energy estimate. This term’s name reflects its connection to the kinetic energy used in establishing the bound. The semiclassical residue \(\mathcal {R}_\textrm{s}\) is proven to be small by oscillatory integrals that appear due to the Husimi transform. Similarly, the estimate of mean-field residue \(\mathcal {R}_\textrm{m}\) is transformed into the factorization of two-particle reduced density matrix, which is a characteristic of mean-field behavior. Thus, its name denotes its association with mean-field properties.
Under the assumptions of Theorem 1.1, the estimate of the kinetic residue \(\tilde{\mathcal {R}}\) can be obtained exactly the same as in [13]. (In the updated arXiv version of [13], the oscillation estimates have been corrected, with which the estimates for the residue terms from BBGKY hierarchy as well as the main result still hold true.) In this paper we give a better estimate for \(\tilde{\mathcal {R}}\) in the following proposition.
Proposition 1.1
Under the assumptions 1, 2, and 3 in H1, the following estimate holds
As a consequence, the following holds:
where C depends on \(\varphi , \phi , f\).
The proof will be listed in Sect. 2.
In addition, we obtain the following two propositions for the other two residues:
Proposition 1.2
Under the assumptions 1, 2, and 3 in H1 and let \(\phi , \varphi \) be test functions, then the following inequality holds:
where \(\frac{1}{2}< \alpha _2 < 1\), and C depends on \(\varphi , \phi , f\).
Detailed proof will be given in Sect. 3.
Proposition 1.3
Assuming H1, let \(\varphi , \phi \in C_0^\infty (\mathbb {R}^3)\) be test functions, then the following inequality holds:
where \(\frac{1}{2}< \alpha _1 < 1\), C depends on \(\phi , \varphi , f\).
Detailed proof will be given in Sect. 4.
By using a similar idea in the estimate of the semi-classical residue term, as shown in Lemma 4.1, the estimate of the mean-field residue term can be reduced to the corresponding quantities involving
where \(\omega _{N,t}\) is the solution to Hartree–Fock equation. The terms with \(T_2\) and \(T_3\) can be estimated by the trace norm and Hilbert-Schmidt norm of \(\gamma _{N,t}^{(1)} - \omega _{N,t}\), respectively. To control \(T_1\), we provide a bound of the following mixed norm estimate for two particle density matrix
In [8], the convergence with respect to the trace norm and Hilbert-Schmidt norm of the difference between \(\gamma _{N,t}^{(k)}\) and \(\omega _{N,t}^{(k)}\) are obtained separately with the help of Wick’s theorem for \(k \geqslant 2\). However, in the current framework the mixed norm estimate as listed in (1.18) needs extra efforts. We trace the strategies given in [8] to reduce it to the estimate of the expectation of the number operator \(\mathcal {N}\) along the quantum fluctuation.
The above estimates show that in the sense of distribution \(\mathcal {R}_\textrm{s}\sim \hbar ^{\frac{1}{2}-}\) and \(\mathcal {R}_\textrm{m}\sim \hbar ^{\frac{9}{4}-}\sim N^{-\frac{3}{4}-}\), from which one can observe that the semi-classical and mean field residue terms are not of the same order in the combined limit \(N^{-1}=\hbar ^3\) argument.
This paper is arranged as follows: we prove the estimate for semi-classical residue in Sect. 3, followed by the estimate for mean-field residue in Sect. 4. Then we conclude the proof of Theorem 1.1 in Sect. 5. In the appendices, for reader’s convenience, we list some basic notations and known estimates.
2 Estimate for Kinetic Residue
In this section, we provide the estimate for \(\widetilde{\mathcal {R}}\) in Proposition 1.1.
Proof of Proposition 1.1
Note that
This implies, by using Hödler inequality, that
In the above estimate, we have used the fact that \(\Vert \rho _{N,t}\Vert _{L^\infty (0,T;L^\frac{5}{3}(\mathbb {R}^3))}\leqslant C\), which is a direct result from Appendix B. Moreover, we obtain
\(\square \)
3 Estimate for Semi-classical Residue
In this section, we will estimate the semi-classical residual term under the assumption with \(V\in W^{2,\infty }(\mathbb {R}^3)\) in Proposition 1.2, with which give us the insight to compare the rate between semi-classical and mean-field residuals.
