Abstract
We prove that the solutions to the discrete nonlinear Schrödinger equation with non-local algebraically decaying coupling converge strongly in \(L^2({\mathbb {R}}^2)\) to those of the continuum fractional nonlinear Schrödinger equation, as the discretization parameter tends to zero. The proof relies on sharp dispersive estimates that yield the Strichartz estimates that are uniform in the discretization parameter. An explicit computation of the leading term of the oscillatory integral asymptotics is used to show that the best constants of a family of dispersive estimates blow up as the non-locality parameter \(\alpha \in (1,2)\) approaches the boundaries.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
The mathematical description of physical phenomena, in many instances, results in the formulation of partial differential equations (PDEs) describing state variables in continuum media. Despite the fact that it is highly unlikely to find exact solutions of many linear or nonlinear PDEs, advances in numerical analysis and scientific computing open the door to find approximate solutions to complex problems. In particular, numerical approximations based on finite difference schemes are constructed by discretizing spatial variables, leading to a system of coupled ordinary differential equations. In this line of research, the objective is then to determine how well the approximate solution evaluated in the grid approximates the solutions of the corresponding PDE.
On the other hand, there are well-known universal models that are inherently discrete. Generically referred to as coupled oscillator systems, they describe phenomena such as localization or synchronization, characteristic of its discrete nature. Best-known examples are the Fermi–Pasta–Ulam–Tsingou model, the discrete nonlinear Schrödinger equation and the Kuramoto model. The first two describe dynamics in a lattice with nearest neighbor interactions, whereas the Kuramoto model addresses synchronization for globally coupled oscillators. These and similar models continue to be studied given their applicability in photonics, lasers, and networks such as the power grid to name some. For such models, a suitable approximation named the long-wave approximation assumes a “smooth” variation of the state variable among neighbor lattices. Specifically in a one-dimensional lattice, this means \(u_{n\pm 1} \approx u_n\). In this regime, it is reasonable to consider continuum approximation. For a 1-d lattice model, the continuum approximation \(u_{n \pm 1} \rightarrow U(x \pm h)\), where \(h>0\) is small, with nearest neighbor coupling \(C(u_{n+1} + u_{n-1})\) leads to a term proportional to \(\frac{\partial ^2U}{\partial x^2}\) and in return, the system of ODEs is then approximated by a PDE.
Recently, there has been an increased interest in the models based on FNLSE. While most of the research deals with continuum models, including numerical computations of solutions in the nonlinear regime, less is known about discrete systems showing global coupling with algebraic decay on the coupling strength with respect to the distance between nodes in the lattice. This work considers such a case in a two-dimensional lattice and centers on the question of the validity of a suitable continuum approximation. This is not always a trivial task as, for instance, invariances and symmetries may arise or be lost. In contrast to the (continuum) nonlinear Schrödinger equation that admits the Galilean boost from which traveling wave solutions emerge, many lattice systems lack translational invariance. It is known that highly localized solutions in a lattice system do not propagate due to the presence of the Peirels–Nabarro potential [10, 26]; for a recent work on FNLSE in this context, see [22]. All this is to point out the challenges and open problems that need to be studied by a combination of analytical and numerical tools. In this contribution, we report what we think are first analytic results on the underlying fundamental question of determining the continuum approximation on the FNLSE in more than one dimension.
2 Statement of the problem
This work concerns the continuum limit of the discrete fractional nonlinear Schrödinger equation (FNLSE)
to the continuum FNLS
as \(h\rightarrow 0+\) where \(\alpha \in (0,2]{\setminus } \{1\},\, p>1,\, \mu = \pm 1,\) and \(u: {\mathbb {R}}^{2+1}\rightarrow {\mathbb {C}},\,u_h:h{\mathbb {Z}}^2\times {\mathbb {R}}\rightarrow {\mathbb {C}}\). Let (2.1), (2.2) be well-posed in some Banach spaces \(X,X_h\), respectively, where \(0<h\le 1\) denotes a discretization parameter. Suppose \(u_{0,h}\in X_h\) is the discretized \(u_0\in X\). Given an interpolation operator \(p_h:X_h \rightarrow X\) and \(T>0\) such that \(u(t),u_h(t)\) denote the well-posed solutions on [0, T], the main problem then reduces to identifying values of \(\alpha ,p\) that allows
The study of evolution equations on \({\mathbb {R}}\) with a general class of interaction kernel was done in [25] where the continuum limit was proved in the weak sense. By applying the analytic tools in [15] that yield dispersive estimates for the discrete Schrödinger evolution that are uniform in h, [16] extended the aforementioned weak convergence to strong convergence in the \(L^2\)-setting (with convergence rates) for \(\alpha =2\) in \({\mathbb {R}}^d,\,d=1,2,3\) and \(\alpha \in (0,2)\setminus \{1\}\) on \({\mathbb {R}}\). The central perspective in [16], upon which we develop, that sharp dispersive estimates that are uniform in h control the difference \(p_h u_h - u\), at least in the scaling-subcritical regime, proved to be fruitful as can be illustrated in various works such as [11] that studied the case \(\alpha =2\) on \({\mathbb {T}}^2\) as the spatial domain, [12] that studied the large box limit for \(\alpha =2\) in \({\mathbb {R}}^d,\, d=2,3\), and [13] that showed the rigorous derivation of the KdV equation from the FPU system. Using a similar idea, the continuum limit of the space-time FNLS was investigated in [7]. Furthermore, see the works of Ignat and Zuazua [17,18,19,20, 31] where novel approaches such as the Fourier filtering and the two-grid algorithm were used.
In practice, obtaining appropriate dispersive estimates reduces to oscillatory integral estimations, which is of central concern in our approach. Unlike the continuum case, the dispersion relation for the discrete evolution has degenerate critical points, which results in weaker dispersion than the continuum Schrödinger evolution. This in return admits weaker Strichartz estimates, which limits the class of nonlinearities that leads to the well-posedness of the corresponding nonlinear equation via the contraction mapping argument. To be more quantitative, let \(U(t) = e^{-it(-\Delta )^{\frac{\alpha }{2}}},\, U_h(t) = e^{-it(-\Delta _h)^{\frac{\alpha }{2}}}\) and \(\Vert f \Vert _{L^p_h}:= h^{\frac{d}{p}}(\sum \limits _{x\in h{\mathbb {Z}}^2}|f(x)|^p)^{\frac{1}{p}}\) for \(p<\infty \) with \(\Vert f \Vert _{L^\infty _h} = \Vert f \Vert _{L^\infty (h{\mathbb {Z}}^2)}\); see Sect. 3 for notations. For \(\alpha =2\), [28, Theorem 1] establishedFootnote 1
where the implicit constant blows up as \(h\rightarrow 0+\), which contrasts with \(\Vert U(t) \Vert _{L^1({\mathbb {R}}^d)\rightarrow L^\infty ({\mathbb {R}}^d)} \simeq |t|^{-\frac{d}{2}}\). Our objective is to obtain Strichartz estimates for the discrete evolution that are uniform in h.
For \(\alpha <2\), [16, Proposition 3.1] obtained
for all \(N\in 2^{\mathbb {Z}}\) with \(N\le 1\) on \(h{\mathbb {Z}}\) where the \(P_N\) denotes the Littlewood–Paley operator. Our goal is to obtain a two-dimensional analog of (2.3). The proof in [16] cannot be directly generalized, since the set of degenerate critical points on \(h{\mathbb {Z}}\) consists of isolated points whose corresponding oscillatory integrals cannot be estimated directly by the Van der Corput lemma. In higher dimensions, the set of degeneracy is geometrically more complicated. In fact, our analysis shows that the degenerate critical points define a one-dimensional embedded smooth submanifold in the torus \([-\pi ,\pi ]^2\) where each singular point admits a unique direction along which the third derivative does not vanish (fold) except at four points (cusp) at which the fourth derivative does not vanish. This observation that a singular point is at worst a cusp is consistent with [1]. It is expected that more severe singularities exist in higher dimensions as the structure of the Hessian of the dispersion relation becomes more complicated. This dimension-dependent geometric complication is purely a remnant of non-locality since the linear evolution of classical Schrödinger operator on \(h{\mathbb {Z}}^d\) splits as the d-fold tensor product on each dimension.
Consider the dispersion relation
and the quantity of interest
where \({\mathbb {T}} =\frac{{\mathbb {R}}}{2\pi {\mathbb {Z}}}= [-\pi ,\pi ]\) and \(\eta \in C^\infty _c(\frac{{\mathbb {T}}^2}{h})\); the dispersion relation of (2.2) is \(w_{h,0}\). [27] showed that when \(m=0,\,\alpha =1\), which corresponds to the dispersion relation of the discrete wave equation, then the quantity of interest decays as \(O(t^{-\frac{2}{3}})\) in \(d=2\) and \(O(t^{-\frac{7}{6}})\) in \(d=3\). When \(m>0,\,\alpha =1\), which corresponds to the discrete Klein–Gordon equation, [1] showed that the quantity of interest decays as \(O(t^{-\frac{3}{4}})\) in \(d=2\), and the result was extended to higher dimensions (\(d=3,4\)) in [5]. When \(m=0,\,\alpha =2\), the time decay of the fundamental solution of the classical discrete Schrödinger equation was shown to be \(O(t^{-\frac{d}{3}})\) in [28].
Our objective is to obtain the sharp time decay of the quantity of interest for \(m=0,\,\alpha \in (1,2)\) in \(d=2\). In particular, it is shown that the oscillatory integral decays as \(O(t^{-\frac{3}{4}})\). The main tool that we adopt is the analysis of Newton polyhedron generated by the Taylor expansion of the phase function \(x\cdot \xi - t w_{h,0}(\xi )\) in an adapted coordinate system, a method pioneered in [30]. Furthermore, the asymptotics in both regimes \(\alpha \rightarrow 1+\) (wave limit) and \(\alpha \rightarrow 2-\) (Schrödinger limit) are studied. To our knowledge, the dependence on the non-local parameter has not been clearly investigated in previous works. To obtain the asymptotics of the leading term of \(O(t^{-\frac{3}{4}})\) as a function of \(\alpha \), we represent the phase function in a superadapted coordinate system to apply results of [8].
