This paper revisits the issue of stability analysis of neural networks subjected to time-varying delays. A novel approach, termed a compound-matrix-based integral inequality (CPBII), which accounts for delay derivatives using two adjustable parameters, is introduced. By appropriately adjusting these parameters, the CPBII efficiently incorporates coupling information along with delay derivatives within integral inequalities. By using CPBII, a novel stability criterion is established for neural networks with time-varying delays. The effectiveness of this approach is demonstrated through a numerical illustration.
Citation: Wenlong Xue, Zhenghong Jin, Yufeng Tian. Stability analysis of delayed neural networks via compound-parameter -based integral inequality[J]. AIMS Mathematics, 2024, 9(7): 19345-19360. doi: 10.3934/math.2024942
[1] | Chao Shi . Existence of stable standing waves for the nonlinear Schrödinger equation with mixed power-type and Choquard-type nonlinearities. AIMS Mathematics, 2022, 7(3): 3802-3825. doi: 10.3934/math.2022211 |
[2] | Yali Meng . Existence of stable standing waves for the nonlinear Schrödinger equation with attractive inverse-power potentials. AIMS Mathematics, 2022, 7(4): 5957-5970. doi: 10.3934/math.2022332 |
[3] | Meixia Cai, Hui Jian, Min Gong . Global existence, blow-up and stability of standing waves for the Schrödinger-Choquard equation with harmonic potential. AIMS Mathematics, 2024, 9(1): 495-520. doi: 10.3934/math.2024027 |
[4] | Jiayin Liu . On stability and instability of standing waves for the inhomogeneous fractional Schrodinger equation. AIMS Mathematics, 2020, 5(6): 6298-6312. doi: 10.3934/math.2020405 |
[5] | Qiuying Li, Xiaoxiao Zheng, Zhenguo Wang . Orbital stability of periodic standing waves of the coupled Klein-Gordon-Zakharov equations. AIMS Mathematics, 2023, 8(4): 8560-8579. doi: 10.3934/math.2023430 |
[6] | Mohammad Alqudah, Manoj Singh . Applications of soliton solutions of the two-dimensional nonlinear complex coupled Maccari equations. AIMS Mathematics, 2024, 9(11): 31636-31657. doi: 10.3934/math.20241521 |
[7] | Aly R. Seadawy, Bayan Alsaedi . Contraction of variational principle and optical soliton solutions for two models of nonlinear Schrödinger equation with polynomial law nonlinearity. AIMS Mathematics, 2024, 9(3): 6336-6367. doi: 10.3934/math.2024309 |
[8] | Yanxia Hu, Qian Liu . On traveling wave solutions of a class of KdV-Burgers-Kuramoto type equations. AIMS Mathematics, 2019, 4(5): 1450-1465. doi: 10.3934/math.2019.5.1450 |
[9] | Lingxiao Li, Jinliang Zhang, Mingliang Wang . The travelling wave solutions of nonlinear evolution equations with both a dissipative term and a positive integer power term. AIMS Mathematics, 2022, 7(8): 15029-15040. doi: 10.3934/math.2022823 |
[10] | Lulu Tao, Rui He, Sihua Liang, Rui Niu . Existence and multiplicity of solutions for critical Choquard-Kirchhoff type equations with variable growth. AIMS Mathematics, 2023, 8(2): 3026-3048. doi: 10.3934/math.2023156 |
This paper revisits the issue of stability analysis of neural networks subjected to time-varying delays. A novel approach, termed a compound-matrix-based integral inequality (CPBII), which accounts for delay derivatives using two adjustable parameters, is introduced. By appropriately adjusting these parameters, the CPBII efficiently incorporates coupling information along with delay derivatives within integral inequalities. By using CPBII, a novel stability criterion is established for neural networks with time-varying delays. The effectiveness of this approach is demonstrated through a numerical illustration.
Gradient estimates are very powerful tools in geometric analysis. In 1970s, Cheng-Yau [3] proved a local version of Yau's gradient estimate (see [25]) for the harmonic function on manifolds. In [16], Li and Yau introduced a gradient estimate for positive solutions of the following parabolic equation,
(Δ−q(x,t)−∂t)u(x,t)=0, | (1.1) |
which was known as the well-known Li-Yau gradient estimate and it is the main ingredient in the proof of Harnack-type inequalities. In [10], Hamilton proved an elliptic type gradient estimate for heat equations on compact Riemannian manifolds, which was known as the Hamilton's gradient estimate and it was later generalized to the noncompact case by Kotschwar [15]. The Hamilton's gradient estimate is useful for proving monotonicity formulas (see [9]). In [22], Souplet and Zhang derived a localized Cheng-Yau type estimate for the heat equation by adding a logarithmic correction term, which is called the Souplet-Zhang's gradient estimate. After the above work, there is a rich literature on extensions of the Li-Yau gradient estimate, Hamilton's gradient estimate and Souplet-Zhang's gradient estimate to diverse settings and evolution equations. We only cite [1,8,11,12,18,19,24,28,31] here and one may find more references therein.
An important generalization of the Laplacian is the following diffusion operator
ΔV⋅=Δ+⟨V,∇⋅⟩ |
on a Riemannian manifold (M,g) of dimension n, where V is a smooth vector field on M. Here ∇ and Δ are the Levi-Civita connenction and Laplacian with respect to metric g, respectively. The V-Laplacian can be considered as a special case of V-harmonic maps introduced in [5]. Recall that on a complete Riemannian manifold (M,g), we can define the ∞-Bakry-Émery Ricci curvature and m-Bakry-Émery Ricci curvature as follows [6,20]
RicV=Ric−12LVg, | (1.2) |
RicmV=RicV−1m−nV⊗V, | (1.3) |
where m≥n is a constant, Ric is the Ricci curvature of M and LV denotes the Lie derivative along the direction V. In particular, we use the convention that m=n if and only if V≡0. There have been plenty of gradient estimates obtained not only for the heat equation, but more generally, for other nonlinear equations concerning the V-Laplacian on manifolds, for example, [4,13,20,27,32].