Proof of Proposition 1.2
First, recall that
Since \(\phi (q)\), \(\varphi (p)\) are test functions, we see that
where we applied the fact that \( (2\pi \hbar )^3 \delta _x(y) = \int e^{\frac{\textrm{i}}{\hbar } p \cdot (x-y)} \textrm{d}{p}\). Then, inserting \(\pm \nabla V(q-w_2)\) and we have
where we used integration by part in the second to last equality.
Before advancing, recalling (B.4), we split the integral and obtain the following estimate, \(\forall \alpha _2\in \big (\frac{1}{2},1 \big )\),
where \(\tilde{C}\) depends on \(\Vert \phi \Vert _{W^{s+1,\infty }}\) and \({{\,\textrm{supp}\,}}\phi \).
Now we want to estimate the term \(I_\textrm{s}\) and \(J_\textrm{s}\) separately. We begin by estimating \(I_\textrm{s}\),
Using \( \Vert D^2 V\Vert _{L^\infty }\leqslant C\), (3.2), the definition of \(\gamma ^{(2)}\) with Cauchy-Schwarz inequality, we have
where we use \(i_{s,1}\) to be the term with \( \upchi _{(w_1-u_1)\in \Omega _\hbar ^\alpha }\), and \(i_{s,2}\) to be the other one.
Due to the symmetric property, we can reduce the estimate for \(i_{\textrm{s},1}\) into the following
Similarly, noticing that \(\phi (q)\) has compact support, one obtains the estimate for the term \(i_{\textrm{s},2}\),
To balance the order between \(i_{\textrm{s},1}\) and \(i_{\textrm{s},2}\), the term s is chosen to be
where \(\alpha _2 \in \big (\frac{1}{2},1\big )\). Therefore, we have
Now, to estimate \(J_\textrm{s}\), we recall the estimate in (3.2) and obtain
The estimate for \(j_{\textrm{s},1}\) can be exactly done as in (3.3) for \(i_{\textrm{s},1}\), the same for \(j_{\textrm{s},2}\) as in \(i_{\textrm{s},2}\). Therefore we obtain the same rate for \(J_{\textrm{s}}\) as in (3.4) for \(I_{\textrm{s}}\). This completes the proof. \(\square \)
Remark 3.1
The key step in the estimates of semi-classical residue is in (3.3), with which the computational bugs appeared in [13] can both be fixed by the same technique.
4 Estimate for Mean-Field Residue
In this section, we will estimate the mean-field residue by first showing in Lemma 4.1 that the estimate for mean-field residue term can be reduced to the estimate for the term
where we denote
Then, we prove the estimate for (4.1) in Proposition 4.1 and finally summary the estimation for the mean-field residue in Proposition 1.3.
Lemma 4.1
Let \(\varphi , \phi \in C_0^\infty (\mathbb {R}^3)\). Then, for \(\frac{1}{2}< \alpha _1 < 1\) and \(s ={\Bigg \lceil \frac{3(\alpha _1 {-} \frac{1}{2})}{2(1-\alpha _1)} \Bigg \rceil }\), we have
where the constant C depends on \(\Vert \varphi \Vert _\infty \), \(\Vert \nabla \phi \Vert _{W^{s,\infty }}\), \({{\,\textrm{supp}\,}}\phi \), \(\Vert f\Vert _{L^\infty \cap H^1}\), \({{\,\textrm{supp}\,}}f\).
