betaFeedback

WorksMathematics

Stabilization of an Euler–Bernoulli beam equation with distributed damping and interior time delay

Abstract

In this paper, we investigate the effect of a fixed time delay on the internal stabilization of an Euler–Bernoulli beam clamped at one end with distributed damping. Under appropriate assumptions on the feedback weights, we first establish that the problem is well-posed using semigroup theory. Furthermore, using the multiplier method combined with a suitable Lyapunov functional, we prove the exponential stability of the system.

Introduction

This study aims to analyze the impact of a fixed internal time delay on the internal stabilization of a uniform Euler–Bernoulli beam, characterized by its mass per unit length mm and flexural rigidity EI>0EI > 0. The vibratory dynamics of the system are governed by the following initial-boundary value problem:

{mytt+EIyxxxx+uτ(x,t)=0,x∈(0,ℓ), t>0,y(0,t)=yx(0,t)=yxx(ℓ,t)=yxxx(ℓ,t)=0,t>0,y(x,0)=y0(x), yt(x,0)=y1(x),x∈[0,ℓ],yt(x,t−τ)=f0(x,t−τ),x∈(0,ℓ), t∈(0,τ).(1)\begin{cases} my_{tt} + EI y_{xxxx} + u_{\tau}(x,t) = 0, & x \in(0,\ell),\ t > 0,\\ y(0,t) = y_x(0,t) = y_{xx}(\ell,t) = y_{xxx}(\ell,t) = 0, & t > 0,\\ y(x,0) = y_0(x),\ y_t(x,0) = y_1(x), & x \in[0,\ell],\\ y_t(x,t-\tau) = f_0(x,t-\tau), & x \in(0,\ell),\ t \in(0,\tau). \tag*{(1)} \end{cases}

where y(x,t)y(x,t) represents the deflection of the beam at position xx and time tt, while uτ(x,t)u_{\tau}(x,t) denotes the internal control law, which depends on a fixed time delay τ>0\tau> 0. The boundary condition (1)2_2 means that the beam is clamped at its end x=0x = 0 and free at the opposite end x=ℓx = \ell. Finally, (y0,y1)(y_0,y_1) is the initial state of the beam, and f0f_0 indicates the memory of controller.

2020 Mathematics Subject Classification. 35B35, 93C43, 93D15, 93D23.

Key words and phrases. Euler–Bernoulli beam equation, time delay, distributed damping, exponential stability, multiplier method, Lyapunov functional.

The stabilization of Euler–Bernoulli beam equation is central to control theory, driven by both its engineering applications and its mathematical significance [2, 5, 9, 14, 25]. This challenge intensifies significantly when time delays are introduced, as even a small latency in the feedback loop can trigger instability [6, 7, 8, 12]. For this reason, designing robust control strategies to neutralize delay-induced instabilities remains a top priority in this field.

In this perspective, two main control strategies are widely distinguished in the literature for time-delay Euler–Bernoulli beam equations, depending on where the control action is applied. On the one hand, boundary control consists in applying feedback directly at the boundaries of the structure. In their pioneering work [13], Guo et al. considered an output feedback stabilization problem for the system:

{wtt+wxxxx=0,0<x<1, t>0,w(0,t)=wx(0,t)=wxx(1,t)=0,wxxx(1,t)=u(t),y(t)=wt(1,t−τ),t≥τ,w(x,0)=w0(x),wt(x,0)=w1(x),(2)\begin{cases} w_{tt}+w_{xxxx}=0, & 0<x<1,\ t>0,\\ w(0,t)=w_x(0,t)=w_{xx}(1,t)=0,\\ w_{xxx}(1,t)=u(t),\\ y(t)=w_t(1,t-\tau), & t\ge\tau,\\ w(x,0)=w_0(x),\quad w_t(x,0)=w_1(x), \tag*{(2)} \end{cases}

where uu is the control, (w0,w1)(w_0,w_1) the initial state, and yy the delayed measurement. The authors exponentially stabilize the system using the control law

u(t)={k2w~t(1,t)=k2w^t(1,t,t),t>τ,0,t∈[0,τ],u(t)= \begin{cases} k_2\widetilde{w}_t(1,t)=k_2\widehat{w}_t(1,t,t), & t>\tau,\\ 0, & t\in[0,\tau], \end{cases}

designed via an observer-predictor scheme based on the estimated state (w~,w~t)(\widetilde{w},\widetilde{w}_t) with k2>0k_2>0. Subsequently, numerous contributions have deepened this direction, proposing various boundary control laws to effectively suppress the vibrations of such structures. A detailed review of these approaches and their recent developments can be found in [1, 3, 4, 15, 16, 19, 20, 21, 22, 23].

On the other hand, the control action can be distributed inside the spatial domain: this is referred to as internal (or distributed) control. The internal stabilization of Euler–Bernoulli beam models was pioneered by Feng et al. In [11], the state-feedback stabilization of the following system was addressed:

{wtt(x,t)+wxxxx(x,t)+u(x,t−τ)=0,x∈(0,1), t>0,w(0,t)=wx(1,t)=0,wxx(0,t)=wxxx(1,t)=0,w(x,0)=w0(x),wt(x,0)=w1(x),x∈(0,1),u(x,s−τ)=h(x,s),x∈(0,1), s∈(0,τ),(3)\begin{cases} w_{tt}(x,t)+w_{xxxx}(x,t)+u(x,t-\tau)=0, & x\in(0,1),\ t>0,\\ w(0,t)=w_x(1,t)=0,\\ w_{xx}(0,t)=w_{xxx}(1,t)=0,\\ w(x,0)=w_0(x),\quad w_t(x,0)=w_1(x), & x\in(0,1),\\ u(x,s-\tau)=h(x,s), & x\in(0,1),\ s\in(0,\tau), \tag*{(3)} \end{cases}

where h(x,s)h(x,s) stands for the memory function of the controller. Employing the backstepping method [24], they designed a feedback control law

u(x,t)=∫0τ∫01q(x,τ;y,r)z(y,r,t) dy dr+∫01γ(x,τ,r)w(y,t) dy+∫01η(x,τ,y)wt(y,t) dy,x∈(0,1), t≥0,(4)\begin{aligned} u(x,t)={}&\int_0^\tau\int_0^1 q(x,\tau;y,r)z(y,r,t)\,dy\,dr+\int_0^1\gamma(x,\tau,r)w(y,t)\,dy\\ &+\int_0^1\eta(x,\tau,y)w_t(y,t)\,dy,\qquad x\in(0,1),\ t\ge0, \tag*{(4)} \end{aligned}

via a bounded, linear, and invertible integral operator-structured using appropriate kernel functions such as q(x,τ;y,r)q(x,\tau;y,r), γ(x,τ,r)\gamma(x,\tau,r), and η(x,τ,y)\eta(x,\tau,y), which maps the original plant into an exponentially stable target system.

However, internal feedback mechanisms remain considerably less explored than their boundary counterparts. Yet, distributed controllers provide greater flexibility and allow for a more effective mitigation of localized physical phenomena [10], such as periodic disturbances, structural resonances, or spatial constraints. To achieve the exponential stabilization of system (1), we propose the following localized internal feedback control law:

uτ(x,t)=a(x)(μ1yt(x,t)+μ2yt(x,t−τ)),x∈(0,ℓ), t≥0.(5)u_{\tau}(x,t)=a(x)\left(\mu_{1}y_{t}(x,t)+\mu_{2}y_{t}(x,t-\tau)\right),\quad x\in(0,\ell),\ t\geq0. \tag*{(5)}

inspired by the framework of Nicaise and Pignotti [17], where μ1>0\mu_{1}>0, μ2≠0\mu_{2}\ne0, and a∈L∞(0,ℓ)a\in L^{\infty}(0,\ell) is a function satisfying

{a(x)≥0a.e. in (0,ℓ),a(x)>a0>0a.e. in ω,(6)\begin{cases} a(x)\geq0 & \text{a.e. in }(0,\ell),\\ a(x)>a_{0}>0 & \text{a.e. in }\omega, \tag*{(6)} \end{cases}

with ω⊂(0,ℓ)\omega\subset(0,\ell) being a non-empty open neighbourhood of ℓ\ell. To the best of our knowledge, the internal stabilization of an Euler–Bernoulli beam under such a localized, delayed distributed feedback law remains entirely open in the literature. First, we reformulate system (1) by introducing a new state variable to handle the time delay. Then, under the dominance condition

∣μ2∣<μ1,(7)|\mu_{2}|<\mu_{1}, \tag*{(7)}

we first prove that the problem is well-posed by invoking the semigroup theory [18]. Secondly, we establish the exponential decay of the system’s energy via a suitable Lyapunov functional coupled with the multiplier technique, departing from the observability-based approach of [17].

The rest of this paper is organized as follows. § 2 is devoted to preliminary reminders and tools necessary for establishing our main results. In § 3, we study the well-posedness of the problem (1) using semigroup theory. Finally, § 4 is entirely dedicated to proving the exponential stability of the system.

Preliminaries

This section gathers the essential tools for our main analysis. Throughout this paper, CC denotes a generic constant.