In [7], Chen and Zhao proved Li-Yau type gradient estimates and Souplet-Zhang type gradient estimates for positive solutions to a general parabolic equation
(ΔV−q(x,t)−∂t)u(x,t)=A(u(x,t)) | (1.4) |
on M×[0,T] with m-Bakry-Émery Ricci tensor bounded below, where q(x,t) is a function on M×[0,T] of C2 in x-variables and C1 in t-variable, and A(u) is a function of C2 in u. In the present paper, by studying the evolution of quantity u13 instead of lnu, we derive localised Hamilton type gradient estimates for |∇u|√u. Most previous studies cited in the paper give the gradient estimates for |∇u|u. The main theorems are below.
Theorem 1.1. Let (Mn,g) be a complete Riemannian manifold with
RicmV≥−(m−1)K1 |
for some constant K1>0 in B(¯x,ρ), some fixed point ¯x in M and some fixed radius ρ. Assume that there exists a constant D1>0 such that u∈(0,D1] is a smooth solution to the general parabolic Eq (1.4) in Q2ρ,T1−T0=B(¯x,2ρ)×[T0,T1], where T1>T0. Then there exists a universal constant c(n) that depends only on n so that
|∇u|√u≤c(n)√D1((m−1)√K1ρ+2m−1ρ2+1t−T0+maxQρ,T1−T0|∇q|23)12+3√D1((m−1)K1+maxQρ,T1−T0|q|2−min{0,minQρ,T1−T0(A′(u)−A(u)2u)})12 | (1.5) |
in Qρ2,T1−T0 with t≠T0.
Remark 1.1. Hamilton [10] first got this gradient estimate for the heat equation on a compact manifold. We also have Hamilton type estimates if we assume that RicV≥−(m−1)K1 for some constant K1>0, and notice that RicV≥−(m−1)K1 is weak than RicmV≥−(m−1)K1. Since we do not have a good enough V-Laplacian comparison for general smooth vector field V, we need the condition that |V| is bounded in this case. Nevertheless, when V=∇f, we can use the method given in [23] to obtain all results in this paper, without assuming that |V| is bounded.
If q=0 and A(u)=aulnu, where a is a constant, then following the proof of Theorem 1.1 we have
Corollary 1.2. Let (Mn,g) be a complete Riemannian manifold with
RicmV≥−(m−1)K1 |
for some constant K1>0 in B(¯x,ρ), some fixed point ¯x in M and some fixed radius ρ. Assume that u is a positive smooth solution to the equation
∂tu=ΔVu−aulnu | (1.6) |
in Q2ρ,T1−T0=B(¯x,2ρ)×[T0,T1], where T1>T0.
(1) When a>0, assuming that 1≤u≤D1 in Q2ρ,T1−T0, there exists a constant c=c(n) such that
|∇u|√u≤c√D1(4√(m−1)2K1ρ2+√2m−1ρ+1√t−T0+|2(m−1)K1−2a|12) | (1.7) |
in Qρ2,T1−T0 with t≠T0.
(2) When a<0, assuming that 0<u≤D1 in Q2ρ,T1−T0, there exists a constant c=c(n) such that
|∇u|√u≤c√D1(4√(m−1)2K1ρ2+√2m−1ρ+1√t−T0+|2(m−1)K1−a(2+lnD1)|12) | (1.8) |
in Qρ2,T1−T0 with t≠T0.
Using the corollary, we get the following Liouville type result.
Corollary 1.3. Let (Mn,g) be a complete Riemannian manifold with RicmV≥−(m−1)K1 for some constant K1>0. Assume that u is a positive and bounded solution to the Eq (1.6) and u is independent of time.
(1) When a>0, assume that 1≤u≤D1. If a=(m−1)K1, then u≡1.
(2) When a<0, assume that 0<u≤exp{−2+2(m−1)K1a}, then u does not exist.
Remark 1.2. When V=0, ΔV and RicmV become Δ and Ric, respectively. It is clear that Corollary 1.2–1.3 generalize Theorem 1.3 and Corollary 1.1 in [14].
We can obtain a global estimate from Theorem 1.1 by taking ρ→0.
Corollary 1.4. Let (Mn,g) be a complete Riemannian manifold with RicmV≥−(m−1)K1 for some constant K1>0. u is a positive smooth solution to the general parabolic Eq (1.4) on Mn×[T0,T1]. Suppose that u≤D1 on Mn×[T0,T1]. We also suppose that
|A′(u)−A(u)2u|≤D2,|q|2≤D3,|∇q|23≤D4 |
on Mn×[T0,T1]. Then there exists a universal constant c that depends only on n so that
|∇u|√u≤c(n)√D1(1t−T0+D4)12+3√D1((m−1)K1+D3+D2)12 | (1.9) |
in Mn×[T0,T1] with t≠T0.
Let A(u)=a(u(x,t))β in Corollary 1.4, we obtain Hamilton type gradient estimates for bounded positive solutions of the equation
(ΔV−q(x,t)−∂t)u(x,t)=a(u(x,t))β,a∈R,β∈(−∞,0]∪[12,+∞). | (1.10) |
Corollary 1.5. Let (Mn,g) be a complete Riemannian manifold with RicmV≥−(m−1)K1 for some constant K1>0. u is a positive smooth solution to (1.10) on Mn×[T0,T1]. Suppose that u≤D1 on Mn×[T0,T1]. We also suppose that
|q|2≤D3,|∇q|23≤D4 |
on Mn×[T0,T1]. Then in Mn×[T0,T1] with t≠T0, there exists a universal constant c that depends only on n so that
|∇u|√u≤c(n)√D1(1t−T0+D4)12+3√D1((m−1)K1+D3+Λ0)12, | (1.11) |
where
Λ0={0,a≥0,β≥12,−a(β−12)Dβ−11,a≤0,β≥12,0,a≤0,β≤0,a(12−β)(minMn×[T0,T1]u)β−1,a≥0,β≤0. |
In the next part, our result concerns gradient estimates for positive solutions of
(Δt−q(x,t)−∂t)u(x,t)=A(u(x,t)) | (1.12) |
on (Mn,g(t)) with the metric evolving under the geometric flow:
∂∂tg(t)=2S(t), | (1.13) |
where Δt depends on t and it denotes the Laplacian of g(t), and S(t) is a symmetric (0,2)-tensor field on (Mn,g(t)). In [31], Zhao proved localised Li-Yau type gradient estimates and Souplet-Zhang type gradient estimates for positive solutions of (1.12) under the geometric flow (1.13). In this paper, we have the following localised Hamilton type gradient estimates for positive solutions to the general parabolic Eq (1.12) under the geometric flow (1.13).