Proof
Recall that, in (1.12), we defined the mean-field residue such that
Then one obtains
where we use the weighted Dirac-Delta function in the last equality, i.e.,
Now, splitting the domains of \(w_1\) and \(u_1\) into two, namely, with the characteristic functions \(\upchi _{(w_1 - u_1)\in \Omega _\hbar ^{\alpha _1}}\) and \(\upchi _{(w_1 - u_1)\in (\Omega _\hbar ^{\alpha _1})^c}\) whose domain \(\Omega _\hbar ^{\alpha _1}\) is defined in (B.3), we have
First, considering the term \(\text {J}_{\textrm{m}}\), by the change of variable \(\sqrt{\hbar }\,\tilde{q}_2 = w_2 - q_2\), we obtain
where we have used that supp\(f\subset B_{R_1}\). Recall again from Lemma B.4 that we have
for s to be chosen later. Hence, together with Hölder’s inequality we get
Now, we focus on \(\text {I}_{m}\). Using the fact that \( \left| \int \textrm{d}p\, e^{\frac{\textrm{i}}{\hbar } p\cdot (w_1-u_1)} \nabla \phi (p)\right| \leqslant \Vert \nabla \phi \Vert _{L^1}, \) we obtain the following estimate:
which together with
implies immediately that
To balance the order between \(\text {I}_{\textrm{m}}\) and \(\text {J}_{\textrm{m}}\), s is chosen to be
for \(\alpha _1 \in \left[ 0,1\right) \). Therefore, we obtained the desired result:
\(\square \)
Next, we want to bound the term with the ‘mixed’-norm, i.e.,
The following proposition provides the estimate of (4.6):
Proposition 4.1
Let \(\gamma ^{(k)}_{N,t}\) be k-particle reduced density matrix associated with \(\Psi _{N,t}\), \(\omega _{N,t}\) be the solution of the Hartree–Fock equation in (1.3). Suppose the assumption for Theorem 1.1 holds. Then the following inequalities hold for all \(t\in \mathbb {R}\):
and
Furthermore, it holds that
where C depends on t but is independent of N.
The proof of Proposition 4.1 requires the following results from [8], namely:
Lemma 4.2
[8, Lemma 3.1] Let \(\textrm{d}{\Gamma }(O)\) be the second quantization of any bounded operator O on \(L^2(\mathbb {R}^3)\), i.e.
For any \(\Psi \in \mathcal {F}_a\), the following inequalities hold
If furthermore O is a Hilbert-Schmidt operator, we have the following bounds:
Finally, if O is a trace class operator, we obtain
where \(\Vert O\Vert _{\text {Tr}}:= \text {Tr}|O| = \text {Tr}\sqrt{O^*O}\).
Lemma 4.3
[8, Proposition 3.4] Suppose the assumption for Theorem 1.1 holds. Then, there exist constants \(K,c>0\) depending only on potential V such that
Lemma 4.4
[8, Theorem 3.2] Let \(\mathcal {U}_{N}(t;s)\) be the quantum fluctuation dynamics defined in (A.6) and \(\mathcal {N}\) be the number operator. If the assumptions in Lemma 4.3 hold. Then for \(\xi _N\in \mathcal {F}_a\) with \(\langle \xi _N, \mathcal {N}^k \xi _N \rangle \leqslant C\) for any \(k \geqslant 1\), we have the following inequality:
Remark 4.1
Here in this paper we only need the result for initial data \(\xi _N=\Omega \) where \(\Omega \) is the vacuum state give in Appendix A.
Now, we are ready to provide the proof of Proposition 4.1.
Proof of Proposition 4.1
The proof of the inequalities (4.7) and (4.8) follows by modifying Theorem 2.1 of [8]. In particular, from Eq. (4.3) in [8], we obtain
by choosing the appropriate operator O as discussed in [8]. Our results for (4.7) and (4.8) are obtained by applying Lemma 4.4 and taking the assumption that \(\Vert (\mathcal {N}+1)\xi _N\Vert \leqslant C\).
Therefore, it remains to prove for (4.9). As remarked previously, the trace norm and Hilbert-Schmidt norm of the difference between \(\gamma _{N,t}^{(k)}\) and \(\omega _{N,t}^{(k)}\) are obtained separately with the help of Wick’s theorem for \(k \geqslant 2\) in [8]. For our term, however, we do not directly use Wick’s theorem to compute (4.9) as each terms requires similar but still unique method when taking the estimation.