The relation of our work to the theory of stability of degenerate oscillatory integrals is subtle. A cursory observation might suggest that a degenerate integral (our quantity of interest) would be stable under a small perturbation in the non-local parameter. However, the phase fails to be smooth for \(\alpha <2\) and therefore becomes large in appropriate norm(s) as the support of \(\eta \) becomes arbitrarily close to the origin. In our approach, it suffices to invoke the stability result [21] under linear perturbations in phase. For more general stability results under analytic or smooth perturbations, see [9, 23]. For the support of \(\eta \) close to the origin, \(\sin z \sim z\) by the small angle approximation, after which one might wish to invoke [3] that obtained sharp dispersive estimates for radial dispersion relations. However, such approximation is not a linear perturbation and hence we handle that case by direct computation.
The paper is organized as follows. Notations and main results are presented in Sect. 3. Assuming the results hold, the desired continuum limit is shown in Sect. 4. The proof of our main proposition is in Sect. 5, followed by a concluding remark in Sect. 6.
3 Main results
To discuss continuum limit, the parameters that yield the well-posedness of (2.1), (2.2) must be identified. For the discrete equation, the linear operator
defines a bounded, nonnegative, self-adjoint operator on \(L^2_h\), and so are its fractional powers given by functional calculus. Equivalently \((-\Delta _h)^{\frac{\alpha }{2}}\) is given by the Fourier multiplier
where the discrete Fourier transform is defined as
for \(\xi \in \frac{{\mathbb {T}}^d}{h}\). Recall the Sobolev space on \(h{\mathbb {Z}}^d\) for \(s \in {\mathbb {R}},\, p \in (1,\infty )\) given by
where
and \(\langle \xi \rangle = (1+|\xi |^2)^{\frac{1}{2}}\) for \(\xi \in \frac{{\mathbb {T}}^d}{h}\). The nonlinearity \(u_h\mapsto |u_h|^{p-1}u_h\) is locally Lipschitz continuous due to \(L^2_h \hookrightarrow L^\infty _h\), which yields an immediate well-posedness of (2.1) in \(L^2_h\) via the contraction mapping argument. For the continuum case, consider the family of self-similar solutions
and observing that \(\{u_\lambda (\cdot ,t)\}_{\lambda >0}\) leaves \({\dot{H}}^{s_c}({\mathbb {R}}^d)\) invariant for all t, one obtains the Sobolev-critical regularity
Our analysis is in the scaling-subcritical regime where the time of existence depends on the Sobolev norm of data. Moreover, suppose the power of nonlinearity is at least cubic.
Lemma 3.1
FNLSE (2.2) is locally well-posed in \(H^s({\mathbb {R}}^2)\) for \(s>s_c\) and \(p \ge 3\) in the subcritical sense. For any \(\alpha>0,\,p>1,\,d\in {\mathbb {N}}\), DNLSE (2.1) is globally well-posed in \(L^2_h\). Moreover, they admit conserved mass and energy functionals given by
Proof
See [14, Theorem 1.1] and [25, Proposition 4.1] for the first and second statement, respectively. \(\square \)
More specifically, our setup is in the mass supercritical and energy subcritical regime, or equivalently,
in which every \(u_0 \in H^{\frac{\alpha }{2}}({\mathbb {R}}^2)\) has a local solution but not necessarily global; for blow-up criteria in the focusing mass supercritical case via localized virial estimates, see [2, 6].
We specify the discretization described in the introduction. For \(h>0\), define \(d_h:L^2({\mathbb {R}}^d)\rightarrow L^2_h\) by
Conversely define \(p_h:L^2_h\rightarrow L^2({\mathbb {R}}^d)\) by
where \(\{e_i\}_{i=1}^d\) generates \({\mathbb {Z}}^d\). The discretization converges to the continuum solution.
Theorem 3.1
Let \(p \ge 3\) and \(\max (\frac{8}{7}, \frac{2(p-1)}{p+1})<\alpha <2\). For any arbitrary \(u_0 \in H^{\frac{\alpha }{2}}({\mathbb {R}}^2)\), let \(u \in C([0,T];H^{\frac{\alpha }{2}}({\mathbb {R}}^2)),\, u_h \in C([0,T];L^2_h)\) be the well-posed solutions from Lemma 3.1 where \(T = T(\Vert u_0 \Vert _{H^{\frac{\alpha }{2}}})>0\). Then there exists \(C_i = C_i(\Vert u_0 \Vert _{H^{\frac{\alpha }{2}}})>0,\, i=1,2\) independent of \(h>0\) such that
Remark 3.1
To estimate the nonlinear part of \(p_hu_h - u\) uniformly in h, we show that an appropriate space-time Lebesgue norm of \(u_h\) is uniformly bounded in \([0,T(\Vert u_0 \Vert _{H^{\frac{\alpha }{2}}})]\) (see Lemma 4.2). However, our proof is insufficient to conclude that a similar uniform bound holds in the energy-critical case, and therefore our method does not extend, at least directly, when \(\alpha = \frac{2(p-1)}{p+1}\).
Remark 3.2
The result is local in time, and thus it is of interest to extend (43.4) such that the estimate holds for \(t \in [0,T_e)\) where \(0<T_e\le \infty \) is the maximal time of existence of (2.2). This extension is not straightforward due to the existence of finite-time blow-up solutions in the mass supercritical regime. For example if \(T_e<\infty \), then \(\lim \limits _{t\rightarrow T_e -} \Vert u(t) \Vert _{H^{\frac{\alpha }{2}}({\mathbb {R}}^2)}=\infty \). Since \(p_h u_h = O_h(1)\) by Lemma 4.1, for all \(h>0\) we have
Remark 3.3
Suppose (2.2) were discretized by another means. Let \(A_h\) be a self-adjoint linear operator on \(L^2_h\) and let \(v_h\in C([0,T];L^2_h)\) be a solution of
Recall that \(u_h(t),u(t)\) are well-posed \(L^2\)-solutions of (2.1), (2.2). If \(\Vert u_h(t)-v_h(t)\Vert _{L^2_h}\lesssim h^\theta \), then \(\Vert p_h v_h(t) - u(t)\Vert _{L^2({\mathbb {R}}^2)} \lesssim h^{\theta ^\prime }\) for some \(\theta ,\theta ^\prime >0\) by (43.4) and the triangle inequality.
It is expected that our approach would apply to a general class of discrete models governed by \(\{A_h\}\). A priori, \(A_h\) is assumed to act on \(L^2_h\) and thus its extension to \(L^2({\mathbb {R}}^d)\) needs to be defined, after which, the limit of \(A_h\) as \(h \rightarrow 0\), if it exists, is considered. Let \(m_h \in A\left( \frac{{\mathbb {T}}^d}{h}\right) \) and define \({\mathcal {F}}_h( A_h f)(\xi ) = m_h(\xi ){\mathcal {F}}_h f (\xi )\) where
Denote \(\nu _h = {\mathcal {F}}_h^{-1}m_h\). Since the Fourier coefficients are absolutely integrable, \(\nu _h\) can be interpreted as a complex Borel measure on \({\mathbb {R}}^d\) given by
where \(\delta _y\) is the Dirac mass at \(y \in h {\mathbb {Z}}^d\). Then for \(f \in L^2_h,\ x \in h{\mathbb {Z}}^d\),
For \(f \in C^\infty _c({\mathbb {R}}^d)\), we have \(f *\delta _{y} (x) = \int _{{\mathbb {R}}^d} f(x-y^\prime ) d\delta _{y}(y^\prime ) = f(x-y)\), and therefore
for \(f \in \bigcup \limits _{p \in [1,\infty ]} L^p({\mathbb {R}}^d)\) and
where \(\Vert \nu _h \Vert _{TV}\) measures the total variation.
Proposition 3.1
Define \(A_h\) as (3.6). Then, \(A_h:L^p({\mathbb {R}}^d)\rightarrow L^p({\mathbb {R}}^d)\) is bounded for all \(p \in [1,\infty ]\) with the operator norms satisfying
Proof
Since \(A_h\) is a convolution against a finite measure with bounded symbol \(m_h\), \(A_h\) is a translation-invariant bounded linear operator on \(L^p({\mathbb {R}}^d)\) for all \(p\in [1,\infty ]\) that satisfies (3.7). Since \(A_h\) is bounded on \(L^p({\mathbb {R}}^d)\), it is bounded on \(L^{p^\prime }({\mathbb {R}}^d)\) by duality. By the Riesz–Thorin theorem, we have
The last equality is given by the fact that any translation-invariant bounded linear operator on \(L^2({\mathbb {R}}^d)\) is given by a bounded multiplier on the Fourier space. \(\square \)
As an example, consider two classes of multipliers
where \(c_{d,\alpha } = \frac{4^{\frac{\alpha }{2}}\Gamma (\frac{d+\alpha }{2})}{\pi ^{\frac{d}{2}}|\Gamma (-\frac{\alpha }{2})|}\). It can be verified that \(m_h\) defines
Proposition 3.2
Both \((-\Delta _h)^{\frac{\alpha }{2}}\) and \(A_h\), where \(A_h\) is as (3.9), are divergent as \(h\rightarrow 0\) in the space of bounded linear operators on \(L^2({\mathbb {R}}^d)\) in the uniform operator topology. On the other hand, \((-\Delta _h)^{\frac{\alpha }{2}},\ A_h\) converge to \((-\Delta )^{\frac{\alpha }{2}}\) strongly in \(L^2\) in the Schwartz class.
Proof
By direct computation,
and hence the divergence by Proposition 3.1. The statements on strong convergence in \(L^2\) follows as [25, Lemma 3.9], noting that
followed by the Dominated Convergence Theorem to interchange the limit as h tends to zero and the integral in \(\xi \), justified by
independent of \(h>0\). \(\square \)
A potential issue with \(m_h\) in (3.8) is that the Euclidean metric \(|\cdot |\) does not yield the physically relevant distance between two points on \(h{\mathbb {Z}}^d\) when \(d \ge 2\). Define \(m_h^q(\xi ) = h^d \sum \limits _{z \in h{\mathbb {Z}}^d{\setminus } \{0\}} \frac{1-\cos \xi \cdot z}{|z|_q^{d+\alpha }}\) where \(|\cdot |_q\) denotes the \(l^q\) norm for \(q \in [1,\infty ]\).
Proposition 3.3
For all \(\xi \in {\mathbb {R}}^d\), \(m_h^q(\xi )\xrightarrow [h\rightarrow 0]{} c |\xi |^\alpha \) for some nonzero constant c if and only if \(q=2\). For all \(q \in [1,\infty ]\setminus \{2\}\), an operator \(A_h^q\) defined by \(m_h^q\) is nonnegative and self-adjoint on \(L^2({\mathbb {R}}^d)\), defined on the dense domain \(H^\alpha ({\mathbb {R}}^d)\).