Theorem 1.6. Let (Mn,g(t))t∈[0,T] be a complete solution to the geometric flow (1.13) on Mn with
Ricg(t)≥−K2g(t),|Sg(t)|g(t)≤K3 |
for some K2, K3>0 on Qρ,T=B(¯x,ρ)×[0,T]. Assume that there exists a constant L1>0 such that u∈(0,L1] is a smooth solution to the general parabolic Eq (1.12) in Q2ρ,T. Then there exists a universal constant c(n) that depends only on n so that
|∇u|√u≤c(n)√L1(√K2ρ+1ρ2+1t+K3+maxQρ,T|∇q|23)12+3√L1(K2+K3+maxQρ,T|q|2−min{0,minQρ,T(A′(u)−A(u)2u)})12 | (1.14) |
in Qρ2,T.
Remark 1.3. Recently, some Hamilton type estimates have been achieved to positive solutions of
(Δt−q−∂t)u=au(lnu)α |
under the Ricci flow in [26], and for
(Δt−q−∂t)u=pub+1 |
under the Yamabe flow in [29], where p, q∈C2,1(Mn×[0,T]), b is a positive constant and a,α are real constants. Our results generalize many previous well-known gradient estimate results.
The paper is organized as follows. In Section 2, we provide a proof of Theorem 1.1 and a proof of Corollary 1.3 and Corollary 1.5. In Section 3, we study gradient estimates of (1.12) under the geometric flow (1.13) and give a proof of Theorem 1.6.
We first give some notations for the convenience of writing throughout the paper. Let h:=u13 and ˆA(h):=A(u)3u23. Then ˆAh=A′(u)−2A(u)3u. To prove Theorem 1.1 we need two basic lemmas. First, we derive the following lemma.
Lemma 2.1. Let (Mn,g) be a complete Riemannian manifold with RicmV≥−(m−1)K1 for some constant K1>0. u is a positive smooth solution to the general parabolic Eq (1.4) in Q2ρ,T1−T0. If h:=u13, and μ:=h⋅|∇h|2, then we have
(ΔV−∂t)μ≥4h−3μ2−2h−1⟨∇h,∇μ⟩−2(m−1)K1μ+qμ−23h32|∇q|√μ+2ˆAhμ+h−1ˆA(h)μ. | (2.1) |
Proof. Since h:=u13, by a simple computation, we can derive the following equation from (1.4):
(ΔV−∂t)h=−2h−1|∇h|2+qh3+ˆA(h). | (2.2) |
By direct computations, we have
∇iμ=|∇h|2∇ih+2h∇i∇jh∇jh, |
and
ΔVμ=Δμ+⟨V,∇μ⟩=2h|∇2h|2+2h∇i∇i∇jh∇jh+4∇2h(∇h,∇h)+|∇h|2Δh+2h∇i∇jh∇jhVi+⟨V,∇h⟩|∇h|2=2h|∇2h|2+2h⟨∇Δh,∇h⟩+2hRic(∇h,∇h)+4∇2h(∇h,∇h)+|∇h|2ΔVh+2h∇i∇jh∇jhVi=2h|∇2h|2+2h⟨∇Δh,∇h⟩+2hRicV(∇h,∇h)+2h∇jVi∇ih∇jh+4∇2h(∇h,∇h)+|∇h|2ΔVh+2h∇i∇jh∇jhVi=2h|∇2h|2+2h⟨∇ΔVh,∇h⟩−2h(∇jVi∇ih∇jh+∇i∇jh∇jhVi)+2hRicV(∇h,∇h)+2h∇jVi∇ih∇jh+4∇2h(∇h,∇h)+|∇h|2ΔVh+2h∇i∇jh∇jhVi=2h|∇2h|2+2h⟨∇ΔVh,∇h⟩+2hRicV(∇h,∇h)+4∇2h(∇h,∇h)+|∇h|2ΔVh. | (2.3) |
By the following fact:
0≤(√m−nmnΔh−√nm(m−n)⟨∇h,V⟩)2=(1n−1m)(Δh)2−2m⟨∇h,V⟩Δh+(1m−n−1m)⟨∇h,V⟩2=1n(Δh)2−1m((Δh)2+2⟨∇h,V⟩Δh+⟨∇h,V⟩2)+1m−n⟨∇h,V⟩2≤|∇2h|2−1m(ΔVh)2+1m−n⟨∇h,V⟩2, |
it yields
|∇2h|2≥1m(ΔVh)2−1m−n⟨∇h,V⟩2. | (2.4) |
Plugging (2.4) into (2.3), we have
ΔVμ≥2h(1m(ΔVh)2−1m−n⟨∇h,V⟩2)+2h⟨∇ΔVh,∇h⟩+2hRicV(∇h,∇h)+4∇2h(∇h,∇h)+|∇h|2ΔVh=2mh(ΔVh)2+2hRicmV(∇h,∇h)+2h⟨∇ΔVh,∇h⟩+4∇2h(∇h,∇h)+|∇h|2ΔVh. | (2.5) |
The partial derivative of μ with respect to t is given by
∂tμ=|∇h|2∂th+2h∇i(∂th)∇ih=|∇h|2∂th+2h∇i(ΔVh+2h−1|∇h|2−qh3−ˆA(h))∇ih=|∇h|2∂th+2h⟨∇ΔVh,∇h⟩+8∇2h(∇h,∇h)−4h−1|∇h|4−2qh|∇h|23−2h2⟨∇q,∇h⟩3−2hˆAh|∇h|2. | (2.6) |
It follows from (2.2), (2.5) and (2.6) that
(ΔV−∂t)μ=2mh(ΔVh)2+2hRicmV(∇h,∇h)−4∇2h(∇h,∇h)+|∇h|2(ΔV−∂t)h+4h−1|∇h|4+2qh3|∇h|2+2h2⟨∇q,∇h⟩3+2hˆAh|∇h|2=2mh(ΔVh)2+2hRicmV(∇h,∇h)−4∇2h(∇h,∇h)+|∇h|2(−2h−1|∇h|2+qh3+ˆA(h))+4h−1|∇h|4+2qh3|∇h|2+2h2⟨∇q,∇h⟩3+2hˆAh|∇h|2=2mh(ΔVh)2+2hRicmV(∇h,∇h)−4∇2h(∇h,∇h)+2h−1|∇h|4+qh|∇h|2+2h2⟨∇q,∇h⟩3+2hˆAh|∇h|2+ˆA(h)|∇h|2≥−2(m−1)K1h|∇h|2−4∇2h(∇h,∇h)+2h−1|∇h|4+qh|∇h|2+2h2⟨∇q,∇h⟩3+2hˆAh|∇h|2+ˆA(h)|∇h|2. | (2.7) |
Note that
−4∇2h(∇h,∇h)=2h−1|∇h|4−2h−1⟨∇h,∇(h|∇h|2)⟩=2h−3μ2−2h−1⟨∇h,∇μ⟩. | (2.8) |
Therefore,
(ΔV−∂t)μ≥−2(m−1)K1h|∇h|2+4h−3μ2−2h−1⟨∇h,∇μ⟩+qh|∇h|2+2h2⟨∇q,∇h⟩3+2hˆAh|∇h|2+ˆA(h)|∇h|2≥−2(m−1)K1μ+4h−3μ2−2h−1⟨∇h,∇μ⟩+qμ−23h32|∇q|√μ+2ˆAhμ+h−1ˆA(h)μ, | (2.9) |
which is the desired estimate.