Simplifying the notation \(\mathcal {R}_{t}:= \mathcal {R}_{\mathcal {V}_{N,t}}\), where \(\mathcal {R}_{\mathcal {V}_{N,t}}\) is the Bogoliubov transformation given in (A.4), we have, from the definition of a 2-particle reduced density matrix and (A.4). that
where we use (A.4) in the third equality. Therefore, we obtain, using the fact that \( \langle \bar{\textrm{v}}_{t,x}, {\textrm{u}}_{t,y} \rangle = 0\), \(\langle \textrm{u}_{t,x}, \bar{\textrm{v}}_{t,y} \rangle = 0\), \(\langle \bar{\textrm{v}}_{t,x}, \bar{\textrm{v}}_{t,y}\rangle = {\omega _{N,t}(y;x)}\), and CAR,
Using the fact that \(\Vert \textrm{u}_t\Vert _{\textrm{op}},\Vert \textrm{v}_t\Vert _{\textrm{op}} \leqslant 1\), \(\Vert \textrm{v}_t\Vert _{\textrm{HS}} \leqslant \sqrt{N}\), \(\Vert \omega _{N,t}\Vert _{\text {Tr}} = N\) and the assumption \(\Vert \xi _N\Vert \leqslant 1\), we do the following estimates for the first term from \(\{A_i\}_{i=1}^{16}\) and \(\{B_i\}_{i=1}^{16}\) separately.
Additionally, we have
The estimates for the rest of the terms can be done with similar steps After getting the bound of each terms, we have obtained the following estimates:
Lastly, the final term is estimated as follows:
As a summary, we have
which implies that, for \(O_1\) and \(O_2\) being Hilbert-Schmidt and trace class operators, we get
Applying Lemma 4.4, we obtain the inequalities in Proposition 4.1 as desired. \(\square \)
Finally, we have the following estimate for the mixed-norm. Since it is one of the main contributions of this paper, we write it as a theorem.
Theorem 4.1
Suppose the assumptions given in Proposition 4.1 hold. Then, we have the following estimate
where the constant \(C_t\) depends on potential V and time t.
Proof of Theorem 4.1
Inserting the intermediate terms
the estimate in (4.23) is then reduced into the estimates of the following terms.
-
\(\left( \displaystyle \int \textrm{d}w_1\textrm{d}u_1\ \left[ \mathbb {T}^{(1)} \big |\gamma _{N,t}^{(2)} - \omega ^{(1)}_{N,t} \otimes \omega ^{(1)}_{N,t} \big | (u_1;w_1) \right] ^2 \right) ^\frac{1}{2}\),
-
\( \Vert \omega _{N,t} - \gamma _{N,t}^{(1)}\Vert _{\textrm{HS}} \Vert \omega _{N,t}\Vert _{\text {Tr}}\),
-
\(\Vert \gamma _{N,t}^{(1)}\Vert _{\textrm{HS}} \Vert \omega _{N,t} - \gamma _{N,t}^{(1)}\Vert _{\text {Tr}}\).
Proposition 4.1 implies immediately (4.23) considering the fact that \(\Vert \omega _{N,t}\Vert _{\text {Tr}} \leqslant N\). \(\square \)
Proof of Proposition 1.3
Direct corollary from Theorem 4.1. \(\square \)
5 Proof of Theorem 1.1
In this section, we prove the main theorem. As have been mentioned in the introduction. We will show that, for any \(T>0\), the sequence \(m_{N,t}\) is weakly compact and that any accumulation point \(m_t\) is exactly the solution of the Vlasov equation. For this purpose, we need the following lemma.
Lemma 5.1
Let \(m_{N,t}\) be weak solution of the reformulated Schrödinger equation (1.11) and \(\rho _{N,t}(q):=\int dp\, m_{N,t}(p,q)\). Then there exists a subsequence of \(m_{N,t}\) (without relabeling for convenience) and function \(m_t\) such that as \(N\rightarrow \infty \)
where \(\rho _t=\int \textrm{d}p\, m_t\).
Proof
The estimates in Appendix B imply
where C appeared in this section denotes a positive constant independent of N. And combining interpolation inequality, we have
Therefore, (5.1) is a direct consequence of the above inequality and the moment estimates in Proposition B.1.