Augmented system

To handle the time delay, we follow [17] and introduce the auxiliary variable z(x,ρ,t)z(x,\rho,t) defined as

z(x,s,t):=yt(x,t−sτ),x∈(0,ℓ), s∈(0,1), t>0.(8)z(x,s,t):=y_{t}(x,t-s\tau),\quad x\in(0,\ell),\ s\in(0,1),\ t>0. \tag*{(8)}

By direct computation, zz satisfies the transport equation:

{τzt+zs=0,x∈(0,ℓ), s∈(0,1), t>0,z(x,0,t)=yt(x,t),x∈(0,ℓ), t>0,z(x,1,t)=yt(x,t−τ),x∈(0,ℓ), s∈(0,1), t>0,z(x,s,0)=f0(x,−sτ),x∈(0,ℓ), s∈(0,1).(9)\left\{ \begin{aligned} \tau z_{t}+z_{s}&=0,\quad x\in(0,\ell),\ s\in(0,1),\ t>0,\\ z(x,0,t)&=y_{t}(x,t),\quad x\in(0,\ell),\ t>0,\\ z(x,1,t)&=y_{t}(x,t-\tau),\quad x\in(0,\ell),\ s\in(0,1),\ t>0,\\ z(x,s,0)&=f_{0}(x,-s\tau),\quad x\in(0,\ell),\ s\in(0,1). \end{aligned} \right. \tag*{(9)}

Consequently, system (1) takes the augmented form:

{mytt+EIyxxxx+a(x)(μ1yt(x,t)+μ2z(x,1,t))=0,x∈(0,ℓ), t>0,τzt+zs=0,x∈(0,ℓ), s∈(0,1), t>0,y(0,t)=yx(0,t)=yxx(ℓ,t)=yxxx(ℓ,t)=0,t>0,z(x,0,t)=yt(x,t),x∈(0,ℓ), t>0,y(x,0)=y0(x),yt(x,0)=y1(x),x∈[0,ℓ],z(x,s,0)=f0(x,−sτ),x∈(0,ℓ), s∈(0,1).(10)\begin{cases} m y_{tt} + EI y_{xxxx} + a(x)\left(\mu_1 y_t(x,t) + \mu_2 z(x,1,t)\right) = 0,\\ x \in(0,\ell),\ t > 0,\\ \tau z_t + z_s = 0,\quad x \in(0,\ell),\ s \in(0,1),\ t > 0,\\ y(0,t) = y_x(0,t) = y_{xx}(\ell,t) = y_{xxx}(\ell,t) = 0,\quad t > 0,\\ z(x,0,t) = y_t(x,t),\quad x \in(0,\ell),\ t > 0,\\ y(x,0) = y_0(x),\quad y_t(x,0) = y_1(x),\quad x \in[0,\ell],\\ z(x,s,0) = f_0(x,-s\tau),\quad x \in(0,\ell),\ s \in(0,1). \tag*{(10)} \end{cases}

Energy of the system

Let E(t)E(t) be the energy of system (1) defined by:

E(t):=m2∫0ℓyt2 dx+EI2∫0ℓyxx2 dx+ξ2∫0ℓa(x)∫01yt2(x,t−sτ) ds dx(11)E(t) := \frac{m}{2}\int_0^\ell y_t^2\,dx + \frac{EI}{2}\int_0^\ell y_{xx}^2\,dx + \frac{\xi}{2}\int_0^\ell a(x)\int_0^1 y_t^2(x,t-s\tau)\,ds\,dx \tag*{(11)}

where ξ\xi is a positive constant.

Lemma 1. Assume that (7) holds. If the parameter ξ\xi satisfies the condition

τ∣μ2∣≤ξ≤τ(2μ1−∣μ2∣)(12)\tau|\mu_2| \le\xi\le\tau(2\mu_1-|\mu_2|) \tag*{(12)}

then, the energy E(t)E(t) is non-increasing over time, i.e.:

ddtE(t)≤−(μ1−∣μ2∣2−ξτ−12)∫0ℓa(x)yt2 dx−(−∣μ2∣2+ξτ−12)∫0ℓa(x)yt2(x,t−τ) dx.(13)\frac{d}{dt}E(t) \le-\left(\mu_1-\frac{|\mu_2|}{2}-\frac{\xi\tau^{-1}}{2}\right)\int_0^\ell a(x)y_t^2\,dx -\left(-\frac{|\mu_2|}{2}+\frac{\xi\tau^{-1}}{2}\right)\int_0^\ell a(x)y_t^2(x,t-\tau)\,dx. \tag*{(13)}

Proof. First, multiplying (10)1_1 by yty_t and integrating over (0,ℓ)(0,\ell) with respect to xx, we obtain

∫0ℓ(mytt+EIyxxxx+a(x)(μ1yt(x,t)+μ2z(x,1,t)))yt dx=0.(14)\int_0^\ell\left(m y_{tt}+EI y_{xxxx}+a(x)\left(\mu_1y_t(x,t)+\mu_2z(x,1,t)\right)\right)y_t\,dx=0. \tag*{(14)}

Integrating by parts, we get

∫0ℓyttyt dx=ddt(12∫0ℓyt2 dx)(15)\int_0^\ell y_{tt}y_t\,dx=\frac{d}{dt}\left(\frac{1}{2}\int_0^\ell y_t^2\,dx\right) \tag*{(15)}
∫0ℓyxxxxyt dx=[yxxxyt−yxxytx]x=0x=ℓ+∫0ℓyxxytx dx=ddt(12∫0ℓyxx2 dx),(16)\int_0^\ell y_{xxxx}y_t\,dx=\left[y_{xxx}y_t-y_{xx}y_{tx}\right]_{x=0}^{x=\ell}+\int_0^\ell y_{xx}y_{tx}\,dx =\frac{d}{dt}\left(\frac{1}{2}\int_0^\ell y_{xx}^2\,dx\right), \tag*{(16)}

where we used the boundary condition (10)3_3. Therefore, relation (14) simplifies to

ddt(m2∫0ℓyt2 dx+EI2∫0ℓyxx2 dx)=−μ1∫0ℓa(x)yt2 dx−μ2∫0ℓa(x)z(x,1,t)yt dx.(17)\frac{d}{dt}\left(\frac{m}{2}\int_0^\ell y_t^2\,dx+\frac{EI}{2}\int_0^\ell y_{xx}^2\,dx\right) =-\mu_1\int_0^\ell a(x)y_t^2\,dx-\mu_2\int_0^\ell a(x)z(x,1,t)y_t\,dx. \tag*{(17)}

Next, equation (10)2_2 implies

ξ∫0ℓa(x)∫01ztz ds dx=−ξτ−1∫0ℓa(x)∫01zsz ds dx.(18)\xi\int_0^\ell a(x)\int_0^1 z_tz\,ds\,dx =-\xi\tau^{-1}\int_0^\ell a(x)\int_0^1 z_sz\,ds\,dx. \tag*{(18)}

An integration by parts with respect to ss, together with the boundary condition (10), yields

∫01ztzs ds=ddt(12∫01z2 ds);(19)\int_0^1 z_t z_s\,ds=\frac{d}{dt}\left(\frac{1}{2}\int_0^1 z^2\,ds\right); \tag*{(19)}
∫01zsz ds=12(z(x,1,t)2−yt2).(20)\int_0^1 z_s z\,ds=\frac{1}{2}\left(z(x,1,t)^2-y_t^2\right). \tag*{(20)}

Combining identities (19)–(20), it follows that

ddt(ξ∫0ℓa(x)∫01z2 ds dx)=ξτ∫0ℓa(x)(yt2−z(x,1,t)2) dx.(21)\frac{d}{dt}\left(\xi\int_0^\ell a(x)\int_0^1 z^2\,ds\,dx\right)=\xi\tau\int_0^\ell a(x)\left(y_t^2-z(x,1,t)^2\right)\,dx. \tag*{(21)}

Finally, adding (17) to (21) leads to

ddt(m2∫0ℓyt2 dx+EI2∫0ℓyxx2 dx+ξ2∫0ℓa(x)∫01z2 ds dx)=−μ1∫0ℓa(x)yt2 dx+ξτ−12∫0ℓa(x)(yt2−z(x,1,t)2) dx−μ2∫0ℓa(x)z(x,1,t)yt dx.(22)\begin{aligned} \frac{d}{dt}\left(\frac{m}{2}\int_0^\ell y_t^2\,dx+\frac{EI}{2}\int_0^\ell y_{xx}^2\,dx+\frac{\xi}{2}\int_0^\ell a(x)\int_0^1 z^2\,ds\,dx\right) ={}&-\mu_1\int_0^\ell a(x)y_t^2\,dx+\frac{\xi\tau^{-1}}{2}\int_0^\ell a(x)\left(y_t^2-z(x,1,t)^2\right)\,dx\\ &-\mu_2\int_0^\ell a(x)z(x,1,t)y_t\,dx. \tag*{(22)} \end{aligned}

In addition, by Young’s inequality, it follows that

∣μ2z(x,1,t)yt∣≤∣μ2∣2(z(x,1,t)2+yt2).(23)\left|\mu_2z(x,1,t)y_t\right|\leq\frac{|\mu_2|}{2}\left(z(x,1,t)^2+y_t^2\right). \tag*{(23)}

Substituting estimate (23) into (22) directly implies (13). The non-positivity of the right-hand side is guaranteed whenever ξ\xi satisfies (12). □

Existence and uniqueness of solution

This section is devoted to the well-posedness of problem (1) via semigroup theory. Specifically, we recast the augmented system (10) into an abstract Cauchy problem and apply the Lumer–Phillips theorem.