The following cut-off function will be used in the proof of Theorem 1.1 (see [2,16,22,30]).
Lemma 2.2. Fix T0, T1∈R and T0<T1. Given τ∈(T0,T1], there exists a smooth function ¯Ψ:[0,+∞)×[T0,T1]→R satisfying the following requirements.
1). The support of ¯Ψ(s,t) is a subset of [0,ρ]×[T0,T1], and 0≤¯Ψ≤1 in [0,ρ]×[T0,T1].
2). The equalities ¯Ψ(s,t)=1 in [0,ρ2]×[τ,T1] and ∂¯Ψ∂s(s,t)=0 in [0,ρ2]×[T0,T1].
3). The estimate |∂t¯Ψ|≤¯Cτ−T0¯Ψ12 is satisfied on [0,+∞]×[T0,T1] for some ¯C>0, and ¯Ψ(s,T0)=0 for all s∈[0,+∞).
4). The inequalities −Cb¯Ψbρ≤∂¯Ψ∂s≤0 and |∂2¯Ψ∂s2|≤Cb¯Ψbρ2 hold on [0,+∞]×[T0,T1] for every b∈(0,1) with some constant Cb that depends on b.
Throughout this section, we employ the cut-off function Ψ:Mn×[T0,T1]→R by
Ψ(x,t)=¯Ψ(r(x),t), |
where r(x):=d(x,¯x) is the distance function from some fixed point ¯x∈Mn.
From Lemma 2.1, we have
(ΔV−∂t)(Ψμ)=μ(ΔV−∂t)Ψ+Ψ(ΔV−∂t)μ+2⟨∇μ,∇Ψ⟩≥μ(ΔV−∂t)Ψ+2⟨∇μ,∇Ψ⟩−2Ψh−1⟨∇h,∇μ⟩+Ψ[−2(m−1)K1μ+4h−3μ2−23h32|∇q|√μ+qμ+2ˆAhμ+h−1ˆA(h)μ]=μ(ΔV−∂t)Ψ+2Ψ⟨∇Ψ,∇(Ψμ)⟩−2|∇Ψ|2Ψμ−2h−1⟨∇h,∇(Ψμ)⟩+2h−1μ⟨∇h,∇Ψ⟩+Ψ[−2(m−1)K1μ+4h−3μ2−23h32|∇q|√μ+qμ+2ˆAhμ+h−1ˆA(h)μ]. | (2.10) |
For fixed τ∈(T0,T1], let (x1,t1) be a maximum point for Ψμ in Qρ,τ−T0. Obviously at (x1,t1), we have the following facts: ∇(Ψμ)=0, ΔV(Ψμ)≤0, and ∂t(Ψμ)≥0. It follows from (2.10) that at such point
0≥h3μ(ΔV−∂t)Ψ−2h3|∇Ψ|2Ψμ−2h32μ32|∇Ψ|+Ψ[−2(m−1)K1h3μ+4μ2−23h92|∇q|√μ+qμh3+2h3ˆAhμ+h2ˆA(h)μ]. | (2.11) |
In other words, we have just proved that
4Ψμ2≤−h3μ(ΔV−∂t)Ψ+2h32μ32|∇Ψ|+23h92|∇q|√μΨ+2h3|∇Ψ|2Ψμ+Ψμ[2(m−1)K1h3−qh3−2h3ˆAh−h2ˆA(h)] | (2.12) |
at (x1,t1).
Next, to realize the theorem, it suffices to bound each term on the right-hand side of (2.12). To deal with ΔVΨ(x1,t1), we divide the arguments into two cases.
Case 1. If d(¯x,x1)<ρ2, then it follows from Lemma 2.2 that Ψ(x,t)=1 around (x1,t1) in the space direction. Therefore, ΔVΨ(x1,t1)=0.
Case 2. Suppose that d(¯x,x1)≥ρ2. Since RicmV≥−(m−1)K1, we can apply the generalized Laplace comparison theorem (see Corollary 3.2 in [20]) to get
ΔVr≤(m−1)√K1coth(√K1r)≤(m−1)(√K1+1r). |
Using the generalized Laplace comparison theorem and Lemma 2.2, we have
ΔVΨ=∂¯Ψ∂rΔVr+∂2¯Ψ∂r2|∇r|2≥−C1/2Ψ12ρ(m−1)(√K1+2ρ)−C1/2Ψ12ρ2 |
at (x1,t1), which agrees with Case 1. Therefore, we have
−h3μ(ΔV−∂t)Ψ=−uμ(ΔV−∂t)Ψ≤(C1/2Ψ12ρ(m−1)(√K1+2ρ)+C1/2Ψ12ρ2+¯CΨ12τ−T0)uμ≤cD1Ψ12μ((m−1)√K1ρ+2m−1ρ2+1τ−T0)≤12Ψμ2+cD21((m−1)√K1ρ+2m−1ρ2+1τ−T0)2 | (2.13) |
for some universal constant c>0. Here we used Lemma 2.2, 0≤Ψ≤1 and Cauchy's inequality.