Furthermore, Proposition B.1 implies \( \Vert |p|^2 m_{N,t}\Vert _{L^\infty (0,T;L^1(\mathbb {R}^3\times \mathbb {R}^3))}\leqslant C, \) together with (5.3), we arrive at \( \Vert \rho _{N,t}\Vert _{L^\infty (0,T;L^s(\mathbb {R}^3))}\leqslant C \) for \(s\in [1,\frac{5}{3} ]\). Hence there exists a subsequence of \(rho_{N,t}\) (which is not re-labeled for convenience) such that
Owing to \(V\in W^{2,\infty }(\mathbb {R}^3)\) and Young’s convolution inequality, we have for a.e. \(t\in (0,T)\),
Similarly, we obtain that
By means of (1.11), we get
It is easy to see that \(\partial _t(\nabla V*\rho _{N,t}\) satisfies the following equation:
where \((u\otimes _*v)_{ij}=u_i*v_j\) for \((u,v)\in \mathbb {R}^3\times \mathbb {R}^3\). Noticing that
we derive for any test function \(\tilde{\varphi }(q)\in W^{1,3}(\mathbb {R}^3)\) and a.e. \(t\in (0,T)\),
For the second term, applying Proposition 1.1, we have
The estimates above show that
The inequalities (5.4) and (5.5) allow us to apply Aubin–Lions lemma (e.g. in [4, 31]) to infer that (5.2). We mention here that the application of Aubin-Lions lemma is proceeded in a sequence of growing balls, and the convergent subsequence is obtained through diagonal rule. \(\square \)
Proof of Theorem 1.1
With the help of (5.1), (5.2), Propositions 1.1, 1.2 and 1.3, we can take limit \(N\rightarrow \infty \) in the weak formulation of the reformulated Schrödinger equation (1.11). More precisely, for any \(\phi ,\varphi \in C_0^\infty (\mathbb {R}^3)\) and \(\eta \in C_0^\infty [0,T)\), \(m_{N,t}\) satisfies the following equation
Since the sums and products of functions of the form \(\eta (t)\varphi (q)\phi (p)\) are dense in \(C_0^\infty ([0,T)\times \mathbb {R}^3\times \mathbb {R}^3)\), we have showed that the limit of the subsequence is a weak solution of the Vlasov equation. On the other hand, the assumption 1 in H1 implies that \(V\in W^{2,\infty }\), from which one obtains that the Vlasov equation has a unique weak solution as in [21, Theorem 1.1]. Therefore, the whole sequence \(m_{N,t}\) converges weakly. Hence the proof of Theorem 1.1 is completed. \(\square \)
Data Availability
Data sharing not applicable to this article as no datasets were generated or analyzed during the current study.
Notes
Definitions of creation- and annihilation-operator is provided in Appendix A.
See [35] for more pedagogic treatment on the topics.
References
Amour, L., Khodja, M., Nourrigat, J.: The classical limit of the Heisenberg and time-dependent Hartree–Fock equations: the Wick symbol of the solution. Math. Res. Lett. 20(1), 119–139 (2013)
Amour, L., Khodja, M., Nourrigat, J.: The semiclassical limit of the time dependent Hartree–Fock equation: the Weyl symbol of the solution. Anal. PDE 6(7), 1649–1674 (2013)
Athanassoulis, A., Paul, T., Pezzotti, F., Pulvirenti, M.: Strong semiclassical approximation of Wigner functions for the Hartree dynamics. In: Atti della Accademia Nazionale dei Lincei, Classe di Scienze Fisiche, Matematiche e Naturali, Rendiconti Lincei Matematica E Applicazioni, 22 (2011)
Aubin, J.-P.: Analyse mathematique-un theoreme de compacite. C. R. Hebdomad. Seanc. Acad. Sci. 256(24), 5042 (1963)
Bach, V., Breteaux, S., Petrat, S., Pickl, P., Tzaneteas, T.: Kinetic energy estimates for the accuracy of the time-dependent Hartree–Fock approximation with Coulomb interaction. J. Math. Pures Appl. 105(1), 1–30 (2016)
Benedikter, N.: Effective dynamics of interacting fermions from semiclassical theory to the random phase approximation. J. Math. Phys. 63(8), 081101 (2022)