Abstract formulation

2.1. Abstract formulation. Let He2(0,ℓ):={y∈H2(0,ℓ)∣y(0)=y′(0)=0}H^2_e(0,\ell):=\{y\in H^2(0,\ell)\mid y(0)=y'(0)=0\}. We define the state space H\mathcal{H} as

H:=He2(0,ℓ)×L2(0,ℓ)×L2((0,ℓ)×(0,1)),(24)\mathcal{H}:=H^2_e(0,\ell)\times L^2(0,\ell)\times L^2((0,\ell)\times(0,1)), \tag*{(24)}

equipped with the inner product

⟨U1,U2⟩=m∫0ℓv1v2 dx+EI∫0ℓy1′′‾y2′′ dx+ξ∫0ℓa(x)∫01z1‾z2 ds dx,(25)\langle U_1,U_2\rangle=m\int_0^\ell v_1v_2\,dx+EI\int_0^\ell\overline{y_1''}y_2''\,dx+\xi\int_0^\ell a(x)\int_0^1\overline{z_1}z_2\,ds\,dx, \tag*{(25)}

for all Ui=(yi,vi,zi)∈HU_i=(y_i,v_i,z_i)\in\mathcal{H}, i=1,2i=1,2. It is straightforward to check that H\mathcal{H} is a Hilbert space. Next, on the space H\mathcal{H}, we define the linear operator A\mathcal{A} with domain D(A)⊂H\mathcal{D}(\mathcal{A}) \subset\mathcal{H} given by

D(A):={U∈H:y∈H4(0,ℓ), v∈He2(0,ℓ),z∈L2(0,ℓ;H1(0,1)),z(⋅,0)=v, y′′(ℓ)=y(3)(ℓ)=0}(26)\mathcal{D}(\mathcal{A}) := \left\{ U \in\mathcal{H} : \begin{array}{l} y \in H^{4}(0,\ell),\ v \in H_{e}^{2}(0,\ell),\\ z \in L^{2}(0,\ell;H^{1}(0,1)),\\ z(\cdot,0)=v,\ y^{\prime\prime}(\ell)=y^{(3)}(\ell)=0 \end{array} \right\} \tag*{(26)}
AU:=(v−EImy(4)−am(μ1v+μ2z(⋅,1))−τ−1zs).(27)\mathcal{A}U := \begin{pmatrix} v\\ -\dfrac{EI}{m}y^{(4)}-\dfrac{a}{m}\left(\mu_{1}v+\mu_{2}z(\cdot,1)\right)\\ -\tau^{-1}z_{s} \end{pmatrix}. \tag*{(27)}

With these definitions at hand, the system (10) can be written into the abstract Cauchy problem

{ddtU(t)=AU(t),U(0)=U0=(y0,y1,f0).(28)\begin{cases} \dfrac{d}{dt}U(t)=\mathcal{A}U(t),\\ U(0)=U_{0}=(y_{0},y_{1},f_{0}). \tag*{(28)} \end{cases}

Well-posedness of system

Before establishing the well-posedness of system (1), we prove the following lemma regarding the norm equivalence on He2(0,ℓ)H_{e}^{2}(0,\ell).

Lemma 2. The mapping ∣⋅∣2:He2(0,ℓ)→R\lvert\cdot\rvert_{2}:H_{e}^{2}(0,\ell)\to\mathbb{R} defined by

∣w∣2:=(∫0ℓ∣w′′(x)∣2 dx+∫0ℓ∣w(x)∣2 dx)12,∀w∈He2(0,ℓ).(29)\lvert w\rvert_{2}:=\left(\int_{0}^{\ell}\lvert w^{\prime\prime}(x)\rvert^{2}\,dx+\int_{0}^{\ell}\lvert w(x)\rvert^{2}\,dx\right)^{\frac{1}{2}},\quad\forall w\in H_{e}^{2}(0,\ell). \tag*{(29)}

defines a norm on He2(0,ℓ)H_{e}^{2}(0,\ell) equivalent to the standard H2(0,ℓ)H^{2}(0,\ell)–norm. That is, there exist two positive constants C1,C2C_{1},C_{2} such that

C1∥w∥H2(0,ℓ)≤∣w∣2≤C2∥w∥H2(0,ℓ),∀w∈He2(0,ℓ).(30)C_{1}\lVert w\rVert_{H^{2}(0,\ell)}\leq\lvert w\rvert_{2}\leq C_{2}\lVert w\rVert_{H^{2}(0,\ell)},\quad\forall w\in H_{e}^{2}(0,\ell). \tag*{(30)}

Proof. The verification that ∣⋅∣2\lvert\cdot\rvert_{2} satisfies the norm axioms on He2(0,ℓ)H_{e}^{2}(0,\ell) is standard and omitted here. Next, we prove the equivalence property. Let w∈He2(0,ℓ)w\in H_{e}^{2}(0,\ell). By definition, we immediately have ∣w∣2≤∥w∥H2(0,ℓ)\lvert w\rvert_{2}\leq\lVert w\rVert_{H^{2}(0,\ell)}. Furthermore, using the triangle and Cauchy–Schwarz inequalities, we get

∫0ℓ∣w′(x)∣2 dx≤∫0ℓx∫0x∣w′′(σ)∣2 dσ dx≤ℓ22∥w′′∥L2(0,ℓ)2.(31)\int_{0}^{\ell}\lvert w^{\prime}(x)\rvert^{2}\,dx\leq\int_{0}^{\ell}x\int_{0}^{x}\lvert w^{\prime\prime}(\sigma)\rvert^{2}\,d\sigma\,dx \leq\frac{\ell^{2}}{2}\lVert w^{\prime\prime}\rVert_{L^{2}(0,\ell)}^{2}. \tag*{(31)}

Consequently, we arrive at

∥w∥H2(0,ℓ)2≤(1+ℓ22)∣w∣22,(32)\lVert w\rVert_{H^{2}(0,\ell)}^{2}\leq\left(1+\frac{\ell^{2}}{2}\right)\lvert w\rvert_{2}^{2}, \tag*{(32)}

which completes the proof. □\square

Theorem 1. Assume that conditions (7) and (12) holds. For any U0∈HU_{0}\in\mathcal{H}, problem (10) admits a unique mild solution U∈C(R+,H)U\in C(\mathbb{R}_{+},\mathcal{H}). Moreover, if U0∈D(A)U_{0}\in\mathcal{D}(\mathcal{A}), then UU is a classical solution satisfying U∈C(R+,D(A))∩C1(R+,H)U\in C(\mathbb{R}_{+},\mathcal{D}(\mathcal{A}))\cap C^{1}(\mathbb{R}_{+},\mathcal{H}).

Proof. First, we prove that the operator A\mathcal{A} is dissipative. Let U∈D(A)U \in\mathcal{D}(\mathcal{A}). By definition of the inner product in H\mathcal{H}, we have

⟨AU,U⟩=−μ1∫0ℓa(x)∣v(x)∣2 dx−EI∫0ℓy(4)v‾ dx+EI∫0ℓv′′y′′‾ dx\langle\mathcal{A}U,U\rangle=-\mu_{1}\int_{0}^{\ell}a(x)|v(x)|^{2}\,dx-EI\int_{0}^{\ell}y^{(4)}\overline{v}\,dx+EI\int_{0}^{\ell}v^{\prime\prime}\overline{y^{\prime\prime}}\,dx
−μ2∫0ℓa(x)z(x,1)v‾ dx−ξτ−1∫0ℓa(x)∫01zsz‾ ds dx.(33)-\mu_{2}\int_{0}^{\ell}a(x)z(x,1)\overline{v}\,dx-\xi\tau^{-1}\int_{0}^{\ell}a(x)\int_{0}^{1}z_{s}\overline{z}\,ds\,dx. \tag*{(33)}

Performing several integrations by parts and taking into account the boundary conditions (10)₃ – (10)₄, the following holds:

∫0ℓy(4)v‾ dx=[y(3)v‾−y′′v′‾]x=0x=ℓ+∫0ℓy′′v′′‾ dx=∫0ℓy′′v′′‾ dx;(34)\int_{0}^{\ell}y^{(4)}\overline{v}\,dx=\left[y^{(3)}\overline{v}-y^{\prime\prime}\overline{v^{\prime}}\right]_{x=0}^{x=\ell}+\int_{0}^{\ell}y^{\prime\prime}\overline{v^{\prime\prime}}\,dx=\int_{0}^{\ell}y^{\prime\prime}\overline{v^{\prime\prime}}\,dx; \tag*{(34)}
∫01zsz‾ ds=[∣z∣2]s=0s=1−∫01zzs‾ ds=∣z(⋅,1)∣2−∣v∣2−∫01zzs‾ ds\int_{0}^{1}z_{s}\overline{z}\,ds=\left[|z|^{2}\right]_{s=0}^{s=1}-\int_{0}^{1}z\overline{z_{s}}\,ds=|z(\cdot,1)|^{2}-|v|^{2}-\int_{0}^{1}z\overline{z_{s}}\,ds
Re⁡(∫01zsz‾ ds)=12(∣z(⋅,1)∣2−∣v∣2).(35)\operatorname{Re}\left(\int_{0}^{1}z_{s}\overline{z}\,ds\right)=\frac{1}{2}\left(|z(\cdot,1)|^{2}-|v|^{2}\right). \tag*{(35)}