On the other hand, by Young's inequality and Lemma 2.2, we obtain
2h32μ32|∇Ψ|=2Ψ34μ32u12|∇Ψ|Ψ34≤12Ψμ2+cD21|∇Ψ|4Ψ3≤12Ψμ2+cD211ρ4, | (2.14) |
23Ψh92|∇q|√μ=23Ψ14μ12u32Ψ34|∇q|≤12Ψμ2+cu2Ψ|∇q|43≤12Ψμ2+cD21Ψ|∇q|43, | (2.15) |
2|∇Ψ|2Ψh3μ=2|∇Ψ|2Ψ−32Ψ12uμ≤2D1|∇Ψ|2Ψ−32Ψ12μ≤12Ψμ2+cD21|∇Ψ|4Ψ3≤12Ψμ2+cD21ρ4 | (2.16) |
and
Ψμ[2(m−1)K1h3−qh3−2h3ˆAh−h2ˆA(h)]=Ψ12μΨ12u[2(m−1)K1−q−2ˆAh−h−1ˆA(h)]≤12Ψμ2+12Ψu2[2(m−1)K1−q−2ˆAh−h−1ˆA(h)]2≤12Ψμ2+D212Ψ[2(m−1)K1−q−2ˆAh−h−1ˆA(h)]2. | (2.17) |
Now, we plug (2.13)–(2.17) into (2.12) to get
(Ψμ)2(x1,t1)≤(Ψμ2)(x1,t1)≤cD21((m−1)√K1ρ+2m−1ρ2+1τ−T0+|∇q|23)2+D21((m−1)K1+|q|2−min{0,minQρ,T1−T0(ˆAh+12h−1ˆA(h))})2≤cD21((m−1)√K1ρ+2m−1ρ2+1τ−T0+maxQρ,T1−T0|∇q|23)2+D21((m−1)K1+maxQρ,T1−T0|q|2−min{0,minQρ,T1−T0(A′(u)−A(u)2u)})2. | (2.18) |
The finally, since Ψ(x,τ)=1 in B(¯x,ρ2), it follows from (2.18) that
μ(x,τ)≤Ψμ(x1,t1)≤cD1((m−1)√K1ρ+2m−1ρ2+1τ−T0+maxQρ,T1−T0|∇q|23)+D1((m−1)K1+maxQρ,T1−T0|q|2−min{0,minQρ,T1−T0(A′(u)−A(u)2u)}). | (2.19) |
Since τ∈(T0,T1] is arbitrary and μ=|∇u|29u, we have
|∇u|√u≤c√D1((m−1)√K1ρ+2m−1ρ2+1t−T0+maxQρ,T1−T0|∇q|23)12+3√D1((m−1)K1+maxQρ,T1−T0|q|2−min{0,minQρ,T1−T0(A′(u)−A(u)2u)})12 | (2.20) |
in Qρ2,T1−T0. We complete the proof.
Proof of Corollary 1.3.
(1) When a>0, for a=(m−1)K1, using the inequality (1.7), we have
|∇u|√u≤c√D1(4√(m−1)2K1ρ2+√2m−1ρ+1√t−T0). | (2.21) |
Letting ρ→+∞, t→+∞ in (2.21), we get u is a constant. Using ΔVu−aulnu=0, we get u=1.
(2) When a<0, for D1=exp{−2+2(m−1)K1a}, using the inequality (1.8), we have
|∇u|√u≤c√D1(4√(m−1)2K1ρ2+√2m−1ρ+1√t−T0). | (2.22) |
Letting ρ→+∞, t→+∞ in (2.22), we get u is a constant. Using ΔVu−aulnu=0, we get u=1, but 0<u≤D1=exp{−2+2(m−1)K1a}<1. So u does not exist.
Proof of Corollary 1.5.
Let
Λ:=−min{0,minMn×[T0,T1](a(β−12)uβ−1)}. |
From Corollary 1.4, we just have to compute Λ. By the definition, we have
Λ0={0,a≥0,β≥12,−a(β−12)Dβ−11,a≤0,β≥12,0,a≤0,β≤0,a(12−β)(minMn×[T0,T1]u)β−1,a≥0,β≤0. |
In this section, we consider positive solutions of the nonlinear parabolic Eq (1.12) on (Mn,g) with the metric evolving under the geometric flow (1.13). To prove Theorem 1.6, we follow the procedure used in the proof of Theorem 1.1.
We first derive a general evolution equation under the geometric flow.
Lemma 3.1. ([21]) Suppose the metric evolves by (1.13). Then for any smooth function f, we have
(|∇f|2)t=−2S(∇f,∇f)+2⟨∇f,∇(ft)⟩. |
Next, we derive the following lemma in the same fashion of Lemma 2.1.