Benedikter, N., Porta, M., Schlein, B.: Mean-field dynamics of fermions with relativistic dispersion. J. Math. Phys. 55(2), 021901 (2014)
Benedikter, N., Porta, M., Schlein, B.: Mean-field evolution of fermionic systems. Commun. Math. Phys. 331(3), 1087–1131 (2014)
Benedikter, N., Porta, M., Schlein, B.: Effective Evolution Equations from Quantum Dynamics, vol. 7. Springer, New York (2016)
Benedikter, N., Jakšić, V., Porta, M., Saffirio, C., Schlein, B.: Mean-field evolution of fermionic mixed states. Commun. Pure Appl. Math. 69(12), 2250–2303 (2016)
Benedikter, N., Porta, M., Saffirio, C., Schlein, B.: From the Hartree dynamics to the Vlasov equation. Arch. Ration. Mech. Anal. 221(1), 273–334 (2016)
Bröcker, T., Werner, R.F.: Mixed states with positive Wigner functions. J. Math. Phys. 36(1), 62–75 (1995)
Chen, L., Lee, J., Liew, M.: Combined mean-field and semiclassical limits of large fermionic systems. J. Stat. Phys. 182(2), 1–41 (2021). arXiv:1910.09892v4
Chong, J.J., Lafleche, L., Saffirio, C.: From many-body quantum dynamics to the Hartree–Fock and Vlasov equations with singular potentials (2021). arXiv:2103.10946
Combescure, M., Robert, D.: Coherent States and Applications in Mathematical Physics. Springer, Heidelberg (2012)
Dietler, E., Rademacher, S., Schlein, B.: From Hartree dynamics to the relativistic Vlasov equation. J. Stat. Phys. 172(2), 398–433 (2018)
Elgart, A., Erdős, L., Schlein, B., Yau, H.-T.: Nonlinear Hartree equation as the mean field limit of weakly coupled fermions. J. Math. Pures Appl. 83(10), 1241–1273 (2004)
Fournais, S., Lewin, M., Solovej, J.P.: The semi-classical limit of large fermionic systems. Calc. Variat. Partial Diff. Equ. 57(4), 105 (2018)
Fröhlich, J., Knowles, A.: A microscopic derivation of the time-dependent Hartree–Fock equation with Coulomb two-body interaction. J. Stat. Phys. 145(1), 23 (2011)
Gasser, I., Illner, R., Markowich, P.A., Schmeiser, C.: Semiclassical, \(t\rightarrow \infty \) asymptotics and dispersive effects for Hartree–Fock systems. ESAIM Math. Model. Numer. Anal. Mod. Math. Anal. Num. 32(6), 699–713 (1998)
Golse, F.: Mean-field limits in statistical dynamics. (2022). arXiv:2201.02005
Golse, F., Paul, T.: The Schrödinger equation in the mean-field and semiclassical regime. Arch. Ration. Mech. Anal. 223(1), 57–94 (2017)
Golse, F., Paul, T.: Mean-field and classical limit for the n-body quantum dynamics with Coulomb interaction. Commun. Pure Appl. Math. (2021)
Golse, F., Paul, T., Pulvirenti, M.: On the derivation of the Hartree equation in the mean field limit: Uniformity in the Planck constant. J. Funct. Anal. 275(7), 1603–1649 (2018)
Graffi, S., Martinez, A., Pulvirenti, M.: Mean-field approximation of quantum systems and classical limit. Math. Models Methods Appl. Sci. 13(01), 59–73 (2003)
Hudson, R.: When is the Wigner quasi-probability density non-negative? Rep. Math. Phys. 6(2), 249–252 (1974)
Kenfack, A., Życzkowski, K.: Negativity of the Wigner function as an indicator of non-classicality. J. Opt. B 6(10), 396 (2004)
Lafleche, L.: Global semiclassical limit from Hartree to Vlasov equation for concentrated initial data. arXiv:1902.08520 (2019)
Lafleche, L.: Propagation of moments and semiclassical limit from Hartree to Vlasov equation. J. Stat. Phys. 177(1), 20–60 (2019)
Laflèche, L., Saffirio, C.: Strong semiclassical limit from Hartree and Hartree–Fock to Vlasov–Poisson equation. arXiv:2003.02926 (2020)
Lions, J.-L.: Quelques Méthodes de Résolution des Problemes aux Limites non Linéaires. Dunod, Paris (1969)
Lions, P.-L., Paul, T.: Sur les measures de Wigner. Rev. Mat. Iberoam. 9(3), 553–618 (1993)