Proof
Let \(\xi = |\xi |\xi ^\prime \) where \(\xi ^\prime \in S^{d-1}\), the unit sphere. Let \(\rho (\xi ^\prime ) \in {\mathbb {R}}^{d \times d}\) be a rotation operator that takes the \(d^{\text {th}}\) standard basis vector to \(\xi ^\prime \), i.e., \(\xi ^\prime = \rho (\xi ^\prime )e_d\). Then, \(\alpha <2\) justifies the limit of Riemann sum, and by changing variables,
A priori, the integral in the last expression, call it \(I(\xi )\), reduces to a continuous function on \(S^{d-1}\) that is constant if and only if \(q=2\), observing the norm-invariance under rotation if and only if \(q=2\). The domain of \(A_q^h\) consists of \(f \in L^2({\mathbb {R}}^d)\) such that \(|\xi |^\alpha I(\xi ){\widehat{f}} \in L^2\). Observing that \(\inf \limits _{\xi \in {\mathbb {R}}^d}|I(\xi )|>0\) due to the norm equivalence of \(\{|\cdot |_q\}\) and
it follows that \(D(A_h^q) = H^\alpha ({\mathbb {R}}^d)\). That \(A_h^q\) is nonnegative and self-adjoint follows from \(I \ge 0\).
\(\square \)
The continuum limit in higher dimensions, therefore, depends on the geometry of the underlying discrete model. This would potentially lead to complications on a spatial domain with an irregular lattice structure, which we leave as an open-ended thought. To this end, our analysis is restricted to \((-\Delta _h)^{\frac{\alpha }{2}}\).
To show the main result, linear dispersive estimates of the discrete evolution are developed. Let \(\psi \in C^\infty _c((-2\pi ,2\pi );[0,1])\) be an even function where \(\psi = 1\) for \(\xi \in [-\pi ,\pi ]\) and let \(\eta (\xi ):= \psi (|\xi |)-\psi (2|\xi |)\). For dyadic integers \(N \le 1\), define Littlewood–Paley projections given by
where \({\mathcal {F}}\) is the Fourier transform on \({\mathbb {R}}^d\). Since \(\xi \in \frac{{\mathbb {T}}^d}{h}\), \(P_N\) is a smooth projector onto \(\frac{\pi }{2}\frac{N}{h}\le |\xi |\le 2\pi \frac{N}{h}\) and altogether resolves the identity
The sum has an upper bound in N since \(h\xi = O_d(1)\).
Adopting the notations in [1], define subsets of \(M:={\mathbb {T}}^2\setminus \{0\}\) given by
where \(w(\xi ) = \left( \sum \limits _{i=1}^d \sin ^2(\frac{\xi _i}{2})\right) ^{\frac{\alpha }{2}}\). The main proposition concerns a family of frequency-localized dispersive estimates with sharp time decay. Furthermore, the lower bounds of implicit constants blow up both in the wave and in the Schrödinger limit.
Proposition 3.4
For all \(t>0,\, N \le 1,\, 0<h\le 1,\, 1<\alpha <2\), there exists \(0<C_i(\alpha )<\infty ,\, i=1,2,3\) such that
and
For more details on the domain of \(N\in 2^{\mathbb {Z}}\) that satisfy (103.10), see (5.18). By interpolating the estimates in (103.10), one obtains
Corollary 3.1
Assume the hypotheses of Proposition 3.4. Then,
Remark 3.4
Assuming Proposition 3.4, it is straightforward to obtain the Strichartz estimates for the linear evolution by averaging in t, N, which we briefly describe. Suppose \(\Vert U_h(t)P_N \Vert _{L^1_h \rightarrow L^\infty _h} \lesssim _\alpha (\frac{N}{h})^{\beta }|t|^{-\sigma }\) for some \(\beta ,\sigma >0\). Define the Strichartz pair \((q,r) \in [2,\infty ]^2\) by the relation
As in [3, p.1127], define \(\tilde{U}(t) = P_N U_h\left( (\frac{N}{h})^{\frac{\beta }{\sigma }}t\right) P_{\sim N}\) where \(P_{\sim N}:= P_{N/2}+P_N+P_{2N}\). Then, \(\{\tilde{U}(t)\}_{t\in {\mathbb {R}}}\) satisfies the hypotheses of [24, Theorem 1.2] from which follows
Squaring both sides and summing in N,
for \(r \in [2,\infty )\) where the first inequality follows from the Littlewood–Paley inequality on the lattice [15, Theorem 4.2]. As an example, Corollary 3.1 asserts \((\beta ,\sigma ) = (2-\frac{3}{4}\alpha ,\frac{3}{4})\) and hence by (133.13),
The derivative loss occurs for \(\alpha <\frac{8}{3}\) or \(\alpha <2\) on \(h{\mathbb {Z}}^2\) (for all \(h>0\)) or \({\mathbb {R}}^2\), respectively.
For \(v \in {\mathbb {R}}^d\), define \(\Phi _v(\xi ) = v\cdot \xi - w(\xi )\) for \(\xi \in {\mathbb {R}}^d\) and let \(\zeta \in C^\infty _c({\mathbb {R}}^d)\) be a test function. Consider
where \(\tau >0\) without loss of generality, for \(\tau <0\) amounts to taking the complex conjugate of \(J_{\Phi _v,\zeta }\). To show (103.10), observe that
by the Young’s inequality applied to the convolution in \(h{\mathbb {Z}}^d\) where
Change variables \(\xi \mapsto \frac{\xi }{h}\) and define \(\tau = \frac{2^\alpha t}{h^\alpha },\, v=\frac{x}{h\tau }\) to obtain
A priori since \(x \in h{\mathbb {Z}}^d\), it follows that \(v \in \tau ^{-1}{\mathbb {Z}}^d\), which we consider as a subset of \({\mathbb {R}}^d\). If \(\sigma _0>0\) is the sharp decay rate for \(J_{\Phi _v,\zeta }\) in the sense that
holds for all \(\tau \in {\mathbb {R}}\setminus \{0\}\) for some \(C(\zeta )>0\) and no bigger \(\sigma ^\prime >0\) satisfies (3.14), then
Hence, our goal reduces to obtaining (3.14) for a dyadic family of Littlewood–Paley functions. Outside of a neighborhood of the origin, \(\Phi _v\) is analytic, and therefore the major contributions to J are due to critical points \(\xi \in {\mathbb {R}}^d\) that satisfy \(\nabla \Phi _v(\xi )=0\), or equivalently, \(v = \nabla w (\xi )=: v_\xi \). In any arbitrary dimension,
For any \(\alpha >1\), \(|\nabla w(\xi )|\xrightarrow [\xi \rightarrow 0]{}0\) and \(\sup \limits _{\xi \in {\mathbb {T}}^d}|\nabla w(\xi )| = C(\alpha ,d) \in (0,\infty )\). With a slight abuse of notation, define \(\nabla w:{\mathbb {T}}^d\rightarrow {\mathbb {R}}^d\) where \(\nabla w (0)=0 \). Then, \(\nabla w\) is continuous and its compact image defines a light cone. If \(|v|\gg _{\alpha ,d,h} 1\) (spacelike event), then J decays faster than \(\tau ^{-n}\) for any \(n \in {\mathbb {N}}\), by integration by parts, with the implicit constant dependent on n and the distance between v and the light cone (see Lemma 5.1). Inside the light cone (including the boundary), J undergoes an algebraic decay due to critical points. For such v, it is generically true that the corresponding critical point(s) \(\xi \in {\mathbb {T}}^d\) are non-degenerate, and therefore J decays as \(\tau ^{-\frac{d}{2}}\). However, there exists a low-dimensional subset of \({\mathbb {T}}^d\) that retards even further the decay rate of \(\frac{d}{2}\). We consider this problem of resolution of singularities for \(d=2\).
To systematically study the decay and asymptotics of J as a function of v and \(\zeta \), consider the Taylor series expansion of \(\Phi _v\). Let \(\xi \in {\mathbb {T}}^2{\setminus }\{0\}\) and \(v_\xi = \nabla w(\xi )\). Consider \(\Phi _{v_\xi }\) so that \(\nabla \Phi _{v_\xi } (\xi )=0\). Pick \(\zeta \in C^\infty _c\) around \(\xi \) such that \(\xi \) is the unique critical point in the support. Then, \(J_{\Phi _{v_\xi },\zeta }\) has an asymptotic expansion
as \(\tau \rightarrow \infty \) where \(\sigma _0\), or the oscillatory index, is chosen to be the minimal number such that for any neighborhood of \(\xi \), say U, there exists \(\zeta _U \in C^\infty _c(U)\) such that \(d_0(\zeta _U) \ne 0\); in particular, \(\sigma _0\) depends only on the phase, not the smooth bump function. Under some hypotheses, \(\sigma _0,d_0\) are deduced from the higher order Taylor expansion of \(\Phi _{v_\xi }\) (see Lemma 5.6), a process that we briefly describe.
Let \(\Phi \) be a real-valued analytic function on a small neighborhood of the origin. Assume \(\Phi (0)=0,\,\nabla \Phi (0)=0\) and therefore the Taylor expansion of \(\Phi \) at the origin in the multi-index notation is
Define the Taylor support \({\mathcal {T}} = \{\alpha \in {\mathbb {N}}^d: c_\alpha \ne 0\}\) and assume that \(\Phi \) is of finite type, i.e., \({\mathcal {T}} \ne \emptyset \). Define the Newton polyhedron of \(\Phi \), call it \({\mathcal {N}}\), to be the convex hull of
and the Newton diagram \({\mathcal {N}}_d\) to be the union of all compact faces of \({\mathcal {N}}\). Let \({\mathcal {N}}_{pr}\), the principal part of Newton diagram, be the subset of \({\mathcal {N}}_d\) that intersects the bisectrix \(\{x_1 = x_2=\cdots = x_d\}\). Define the principal part of \(\Phi \) (or the normal form) as
Let \(d=d(\Phi ) = \inf \{t: (t,t,\dots ,t)\in {\mathcal {N}}\}\) be the distance from the origin to \({\mathcal {N}}\). Since \(\Phi \) is of finite type, \(0<d<\infty \). Note that \({\mathcal {T}}\) is not invariant under analytic coordinate transformations. Let \(d_x\) be the distance computed in the x coordinate system and define the height of \(\Phi \) as \(h(\Phi )=\sup \limits _{x} d_x\) where the supremum is over all analytic coordinate systems. The coordinate system (x) is adapted if \(d_x = h\). In \({\mathbb {R}}^2\), see [30, Proposition 0.7,0.8] for sufficient conditions for (x) to be adapted. An adapted system need not be unique. To obtain the asymptotics of oscillatory integrals, we work in a superadapted coordinate system defined specifically in dimension two in [8] as a coordinates system in which \(\Phi _{pr}(x,\pm 1)\) have no real roots of order greater than or equal to \(d_x(\Phi )\), possibly except \(x=0\). In particular, if \(d(\Phi )>1\) and \(\Phi _{pr}(x,\pm 1)\) is a quadratic polynomial with no repeated roots, then (x, y) is superadapted.