Thus, combining (34) and (35), we arrive at

⟨AU,U⟩=−μ1∫0ℓa(x)∣v(x)∣2 dx−μ2∫0ℓa(x)z(x,1)v‾ dx\langle\mathcal{A}U,U\rangle=-\mu_{1}\int_{0}^{\ell}a(x)|v(x)|^{2}\,dx-\mu_{2}\int_{0}^{\ell}a(x)z(x,1)\overline{v}\,dx
−ξτ−1∫0ℓa(x)∫01zsz‾ ds dx+2iIm⁡(EI∫0ℓv′′y′′‾ dx).-\xi\tau^{-1}\int_{0}^{\ell}a(x)\int_{0}^{1}z_{s}\overline{z}\,ds\,dx+2i\operatorname{Im}\left(EI\int_{0}^{\ell}v^{\prime\prime}\overline{y^{\prime\prime}}\,dx\right).
Re⁡⟨AU,U⟩=−μ1∫0ℓa(x)∣v(x)∣2 dx−μ2∫0ℓa(x)Re⁡(z(x,1)v‾) dx\operatorname{Re}\langle\mathcal{A}U,U\rangle=-\mu_{1}\int_{0}^{\ell}a(x)|v(x)|^{2}\,dx-\mu_{2}\int_{0}^{\ell}a(x)\operatorname{Re}\left(z(x,1)\overline{v}\right)\,dx
−ξτ−1∫0ℓa(x)Re⁡(∫01zsz‾ ds) dx=(−μ1+ξτ−12)∫0ℓa(x)∣v(x)∣2 dx−ξτ−12∫0ℓa(x)∣z(x,1)∣2 dx\begin{aligned} -\xi\tau^{-1}\int_{0}^{\ell}a(x)\operatorname{Re}\left(\int_{0}^{1}z_{s}\overline{z}\,ds\right)\,dx \\ &=\left(-\mu_{1}+\frac{\xi\tau^{-1}}{2}\right)\int_{0}^{\ell}a(x)|v(x)|^{2}\,dx-\frac{\xi\tau^{-1}}{2}\int_{0}^{\ell}a(x)|z(x,1)|^{2}\,dx \end{aligned}
−μ2∫0ℓa(x)Re⁡(z(x,1)v(x)‾) dx.(36)-\mu_{2}\int_{0}^{\ell}a(x)\operatorname{Re}\left(z(x,1)\overline{v(x)}\right)\,dx. \tag*{(36)}

Applying Young’s inequality to the integrand of the last integral term, we get

Re⁡(μ2z(⋅,1)v‾)≤∣μ2∣2(∣v∣2+∣z(⋅,1)∣2).(37)\operatorname{Re}\left(\mu_{2}z(\cdot,1)\overline{v}\right)\leq\frac{|\mu_{2}|}{2}\left(|v|^{2}+|z(\cdot,1)|^{2}\right). \tag*{(37)}

whose incorporation in (36) leads to:

Re⁡⟨AU,U⟩≤(−μ1+∣μ2∣2+ξτ−12)∫0ℓa(x)∣v(x)∣2 dx+(∣μ2∣2−ξτ−12)∫0ℓa(x)∣z(x,1)∣2 dx≤−C(∫0ℓa(x)∣v(x)∣2 dx+∫0ℓa(x)∣z(x,1)∣2 dx),\begin{aligned} \operatorname{Re}\langle\mathcal{A}U,U\rangle\leq\left(-\mu_{1}+\frac{|\mu_{2}|}{2}+\frac{\xi\tau^{-1}}{2}\right)\int_{0}^{\ell}a(x)|v(x)|^{2}\,dx \\ &+\left(\frac{|\mu_{2}|}{2}-\frac{\xi\tau^{-1}}{2}\right)\int_{0}^{\ell}a(x)|z(x,1)|^{2}\,dx \\ &\leq-C\left(\int_{0}^{\ell}a(x)|v(x)|^{2}\,dx+\int_{0}^{\ell}a(x)|z(x,1)|^{2}\,dx\right), \end{aligned}

where the constant CC is positive by virtue of condition (12) and is defined as

C:=min⁡{−∣μ2∣2+ξτ−12; μ1−∣μ2∣2−ξτ−12}.C := \min\left\{ -\frac{|\mu_2|}{2}+\frac{\xi\tau^{-1}}{2};\ \mu_1-\frac{|\mu_2|}{2}-\frac{\xi\tau^{-1}}{2}\right\}.

Next, to prove the maximality of A\mathcal{A}, it suffices to show that λI−A\lambda I-\mathcal{A} is surjective for a given λ>0\lambda>0. That is, for any Ψ=(f,g,h)T∈H\Psi=(f,g,h)^{\mathsf{T}}\in\mathcal{H}, we seek a solution U=(y,v,z)T∈D(A)U=(y,v,z)^{\mathsf{T}}\in\mathcal{D}(\mathcal{A}) to the equation (λI−A)U=Ψ(\lambda I-\mathcal{A})U=\Psi. Equivalently, this resolvent equation is expressed componentwise as:

{λy−v=f,λv+EImy(4)+am(μ1v+μ2z(⋅,1))=g,λz+τ−1zs=h.(38)\begin{cases} \lambda y-v=f,\\ \lambda v+\dfrac{EI}{m}y^{(4)}+\dfrac{a}{m}\left(\mu_1v+\mu_2z(\cdot,1)\right)=g,\\ \lambda z+\tau^{-1}z_s=h. \tag*{(38)} \end{cases}

Solving equation (38)₁ and equation (38)₃ with respect to xx and ss, respectively, and taking into account the boundary condition (10)₄, we obtain

v=λy−f,(39)v=\lambda y-f, \tag*{(39)}
z(⋅,s)=e−λτs(v+τ−1∫0sh(⋅,r)eτλr dr).(40)z(\cdot,s)=e^{-\lambda\tau s}\left(v+\tau^{-1}\int_0^s h(\cdot,r)e^{\tau\lambda r}\,dr\right). \tag*{(40)}

Consequently, inserting expressions (39)–(40) into (38)₂, we arrive at the following boundary value problem:

{EIy(4)+λGτ,1(λ,⋅)y=mg+Gτ,1(λ,⋅)f+Gτh(λ,⋅),y(0)=y′(0)=y′′(ℓ)=y(3)(ℓ)=0,(41)\begin{cases} EIy^{(4)}+\lambda G_{\tau,1}(\lambda,\cdot)y=mg+G_{\tau,1}(\lambda,\cdot)f+G_\tau^h(\lambda,\cdot),\\ y(0)=y'(0)=y''(\ell)=y^{(3)}(\ell)=0, \tag*{(41)} \end{cases}

where the functions Gτ,1G_{\tau,1}, GτhG_\tau^h are defined by

Gτ,1(λ,x):=mλ+(μ1+μ2e−τλ)a(x),(42)G_{\tau,1}(\lambda,x):=m\lambda+(\mu_1+\mu_2e^{-\tau\lambda})a(x), \tag*{(42)}
Gτh(λ,x):=−mμ2τe−τλ∫01h(⋅,r)eτλr dr,(43)G_\tau^h(\lambda,x):=-m\mu_2\tau e^{-\tau\lambda}\int_0^1h(\cdot,r)e^{\tau\lambda r}\,dr, \tag*{(43)}

We note that μ1+μ2e−τλ>0\mu_1+\mu_2e^{-\tau\lambda}>0 by virtue of (12). We multiply the equation (41)₁ by φ‾\overline{\varphi} and integrate over (0,ℓ)(0,\ell) with respect to xx:

∫0ℓEI(y(4)+λGτ,1(λ,⋅)y)φ‾ dx=∫0ℓ(mg+Gτ,1(λ,⋅)f+Gτh(λ,⋅))φ‾ dx.(44)\int_0^\ell EI\left(y^{(4)}+\lambda G_{\tau,1}(\lambda,\cdot)y\right)\overline{\varphi}\,dx = \int_0^\ell\left(mg+G_{\tau,1}(\lambda,\cdot)f+G_\tau^h(\lambda,\cdot)\right)\overline{\varphi}\,dx. \tag*{(44)}

By performing several integrations by parts together with the boundary conditions (41)₂, we obtain

EI∫0ℓy′′φ′′‾ dx+λ∫0ℓGτ,1(λ,⋅)yφ‾ dx=∫0ℓ(mg+Gτ,1(λ,⋅)f+Gτh(λ,⋅))φ‾ dx.(45)EI\int_0^\ell y''\overline{\varphi''}\,dx+\lambda\int_0^\ell G_{\tau,1}(\lambda,\cdot)y\overline{\varphi}\,dx = \int_0^\ell\left(mg+G_{\tau,1}(\lambda,\cdot)f+G_\tau^h(\lambda,\cdot)\right)\overline{\varphi}\,dx. \tag*{(45)}

This leads to the variational problem which consists in finding y∈He2(0,ℓ)y\in H^2_e(0,\ell) such that

J(y,φ)=L(φ),(46)\mathcal{J}(y,\varphi)=\mathcal{L}(\varphi), \tag*{(46)}

for all φ∈H02(0,ℓ)\varphi\in H^2_0(0,\ell), where the functionals J\mathcal{J} and L\mathcal{L} are given by

J(y,φ)=EI∫0ℓy′′φ′′‾ dx+λ∫0ℓGτ,1(λ,⋅)yφ‾ dx.(47)\mathcal{J}(y,\varphi)=EI\int_0^\ell y''\overline{\varphi''}\,dx+\lambda\int_0^\ell G_{\tau,1}(\lambda,\cdot)y\overline{\varphi}\,dx. \tag*{(47)}
L(φ)=∫0ℓ(mg+Gτ,1(λ,⋅)f+Gτh(λ,⋅))φ‾ dx.(48)\mathcal{L}(\varphi)=\int_0^\ell\left(mg+G_{\tau,1}(\lambda,\cdot)f+G^h_{\tau}(\lambda,\cdot)\right)\overline{\varphi}\,dx. \tag*{(48)}