Mandilara, A., Karpov, E., Cerf, N.J.: Extending Hudson’s theorem to mixed quantum states. Phys. Rev. A 79, 062302 (2009)
Markowich, P.A., Mauser, N.J.: The classical limit of a self-consistent quantum-Vlasov equation in 3d. Math. Models Methods Appl. Sci. 3, 109–124 (1993)
Nam, P.T.: Mathematical Quantum Mechanics II. Lecture Notes (2020)
Narnhofer, H., Sewell, G.L.: Vlasov hydrodynamics of a quantum mechanical model. Commun. Math. Phys. 79(1), 9–24 (1981)
Petrat, S.: Derivation of Mean-field Dynamics for Fermions. PhD thesis, LMU Munich (2014)
Petrat, S.: Hartree corrections in a mean-field limit for fermions with Coulomb interaction. J. Phys. A 50(24), 244004 (2017)
Petrat, S., Pickl, P.: A new method and a new scaling for deriving fermionic mean-field dynamics. Math. Phys. Anal. Geom. 19, 1–51 (2016)
Porta, M., Rademacher, S., Saffirio, C., Schlein, B.: Mean field evolution of fermions with Coulomb interaction. J. Stat. Phys. 166(6), 1345–1364 (2017)
Saffirio, C.: Mean-field evolution of fermions with singular interaction. In: Workshop on Macroscopic Limits of Quantum Systems. Springer, New York, pp. 81–99 (2017)
Saffirio, C.: From the Hartree equation to the Vlasov–Poisson system: strong convergence for a class of mixed states. SIAM J. Math. Anal. 52(6), 5533–5553 (2020)
Saffirio, C.: Semiclassical limit to the Vlasov equation with inverse power law potentials. Commun. Math. Phys. 373(2), 571–619 (2020)
Solovej, J.P.: Many body quantum mechanics (2009)
Soto, F., Claverie, P.: When is the Wigner function of multidimensional systems nonnegative? J. Math. Phys. 24(1), 97–100 (1983)
Spohn, H.: On the Vlasov hierarchy. Math. Methods Appl. Sci. 3(1), 445–455 (1981)
Villani, C.: Topics in Optimal Transportation. American Mathematical Society, London (2003)
Zhang, P.: Wigner Measure and Semiclassical Limits of Nonlinear Schrödinger Equations. American Mathematical Society, London (2008)
Acknowledgements
We acknowledge support from the Deutsche Forschungsgemeinschaft through Grant CH 955/4-1. Yue Li supported by National Natural Science Foundation of China (12071212). Jinyeop Lee received funding from the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy (EXC-2111-390814868).
Funding
Open Access funding enabled and organized by Projekt DEAL.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflicts of interest
The authors declare that they have no conflict of interest.
Additional information
Communicated by Michael Kiessling.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendices
Appendix A: Second Quantization
The Fock space formalism and some results from Bogoliubov theory for the proof of this paper are listed in the following.Footnote 2 In particular, as in [9], we will introduce the fermionic Fock space over Hilbert spaces as the following direct sum:
By convention, we say that the vacuum state, denoted as \(\Omega = \{1,0,0,\dots \}\), belongs to \(\mathbb {C}\). For all \(\Psi = \{\psi ^{(n)}\}_{n\in \mathbb {N}} \in \mathcal {F}_a\) and \(\psi ^{(n)} \in L^2_a (\mathbb {R}^{3n})\), we define the number of particle operator on the n-th sector by \(\big (\mathcal {N}\Psi \big )^{(n)} = n \psi ^{(n)}\).
As in [9], the creation and annihilation operators acting on \(\Psi \in \mathcal {F}_a\) is defined as follows: for any \(f \in L^2(\mathbb {R}^3)\)
where \(\psi ^{(n)} \in L^2(\mathbb {R}^{3n})\) for any \(n\in \mathbb {N}\). Additionally, for convenient purposes, the creation and annihilation operators will be represented by its operator-value distribution, \(a^*_x\) and \(a_x\) respectively, so that
Therefore, the canonical anticommutator relation(CAR) is written as
for any \(x,y \in \mathbb {R}^3\).