Lemma 3.2. Let (Mn,g(t))t∈[0,T] be a complete solution to the geometric flow (1.13) and u be a smooth positive solution to the nonlinear parabolic Eq (1.12) in Q2ρ,T. Suppose that there exists positive constants K2 and K3, such that
Ricg(t)≥−K2g(t),|Sg(t)|g(t)≤K3 |
in Qρ,T. If h:=u13, and μ:=h⋅|∇h|2, then in Qρ,T, we have
(Δt−∂t)μ≥4h−3μ2−4h−1⟨∇h,∇μ⟩−23h32|∇q|√μ−2(K2+K3)μ+qμ+h−1ˆA(h)μ+2ˆAhμ. | (3.1) |
Proof. Since u is a solution to the nonlinear parabolic Eq (1.12), the function h=u13 satisfies
(Δt−∂t)h=−2h−1|∇h|2+qh3+ˆA(h). | (3.2) |
As in the proof of Lemma 2.1, we have that
Δtμ=2h|∇2h|2+2h⟨∇Δth,∇h⟩+2hRic(∇h,∇h)+4∇2h(∇h,∇h)+|∇h|2Δth. |
On the other hand, by the equation ∂tg(t)=2S(t), we have
∂tμ=|∇h|2∂th+2h∇i(∂th)∇ih−2hS(∇f,∇f)=|∇h|2∂th+2h∇i(Δth+2h−1|∇h|2−qh3−ˆA(h))∇ih−2hS(∇f,∇f)=|∇h|2∂th+2h⟨∇Δth,∇h⟩+8∇2h(∇h,∇h)−4h−1|∇h|4−2qh|∇h|23−2h2⟨∇q,∇h⟩3−2hˆAh|∇h|2−2hS(∇f,∇f). | (3.3) |
Therefore,
(Δt−∂t)μ=2h|∇2h|2−4∇2h(∇h,∇h)+2h−1|∇h|4+2h2⟨∇q,∇h⟩3+2hRic(∇h,∇h)+2hS(∇h,∇h)+qh|∇h|2+2hˆAh|∇h|2+ˆA(h)|∇h|2≥−8∇2h(∇h,∇h)−2h(K2+K3)|∇h|2−23h2|∇h||∇q|+(qh+2hˆAh+ˆA(h))|∇h|2=4h−3μ2−4h−1⟨∇h,∇μ⟩−23h32|∇q|√μ−2(K2+K3)μ+qμ+h−1ˆA(h)μ+2ˆAhμ, | (3.4) |
where we used (2.8), the assumption on bound of Ric+S and
h|∇2h|2+2∇2h(∇h,∇h)+h−1|∇h|4=|√h∇2h+dh⊗dh√h|2≥0. |
The proof is complete.
Finally, we employ the cut-off function Ψ:Mn×[T0,T1]→R, with T0=0, T1=T, by
Ψ(x,t)=¯Ψ(r(x,t),t), |
where r(x,t):=dg(t)(x,¯x) is the distance function from some fixed point ¯x∈Mn with respect to the metric g(t).
To prove Theorem 1.6, we follows the same procedure used previously to prove Theorem 1.1; hence in view of Lemma 3.2,
(Δt−∂t)(Ψμ)≥μ(Δt−∂t)Ψ+2Ψ⟨∇Ψ,∇(Ψμ)⟩−2|∇Ψ|2Ψμ−4h−1⟨∇h,∇(Ψμ)⟩+4h−1μ⟨∇h,∇Ψ⟩+Ψ[−2(K2+K3)μ+4h−3μ2−23h32|∇q|√μ+qμ+2ˆAhμ+h−1ˆA(h)μ]. | (3.5) |
For fixed τ∈(0,T], let (x2,t2) be a maximum point for Ψμ in Qρ,τ:=B(¯x,ρ)×[0,τ]⊂Qρ,T. It follows from (3.5) that at such point
0≥h3μ(Δt−∂t)Ψ−2h3|∇Ψ|2Ψμ−4h32μ32|∇Ψ|+Ψ[−2(K2+K3)μh3+4μ2−23h92|∇q|√μ+qμh3+2h3ˆAhμ+h2ˆA(h)μ]. | (3.6) |
At (x2,t2), making use of (3.6), we further obtain
4Ψμ2≤−h3μ(Δt−∂t)Ψ+4h32μ32|∇Ψ|+23h92|∇q|√μΨ+2h3|∇Ψ|2Ψμ+Ψμ[2(K2+K3)h3−qh3−2h3ˆAh−h2ˆA(h)]. | (3.7) |
Next, we need to bound each term on the right-hand side of (3.7). To deal with ΔtΨ(x2,t2), we divide the arguments into two cases:
Case 1: r(x2,t2)<ρ2. In this case, from Lemma 2.2, it follows that Ψ(x,t)=1 around (x2,t2) in the space direction. Therefore, Δt2Ψ(x2,t2)=0.
Case 2: r(x2,t2)≥ρ2. Since Ricg(t)≥−(n−1)K2, the Laplace comparison theorem (see [17]) implies that
Δtr≤(n−1)√K2coth(√K2r)≤(n−1)(√K2+1r). |
Then, it follows that
ΔtΨ=∂¯Ψ∂rΔtr+∂2¯Ψ∂r2|∇r|2≥−C1/2Ψ12ρ(n−1)(√K2+2ρ)−C1/2Ψ12ρ2 |
at (x2,t2), which agrees with Case 1.
Next, we estimate ∂tΨ. ∀x∈B(¯x,ρ), let γ:[0,a]→Mn be a minimal geodesic connecting x and ¯x at time t∈[0,T]. Then, we have
|∂tr(x,t)|g(t)=|∂ta∫0|˙γ(s)|g(t)ds|g(t)≤12a∫0||˙γ(s)|−1g(t)∂tg(˙γ(s),˙γ(s))|g(t)ds≤K3r(x,t)≤K3ρ. |
Thus, together with Lemma 2.2, we find that
∂tΨ≤|∂t¯Ψ|g(t)+|∇¯Ψ|g(t)|∂tr|g(t)≤¯CΨ1/2τ+C1/2K3Ψ1/2. | (3.8) |
Therefore, we have
−h3μ(Δt−∂t)Ψ=−uμ(Δt−∂t)Ψ≤(C1/2Ψ12ρ(n−1)(√K2+2ρ)+C1/2Ψ12ρ2+¯CΨ1/2τ+C1/2K3Ψ1/2)uμ≤cL1Ψ12μ(√K2ρ+1ρ2+1τ+K3)≤12Ψμ2+cL21(√K2ρ+1ρ2+1τ+K3)2 | (3.9) |
at (x2,t2) for some universal constant c>0 that depends only on n. By similar computations as in the proof of Theorem 1.1, we arrive at
4h32μ32|∇Ψ|≤Ψμ2+2cL211ρ4, | (3.10) |
23Ψh92|∇q|√μ≤12Ψμ2+cL21Ψ|∇q|43, | (3.11) |
2|∇Ψ|2Ψh3μ≤12Ψμ2+cL21ρ4 | (3.12) |
and
Ψμ[2(K2+K3)h3−qh3−2h3ˆAh−h2ˆA(h)]≤12Ψμ2+L212Ψ[2(K2+K3)+|q|−min{0,minQρ,T(2ˆAh+h−1ˆA(h))}]2. | (3.13) |
Plugging (3.9)–(3.13) into (3.6), we get
(Ψμ)2(x2,t2)≤(Ψμ2)(x2,t2)≤cL21(√K2ρ+1ρ2+1τ+K3+|∇q|23)2+L21(K2+K3+|q|2−min{0,minQρ,T(ˆAh+12h−1ˆA(h))})2≤cL21(√K2ρ+1ρ2+1τ+K3+maxQρ,T|∇q|23)2+L21(K2+K3+maxQρ,T|q|2−min{0,minQρ,T(A′(u)−A(u)2u)})2. | (3.14) |
Note that Ψ(x,τ)=1 when d(x,¯x)<ρ2, it follows from (2.18) that
μ(x,τ)≤Ψμ(x2,t2)≤cL1(√K2ρ+1ρ2+1τ+K3+maxQρ,T|∇q|23)+L1(K2+K3+maxQρ,T|q|2−min{0,minQρ,T(A′(u)−A(u)2u)}). | (3.15) |
Since τ∈(0,T] is arbitrary and μ=|∇u|29u, we have
|∇u|√u≤c√L1(√K2ρ+1ρ2+1τ+K3+maxQρ,T|∇q|23)12+3√L1(K2+K3+maxQρ,T|q|2−min{0,minQρ,T(A′(u)−A(u)2u)})12 | (3.16) |
in Qρ2,T.