It is straightforward to verify that J\mathcal{J} and L\mathcal{L} are sesquilinear and antilinear, respectively. Furthermore, by applying the Cauchy–Schwarz inequality and taking into account condition (6), we obtain the following estimates:

∣∫0ℓy′′φ′′‾ dx∣≤∥y′′∥L2(0,ℓ)∥φ′′∥L2(0,ℓ)(49)\left|\int_0^\ell y''\overline{\varphi''}\,dx\right|\leq\|y''\|_{L^2(0,\ell)}\|\varphi''\|_{L^2(0,\ell)} \tag*{(49)}
∣∫0ℓGτ,1(λ,⋅)yφ‾ dx∣≤[mλ+(μ1+μ2e−τλ)∥a∥L∞(0,ℓ)]∥y∥L2(0,ℓ)∥φ∥L2(0,ℓ).(50)\left|\int_0^\ell G_{\tau,1}(\lambda,\cdot)y\overline{\varphi}\,dx\right|\leq\left[m\lambda+\left(\mu_1+\mu_2e^{-\tau\lambda}\right)\|a\|_{L^\infty(0,\ell)}\right]\|y\|_{L^2(0,\ell)}\|\varphi\|_{L^2(0,\ell)}. \tag*{(50)}

Combining expressions (49) and (50), it follows that

∣J(y,φ)∣≤EI∥y′′∥L2(0,ℓ)∥φ′′∥L2(0,ℓ)+λ[mλ+(μ1+μ2e−τλ)∥a∥L∞(0,ℓ)]∥y∥L2(0,ℓ)∥φ∥L2(0,ℓ)∣J(y,φ)∣≤C∥y∥H2(0,ℓ)∥φ∥H2(0,ℓ),(51)\begin{aligned} |\mathcal{J}(y,\varphi)|&\leq EI\|y''\|_{L^2(0,\ell)}\|\varphi''\|_{L^2(0,\ell)}\\ &\quad+\lambda\left[m\lambda+\left(\mu_1+\mu_2e^{-\tau\lambda}\right)\|a\|_{L^\infty(0,\ell)}\right]\|y\|_{L^2(0,\ell)}\|\varphi\|_{L^2(0,\ell)}\\ |\mathcal{J}(y,\varphi)|&\leq C\|y\|_{H^2(0,\ell)}\|\varphi\|_{H^2(0,\ell)}, \tag*{(51)} \end{aligned}

where the positive constant CC is given by:

C:=2max⁡{EI,λ(mλ+(μ1+μ2e−τλ)∥a∥L∞(0,ℓ))}.(52)C:=2\max\left\{EI,\lambda\left(m\lambda+\left(\mu_1+\mu_2e^{-\tau\lambda}\right)\|a\|_{L^\infty(0,\ell)}\right)\right\}. \tag*{(52)}

Thus, J\mathcal{J} is continuous. Next, by the Cauchy–Schwarz inequality, we have

∣∫0ℓ(∫01h(⋅,r)eτλr dr)φ‾ dx∣≤∥∫01h(⋅,r)eτλr dr∥L2(0,ℓ)∥φ∥L2(0,ℓ)≤eτλ∥h∥L2((0,ℓ)×(0,1))∥φ∥L2(0,ℓ).(53)\begin{aligned} \left|\int_0^\ell\left(\int_0^1h(\cdot,r)e^{\tau\lambda r}\,dr\right)\overline{\varphi}\,dx\right|&\leq\left\|\int_0^1h(\cdot,r)e^{\tau\lambda r}\,dr\right\|_{L^2(0,\ell)}\|\varphi\|_{L^2(0,\ell)}\\ &\leq e^{\tau\lambda}\|h\|_{L^2((0,\ell)\times(0,1))}\|\varphi\|_{L^2(0,\ell)}. \tag*{(53)} \end{aligned}

Similarly to estimate (51), using (53) yields

∣L(φ)∣≤[m∥g∥L2(0,ℓ)+mμ2τ∥h∥L2((0,ℓ)×(0,1))+(mλ+(μ1+μ2e−τλ)∥a∥L∞(0,ℓ))∥f∥L2(0,ℓ)]∥φ∥L2(0,ℓ)≤C∥φ∥H2(0,ℓ),(54)\begin{aligned} |\mathcal{L}(\varphi)|&\leq\left[m\|g\|_{L^2(0,\ell)}+m\mu_2\tau\|h\|_{L^2((0,\ell)\times(0,1))}\right.\\ &\quad\left.+\left(m\lambda+\left(\mu_1+\mu_2e^{-\tau\lambda}\right)\|a\|_{L^\infty(0,\ell)}\right)\|f\|_{L^2(0,\ell)}\right]\|\varphi\|_{L^2(0,\ell)}\\ &\leq C\|\varphi\|_{H^2(0,\ell)}, \tag*{(54)} \end{aligned}

where the constant C>0C>0 is given by:

C:=m∥g∥L2(0,ℓ)+mμ2τ∥h∥L2((0,ℓ)×(0,1))+(mλ+(μ1+μ2e−τλ)∥a∥L∞(0,ℓ))∥f∥L2(0,ℓ).(55)\begin{aligned} C&:=m\|g\|_{L^2(0,\ell)}+m\mu_2\tau\|h\|_{L^2((0,\ell)\times(0,1))}\\ &\quad+\left(m\lambda+\left(\mu_1+\mu_2e^{-\tau\lambda}\right)\|a\|_{L^\infty(0,\ell)}\right)\|f\|_{L^2(0,\ell)}. \tag*{(55)} \end{aligned}

This proves the continuity of the functional L\mathcal{L}. Moreover, J\mathcal{J} is coercive on He2(0,ℓ)H_e^2(0,\ell) since

Re⁡J(y,y)=EI∫0ℓ∣y′′∣2 dx+λ∫0ℓGτ,1(λ,x)∣y∣2 dx≥EI∫0ℓ∣y′′∣2 dx+mλ2∫0ℓ∣y∣2 dx+(1−e−τλ)μ1λ∫0ℓa(x)∣y∣2 dx\begin{aligned} \operatorname{Re}\mathcal{J}(y,y)=EI\int_0^\ell|y''|^2\,dx+\lambda\int_0^\ell G_{\tau,1}(\lambda,x)|y|^2\,dx \\ &\ge EI\int_0^\ell|y''|^2\,dx+m\lambda^2\int_0^\ell|y|^2\,dx+(1-e^{-\tau\lambda})\mu_1\lambda\int_0^\ell a(x)|y|^2\,dx \end{aligned}
Re⁡J(y,y)≥min⁡{EI;mλ2}∥y∥22≥min⁡{EI;mλ2}C1∥y∥H2(0,ℓ)2.(56)\operatorname{Re}\mathcal{J}(y,y)\ge\min\{EI;m\lambda^2\}\|y\|_2^2\ge\min\{EI;m\lambda^2\}C_1\|y\|_{H^2(0,\ell)}^2. \tag*{(56)}

Consequently, by the Lax–Milgram theorem, the variational problem (46) admits a unique solution y∈He2(0,ℓ)y\in H_e^2(0,\ell).

Conversely, consider a weak solution y∈H2(0,ℓ)y\in H^2(0,\ell). Applying integration by parts, equation (46) becomes:

∫0ℓ(EIy(4)+λGτ,1(λ,⋅)y)φ‾ dx+EIy′′(ℓ)φ′(ℓ)‾−EIy(3)(ℓ)φ(ℓ)‾=∫0ℓ(mg+Gτ,1(λ,⋅)f+Gτh(λ,⋅))φ‾ dx.(57)\int_0^\ell\left(EIy^{(4)}+\lambda G_{\tau,1}(\lambda,\cdot)y\right)\overline{\varphi}\,dx+EIy''(\ell)\overline{\varphi'(\ell)}-EIy^{(3)}(\ell)\overline{\varphi(\ell)} =\int_0^\ell\left(mg+G_{\tau,1}(\lambda,\cdot)f+G_\tau^h(\lambda,\cdot)\right)\overline{\varphi}\,dx. \tag*{(57)}

Taking first φ∈D(0,ℓ)\varphi\in\mathcal{D}(0,\ell), the preceding identity reduces to

∫0ℓ(EIy(4)+λGτ,1(λ,⋅)y)φ‾ dx=∫0ℓ(mg+Gτ,1(λ,⋅)f+Gτh(λ,⋅))φ‾ dx.(58)\int_0^\ell\left(EIy^{(4)}+\lambda G_{\tau,1}(\lambda,\cdot)y\right)\overline{\varphi}\,dx=\int_0^\ell\left(mg+G_{\tau,1}(\lambda,\cdot)f+G_\tau^h(\lambda,\cdot)\right)\overline{\varphi}\,dx. \tag*{(58)}

This implies that EIy(4)+λGτ,1(λ,⋅)y=mg+Gτ,1(λ,⋅)f+Gτ,2(λ)EIy^{(4)}+\lambda G_{\tau,1}(\lambda,\cdot)y=mg+G_{\tau,1}(\lambda,\cdot)f+G_{\tau,2}(\lambda) almost everywhere on (0,ℓ)(0,\ell). Consequently, y∈H4(0,ℓ)y\in H^4(0,\ell) and satisfies (41)1_1 by virtue of the density of D(0,ℓ)\mathcal{D}(0,\ell) in L2(0,ℓ)L^2(0,\ell).