Observe that for given any \(\Psi , \Phi \in \mathcal {F}_a\), it holds that
Therefore, we write the number of particles operator as \(\mathcal {N}= \int \textrm{d}{x} a^*_x a_x\). Similarly, the integral kernel of the k-particle reduced density matrix is written as follows:
Moreover, the Hamiltonian acting on \(\Psi \in \mathcal {F}_a\) can be written as
where V is the interaction potential. We will denote the operator of the kinetic term as
As presented in [8, 44], for any \(t \geqslant 0\), there exists a unitary transformation \(\mathcal {R}_{\mathcal {V}_{N,t}}: \mathcal {F}_a \rightarrow \mathcal {F}_a\) such that
where \(\textrm{v}_{t,x}:= \sum _{j=1}^N |\overline{\textrm{e}}_{j,t} \rangle \langle \textrm{e}_{j,t}|\) and \(\textrm{u}_{t,x}:= \mathbb {1} - \sum _{j=1}|{\textrm{e}}_{j,t} \rangle \langle \textrm{e}_{j,t}|\), for any orthonormal basis \(\{\textrm{e}_{j,t}\}_{j=1}^N \subset L^2(\mathbb {R}^3)\).
Then, for \(t \geqslant 0\), the solution of the Schrödinger equation is given as
where \(\mathcal {R}_{\mathcal {V}_{N,t}}\) is a unitary Bogoliubov mapping and \(\mathcal {U}_{N}\) is the quantum fluctuation dynamics defined as follows,
Appendix B: A Priori Estimates
In this appendix, we present in this section a sequence of estimates from [13] that will prove useful to our calculation. First, we have the following properties of k-particle Husimi measures from [13, Lemma 2.2]
Lemma B.1
Let \(m^{(k)}_{N,t}\) be the k-particle Husimi measure as defined in (1.6). Then, the following properties hold:
-
1.
\(m^{(k)}_{N,t}(q,p,\dots ,q_k,p_k)\) is symmetric,
-
2.
\(\frac{1}{(2\pi )^{3k}} \int (\textrm{d}q\textrm{d}p)^{\otimes k} m^{(k)}_{N,t}(q,p,\dots ,q_k,p_k) = \frac{N(N-1)\cdots (N-k+1)}{N^k}\),
-
3.
\(\frac{1}{(2\pi \hbar )^{3}} \int \textrm{d}q_k \textrm{d}p_k\ m^{(k)}_{N,t}(q,p,\dots ,q_k,p_k) = (N-k+1) m^{(k-1)}_{N,t}(q,p,\dots ,q_{k-1},p_{k-1}) \),
-
4.
\( 0 \leqslant m^{(k)}_{N,t}(q,p,\dots ,q_k,p_k) \leqslant 1\) a.e.,
where \(1 \leqslant k \leqslant N\).
From [13, Lemma 2.6] and [13, Proposition 2.3], we have the following estimate for the kinetic energy as well as the moment estimate of the 1-particle Husimi measure respectively:
Lemma B.2
Assume \(V \in W^{1,\infty } (\mathbb {R}^3)\), then the kinetic energy is bounded as follows:
where \(\mathcal {K}\) is defined in (A.3) and the constant C depends on \(\Vert \nabla V\Vert _\infty \).
Proposition B.1
For \(t \geqslant 0\), we have the following finite moments:
where \(C>0\) is a constant that depends on initial data \(\int \textrm{d}q\textrm{d}p\, (|q| + |p|^2) m_{N}(q, p) \).
Next, we will present the oscillation estimate from [13, Lemma 2.5] which will be used frequently in our proof:
Lemma B.3
(Bound on localized number operator) Let \(\psi _N \in \mathcal {F}^{(N)}_a\) such that \(\Vert \psi _N\Vert = 1\), and R be the radius of a ball such that the volume is 1. Then we have
where \(\upchi \) is a characteristic function.
Lemma B.4
(Estimate of oscillation) For \(\varphi \in C^\infty _0 (\mathbb {R}^3)\) and
it holds for every \(\alpha \in (0,1)\), \(s \in \mathbb {N}\), and \(x \in \mathbb {R}^3\backslash \Omega _\hbar ^\alpha \),
where C depends on the compact support and the \(C^s\)-norm of \(\varphi \).
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
Chen, L., Lee, J., Li, Y. et al. A Mixed-Norm Estimate of the Two-Particle Reduced Density Matrix of Many-Body Schrödinger Dynamics for Deriving the Vlasov Equation. J Stat Phys 190, 109 (2023). https://doi.org/10.1007/s10955-023-03123-5
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10955-023-03123-5