We complete the proof.
Remark 3.1. We also obtain the corresponding applications similar to Corollary 1.4 and Corollary 1.5, which we will not write them down here.
The author would like to thank the referee for helpful suggestion which makes the paper more readable. This work was supported by NSFC (Nos. 11971415, 11901500), Henan Province Science Foundation for Youths (No.212300410235), and the Key Scientific Research Program in Universities of Henan Province (No.21A110021) and Nanhu Scholars Program for Young Scholars of XYNU (No. 2019).
The author declares no conflict of interest.
[1] |
N. Huo, B. Li, Y. Li, Global exponential stability and existence of almost periodic solutions in distribution for Clifford-valued stochastic high-order Hopfield neural networks with time-varying delays, AIMS Math., 7 (2022), 3653–3679. http://dx.doi.org/10.3934/math.2022202 doi: 10.3934/math.2022202
![]() |
[2] |
H. Qiu, L. Wan, Z. Zhou, Q. Zhang, Q. Zhou, Global exponential periodicity of nonlinear neural networks with multiple time-varying delays, AIMS Math., 8 (2023), 12472–12485. http://dx.doi.org/10.3934/math.2023626 doi: 10.3934/math.2023626
![]() |
[3] |
R. Wei, J. Cao, W. Qian, C. Xue, X. Ding, Finite-time and fixed-time stabilization of inertial memristive Cohen-Grossberg neural networks via non-reduced order method, AIMS Math., 6 (2021), 6915–6932. http://dx.doi.org/10.3934/math.2021405 doi: 10.3934/math.2021405
![]() |
[4] |
J. Kim, Further improvement of Jensen inequality and application to stability of time-delayed systems, Automatica, 64 (2016), 121–125. https://doi.org/10.1016/j.automatica.2015.08.025 doi: 10.1016/j.automatica.2015.08.025
![]() |
[5] |
O. M. Kwon, M. J. Park, S. M. Lee, J. H. Park, E. J. Cha, Stability for neural networks with time-varying delays via some new approaches, IEEE T. Neur. Net. Lear., 24 (2013), 181–193. https://doi.org/10.1109/TNNLS.2012.2224883 doi: 10.1109/TNNLS.2012.2224883
![]() |
[6] |
X. M. Zhang, Q. L. Han, Z. Zeng, Hierarchical type stability criteria for delayed neural networks via canonical Bessel-Legendre inequalities, IEEE T. Cybern., 48 (2018), 1660–1671. https://doi.org/10.1109/TCYB.2017.2776283 doi: 10.1109/TCYB.2017.2776283
![]() |
[7] |
T. Wu, S. Gorbachev, H. Lam, J. Park, L. Xiong, J. Cao, Adaptive event-triggered space-time sampled-data synchronization for fuzzy coupled RDNNs under hybrid random cyberattacks, IEEE T. Fuzzy Syst., 31 (2023), 1855–1869. https://doi.org/10.1109/TFUZZ.2022.3215747 doi: 10.1109/TFUZZ.2022.3215747
![]() |
[8] |
T. Wu, J. Cao, J. Park, K. Shi, L. Xiong, T. Huang, Attack-resilient dynamic event-triggered synchronization of fuzzy reaction-diffusion dynamic networks with multiple cyberattacks, IEEE T. Fuzzy Syst., 32 (2024), 498–509. https://doi.org/10.1109/TFUZZ.2023.3300882 doi: 10.1109/TFUZZ.2023.3300882
![]() |
[9] |
T. Wu, J. Cao, L. Xiong, J. Park, X. Tan, Adaptive event-triggered mechanism to synchronization of reaction-diffusion CVNNs and its application in image secure communication, IEEE T. Syst. Man Cy-S., 53 (2023), 5307–5320. https://doi.org/10.1109/TSMC.2023.3258222 doi: 10.1109/TSMC.2023.3258222
![]() |
[10] |
C. K. Zhang, Y. He, L. Jiang, M. Wu, Notes on stability of time-delay systems: Bounding inequalities and augmented Lyapunov-Krasovskii functionals, IEEE T. Autom. Control, 62 (2017), 5331–5336. https://doi.org/10.1109/TAC.2016.2635381 doi: 10.1109/TAC.2016.2635381
![]() |
[11] |
W. Lin, Y. He, C. Zhang, M. Wu, J. Shen, Extended dissipativity analysis for Markovian jump neural networks with time-varying delay via delay-product-type functionals, IEEE T. Neur. Net. Lear., 30 (2019), 2528–2537. https://doi.org/10.1109/TNNLS.2018.2885115 doi: 10.1109/TNNLS.2018.2885115
![]() |
[12] |
Y. Tian, Z. Wang, Extended dissipativity analysis for Markovian jump neural networks via double-integral-based delay-product-type Lyapunov functional, IEEE T. Neur. Net. Lear., 32 (2020), 3240–3246. https://doi.org/10.1109/TNNLS.2020.3008691 doi: 10.1109/TNNLS.2020.3008691
![]() |
[13] | K. Gu, V. L. Kharitonov, J. Chen, Stability of Time-Delay Systems, 2003. |
[14] |
A. Seuret, F. Gouaisbaut, Wirtinger-based integral inequality: application to time-delay systems, Automatica, 49 (2013), 2860–2866. https://doi.org/10.1016/j.automatica.2013.05.030 doi: 10.1016/j.automatica.2013.05.030
![]() |
[15] |
H.B. Zeng, Y. He, M. Wu, J. She, Free-matrix-based integral inequality for stability analysis of systems with time-varying delay, IEEE T. Autom. Control, 60 (2015), 2768–2772. https://doi.org/10.1109/TAC.2015.2404271 doi: 10.1109/TAC.2015.2404271
![]() |
[16] |
H.B. Zeng, Y. He, M. Mu, J. She, New results on stability analysis for systems with discrete distributed delay, Automatica, 60 (2015), 189–192. https://doi.org/10.1016/j.automatica.2015.07.017 doi: 10.1016/j.automatica.2015.07.017
![]() |
[17] |
X. M. Zhang, W. J. Lin, Q. L. Han, Y. He, M. Wu, Global asymptotic stability for delayed neural networks using an integral inequality based on nonorthogonal polynomials, IEEE T. Neur. Net. Lear., 29 (2018), 4487–4493. https://doi.org/10.1109/TNNLS.2017.2750708 doi: 10.1109/TNNLS.2017.2750708
![]() |
[18] |
J. Chen, S. Xu, B. Zhang, Single/Multiple integral inequalities with applications to stability analysis of time-delay systems, IEEE T. Autom. Control, 62 (2017), 3488–3493. https://doi.org/10.1109/TAC.2016.2617739 doi: 10.1109/TAC.2016.2617739
![]() |
[19] |
C. K. Zhang, Y. He, L. Jiang, W. J. Lin, M. Wu, Delay-dependent stability analysis of neural networks with time-varying delay: A generalized free-weighting-matrix approach, Appl. Math. Comput., 294 (2017), 102–120. https://doi.org/10.1016/j.amc.2016.08.043 doi: 10.1016/j.amc.2016.08.043
![]() |
[20] |
A. Seuret, F. Gouaisbaut, Hierarchy of LMI conditions for the stability analysis of time-delay systems, Syst. Control Lett., 81 (2015), 1–8. https://doi.org/10.1016/j.sysconle.2015.03.007 doi: 10.1016/j.sysconle.2015.03.007
![]() |
[21] |
A. Seuret, F. Gouaisbaut, Stability of linear systems with time-varying delays using Bessel-Legendre inequalities, IEEE T. Autom. Control, 63 (2018), 225–232. https://doi.org/10.1109/TAC.2017.2730485 doi: 10.1109/TAC.2017.2730485
![]() |
[22] |
Y. Huang, Y. He, J. An, M. Wu, Polynomial-type Lyapunov-Krasovskii functional and Jacobi-Bessel inequality: Further results on stability analysis of time-delay systems, IEEE Trans. Autom. Control, 66 (2021), 2905–2912. https://doi.org/10.1109/TAC.2020.3013930 doi: 10.1109/TAC.2020.3013930
![]() |
[23] |
P. Park, J. W. Ko, C. Jeong, Reciprocally convex approach to stability of systems with time-varying delays, Automatica, 47 (2011), 235–238. https://doi.org/10.1016/j.automatica.2010.10.014 doi: 10.1016/j.automatica.2010.10.014
![]() |
[24] |
X. M. Zhang, Q. L. Han, A. Seuret, F. Gouaisbaut, An improved reciprocally convex inequality and an augmented Lyapunov-Krasovskii functional for stability of linear systems with time-varying delay, Automatica, 84 (2017), 221–226. https://doi.org/10.1016/j.automatica.2017.04.048 doi: 10.1016/j.automatica.2017.04.048
![]() |
[25] |
C. Zhang, Y. He, L. Jiang, M. Wu, Q. Wang, An extended reciprocally convex matrix inequality for stability analysis of systems with time-varying delay, Automatica, 85 (2017), 481–485. https://doi.org/10.1016/j.automatica.2017.07.056 doi: 10.1016/j.automatica.2017.07.056
![]() |
[26] |
W. I. Lee, S. Y. Lee, P. G. Park, Affine Bessel-Legendre inequality: Application to stability analysis for systems with time-varying delays, Automatica, 93 (2018), 535–539. https://doi.org/10.1016/j.automatica.2018.03.073 doi: 10.1016/j.automatica.2018.03.073
![]() |
[27] |
J. Chen, J. H. Park, S. Xu, Stability analysis for delayed neural networks with an improved general free-matrix-based integral inequality, IEEE T. Neur. Net. Lear. Syst., 31 (2020), 675–684. https://doi.org/10.1109/TNNLS.2019.2909350 doi: 10.1109/TNNLS.2019.2909350
![]() |
[28] |
Y. Tian, Y. Yang, X. Ma, X. Su, Stability of discrete-time delayed systems via convex function-based summation inequality, Appl. Math. Lett., 145 (2023), 108764, https://doi.org/10.1016/j.aml.2023.108764 doi: 10.1016/j.aml.2023.108764
![]() |
1. | Chao Shi, Existence of stable standing waves for the nonlinear Schrödinger equation with mixed power-type and Choquard-type nonlinearities, 2022, 7, 2473-6988, 3802, 10.3934/math.2022211 | |
2. | Qiuying Li, Xiaoxiao Zheng, Zhenguo Wang, Orbital stability of periodic standing waves of the coupled Klein-Gordon-Zakharov equations, 2023, 8, 2473-6988, 8560, 10.3934/math.2023430 | |
3. | Jianlun Liu, Ziheng Zhang, Rui Guo, Ground state normalized solutions to Schrödinger equation with inverse-power potential and HLS lower critical Hartree-type nonlinearity, 2025, 44, 2238-3603, 10.1007/s40314-025-03131-z |