Returning to φ∈He2(0,ℓ)\varphi\in H_e^2(0,\ell) and from the above, equation (57) leads to:

EIy′′(ℓ)φ′(ℓ)‾−EIy(3)(ℓ)φ(ℓ)‾=0.(59)EIy''(\ell)\overline{\varphi'(\ell)}-EIy^{(3)}(\ell)\overline{\varphi(\ell)}=0. \tag*{(59)}

By the arbitrariness of φ\varphi, yy fulfills the boundary condition (41)2_2, and therefore solves (41). Thus, the operator λI−A\lambda I-A is surjective, which proves that AA is maximal. Finally, the Lumer–Phillips theorem ensures that AA generates a C0C_0-semigroup of contractions on H\mathcal{H}, which guarantees the existence and uniqueness of solutions to the Cauchy problem (28). □

Exponential stability

This section is dedicated to the statement and proof of the exponential stability of system (1) based on Lyapunov method combined with multiplier techniques. To this end, we first establish several technical lemmas.

We begin by introducing the auxiliary functional Q(t)Q(t), defined as

Q(t):=Q1(t)−Q2(t),t≥0,(60)Q(t):=Q_1(t)-Q_2(t),\qquad t\ge0, \tag*{(60)}

where the functionals QiQ_i, i=1,2i=1,2 are given by

Q1(t):=m∫0ℓyyt dx(61)Q_1(t):=m\int_0^\ell y y_t\,dx \tag*{(61)}
Q2(t):=∫0ℓa(x)∫01syt2(x,t−τs) ds dx.(62)Q_2(t):=\int_0^\ell a(x)\int_0^1 sy_t^2(x,t-\tau s)\,ds\,dx. \tag*{(62)}

Prior to estimating the time derivative of Q(t)Q(t), we state and prove a few Poincaré-type inequalities.

Lemma 3. Let y∈He2(0,ℓ)y \in H_e^2(0,\ell). Then, the following estimates hold:

∫0ℓy2(x) dx≤ℓ48∥y′′∥L2(0,ℓ)2,(63)\int_0^\ell y^2(x)\,dx \le\frac{\ell^4}{8}\lVert y''\rVert_{L^2(0,\ell)}^2, \tag*{(63)}
∫0ℓa(x)y2(x) dx≤ℓ32∫0ℓa(x) dx∥y′′∥L2(0,ℓ)2.(64)\int_0^\ell a(x)y^2(x)\,dx \le\frac{\ell^3}{2}\int_0^\ell a(x)\,dx\lVert y''\rVert_{L^2(0,\ell)}^2. \tag*{(64)}

Proof. We have for all x∈(0,ℓ)x \in(0,\ell):

y2(x)=(∫0xy′(r) dr)2≤x∫0xy′(r)2 dr.(65)y^2(x)=\left(\int_0^x y'(r)\,dr\right)^2\le x\int_0^x y'(r)^2\,dr. \tag*{(65)}

In the same manner, taking into account y′(0)=0y'(0)=0, we have

y′2(r)≤r∫0ry′′(σ)2 dσ≤r∥y′′∥L2(0,ℓ)2.(66)y'^2(r)\le r\int_0^r y''(\sigma)^2\,d\sigma\le r\lVert y''\rVert_{L^2(0,\ell)}^2. \tag*{(66)}

Consequently, integrating over (0,ℓ)(0,\ell) and combining estimates (65) and (66), we deduce that

∫0ℓy2(x) dx≤∫0ℓx∫0xy′(r)2 dr dx≤∫0ℓx∫0xr dr dx∥y′′∥L2(0,ℓ)2\begin{aligned} \int_0^\ell y^2(x)\,dx &\le\int_0^\ell x\int_0^x y'(r)^2\,dr\,dx\\ &\le\int_0^\ell x\int_0^x r\,dr\,dx\lVert y''\rVert_{L^2(0,\ell)}^2 \end{aligned}
∫0ℓy2(x) dx≤ℓ48∥y′′∥L2(0,ℓ)2;\int_0^\ell y^2(x)\,dx\le\frac{\ell^4}{8}\lVert y''\rVert_{L^2(0,\ell)}^2;
∫0ℓa(x)y2(x) dx≤∫0ℓxa(x)∫0xr dr dx∥y′′∥L2(0,ℓ)2≤∫0ℓx32a(x) dx∥y′′∥L2(0,ℓ)2\begin{aligned} \int_0^\ell a(x)y^2(x)\,dx &\le\int_0^\ell xa(x)\int_0^x r\,dr\,dx\lVert y''\rVert_{L^2(0,\ell)}^2\\ &\le\int_0^\ell\frac{x^3}{2}a(x)\,dx\lVert y''\rVert_{L^2(0,\ell)}^2 \end{aligned}
∫0ℓa(x)y2(x) dx≤ℓ32∫0ℓa(x) dx∥y′′∥L2(0,ℓ)2.\int_0^\ell a(x)y^2(x)\,dx\le\frac{\ell^3}{2}\int_0^\ell a(x)\,dx\lVert y''\rVert_{L^2(0,\ell)}^2.

This completes the proof. □

Lemma 4. Let yy be the solution of system (1). For every ε1>0\varepsilon_1>0, the time derivative of Q(t)Q(t) satisfies for all t≥0t\ge0,

Q′(t)≤∫0ℓ(m2+μ122ε1a(x))yt2 dx+∫0ℓ(−EI+ε1ℓ32∫0ℓa(x) dx)yxx2 dx+(τ−1+μ222ε1)∫0ℓa(x)yt2(x,t−τ) dx−τ−1∫0ℓa(x)∫01yt2(x,t−τs) ds dx.(67)\begin{aligned} Q'(t)\le{}&\int_0^\ell\left(\frac{m}{2}+\frac{\mu_1^2}{2\varepsilon_1}a(x)\right)y_t^2\,dx+\int_0^\ell\left(-EI+\varepsilon_1\frac{\ell^3}{2}\int_0^\ell a(x)\,dx\right)y_{xx}^2\,dx\\ &+\left(\tau^{-1}+\frac{\mu_2^2}{2\varepsilon_1}\right)\int_0^\ell a(x)y_t^2(x,t-\tau)\,dx-\tau^{-1}\int_0^\ell a(x)\int_0^1 y_t^2(x,t-\tau s)\,ds\,dx. \tag*{(67)} \end{aligned}

Proof. First, we compute the time derivative of Q1(t)Q_1(t). Using equation (1)₁, we have

Q1′(t)=m∫0ℓ(yt2+yytt) dx=m∫0ℓyt2 dx−EI∫0ℓyyxxxx dx−∫0ℓa(x)(μ1yt+μ2yt(x,t−τ))y dx.(68)\begin{aligned} Q_1'(t)=m\int_0^\ell\left(y_t^2+yy_{tt}\right)\,dx \\ &=m\int_0^\ell y_t^2\,dx-EI\int_0^\ell yy_{xxxx}\,dx-\int_0^\ell a(x)\left(\mu_1y_t+\mu_2y_t(x,t-\tau)\right)y\,dx. \tag*{(68)} \end{aligned}

Next, using integrations by parts together with the boundary condition (1)₂, the second integral term in (68) can be rewritten as

EI∫0ℓyxxxxy dx=EI[yyxxx−yxyxx]x=0x=ℓ+EI∫0ℓyxx2 dx=EI∫0ℓyxx2 dx.(69)\begin{aligned} EI\int_0^\ell y_{xxxx}y\,dx=EI\left[yy_{xxx}-y_xy_{xx}\right]_{x=0}^{x=\ell}+EI\int_0^\ell y_{xx}^2\,dx \\ &=EI\int_0^\ell y_{xx}^2\,dx. \tag*{(69)} \end{aligned}

It follows that:

Q1′(t)=m∫0ℓyt2 dx−EI∫0ℓyxx2 dx−∫0ℓa(x)(μ1yt+μ2yt(x,t−τ))y dx.(70)Q_1'(t)=m\int_0^\ell y_t^2\,dx-EI\int_0^\ell y_{xx}^2\,dx-\int_0^\ell a(x)\left(\mu_1y_t+\mu_2y_t(x,t-\tau)\right)y\,dx. \tag*{(70)}

Furthermore, applying Young’s inequality to the last term of (70) and invoking estimate (64) from Lemma 3, we obtain for any ε1>0\varepsilon_1>0:

∣∫0ℓxa(x)(μ1yt+μ2yt(x,t−τ))y dx∣≤ε1∫0ℓa(x)y2 dx+μ122ε1∫0ℓa(x)yt2 dx+μ222ε1∫0ℓa(x)yt2(x,t−τ) dx≤ℓ32ε1∫0ℓa(x) dx ∥y′′∥L2(0,ℓ)2+μ122ε1∫0ℓa(x)yt2 dx+μ222ε1∫0ℓa(x)yt2(x,t−τ) dx.(71)\begin{aligned} \left|\int_0^\ell xa(x)\left(\mu_1y_t+\mu_2y_t(x,t-\tau)\right)y\,dx\right| \leq\varepsilon_1\int_0^\ell a(x)y^2\,dx+\frac{\mu_1^2}{2\varepsilon_1}\int_0^\ell a(x)y_t^2\,dx+\frac{\mu_2^2}{2\varepsilon_1}\int_0^\ell a(x)y_t^2(x,t-\tau)\,dx \\ &\leq\frac{\ell^3}{2}\varepsilon_1\int_0^\ell a(x)\,dx\,\lVert y''\rVert_{L^2(0,\ell)}^2+\frac{\mu_1^2}{2\varepsilon_1}\int_0^\ell a(x)y_t^2\,dx+\frac{\mu_2^2}{2\varepsilon_1}\int_0^\ell a(x)y_t^2(x,t-\tau)\,dx. \tag*{(71)} \end{aligned}