See the introductions of [1, 8, 21, 30], from which this paper adopts all relevant terminologies, for a brief survey of the relationship between oscillatory integrals and Newton polyhedra. To illustrate these ideas, consider an example. Let \(\Phi (x,y) = x^2+y^2+x^3\). Then,
Since \(\Phi _{pr}(x,\pm 1) = x^2+1\) has no real root, the given coordinates system is superadapted.
4 Continuum limit
The proof of Theorem 3.1 is given.
Lemma 4.1
Let \(\beta \in [0,1],\, p>1,\, d \in {\mathbb {N}}\). The implicit constants in the following estimates are independent of \(h>0\) and dependent only on \(\beta ,d\).
-
1.
\(\Vert d_h f \Vert _{H^\beta _h} \lesssim \Vert f \Vert _{H^\beta ({\mathbb {R}}^d)}\).
-
2.
\(\Vert p_h f_h \Vert _{H^\beta ({\mathbb {R}}^d)} \lesssim \Vert f_h \Vert _{H^\beta _h}\).
-
3.
\(\Vert p_h d_h f - f \Vert _{L^2({\mathbb {R}}^d)} \lesssim h^\beta \Vert f \Vert _{H^\beta ({\mathbb {R}}^d)}\).
-
4.
\(\Vert p_h U_h(t) f_h - U(t)u_0 \Vert _{L^2({\mathbb {R}}^d)} \lesssim h^{\frac{\beta }{1+\beta }}|t|(\Vert f_h \Vert _{H^\beta _h} + \Vert u_0 \Vert _{H^\beta ({\mathbb {R}}^d)})+ \Vert p_h f_h-u_0 \Vert _{L^2({\mathbb {R}}^d)}\). If \(f_h\) is the discretization of \(u_0\), i.e., \(f_h= u_{0,h}\), then \(\Vert p_h U_h(t) u_{0,h} - U(t)u_0\Vert _{L^2({\mathbb {R}}^d)} \lesssim \langle t \rangle h^{\frac{\beta }{1+\beta }}\Vert u_0 \Vert _{H^\beta ({\mathbb {R}}^d)}\).
-
5.
\(\Vert p_h (|u_h|^{p-1}u_h) - |p_h u_h|^{p-1}p_h u_h \Vert _{L^2({\mathbb {R}}^d)} \lesssim h^\beta \Vert u_h \Vert _{L^\infty _h}^{p-1} \Vert u_h \Vert _{H^\beta _h}\).
Proof
Lemma 4.2
Let \(p \ge 3\) and \(\max (\frac{8}{7}, \frac{2(p-1)}{p+1})<\alpha <2\). There exists \(\delta = \delta (\alpha ,p)>0\) sufficiently small such that a Strichartz pair (q, r), defined by (123.12) with \(\sigma = \frac{3}{4}\) and \(q:= \frac{2\alpha }{2-\alpha +\delta }\), satisfies \(p-1<q\) and yields the uniform \(L^\infty _h\) estimate given by
Proof
Let \(s = \frac{\alpha }{2}-(2-\frac{3}{4}\alpha )(\frac{1}{2}-\frac{1}{r})\). Then, it can be verified by direct computation that \(s>\frac{2}{r}\) using \(\delta >0\). Hence, by the Sobolev embedding and (133.13), respectively,
Furthermore, it can be directly verified that \((q,r) \in [2,\infty ]^2\). From the Strichartz pair relation and the definition of q, we have \(r \le \infty \) iff \(\alpha > \frac{8}{7}\). Lastly, by choosing \(\delta >0\) sufficiently small, \(p-1<q\) is satisfied. \(\square \)
Remark 4.1
From (3.3), we have
from which the definition of q in Lemma 4.2 is motivated.
Lemma 4.3
Assume the hypothesis of Theorem 3.1 and let q be given by Lemma 4.2. Given \(u_0 \in H^{\frac{\alpha }{2}}({\mathbb {R}}^2)\), let \(T \simeq \Vert u_0 \Vert _{H^{\frac{\alpha }{2}}}^{-\frac{1}{\frac{1}{p-1}-\frac{1}{q}}}\). By Lemma 3.1, let \(u,u_h\) be the well-posed solutions corresponding to initial data \(u_0,u_{0,h}\). Then, \(u,u_h\) satisfy
Proof
The estimate for u is derived from the proof of local well-posedness by the contraction mapping argument. Similarly the time of existence for the discrete evolution that ensures the estimate (14.1) is \(T_h \sim \Vert u_{0,h} \Vert _{H^{\frac{\alpha }{2}}_h}^{-\frac{1}{\frac{1}{p-1}-\frac{1}{q}}}\). By Lemma 4.1, \(T_h \gtrsim T\) uniformly in h, and therefore \(u,u_h\) are well defined on [0, T]. \(\square \)
Proof of Theorem 3.1
Let \(u_0 \in H^{\frac{\alpha }{2}}({\mathbb {R}}^2)\). There exists \(T \sim \Vert u_0 \Vert _{H^{\frac{\alpha }{2}}}^{-\frac{1}{\frac{1}{p-1}-\frac{1}{q}}}\) and \(u \in C([0,T];H^{\frac{\alpha }{2}}({\mathbb {R}}^2))\), unique in a (smaller) Strichartz space (see [14, Theorem 1.1]), satisfying
For \(h>0\), consider \(u_{0,h} = d_h u_0 \in L^2_h\) and the global solution \(u_h \in C^1([0,\infty );L^2_h)\) given by Lemma 3.1. Similarly as above,
The difference \(p_h u_h(t) - u(t)\) is given by
Following the proof of [16, Theorem 1.1], we have
and altogether,
By the Gronwall’s inequality,
Applying (14.1) to (4.2), we obtain (43.4) where \(C_1,C_2>0\) depend on various parameters including \(\Vert u_0 \Vert _{H^{\frac{\alpha }{2}}}\), but not h. This completes the proof. \(\square \)
5 Proof of Proposition 3.4
Let \(\alpha \in (1,2)\) unless otherwise specified. For \(\zeta \in C^\infty _c(M)\), the quantity \(J_{\Phi _v,\zeta }\) is at worst a non-degenerate integral almost everywhere with respect to the Lebesgue measure on \({\mathbb {R}}^2_v\).
Lemma 5.1
Let \(\zeta \in C^\infty _c(U)\) where \(U \subseteq {\mathbb {R}}^d {\setminus } \{0\}\). For \(v \in {\mathbb {R}}^d\), suppose \(\inf \limits _{\xi \in U}|v-\nabla w(\xi )| \ge m>0\) on U. Then, \(|J_{\Phi _v,\zeta }| \lesssim _{n,m,\zeta } |\tau |^{-n}\) for all \(\tau \in {\mathbb {R}}{\setminus }\{0\},\, n \in {\mathbb {N}}\).
Non-degenerate critical points are treated by the method of stationary phase. A well-known asymptotics [29, Chapter 8, Proposition 6] is given.
Lemma 5.2
Let \(\xi \in M,\, v \in {\mathbb {R}}^2\) satisfy \(v = \nabla w(\xi )\) and \(\det D^2 w(\xi ) \ne 0\). Then, there exists a small neighborhood around \(\xi \) such that for all \(\zeta \in C^\infty _c\) supported in the neighborhood,
as \(\tau \rightarrow \infty \) where
If \(D^2w(\xi )\) is singular, i.e., if \(\xi \in M\) satisfies \(H(\xi ,\alpha ):=\det D^2w(\xi )=0\), then the asymptotic formula of Lemma 5.2 is not applicable; \(\xi =0\) is not considered since the oscillatory integrals from Proposition 3.4 have test functions supported outside of the origin. Note that
is well-defined for \(\xi \in M\) and blows up (in the sense of determinant) as \(\xi \rightarrow 0\). It can be shown directly that \(D^2 w(\xi )\) is not the zero matrix for any \(\xi \in M\). If \(D^2w(\xi )\) is of full rank, then the decay of \(J_{\Phi _{v_\xi },\zeta }\) can be analyzed via Lemmas 5.1 and 5.2, and henceforth suppose \(rank(D^2w(\xi ))=1\). Then, \(H(\xi ,\alpha ) = \tilde{h}(\xi ,\alpha ) h(\xi ,\alpha )\) where
Since \(h,\tilde{h}\) are symmetric under \(\xi _1\mapsto \pm \xi _1,\,\xi _2 \mapsto \pm \xi _2,\, (\xi _1,\xi _2)\mapsto (\xi _2,\xi _1)\), the domain of analysis could be restricted to the first quadrant either above or below the identity \(\xi _1=\xi _2\). By definition, \(\tilde{h}\) is nonzero on M and therefore the roots of H correspond to those of h. Following the approach in [1], the representation of h is in polynomials under the change of variables \(a=\cos \xi _1,\, b= \cos \xi _2\), and is given by
For brevity, let
For \(\alpha \in [1,2)\), since \(\nabla _{(a,b)}h \ne 0\) on \([-1,1)^2\), \(E_\alpha \) is a smooth one-dimensional embedded submanifold by the implicit function theorem. This is false for \(\alpha =2\); a cusp \((a,b)=(0,0)\), which corresponds to \(\xi =(\pm \frac{\pi }{2},\pm \frac{\pi }{2})\), appears when \(\alpha =2\) since \(\nabla _{(a,b)}h(0,0) = 0\). By direct computation, \(E_{\alpha _1}\cap E_{\alpha _2} = \{(0,0)\}\) for all \(\alpha _1,\alpha _2 \in [1,2]\). Hence, there exists a smooth map \(E {\setminus } \{0\} \ni (a,b)\mapsto \alpha \in (1,2)\) satisfying \(h(a,b,\alpha )=0\) by the implicit function theorem, observing that \(\partial _\alpha h(a,b,\alpha ) = (a+b)(ab-1)\) is nonvanishing on \(E {\setminus } \{0\}\). Observe that \(E_\alpha \) consists of two connected components; one component, say \(\Gamma ^1_{(a,b)}(\alpha )\), passes through the origin whereas the other, say \(\Gamma ^2_{(a,b)}(\alpha )\), does not. As \(\alpha \rightarrow 1+\), \(\Gamma ^2_{(a,b)}(\alpha )\) becomes arbitrarily close to \((a,b)=(1,1)\) whose corresponding point in \({\mathbb {T}}^2\), the origin, is not in M. In fact,
See Fig. 1 for the contour plots of \(E_\alpha \).