Thus, inserting estimate (71) into (70) leads to

Q1′(t)≤∫0ℓ(m+μ122ε1a(x))yt2 dx+(−EI+ε1ℓ32∫0ℓa(x) dx)∫0ℓyxx2 dx+μ222ε1∫0ℓa(x)yt2(x,t−τ) dx.(72)\begin{aligned} Q_1'(t)\leq\int_0^\ell\left(m+\frac{\mu_1^2}{2\varepsilon_1}a(x)\right)y_t^2\,dx \\ &+\left(-EI+\varepsilon_1\frac{\ell^3}{2}\int_0^\ell a(x)\,dx\right)\int_0^\ell y_{xx}^2\,dx \\ &+\frac{\mu_2^2}{2\varepsilon_1}\int_0^\ell a(x)y_t^2(x,t-\tau)\,dx. \tag*{(72)} \end{aligned}

Next, differentiating Q2(t)Q_2(t) with respect to tt, we obtain

Q2′(t)=2∫0ℓa(x)∫01syt(x,t−τs)ytt(x,t−τs) ds dx=−τ−1∫0ℓa(x)∫01s∂∂s(yt2(x,t−τs)) ds dx=τ−1∫0ℓa(x)∫01yt2(x,t−τs) ds dx−τ−1∫0ℓa(x)yt2(x,t−τ) dx.(73)\begin{aligned} Q_2'(t)=2\int_0^\ell a(x)\int_0^1 sy_t(x,t-\tau s)y_{tt}(x,t-\tau s)\,ds\,dx \\ &=-\tau^{-1}\int_0^\ell a(x)\int_0^1s\frac{\partial}{\partial s}\left(y_t^2(x,t-\tau s)\right)\,ds\,dx \\ &=\tau^{-1}\int_0^\ell a(x)\int_0^1y_t^2(x,t-\tau s)\,ds\,dx-\tau^{-1}\int_0^\ell a(x)y_t^2(x,t-\tau)\,dx. \tag*{(73)} \end{aligned}

Finally, combining estimates (72) and (73) directly yields (67). □

We now define the Lyapunov functional V(t)V(t) by

V(t):=E(t)+δQ(t),t≥0,(74)V(t) := E(t) + \delta Q(t), \qquad t \ge0, \tag*{(74)}

where E(t)E(t), Q(t)Q(t) are defined in (11) and (60), respectively, and δ>0\delta> 0 is a positive parameter to be specified later.

Lemma 5. For a sufficiently small δ>0\delta> 0, there exist two positive constants θ1\theta_1 and θ2\theta_2 such that, for all t≥0t \ge0:

θ1E(t)≤V(t)≤θ2E(t).(75)\theta_1 E(t) \le V(t) \le\theta_2 E(t). \tag*{(75)}

Proof. Using estimate (63) together with Young’s inequality, we have for any ε2>0\varepsilon_2 > 0:

∣Q1(t)∣≤m∫0ℓ∣yyt∣ dx≤m(ε22∫0ℓy2 dx+12ε2∫0ℓyt2 dx)≤m(ε2ℓ416∥yxx∥L2(0,ℓ)2+12ε2∥yt∥L2(0,ℓ)2).(76)\begin{aligned} \lvert Q_1(t)\rvert&\le m \int_0^\ell\lvert y y_t\rvert\, dx \\ &\le m\left(\frac{\varepsilon_2}{2}\int_0^\ell y^2\, dx+\frac{1}{2\varepsilon_2}\int_0^\ell y_t^2\, dx\right) \\ &\le m\left(\frac{\varepsilon_2\ell^4}{16}\lVert y_{xx}\rVert_{L^2(0,\ell)}^2+\frac{1}{2\varepsilon_2}\lVert y_t\rVert_{L^2(0,\ell)}^2\right). \tag*{(76)} \end{aligned}

Choosing ε2=22ℓ2\varepsilon_2=\frac{2\sqrt{2}}{\ell^2}, the preceding estimate simplifies to:

∣Q1(t)∣≤mℓ242(∥yxx∥L2(0,ℓ)2+∥yt∥L2(0,ℓ)2).(77)\lvert Q_1(t)\rvert\le\frac{m\ell^2}{4\sqrt{2}}\left(\lVert y_{xx}\rVert_{L^2(0,\ell)}^2+\lVert y_t\rVert_{L^2(0,\ell)}^2\right). \tag*{(77)}

On the other hand, Q2(t)Q_2(t) satisfies:

∣Q2(t)∣≤∫0ℓa(x)∫01yt2(x,t−τs) ds dx.(78)\lvert Q_2(t)\rvert\le\int_0^\ell a(x)\int_0^1 y_t^2(x,t-\tau s)\,ds\,dx. \tag*{(78)}

Gathering (77) and (78) leads to

∣Q(t)∣≤mℓ242(∥yxx∥L2(0,ℓ)2+∥yt∥L2(0,ℓ)2)+δ∫0ℓa(x)∫01yt2(x,t−τs) ds dx≤CE(t),(79)\begin{aligned} \lvert Q(t)\rvert&\le\frac{m\ell^2}{4\sqrt{2}}\left(\lVert y_{xx}\rVert_{L^2(0,\ell)}^2+\lVert y_t\rVert_{L^2(0,\ell)}^2\right) \\ &\quad+\delta\int_0^\ell a(x)\int_0^1 y_t^2(x,t-\tau s)\,ds\,dx \\ &\le CE(t), \tag*{(79)} \end{aligned}

where C>0C>0 is defined as

C:=max⁡{2δξ,ℓ222,mℓ222EI}.(80)C:=\max\left\{\frac{2\delta}{\xi},\frac{\ell^2}{2\sqrt{2}},\frac{m\ell^2}{2\sqrt{2}EI}\right\}. \tag*{(80)}

Finally, by applying the triangle inequality on the estimate (79) and by adding the energy E(t)E(t) to each side of the resulting inequality, we arrive at:

0<(1−δC)E(t)≤V(t)≤(1+δC)E(t),0<(1-\delta C)E(t)\le V(t)\le(1+\delta C)E(t),

by choosing δ∈(0,C−1)\delta\in(0,C^{-1}). □

The main result of this paper states as follows:

Theorem 2. Suppose that assumptions (7) and (12) are satisfied. For any regular solution yy of (1), there exist constants υ1,υ2>0\upsilon_1,\upsilon_2>0 such that the energy E(t)E(t) satisfies

E(t)≤υ1E(0)e−υ2t,∀t≥0.(81)E(t)\le\upsilon_1 E(0)e^{-\upsilon_2t}, \qquad\forall t\ge0. \tag*{(81)}

Proof. In view of Lemmas 1 and 4, we obtain (13) and (67), together with condition (6)$_2$, the time derivative of V(t)V(t) satisfies

ddtV(t)=ddtE(t)+ddtQ(t)≤∫0ℓ[−(−μ1+∣μ2∣2+ξτ−12)a(x)+δ(m2+μ122ε1)a(x)]yt2 dx+δ∫0ℓ(−EI+ε1ℓ32∫0ℓa(x) dx)yxx2 dx−δτ−1∫0ℓa(x)∫01yt2(x,t−τs) ds dx+[∣μ2∣2−ξτ−12+δ(τ−1+μ222ε1)]∫0ℓa(x)yt2(x,t−τ) dx≤[−(−μ1+∣μ2∣2+ξτ−12)a0+δ(m2+μ122ε1∥a∥L∞(0,ℓ))]∫0ℓyt2 dx+δ∫0ℓ(−EI+ε1ℓ32∫0ℓa(x) dx)yxx2 dx−δτ−1∫0ℓa(x)∫01yt2(x,t−τs) ds dx+[∣μ2∣2−ξτ−12+δ(τ−1+μ222ε1)]∫0ℓa(x)yt2(x,t−τ) dx.(82)\begin{aligned} \frac{d}{dt}V(t)={}&\frac{d}{dt}E(t)+\frac{d}{dt}Q(t)\\ \le{}&\int_{0}^{\ell}\left[-\left(-\mu_1+\frac{|\mu_2|}{2}+\frac{\xi\tau^{-1}}{2}\right)a(x)+\delta\left(\frac{m}{2}+\frac{\mu_1^2}{2\varepsilon_1}\right)a(x)\right]y_t^2\,dx\\ &+\delta\int_{0}^{\ell}\left(-EI+\varepsilon_1\frac{\ell^3}{2}\int_{0}^{\ell}a(x)\,dx\right)y_{xx}^2\,dx\\ &-\delta\tau^{-1}\int_{0}^{\ell}a(x)\int_{0}^{1}y_t^2(x,t-\tau s)\,ds\,dx\\ &+\left[\frac{|\mu_2|}{2}-\frac{\xi\tau^{-1}}{2}+\delta\left(\tau^{-1}+\frac{\mu_2^2}{2\varepsilon_1}\right)\right]\int_{0}^{\ell}a(x)y_t^2(x,t-\tau)\,dx\\ \le{}&\left[-\left(-\mu_1+\frac{|\mu_2|}{2}+\frac{\xi\tau^{-1}}{2}\right)a_0+\delta\left(\frac{m}{2}+\frac{\mu_1^2}{2\varepsilon_1}\lVert a\rVert_{L^\infty(0,\ell)}\right)\right]\int_{0}^{\ell}y_t^2\,dx\\ &+\delta\int_{0}^{\ell}\left(-EI+\varepsilon_1\frac{\ell^3}{2}\int_{0}^{\ell}a(x)\,dx\right)y_{xx}^2\,dx\\ &-\delta\tau^{-1}\int_{0}^{\ell}a(x)\int_{0}^{1}y_t^2(x,t-\tau s)\,ds\,dx\\ &+\left[\frac{|\mu_2|}{2}-\frac{\xi\tau^{-1}}{2}+\delta\left(\tau^{-1}+\frac{\mu_2^2}{2\varepsilon_1}\right)\right]\int_{0}^{\ell}a(x)y_t^2(x,t-\tau)\,dx. \tag*{(82)} \end{aligned}