Let \(\{e_1,e_2\}\) be the standard basis of \({\mathbb {R}}^2\), and with a slight abuse of notation, consider \(\{e_1,e_2\}\) as a global orthonormal frame of \({\mathbb {T}}^2\). For each \(\xi \in M\), let \(\{k_1(\xi ),k_2(\xi )\}\) be an orthonormal basis of \(T_\xi M\), the tangent space at \(\xi \), with the coordinate system (x, y), i.e., for all \(v \in T_\xi M\), there exists unique \((x,y) \in {\mathbb {R}}^2\) such that \(v=xk_1(\xi )+yk_2(\xi )\). Moreover, assume \(\{k_1(\xi ),k_2(\xi )\}\) diagonalizes \(D^2w(\xi )\) as
where \(\partial _x = k_1(\xi ) \cdot \nabla _\xi ,\, \partial _y = k_2(\xi ) \cdot \nabla _\xi \) are the directional derivatives. Suppose \(\partial _{yy}w(\xi )=0\) for all \(\xi \in E\), or equivalently, \(k_2(\xi )\) is the direction along which \(D^2w(\xi )\) is degenerate. Then, it follows that \(\partial _{xx}w(\xi )\ne 0\) since \(D^2w(\xi )\) is not a zero matrix. Due to diagonalization, \(\partial _{xy}w(\xi )=0\).
To further investigate the higher-order directional derivatives, [1, Lemma 5.4] is extended by induction and the product rule for derivatives whose proof is immediate and hence omitted.
Lemma 5.3
For \(m \ge 2\), let \(f \in C^{m+1}(U)\) where \(\xi _0 \in U \subseteq {\mathbb {R}}^d\). Suppose \(D^2 f(\xi _0)\) has rank \(d-1\) and let \(k_d\) be a normalized eigenvector corresponding to eigenvalue zero. Suppose \((k_d(\xi _0)\cdot \nabla _\xi )^j f(\xi _0)=0\) for \(2 \le j \le m\). Then, \((k_d(\xi _0)\cdot \nabla _\xi )^{m+1} f(\xi _0)=0\) if and only if \((k_d(\xi _0)\cdot \nabla _\xi )^{m-1} \det D^2 f(\xi _0)=0\).
The inflection points of \(\sin ^2(\frac{\xi _i}{2})\) persist to exist as singular points of \(D^2w(\xi )\) even when \(\alpha <2\). By symmetry, the qualitative behavior of \(J_{\Phi _v,\zeta }\) is the same near each point in \(K_3\).
Lemma 5.4
If \(\xi \in E_\alpha {\setminus } K_3\), then \(\partial _y^3 w(\xi ) \ne 0\). Moreover, \(\partial _y^3 w(P) = 0\) for all \(P \in K_3\).
Proof
For \(\xi \in E_\alpha \), since \(\nabla _\xi H(\xi ,\alpha )= \tilde{h}(\xi ) \nabla _\xi h (\xi ,\alpha )\) where \(\tilde{h}(\xi ) \ne 0\), it suffices to show \(v \cdot \nabla _\xi h(\xi ,\alpha ) = 0\) implies \(\xi \in K_3\) where v is any scalar multiple of \(k_2(\xi )\). From (5.1), let
Take the former; the proof for the latter is similar and thus is omitted. First assume \(\sin \xi _2 \ne 0\).
In the (a, b) coordinates, \(\nabla _\xi h = -(\sin \xi _1 \partial _a h,\sin \xi _2 \partial _b h)\), and our task reduces to solving
by applying Lemma 5.3 with \(m=2\). Modulo \(\sin \xi _2\), (5.3) is a polynomial equation of degree 3 in b (or a) and therefore can be solved explicitly. The intersection of \(h(a,b,\alpha )=0\) and (5.3) occurs at \((a,b)=(0,0)\).
If \(\xi \) lies at the intersection of \(\sin \xi _2=0\) and \(h(\xi ,\alpha )=0\), then it can be directly verified that the left vector of (5.2) is zero whereas the right vector is a scalar multiple of \(\begin{pmatrix} 1\\ 0 \end{pmatrix}\). Then, the claim can be proved similarly as before. \(\square \)
The higher-order derivatives at critical points determine the height of the phase function.
Lemma 5.5
For all \(\xi \in M\),
Proof
Let \(\xi \notin E_\alpha \). In any given coordinate system, say \((\tilde{x},\tilde{y})\), if \(\partial _{\tilde{x}\tilde{y}}w(\xi ) \ne 0\), then \(d_{(\tilde{x},\tilde{y})}=1\). If \(\partial _{\tilde{x}\tilde{y}}w(\xi ) = 0\), then both \(\partial _{\tilde{x}}^2 w(\xi ),\partial _{\tilde{y}}^2 w(\xi ) \ne 0\) due to non-degeneracy. In either case, \(d_{(\tilde{x},\tilde{y})}=1\). Taking supremum over all such coordinate systems, the first claim has been shown. The rest follows similarly as [1, Lemma 3.1] using Lemma 5.4. In particular, the Newton diagrams for \(\xi \in E_\alpha {\setminus } K_3\) and \(\xi \in K_3\) are given by
respectively. \(\square \)
The computation of heights depends only on the nonzero Taylor coefficients of \(\Phi _v\) and therefore does not reflect the variations on \(\alpha \). However, the leading terms of asymptotics (3.16) depend on \(\alpha \). Define
As in (3.16), I has an asymptotic expansion as \(\epsilon \rightarrow 0+\),
where \(\{r_j\}\) is an increasing arithmetic sequence of positive rational numbers such that the (minimal) \(r_0\) is determined only by the phase function that renders at least one of \(c_0,c_0^\prime ,C_0,C_0^\prime \) nonzero. For \(0 \le m \le \infty \), let \(-\frac{1}{m}\) be the slope of the subset of \({\mathcal {N}}_d\) that the bisectrix intersects. Define \(\Phi _{v_\xi ,pr}^+(x,\pm 1) = \Phi _{v_\xi ,pr}(x,\pm 1)\) if \(\Phi _{v_\xi ,pr}(x,\pm 1)>0\) and zero otherwise. A summary of [8, Theorem 1.1,1.2] that applies to our case is given.
Lemma 5.6
Let \(\xi \in E_\alpha \). Then, \(\Phi _{v_\xi }=\Phi _{v_\xi }(x,y)\) in the coordinate system defined by \(\{k_1(\xi ),k_2(\xi )\}\) is superadapted. The slowest decay of the asymptotics is given by \(\sigma _0 = r_0 = \frac{1}{h(\Phi _{v_\xi })}\). The leading terms have vanishing logarithmic terms, i.e., \(c_0^\prime = C_0^\prime = 0\), and moreover
where \(C_0\) is computed similarly by replacing \(\Phi _{v_\xi }\) by its negative.
Since \(\partial _y^3 w(P)=0\) for all \(P\in K_3\) (see Lemma 5.4), the decay of J is the slowest on \(K_3\). Recalling that \(K_3 \subseteq \bigcap \limits _{\alpha \in (1,2)}E_\alpha \), it is of interest to determine \(d_0(\alpha )\) on \(K_3\).
Lemma 5.7
Let \(\xi \in K_3\) and let \(\zeta \in C^\infty _c\) be supported in a small neighborhood around \(\xi \) in which \(\xi \) is the unique critical point of \(\Phi _{v_\xi }\). Then,
where \(c \in {\mathbb {C}}\setminus \{0\}\) is independent of \(\zeta \) and \(\alpha \).
Proof
Without loss of generality, let \(\xi = (\frac{\pi }{2},\frac{\pi }{2})\). Define \(k_1 = \frac{e_1(\xi )+e_2(\xi )}{\sqrt{2}},\, k_2 = \frac{-e_1(\xi )+e_2(\xi )}{\sqrt{2}}\) where the linear span of \(\{k_1,k_2\}\) is coordinatized in (x, y). Using \((\xi _1,\xi _2)\) as a coordinate for \(\{e_1(0),e_2(0)\}\), we have
We first claim (x, y) defines a superadapted system for \(\Phi _{v_\xi }\) at \((x,y)=(0,0)\). Note that \(\Phi _{v_\xi ,pr} = -w_{pr}\) and
The second order derivatives are
and the third order derivatives are
By direct computation, it can be verified that \(\partial _y^j w(\xi )=0\) for all \(j \ge 2\). These derivatives determine the Newton polyhedron, Newton diagram, and \(w_{pr}\) given by
The bisectrix intersects \({\mathcal {N}}_d\) at \(x=y=\frac{4}{3}=d(\Phi _{v_\xi })>1\). Since \(\Phi _{v_\xi ,pr}(x,\pm 1) = Ax^2+Bx\) has two real roots, \(\{0,-\frac{B}{A}\}\), our first claim has been shown. It follows immediately from Lemma 5.6 that \(r_0 = \sigma _0 = \frac{1}{d(\Phi _{v_\xi })} = \frac{3}{4}\). Moreover, \(c_0^\prime = C_0^\prime = 0\) and
by using (5.6); the computation of \(C_0\), which amounts to replacing \(\Phi _{v_\xi }\) by \(-\Phi _{v_\xi }\), is similar and thus is omitted. The conclusion of lemma follows immediately from the explicit form of \(d_0\) in (45.4) and (5.7). \(\square \)
Remark 5.1
If \(\alpha =2\), \(\partial _{xx}w(\xi )=0\) by (5.5). Consequently, the quadratic term of \(\Phi _{v_\xi ,pr}\) vanishes and the Newton diagram is given by
By analogy with the proof of Lemma 5.7, it is expected that the reciprocal of the distance of this diagram, \(\frac{2}{3}\), yields the sharp decay rate of the corresponding oscillatory integral. Indeed, this expectation coincides with the result obtained in [28].
Lemma 5.8
For all \(\xi \in E_\alpha \setminus K_3\), let \(\zeta \) be as Lemma 5.7 such that its support does not intersect \(K_3\). Then, we have
where \(c \in {\mathbb {C}}\setminus \{0\}\) is independent of \(\zeta ,\alpha ,\xi \).
Proof
By taking \(\partial _y = k_2(\xi )\cdot \nabla _\xi \) on H,
by \(h(\xi ,\alpha ), \partial _y^2 w(\xi )=0\). Since the trace of a matrix is the sum of eigenvalues, \(Tr D^2w(\xi ) = \partial _x^2 w(\xi )\), and therefore
By Lemma 5.5, \(\partial _x^2 w(\xi ),\partial _y^3 w(\xi ) \ne 0\), which yields \(\sigma _0 = r_0= \frac{5}{6}\) and
and therefore the coordinate system (x, y) is superadapted. By Lemma 5.6, it suffices to compute \(c_0\), given by
where the last equation is by (5.9). \(\square \)
It is insightful to apply (85.8) to obtain the series expansion of \(d_0\). For \(\{(a,b):h(a,b,\alpha )=0\}\), it suffices to consider \(a \ge b\) or \(a \le b\) by the symmetry of h under \((a,b) \mapsto (b,a)\). Define the two roots of \(h(a,b,\alpha )=0\) in terms of \(a,\alpha \) as
Several comments regarding \(B_P,B\) are summarized below. The following lemma can be verified by direct computation using (5.10).
Lemma 5.9
For all \(\alpha \in (1,2)\), the curve \(a\mapsto B_P(a,\alpha )\) parametrizes \(\Gamma ^1_{(a,b)}(\alpha ) \cap \{a\le b\}\) and \(a\mapsto B(a,\alpha )\) parametrizes \(\Gamma ^2_{(a,b)}(\alpha ) \cap \{a \ge b\}\). The pointwise convergence \(\lim \limits _{a\rightarrow 0-}B_P(a,\alpha )=0,\, \lim \limits _{\alpha \rightarrow 2-}B_P(a,\alpha )=0\) holds. Furthermore, \(B(\cdot ,\alpha )\) obtains the global maxima on \([\frac{2-\alpha }{\alpha },1]\) at the boundary where \(B(\frac{2-\alpha }{\alpha },\alpha )=B(1,\alpha )=\frac{2-\alpha }{\alpha }\). The global minimum is obtained at \(a_m=(\frac{2-\alpha }{\alpha })^{\frac{1}{2}}\) and \(B(a_m,\alpha ) \ge 1-(1+\sqrt{2})(\alpha -1)\).
Corollary 5.1
Consider \(d_0 = d_0(a,B_P(a,\alpha ),\alpha ,\zeta )\) defined on \(\Gamma ^1_{(a,b)}(\alpha ) \cap \{a\le b\}\) and let \(\zeta \in C^\infty _c\) be supported in a small neighborhood around \((a,B_P(a,\alpha ))\) excluding \((a,b)=(0,0)\). Then, for some \(c\in {\mathbb {C}}\setminus \{0\}\) independent of \(a,\alpha ,\zeta \),
holds where the series converges absolutely for all \(a \in [-1,0)\). The coefficients \(\{a_j\}\) can be computed explicitly; for example, \(a_0(\alpha ) = c^\prime (2-\alpha )^{-\frac{1}{6}}\alpha ^{-\frac{5}{6}}\) where \(c^\prime \) is a nonzero numerical constant independent of \(\zeta ,\alpha \).
Proof
The series expansion (115.11) is shown by the general formula (85.8). The pointwise absolute convergence on \(a\in [-1,0)\) follows from the analyticity of the RHS of (85.8) on \(\xi \in M\) corresponding to \(\Gamma ^1_{(a,b)}(\alpha ) \cap \{a\le b\}\). \(\square \)
Remark 5.2
By direct computation, \(a_j(\alpha )\) contains the term \((2-\alpha )^{-p_j}\) for all \(j \ge 0\) for some \(p_j>0\), and therefore one obtains a singular behavior of the leading term of J as \(\alpha \rightarrow 2-\). Another interesting regime is when \(\xi \rightarrow P\), or equivalently \(a\rightarrow 0-\), along \(E_\alpha \). A qualitative difference between a cusp (\(\xi =P\)) and a fold (\(\xi \ne P\)) is manifested quantitatively by the blow-up \(|a|^{-\frac{1}{3}}\) as \(a \rightarrow 0-\).
On the other hand, consider the case \(\alpha \rightarrow 1+\). By symmetry, consider \(\Gamma ^2_{(a,b)}(\alpha ) \cap \{a \ge b\}\). The asymptotic behavior of \(d_0(a(\alpha ),B(a(\alpha ),\alpha ),\alpha )\) where \(a(\alpha )\in \left[ \frac{2-\alpha }{\alpha },1 \right] \) is computed as \(\alpha \rightarrow 1+\).
Corollary 5.2
Consider \(d_0 = d_0(a(\alpha ),B(a(\alpha ),\alpha ),\alpha ,\zeta )\) and let \(\zeta \in C^\infty _c\) be supported in a small neighborhood around \((a(\alpha ),B(a(\alpha ),\alpha ))\). Then, there exists \(\alpha _0>1\) such that
whenever \(\alpha \in (1,\alpha _0]\) and \(a \in [\frac{2-\alpha }{\alpha },1]\). Furthermore, the implicit constants depend only on \(\alpha _0\).
Proof
Let \(a=1-\tilde{a},\, b=1-\tilde{b}\). By \(a \in [\frac{2-\alpha }{\alpha },1]\) and Lemma 5.9,
Observe that the image of \(\Gamma ^{2}_{(a,b)}\) under the inverse cosine lies in \(\left[ -\frac{\pi }{2},\frac{\pi }{2}\right] ^2\), and it can be verified that for \(z \in [0,\frac{\pi }{2}]\),
Define \((\xi _1,\xi _2) \in M\) as the inverse cosine of \((a(\alpha ),B(a(\alpha ),\alpha ))\), respectively; for definiteness, let \(\xi _i \ge 0\). In the polar coordinate where \(r^2 = \xi _1^2 + \xi _2^2\), (5.13) is used to obtain
where we neglect any powers of \(\alpha \) since they can be uniformly bounded for \(\alpha \in (1,2)\). To estimate \(Tr D^2w\), define \((\tilde{r},\tilde{\theta })\) such that \(\tilde{r}^2 = \tilde{a}^2 + \tilde{b}^2,\, \tan \tilde{\theta } = \frac{\tilde{b}}{\tilde{a}}\). Then,
We claim
The lower bound is trivial since \(\tilde{a},\tilde{b}\ge 0\). The upper bound is equivalent to
which holds uniformly on \(\alpha \in (1,2)\) if \(\tilde{r} \le \frac{1}{3}\). Hence for all \(\tilde{a},\tilde{b}\) sufficiently small,
by (5.12). Likewise for sufficiently small \(\tilde{a},\tilde{b}\)
For all \(\epsilon >0\), since \(\Gamma ^2_{(a,b)}(\alpha ) \subseteq \{(a,b) \in [1-\epsilon ,1]\times [1-\epsilon ,1]\}\) whenever \(\alpha \in (1,\alpha _0(\epsilon )]\) for some \(\alpha _0(\epsilon )>1\), there exists \(\alpha _0>1\) sufficiently close to 1 such that all small angle approximations are justified (see (5.13)) and
by (85.8), (5.14), (5.15), and (5.16). Combining with
the proof is complete. \(\square \)
Remark 5.3
As can be seen in Fig. 1, the trajectory \(\alpha \mapsto \left( \frac{2-\alpha }{\alpha },\frac{2-\alpha }{\alpha }\right) \) traces the intersection of \(\Gamma ^2_{(a,b)}\) and the bisectrix. For \(a(\alpha ) = \frac{2-\alpha }{\alpha }\), an explicit computation yields
for some \(c \in {\mathbb {C}}\setminus \{0\}\) independent of \(\alpha ,\zeta \). Suppose \(supp(\zeta )\) is sufficiently small such that \(\zeta (a(\alpha ),a(\alpha ))=1\). Then, note that \(d_0 \xrightarrow [\alpha \rightarrow 2-]{}\infty \) as \(\left( \frac{2-\alpha }{\alpha },\frac{2-\alpha }{\alpha }\right) \) approaches the origin, which corresponds to the cusp \(K_3\). Furthermore, \(d_0 \xrightarrow [\alpha \rightarrow 1+]{}0\) as \(\left( \frac{2-\alpha }{\alpha },\frac{2-\alpha }{\alpha }\right) \) approaches \((a,b)=(1,1)\), which corresponds to the origin of \({\mathbb {T}}^2\) where \(w(\xi )\) blows up.
Another example of trajectory, given by \((a,b)=(1,\frac{2-\alpha }{\alpha })\) with the leading term
shows a different qualitative behavior as \(\alpha \rightarrow 2-\).
Proof of Proposition 3.4
For all \(\tau >0\),
by the triangle inequality. Hence \(\tau \ge 1\).
Considering \(supp \left( \eta (\frac{\cdot }{N})\right) = \{|\xi | \in [\frac{\pi }{2}N,2\pi N]\}\) and
obtained at
respectively, define \(N_\alpha \in 2^{\mathbb {Z}}\) to be the largest number satisfying
Note that \(N_\alpha \) increases as \(\alpha \) increases with \(\lim \limits _{\alpha \rightarrow 1+}N_\alpha =0\) and \(N_2 = 2^{-3}\). Using the support condition of \(\eta (\frac{\cdot }{N})\), the set of \(N\in 2^{\mathbb {Z}}\) that satisfies the RHS of (103.10) is given by
Suppose \(N >N_\alpha \). For every \(\xi \in {\mathbb {T}}^2 {\setminus } B(0,\frac{r_\alpha }{2})\), there exists a neighborhood \(\Omega _\xi (\alpha )\) containing \(\xi \) and a constant \(C_\xi (\alpha )>0\) such that for all \(\zeta \in C^\infty _c(\Omega _\xi )\),
by [21, Theorem 1.1]. By compactness, an open cover \(\{\Omega _\xi \}\) of \({\mathbb {T}}^2{\setminus } B(0,\frac{r_\alpha }{2})\) reduces to a finite subcover \(\{\Omega _{\xi _j}\}_{j=1}^{n_0}\), where \(n_0 = n_0(\alpha )\in {\mathbb {N}}\), and let \(\{\phi _j\}_{j=1}^{n_0}\) be a (\(\alpha \)-dependent) partition of unity subordinate to the finite subcover. Note that if \(\xi \in K_1\), then \(U_\xi \cap K_2 = \emptyset \), and if \(\xi \in K_2\), then \(U_\xi \cap K_3 = \emptyset \) since the oscillatory indices of \(\Phi _{v_\xi }\) at \(\xi \) are distinct on each \(K_j\). Consequently each \(\xi \in K_3\) contributes to the finite subcover for all \(\alpha \in (1,2)\).
Let \(\eta _{N,j}(\cdot ) = \eta (\frac{\cdot }{N})\phi _j(\cdot )\). Then, \(\Vert \eta _{N,j}\Vert _{C^3}\lesssim N^{-3}\) where the implicit constant depends on the given partition of unity. Since \(N>N_\alpha \), we have \(N^{-3} \le C(\alpha )N^{2-\frac{3}{4}\alpha }\) where \(C(\alpha ) = N_\alpha ^{-5+\frac{3}{4}\alpha }\).
By Lemma 5.5 and \(\tau \ge 1\), the slowest decay occurs on \(K_3\) with \(h(\Phi _v) = \frac{4}{3}\). By Lemma 5.5, (5.19), and the triangle inequality,
For \(N>N_\alpha \), a similar argument using the partition of unity and (5.19) yields (103.10) with sharp decay rates \(\sigma _0 \in \{\frac{3}{4},\frac{5}{6},1\}\).
Suppose \(N \le N_\alpha \le \frac{1}{8}\). Recalling that \(v=\frac{x}{h\tau }\), do a change of variable \(\xi \mapsto N \xi \) to obtain
By adopting the proof of [3, Proposition 1], we obtain sharp dispersive estimates of a free solution governed by a non-smooth, non-homogeneous dispersion relation. A change of variables
with \(J_c(z) = \left( \left( 1-\left( \frac{Nz_1}{2}\right) ^{2} \right) \left( 1-\left( \frac{Nz_2}{2}\right) ^{2}\right) \right) ^{-\frac{1}{2}}\), yields
where we denote \((r,\theta )\) and \((\rho ,\phi )\) as the polar coordinates for \(x=(x_1,x_2)\) and \(z=(z_1,z_2)\), respectively, and
where \(\lambda = \rho r\) and
By the support condition of \(\eta \),
and the small angle approximation \(z\le \sin ^{-1}z \le 2z\) on \(z\in [0,\frac{1}{\sqrt{2}}]\), one obtains
since \(\frac{N |z_i|}{2} = |\sin (\frac{N \xi _i}{2})| \le \frac{1}{\sqrt{2}}\). When clear in context, we use the same symbols \(\eta ,J_c\) for the representations in different variables. We prove
from which the proof is complete by interpolating with the trivial bound \(|I| \lesssim N^2\).
Let \(r_0>0\), independent of N, to be specified later and suppose \(r \le r_0\). Integration by parts yields
Since \(|G|\lesssim 1\) and the domain of integration is supported away from the origin,
By the chain rule \(\partial _\rho = \cos \phi \partial _{z_1}+\sin \phi \partial _{z_2}\) and the estimate,
one obtains
By repeated applications of the product and chain rule,
for all \(k \ge 0\) and therefore
altogether implying
Suppose \(r>r_0\). From (5.21),
For a fixed \(\rho >0\), denote \(\Phi _G(\phi ) = \Phi _G(\rho ,\phi )\). The critical points correspond to the roots of \(\partial _\phi \Phi _G\), and are the solutions to
By direct computation, \(g(\rho ,\cdot )\) is a strictly positive \(\pi \)-periodic even function such that \(g(\rho ,\frac{\pi }{4}+\frac{k\pi }{2})=1\) for all \(k \in {\mathbb {Z}}\). Furthermore, \(g(\rho ,\phi )>1\) if \(\phi \in [0,\frac{\pi }{4})\), \(g(\rho ,\phi )<1\) if \(\phi \in (\frac{\pi }{4},\frac{\pi }{2}]\), and
Therefore, \(\tan \phi< g(\rho ,\phi )\tan \phi <1\) on \((0,\frac{\pi }{4})\) and \(g(\rho ,\phi )\tan \phi < \tan \phi \) on \((\frac{\pi }{4},\frac{\pi }{2})\). By symmetry, suppose \(x_i \ge 0\) for \(i=1,2\), and therefore \(0\le \tan \theta \le \infty \). By graphing \(\phi \mapsto g(\rho ,\phi )\tan \phi \) on \([0,2\pi )\), there exist two solutions, \(\phi _{\pm }(\rho )\), where \(\phi _+ \in [0,\frac{\pi }{2}]\) and \(\phi _{-} = \phi _+ + \pi \). A crude estimate \(|\phi _{\pm } - \theta _{\pm }|\le \frac{\pi }{4}\) follows from a further inspection of the graph where \(\theta _+ = \theta ,\, \theta _{-}=\theta +\pi \).
The critical points are non-degenerate. If \(\phi \) satisfies \(\partial _\phi ^2 \Phi _G=0\), then \(g(\rho ,\phi )^3 \cot \phi = -\tan \theta \). Assuming \(\partial _{\phi }^2 \Phi _G(\phi _{\pm })=0\) and substituting (5.24), we have \(g(\rho ,\phi _{\pm })^2 + \tan ^2\phi _{\pm }=0\), a contradiction. Since \(\phi _{\pm }\) and \(\theta _{\pm }\) are in the same quadrants, we have \(\cos \theta _{\pm }\cos \phi _{\pm },\,\sin \theta _{\pm }\sin \phi _{\pm }\ge 0\), and therefore
independent of \(N,\rho \).
Construct \(\chi _{\pm } \in C^\infty _c({\mathbb {R}}_\phi )\) given by \(\chi _{+} = 1\) on \([0,\frac{\pi }{2}]\), supported in \((-\frac{\pi }{8},\frac{5\pi }{8})\), \(\chi _{-} = 1\) on \([\pi ,\frac{3\pi }{2}]\), supported in \((\frac{7\pi }{8},\frac{13\pi }{8})\), and define \(\chi _0 = 1-(\chi _{+}+\chi _{-})\). Since \(|\phi _{\pm } - \theta _{\pm }|\le \frac{\pi }{4}\), we have \(\chi _{\pm }(\phi _{\pm })=1\) for all \(N \le N_\alpha \). Let \(\tilde{\chi }_{\pm }:=\eta J_c \chi _{\pm }\) and \(\tilde{\chi }_0:= \eta J_c \chi _{0}\). Note that since
for all \(k\ge 0\) independent of \(N,\rho \), so are the higher-order partial derivatives (in \(\phi \)) of \(\tilde{\chi }_{\pm }\). Define
and similarly for \(G_0\), and hence \(G = G_+ + G_{-}+ G_0\). By [29, Chapter VIII, Proposition 3], \(G_{\pm }\) has the asymptotics as \(\lambda \rightarrow \infty \) given by
More precisely, for all \(k \in {\mathbb {N}}\cup \{0\}\), there exists \(\lambda _0(k),\,C(k)>0\) such that
for all \(\lambda \ge \lambda _0\). Since the estimates (5.25), (5.26) are uniform with respect to \(N,\rho \), the constants \(\lambda _0,C\) can be chosen to be independent of \(N,\rho \). Since \(\rho \le 2\pi \), let \(r_0 \ge \frac{\max (\lambda _0(0),\lambda _0(1))}{2\pi }\).
Away from the critical points, the integral in \(\phi \) yields a rapid decay in \(\lambda \). We claim
for all \(\lambda >0\) and \(k\ge 1\) uniformly in \(N,\rho \). Since
and \(|\partial _\phi \Phi _G|\ge |\sin (\phi -\theta )| \ge \sin (\frac{\pi }{8})\) for all \(\phi \in [\frac{5\pi }{8},\frac{7\pi }{8}]\cup [\frac{13\pi }{8},\frac{15\pi }{8}]\) and \(\theta \in [0,\frac{\pi }{2}] \cup [\pi ,\frac{3\pi }{2}]\), (5.29) is shown for \(k=1\) by the triangle inequality; for \(k\ge 2\), (5.29) is shown by repeated use of integration by parts.
For \(r\ge r_0\), consider the integral in (5.23) with G replaced by \(G_0\). Since by (5.29),
the integration by parts yields an estimate consistent with (5.22). By replacing G by \(\tilde{G}_{\pm }\) in the same integral, the bound (5.22) follows by (5.28).
It remains to show that I with G replaced by the leading term of \(G_+\) in (5.27) satisfies (5.22); the analysis on \(G_{-}\) is similar and therefore is omitted. Consider
where
The region of integration
is included in the case
since \(|\partial _\rho (\Phi _G(\phi _+)\rho )|\simeq 1\) uniformly in \(N,\rho \) as can be observed in (5.31). For \(r,\xi \) satisfying (5.30), the lower bound of the phase function on \(supp(\eta )\subseteq {\mathbb {R}}^2_{\xi }\) is
Let \(E_i = supp(\eta ) \cap \{|\partial _{\xi _i}(\frac{x}{\tau }\cdot \xi - w(N\xi ))|\gtrsim N^\alpha \}\) for \(i=1,2\). By direct computation,
and thus by integration by parts,
It suffices to assume \(\frac{r}{\tau } \simeq N |\nabla w(N\xi )|\simeq N^\alpha (\xi _1^2+\xi _2^2)^{\frac{\alpha -1}{2}}\simeq N^\alpha \). By direct computation,
We claim \(|\partial _\rho ^2(\Phi _G(\phi _+)\rho )|\lesssim N^2\). Since \(N\rho \le \frac{\pi }{4}\) and \(\theta ,\phi _+ \in [0,\frac{\pi }{2}]\),
Since \(|IV| \lesssim |\partial _\rho \phi _+|\), it suffices to show
which follows from implicitly differentiating (5.24), thereby obtaining
By the triangle inequality,
where the last inequality follows from \(N \le N_\alpha \). By the Van der Corput lemma [29, Chapter VIII],
since \(\frac{r}{\tau }\simeq N^\alpha \). By (5.25), \(a \in L^\infty ([\frac{\pi }{4},2\pi ])\) uniformly in N. To estimate \(\partial _\rho a\), the term that needs most care is \(\partial _\rho |\partial _{\phi }^2 \Phi _G(\phi _+)|^{-\frac{1}{2}}\). Since \(\phi _+,\theta \in [0,\frac{\pi }{2}]\), we have \(\partial _{\phi }^2 \Phi _G(\phi _+) \le 0\). By (5.25), (5.32), the chain rule
and the uniform bound
we have \(\partial _\rho a \in L^\infty ([\frac{\pi }{4},2\pi ])\) uniformly in N.
Lastly, we show (113.11). Let \(C_3(\alpha )>0\) satisfy
and define \(C_i(\alpha )\) similarly for \(i=1,2\). By (5.33), (5.20), we have \(\max \limits _{1 \le i \le 3}C_i(\alpha )<\infty \). For \(\sigma _0 \in \{\frac{3}{4},\frac{5}{6},1\}\) and \(\xi \in supp(\eta (\frac{\cdot }{N}))\), we have
The limit above is a constant multiple of the nonzero leading terms given by (3.16) due to the set of critical points of \(\Phi _{v_{\xi }}\) whose cardinality is uniformly bounded above for all \(\alpha \in (1,2)\) by observing (3.15).
For \(i=3,\, \sigma _0 = \frac{3}{4}\), the nonzero contributions to the limit are due to the cusps in \(K_3\). Let \(N \in S_3\). By Lemma 5.7,
where \(c>0\) depends only on \(\eta \).
For \(i=2,\, \sigma _0 = \frac{5}{6}\), let \(N\in S_2\). For \(\alpha \) sufficiently close to 2, we have \(N=2^{-2}\). Since \(\eta (\frac{\xi (\alpha )}{2^{-2}})\xrightarrow [\alpha \rightarrow 2-]{} 0\) for \(\xi = (0,r_\alpha )\), we may replace \(\eta (\frac{\cdot }{2^{-2}})\) by another smooth bump function \(\tilde{\eta }\) supported in \(\{|\xi | \in [\frac{\pi }{8},\frac{\pi }{2}+\epsilon _0]\}\) where \(\epsilon _0>0\) is sufficiently small so that \(supp (\tilde{\eta })\cap K_3 = \emptyset \). Arguing as (5.35) by using (175.17), one obtains
On the contrary, suppose \(\alpha >1\) is not close to 2 such that \(N\in S_2\) satisfies \(N<2^{-2}\). Then, there exists \(N^{(\alpha )}\in S_2\) such that \(\frac{2}{3\pi }r_\alpha \le N^{(\alpha )} \le \frac{4}{3\pi }r_\alpha \). Then, \(|\eta (\frac{r_\alpha }{N^{(\alpha )}})|\ge c>0\) where c is independent of \(\alpha \). Using the same example (175.17), one obtains
Lastly, let \(N \in S_1\) and \(\sigma _0 = 1\). Pick \(\xi \in {\mathbb {T}}^2\) such that \(|\xi | = N\pi \). Arguing as (5.34) and invoking Lemma 5.2, we have
where the last inequality follows from using the small angle approximation (see (5.13), (5.12)) to obtain
6 Conclusion and future work
We have shown, with a convergence rate, the continuum limit of DNLSE on \(h{\mathbb {Z}}^2\) to the FNLSE on \({\mathbb {R}}^2\) as \(h\rightarrow 0\) in the energy subcritical regime for finite time. Our proof employs sharp dispersive estimates that are obtained by studying appropriate degenerate oscillatory integrals. It is of interest to compare the sharp decay rate of \(\sigma _0=\frac{3}{4}\) to that in the discrete classical Schrödinger equation (\(\sigma _0 = \frac{2}{3}\)) and the discrete wave equation (\(\sigma _0 = \frac{2}{3}\)) at the cost of the best constants blowing up as \(\alpha \rightarrow 1+,\,2-\). As for future work, it is of interest to extend to the case of mixed fractional derivatives [4] where (3.2), in dimension two, is replaced by an appropriate discrete analog of
By numerical and asymptotic techniques, we will explore the conditions of highly localized states in the discrete models that may relate to finite-time blow-up solutions in the continuum limit. \(\square \)
Data availability
Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.
Notes
Denote \(A \lesssim B\) when there exists a constant of non-interest \(C>0\) such that \(A \le CB\) and define \(A \simeq B\) if \(A \lesssim B\) and \(B \lesssim A\).
References
V. Borovyk and M. Goldberg. The klein–gordon equation on \({\mathbb{Z}}^2\) and the quantum harmonic lattice. Journal de Mathématiques Pures et Appliquées, 107(6):667–696, 2017.
T. Boulenger, D. Himmelsbach, and E. Lenzmann. Blowup for fractional nls. Journal of Functional Analysis, 271(9):2569–2603, 2016.
Y. Cho, T. Ozawa, and S. Xia. Remarks on some dispersive estimates. Communications on Pure & Applied Analysis, 10(4):1121, 2011.
B. Choi and A. Aceves. Well-posedness of the mixed-fractional nonlinear schrödinger equation on \({\mathbb{R}}^2\). Partial Differential Equations in Applied Mathematics, pp. 100406, 2022.
J.-C. Cuenin and I. A. Ikromov. Sharp time decay estimates for the discrete klein–gordon equation. Nonlinearity, 34(11):7938, 2021.
V. D. Dinh. Blow-up criteria for fractional nonlinear schrödinger equation. arXiv preprint arXiv:1808.07368, 2018.
R. Grande. Continuum limit for discrete nls with memory effect. arXiv preprint arXiv:1910.05681, 2019.
M. Greenblatt. The asymptotic behavior of degenerate oscillatory integrals in two dimensions. Journal of Functional Analysis, 257(6):1759–1798, 2009.
M. Greenblatt. Stability of oscillatory integral asymptotics in two dimensions. Journal of Geometric Analysis, 24(1):417–444, 2014.
B. J. Hocking, H. S. Ansell, R. D. Kamien, and T. Machon. The topological origin of the peierls–nabarro barrier. Proceedings of the Royal Society A, 478(2258):20210725, 2022.
Y. Hong, C. Kwak, S. Nakamura, and C. Yang. Finite difference scheme for two-dimensional periodic nonlinear schrödinger equations. Journal of Evolution Equations, 21:391–418, 2021.
Y. Hong, C. Kwak, and C. Yang. On the continuum limit for the discrete nonlinear schrödinger equation on a large finite cubic lattice. arXiv preprint arXiv:2106.13417, 2021.
Y. Hong, C. Kwak, and C. Yang. On the korteweg–de vries limit for the fermi–pasta–ulam system. Archive for Rational Mechanics and Analysis, 240(2):1091–1145, 2021.
Y. Hong and Y. Sire. On fractional schrödinger equations in sobolev spaces. Communications on Pure & Applied Analysis, 14(6):2265–2282, 2015.
Y. Hong and C. Yang. Uniform strichartz estimates on the lattice. arXiv preprint arXiv:1806.07093, 2018.
Y. Hong and C. Yang. Strong convergence for discrete nonlinear schrödinger equations in the continuum limit. SIAM Journal on Mathematical Analysis, 51(2):1297–1320, 2019.
L. I. Ignat and E. Zuazua. Dispersive properties of a viscous numerical scheme for the schrödinger equation. Comptes Rendus Mathematique, 340(7):529–534, 2005.
L. I. Ignat and E. Zuazua. A two-grid approximation scheme for nonlinear schrödinger equations: dispersive properties and convergence. Comptes Rendus Mathematique, 341(6):381–386, 2005.
L. I. Ignat and E. Zuazua. Numerical dispersive schemes for the nonlinear schrödinger equation. SIAM journal on numerical analysis, 47(2):1366–1390, 2009.
L. I. Ignat and E. Zuazua. Convergence rates for dispersive approximation schemes to nonlinear schrödinger equations. Journal de mathématiques pures et appliquées, 98(5):479–517, 2012.
I. A. Ikromov and D. Müller. Uniform estimates for the fourier transform of surface carried measures in \({\mathbb{R}}^3\) and an application to fourier restriction. Journal of Fourier Analysis and Applications, 17(6):1292–1332, 2011.
M. Jenkinson and M. I. Weinstein. Discrete solitary waves in systems with nonlocal interactions and the peierls-nabarro barrier. Communications in Mathematical Physics, 351:45–94, 2017.
V. Karpushkin. A theorem concerning uniform estimates of oscillatory integrals when the phase is a function of two variables. Journal of Soviet Mathematics, 35(6):2809–2826, 1986.
M. Keel and T. Tao. Endpoint strichartz estimates. American Journal of Mathematics, 120(5):955–980, 1998.
K. Kirkpatrick, E. Lenzmann, and G. Staffilani. On the continuum limit for discrete nls with long-range lattice interactions. Commun. Math. Phys., 317:563—591, 2013.
Y. S. Kivshar and D. K. Campbell. Peierls-nabarro potential barrier for highly localized nonlinear modes. Phys. Rev. E, 48:3077–3081, Oct 1993.
P. Schultz. The wave equation on the lattice in two and three dimensions. Communications on Pure and Applied Mathematics: A Journal Issued by the Courant Institute of Mathematical Sciences, 51(6):663–695, 1998.
A. Stefanov and P. G. Kevrekidis. Asymptotic behaviour of small solutions for the discrete nonlinear schrödinger and klein–gordon equations. Nonlinearity, 18(4):1841, 2005.
E. M. Stein and T. S. Murphy. Harmonic analysis: real-variable methods, orthogonality, and oscillatory integrals, volume 3. Princeton University Press, 1993.
A. N. Varchenko. Newton polyhedra and estimation of oscillating integrals. Functional analysis and its applications, 10(3):175–196, 1976.
E. Zuazua and L. I. Ignat. Convergence of a two-grid algorithm for the control of the wave equation. Journal of the European Mathematical Society, 11(2):351–391, 2009.
Acknowledgements
Both authors are supported by the US National Science Foundation under the grant DMS-1909559. B. Choi was also supported by an NSF/RTG postdoctoral fellowship under the under the RTG Grant No. DMS-1840260.
Funding
Open access funding provided by SCELC, Statewide California Electronic Library Consortium
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
There are no conflicts of interest other than the National Science Foundation fundings written in the Acknowledgment section.
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
Choi, B., Aceves, A. Continuum limit of 2D fractional nonlinear Schrödinger equation. J. Evol. Equ. 23, 30 (2023). https://doi.org/10.1007/s00028-023-00881-3
Accepted:
Published:
DOI: https://doi.org/10.1007/s00028-023-00881-3