We fix ε1>0\varepsilon_1>0 as

ε1=EI(ℓ3∫0ℓa(x) dx)−1.(83)\varepsilon_1=EI\left(\ell^3\int_{0}^{\ell}a(x)\,dx\right)^{-1}. \tag*{(83)}

Then, we select δ\delta small enough so that 0<δ<δ00<\delta<\delta_0, where δ0\delta_0 is defined by

δ0:=min⁡{(−∣μ2∣2+ξ2τ)(τ−1+μ222ε1)−1,(−μ1+∣μ2∣2+ξ2τ)(m2+μ122ε1∥a∥L∞(0,ℓ))−1}.(84)\delta_0:=\min\left\{\left(-\frac{|\mu_2|}{2}+\frac{\xi}{2\tau}\right)\left(\tau^{-1}+\frac{\mu_2^2}{2\varepsilon_1}\right)^{-1},\left(-\mu_1+\frac{|\mu_2|}{2}+\frac{\xi}{2\tau}\right)\left(\frac{m}{2}+\frac{\mu_1^2}{2\varepsilon_1}\lVert a\rVert_{L^\infty(0,\ell)}\right)^{-1}\right\}. \tag*{(84)}

we deduce that

V′(t)≤−CδE(t),(85)V'(t)\le-C_\delta E(t), \tag*{(85)}

where the positive constant CδC_\delta is given by:

Cδ:=min⁡{δ,2δτξ,(−2μ1+∣μ2∣+ξ2τ)a0m−δ(1+μ12ε1m∥a∥L∞(0,ℓ))}.C_\delta:=\min\left\{\delta,\frac{2\delta}{\tau\xi},\left(-2\mu_1+|\mu_2|+\frac{\xi}{2\tau}\right)\frac{a_0}{m}-\delta\left(1+\frac{\mu_1^2}{\varepsilon_1m}\lVert a\rVert_{L^\infty(0,\ell)}\right)\right\}.

Thanks to the functional equivalence established in Lemma 5, we arrive at

V′(t)≤−Cδθ1V(t).(86)V'(t)\le-\frac{C_\delta}{\theta_1}V(t). \tag*{(86)}

applying Gronwall’s inequality together with Lemma 5 directly yields the exponential decay estimate (81).

References

Université Nangui Abrogoua, Abidjan, Côte d’Ivoire

Email address: [email protected]

Université Félix Houphouët–Boigny, Abidjan, Côte d’Ivoire.

Email address: [email protected]

Université Félix Houphouët–Boigny, Abidjan, Côte d’Ivoire

Email address: [email protected]

References

  1. [1]M. Azzaoui, J. Salhi, and M. Segaoui, Mathematical and numerical analysis of the stability of coupled Euler–Bernoulli equations with indirect time-delayed damping. preprint, (2026). https://dx.doi.org/10.2139/ssrn.6899599
  2. [2]A. V. Balakrishnan and L. Taylor, The SCOLE design challenge, Proceedings of the 3rd Annual NASA–SCOLE Workshop "NASA Tech. Memo.", volume 89075, Hampton 1987, 385–412.
  3. [3]A. Camasta, G. Fragnelli, and C. Pignotti, Stability for some classes of degenerate nonlinear hyperbolic equations with time delay. Nonlinear Anal. Real World Appl., 81(2025), no. 104191. https://doi.org/10.1016/j.nonrwa.2024.104191
  4. [4]L. Cao, S.–J. Li, and S. Chai, Stabilization of a degenerate beam on an elastic foundation with axial force under fractional derivative-type delayed boundary feedback. Electron. J. Differ. Equ., 2026(2026), no. 73, pp. 1–22. https://doi.org/10.5897/ejde.2026.73
  5. [5]A. Chakravarthy, K. A. Evans, and J. E. Evers, Sensitivities & functional gains for a flexible aircraft-inspired model, Proceedings of the 2010 American Control Conference, 2010, 4893–4898. https://doi.org/10.1109/ACC.2010.5530914
  6. [6]R. Datko, J. Lagnesе, and M. P. Polis, An example of the effect of time delays in boundary feedback stabilization of wave equations, Proceedings of the 1985 24th IEEE Conference on Decision and Control, 1985, 117–117. https://doi.org/10.1109/CDC.1985.268529
  7. [7]R. Datko, Two questions concerning the boundary control of certain elastic systems. J. Differ. Equations, 921991, no. 1, pp. 27–44. https://doi.org/10.1016/0022-0396(91)90062-E
  8. [8]R. Datko, Two examples of ill-posedness with respect to small time delays in stabilized elastic systems. IEEE Trans. Automat. Contr., 38(1993), no. 1, pp. 163–166. https://doi.org/10.1109/9.1863382
  9. [9]Y. Du, P. Cheng, and F. Zhou, Free vibration analysis of axial-loaded beams with variable cross sections and multiple concentrated elements. J. Vib. Control, 28(2022), no. 17–18, pp. 2197–2211. https://doi.org/10.1177/10775463211012878
  10. [10]C. R. Fanou, K. Ammari, and I. Boussada, Delay is necessary for a potential to achieve exponential stabilization of the wave equation via control. Preprint, (2026). https://arxiv.org/abs/2603.21777
  11. [11]X. Feng, G. Xu, and Y. Chen, Rapid stabilisation of an Euler–Bernoulli beam with the internal delay control. Int. J. Control, 92(2019), no. 1, pp. 42–55. https://doi.org/10.1080/00207179.2017.1286693
  12. [12]I. Gumowski and C. Mira, Optimization in Control Theory and Practice, Cambridge University Press, Cambridge, 1968.DOI
  13. [13]B.–Z. Guo and K.–Y. Yang, Dynamic stabilization of an Euler–Bernoulli beam equation with time delay in boundary observation. Automatica, 45(2009), no. 6, pp. 1468–1475. https://doi.org/10.1016/j.automatica.2009.02.004
  14. [14]W. He, S. Ge, B. How, and Y. S. Choo, Dynamics and Control of Mechanical Systems in Offshore Engineering, Springer London, London, 2014. https://doi.org/10.1007/978-1-4471-5337-5
  15. [15]H. Hlachami and R. Shorten, Boundary input-to-state stabilization of a damped Euler–Bernoulli beam in the presence of a state-delay. Preprint, (2019). https://arxiv.org/abs/1912.01117
  16. [16]N. Mahfoudhi and H. Ayadi, Well-posedness and global stabilization of an Euler–Bernoulli beam with distributed damping and nonlinear boundary delay. Acta Appl. Math., 203(2026), no. 4. https://doi.org/10.1007/s10440-026-00795-5
  17. [17]S. Nicaise and C. Pignotti, Stability and instability results of the wave equation with a delay term in the boundary or internal feedbacks. J. Control Optim., 45(2006), no. 5, pp. 1561–1585. https://doi.org/10.1137/060648891
  18. [18]A. Pazy, Semigroups of Linear Operators and Applications to Partial Differential Equations, Springer New York, New York, 1983. https://doi.org/10.1007/978-1-4612-5561-1
  19. [19]A. J. Rahmtati, G.–Q. Xu, and S. Effati, Stability analysis of the Euler–Bernoulli beam with multi-delay controller. Control Theory Technol., 20(2022), pp. 338–348. https://doi.org/10.1007/s11768-022-00095-x
  20. [20]J. Salhi, M. Azzaoui, and M. Tilioua, Stabilization of a degenerate Euler–Bernoulli beam model under boundary time-delayed damping. Rend. Circ. Mat. Palermo II. Ser., 75(2026), pp. 119. https://doi.org/10.1007/s12215-026-01444-4
  21. [21]B. B. J. Siriki and A. Coulibaly, Boundary stabilization of a degenerate Euler–Bernoulli beam under axial force and time delay. Preprint, (2025). https://arxiv.org/abs/2512.18179
  22. [22]B. B. J. Siriki and A. Coulibaly, Exponential stability of a degenerate Euler–Bernoulli beam with axial force and delayed boundary control. Preprint, (2026). https://arxiv.org/abs/2510.25484
  23. [23]B. B. J. Siriki and A. Coulibaly, Boundary stabilization of an Euler–Bernoulli beam with axial force and internal delay. Preprint, (2026). https://arxiv.org/abs/2605.24315
  24. [24]A. Smyshlyaev, B.-Z. Guo, and M. Krstic, Arbitrary decay rate for Euler–Bernoulli beam by backstepping boundary feedback. IEEE Trans. Autom. Control, 54(2009), no. 5, pp. 1134–1140. https://doi.org/10.1109/TAC.2009.2013038
  25. [25]A. Wright, Modern Control Design for Flexible Wind Turbines, PhD thesis, University of Colorado, 2003. https://doi.org/10.2172/15011696

Paper details

vlnc:W_nhm8q7ph
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 5

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents