数学杂志  2017, Vol. 37 Issue (5): 1075-1080   PDF    
扩展功能
加入收藏夹
复制引文信息
加入引用管理器
Email Alert
RSS
本文作者相关文章
冯艳青
王忠英
姚俊
文传军
一类二阶抛物型方程初边值问题解的存在定理
冯艳青, 王忠英, 姚俊, 文传军    
常州工学院数理与化工学院, 江苏常州 213000
摘要:本文研究了一类二阶非线性抛物型方程解的存在唯一性问题.利用非线性分析中的吸引盆理论和同胚理论,获得了相应的二阶非线性抛物型方程初边值问题解的大范围存在唯一性定理.
关键词二阶抛物型方程    初边值问题    吸引盆    全局同胚    
A EXISTENCE THEOREM FOR SOME SECOND ORDER PARABOLIC INITIAL-BOUNDARY VALUE PROBLEMS
FENG Yan-qing, WANG Zong-ying, YAO Jun, WEN Chuan-jun    
School of Mathematics and Chemical Engineering, Changzhou Institute of Technology, Changzhou 213000, China
Abstract: In this paper, an a priori estimate for a second order linear parabolic operators is established. By using the basin of attraction and homeomorphism, a new sufficient condition of the existence and uniqueness of an initial boundary value problem for a second order parabolic equations is proved. This idea can be applied some semi-linear partial differential equations.
Key words: second order parabolic equation     initial-boundary value problem     the basin of attraction     homeomorphism    
1 引言

二阶抛物型方程

$ \begin{eqnarray} \Delta u - {u_t} = h(x, t, u) \end{eqnarray} $ (1.1)

也被称为热传导方程, 其中$\Delta$表示$n$维拉普拉斯算子.由于在物理、几何中的广泛应用, 许多数学工作者都研究过方程(1) 解的存在性问题, 也得到很多结果(见文献[1-5]).

在抛物型偏微分方程解的存在性问题的研究中, 一般是先建立一个可能的解的先验估计, 然后利用一些非线性分析的方法证明解的存在性.如Elcart和Sigillito在文[1]中先推导出了抛物算子${L_a}u$

$ \begin{aligned} \ {L_a}u= {a_{ij}}{u_{{x_i}{x_j}}} + {b_i}{u_{{x_i}}} - au - c{u_t} \end{aligned} $ (1.2)

的一个先验估计$\begin{aligned} \left\| u \right\|_{2, 1} \le C{\left\| {Lu} \right\|_0}, \end{aligned}$进而得到了解存在唯一性定理.

受到上述思想的启发, 我们将对二阶抛物型算子

$ \begin{aligned} {L_0}u = \Delta u - au - {u_t} \end{aligned} $ (1.3)

建立一个优先估计

$ \begin{aligned} \left\| u \right\|_{2, 1} \le C({\left\| u \right\|_0} + {\left\| {Lu} \right\|_0}), \end{aligned} $ (1.4)

然后利用非线性分析的方法讨论方程(1.1) 解的存在性问题, 并推导出一类二阶抛物型方程初边值问题解的存在唯一性的一个充分条件.我们的证明不同于Elcart和sigillito给出的证明.一个有趣的工具, 吸引盆, 在我们的主要定理2.1的证明中起着重要的作用, 下面先给出有关吸引盆理论的相关知识.

引理1.1[6]  设$E, F$为Banach空间, $H$$E$中连通开集. $f:H\subset E\to F$在H上是局部同胚的$ {C^1}$映射.设${x_0}\in H\ $, 对于任何$x \in H$, 路径提升问题

$ \begin{aligned} \left\{ \begin{aligned} &f({\gamma _x}(t)) = f({x_0}) + e^{ - t}(f(x) - f({x_0})), ~~t \in R, \\ &\gamma _x(0) = x, ~~{\gamma _x}(t) \in H \end{aligned} \right. \end{aligned} $ (1.5)

有唯一一个定义在最大开区间$ {I_x} = ({t_{{x}}^ - }, {t_{{x}}^ + }) , -\infty \le {t_{{x}}^-}, {t_{{x}}^ + } \le + \infty $上的连续解$ t \to {\gamma _x}(t)$, 并且集合$\{ (x, t) \in H \times R:t \in {I_x}\} $$ H \times R$上的开集, 映射$ (x, t) \to {\gamma _x}(t)$是连续的.

定义1.1[6]  在引理1.1的假设条件下, ${x_0}$的吸引盆是指集合$\begin{aligned} B = \{ x \in H:t_x^ + = + \infty \}. \end{aligned}$

定理1.1[8]  设连续映射$f:H\subset E\to F$是局部同胚的, 则$f$是全局同胚的充要条件是对所有的$x \in B$, ${\gamma _x}(t)$都定义在实数R上, 即${\gamma _x}(t)$可以向$ -\infty $延伸.

2 一些不等式

在这一部分, 将推导一些重要的不等式, 它们在主要定理的证明中起着重要作用.

考虑二阶抛物算子

$ \begin{equation} {L_0}u = \Delta u - au - {u_t}, \end{equation} $ (2.1)

其中$a(x)$$t, {x_1}, \cdots, {x_n}$的有界函数.设${W_0}(D)$是以

$ \begin{equation} {({\left\| u \right\|_{2, 1}})^2} = \int {({u^2} + {{\left| {\nabla u} \right|}^2} + {{\left| {{D^2}u} \right|}^2} + {u_t}^2)dxdt} \end{equation} $ (2.2)

为范数的Hilbert空间, 其中梯度是相对于空间变量, 并且$u \in {W_0}(D)$意味着$u$是定义在$D = \Omega \times [0, T]$中的; $\Omega$是紧的有界集, 其边界分段光滑且处处有非负平均曲率; ${\left| {{D^2}u} \right|^2}$表示关于空间变量的所有二阶导数的平方和.

$ \begin{equation} Lu = \Delta u - {u_t}, \end{equation} $ (2.3)

(2.1) 式可以写为

$ \begin{equation} {L_0}u = Lu - au. \end{equation} $ (2.4)

根据上述假设可得${L_0}$是一个由${W_0}(D)$映入${L_2}(D)$的线性算子.

定义$m = \mathop {\inf }\limits_D a$, $M = \mathop {\sup }\limits_D a$.

定理2.1  假设$m = \mathop {\inf }\limits_D a > - \lambda $, 则对于$u \in {W_0}(D)$, 不等式

$ \begin{equation} \int_D {{{(\nabla u)}^2}} dxdt \le \frac{1}{\lambda }\int_D {{{(Lu)}^2}} dxdt + (\frac{{1 + 3\lambda }}{2} + \frac{M}{\lambda })\int_D {{u^2}} dxdt + \frac{1}{2}\int_D {{u_t}^2} dxdt \end{equation} $ (2.5)

成立, 其中$\lambda$${L_2}(D)$上满足齐次边界条件的拉普拉斯算子$ - \Delta $的最小特征值.

  由式(2.3)-(2.5), 对于${\varepsilon _1} > 0$${\varepsilon _2} > 0$, 使用算术几何平均不等式两次, 得到

$ \begin{aligned} &\int_D {{{(\nabla u)}^2}} dxdt + \int_D {a{u^2}} dxdt = - \int_D {u{L_0}u} dxdt + \int_D {u{u_t}} dxdt \\ \le& \frac{1}{{2{\varepsilon _1}}}\int_D {{{({L_0}u)}^2}} dxdt+ \frac{{{\varepsilon _1}}}{2}\int_D {{u^2}dxdt} + \frac{1}{{2{\varepsilon _2}}}\int_D {u_t^2dxdt} + \frac{{{\varepsilon _2}}}{2}\int_D {{u^2}dxdt}, \\ &\int_D {{{(\nabla u)}^2}} dxdt \le \frac{1}{{{\varepsilon _1}}}\int_D {{{(Lu)}^2}} dxdt + (\frac{{{\varepsilon _1} + {\varepsilon _2}}}{2} + \frac{M}{{{\varepsilon _1}}} - m)\int_D {{u^2}} dxdt + \frac{1}{{2{\varepsilon _2}}}\int_D {{u_t}^2} dxdt. \end{aligned} $

(1)$m = 0 < \lambda$并设${\varepsilon _1} = \lambda, 1 = {\varepsilon _2}$, 得

$ \begin{aligned} \int_D {{{(\nabla u)}^2}} dxdt \le \frac{1}{\lambda }\int_D {{{(Lu)}^2}} dxdt + (\frac{{1 + \lambda }}{2} + \frac{M}{\lambda })\int_D {{u^2}} dxdt + \frac{1}{2}\int_D {{u_t}^2} dxdt. \end{aligned} $

(2)$m > 0$, 则有$m = \mathop {\inf }\limits_D a > - \lambda$, 设${\varepsilon _1} = \lambda, 1 = {\varepsilon _2}$, 有

$ \begin{aligned} \int_D {{{(\nabla u)}^2}} dxdt \le \frac{1}{\lambda }\int_D {{{(Lu)}^2}} dxdt + (\frac{{1 + 3\lambda }}{2} + \frac{M}{\lambda })\int_D {{u^2}} dxdt + \frac{1}{2}\int_D {{u_t}^2} dxdt, \end{aligned} $

所以

$ \begin{aligned} \int_D {{{(\nabla u)}^2}} dxdt \le \frac{1}{\lambda }\displaystyle\int_D {{{(Lu)}^2}} dxdt + (\frac{{1 + 3\lambda }}{2} + \frac{M}{\lambda })\displaystyle\int_D {{u^2}} dxdt + \frac{1}{2}\displaystyle\int_D {{u_t}^2} dxdt. \end{aligned} $

下面两个定理的证明类似于文献[9] (Chap.2, §8) 和文献[10]中的相关证明.

定理2.2   $\displaystyle\int_D {{{({u_t})}^2}dxdt \le 2\displaystyle\int_D {{{(Lu)}^2}dxdt} } + 2(1 + M)\displaystyle\int_D {{u^2}dxdt}.$

定理2.3   $\displaystyle\int_D {{{\left| {{D^2}u} \right|}^2}dxdt \le \displaystyle\int_D {{{(\Delta u)}^2}dxdt} }.$

定理2.4   $\displaystyle\int_D {{{(\Delta u)}^2}dxdt \le 4\displaystyle\int_D {{{(Lu)}^2}dxdt} } + (7 + {M^2})\displaystyle\int_D {{u^2}dxdt}. $

$ \begin{eqnarray*}&&\displaystyle\int_D {au\Delta u} dxdt \le \frac{1}{2}\displaystyle\int_D {{{(au)}^2}} dxdt + \frac{1}{2}\displaystyle\int_D {{{(\Delta u)}^2}} dxdt, \\ &&\displaystyle\int_D {[{{(\Delta u)}^2}-{u_t}\Delta u-Lu\Delta u]} dxdt \le \frac{1}{2}\displaystyle\int_D {{{(au)}^2}} dxdt + \frac{1}{2}\displaystyle\int_D {{{(\Delta u)}^2}} dxdt. \end{eqnarray*} $

通过选择适当$\varepsilon$, 利用算术几何平均值不等式, 得到不等式

$ \begin{aligned} \displaystyle\int_D {{{(\Delta u)}^2}dxdt \le 4\displaystyle\int_D {{{(Lu)}^2}dxdt} } + (7 + {M^2})\displaystyle\int_D {{u^2}dxdt}. \end{aligned} $

定理2.5  假设$m = \mathop {\inf }\limits_D a > - \lambda $, 不等式$\begin{aligned} {\left\| u \right\|_{2, 1}} \le C({\left\| u \right\|_0} + {\left\| {Lu} \right\|_0}) \end{aligned}$成立, 并且

$ \begin{aligned} {\rm{C = max\{ }}7 + \frac{1}{\lambda }, {M^2} + 3M + 12 + \frac{{3\lambda }}{2} + \frac{M}{\lambda }{\rm{\} }}. \end{aligned} $ (2.6)

  我们能够得到

$ \begin{aligned} &{({\left\| u \right\|_{2, 1}})^2} = \displaystyle\int {({u^2} + {{\left| {\nabla u} \right|}^2} + {{\left| {{D^2}u} \right|}^2} + {u_t}^2)dxdt}\\ =& \displaystyle\int {{u^2}dxdt + \displaystyle\int {({{\left| {\nabla u} \right|}^2} + {{\left| {{D^2}u} \right|}^2} + {u_t}^2)dxd} t}. \end{aligned} $

根据定理2.1-2.4, 有

$ \begin{aligned} &\displaystyle\int {({{\left| {\nabla u} \right|}^2} + {{\left| {{D^2}u} \right|}^2} + {u_t}^2)dxd} t\\ \le& (7 + \frac{1}{\lambda })\displaystyle\int_D {{{(Lu)}^2}dxdt} + ({M^2} + 3M + 11 + \frac{{3\lambda }}{2} + \frac{M}{\lambda })\displaystyle\int_D {{u^2}dxdt}, \end{aligned} $

所以

$ \begin{aligned} {({\left\| u \right\|_{2, 1}})^2} \le (7 + \frac{1}{\lambda })\displaystyle\int_D {{{(Lu)}^2}dxdt} + ({M^2} + 3M + 12 + \frac{{3\lambda }}{2} + \frac{M}{\lambda })\displaystyle\int_D {{u^2}dxdt}. \end{aligned} $

${\rm{C = max\{ }}7 + \frac{1}{\lambda }, {M^2} + 3M + 12 + \frac{{3\lambda }}{2} + \frac{M}{\lambda }{\rm{\} }}$, 则$\begin{aligned} {\left\| u \right\|_{2, 1}} \le C({\left\| u \right\|_0} + {\left\| {Lu} \right\|_0}). \end{aligned}$

3 主要定理

下面给出方程

$ \begin{aligned} Lu = \Delta u - {u_t} = h(x, t, u), \end{aligned} $ (3.1)

满足初始条件$u(x, 0) = 0$, $(x, t) \in \partial \Omega \times (0, T)$的解的存在性和唯一性的充分条件.

$\partial \Omega \in {C^2}$, 对于所有的$(x, t)$, $h$关于$u$连续, 且对于所有的$u, h$关于$(x, t)$可测, 且有直到三阶连续偏导数.

将式(3.1) 改写为算子形式

$ \begin{aligned} F(u) = Lu - h(x, t, u), \end{aligned} $ (3.2)

于是式(3.1) 等价于

$ \begin{aligned} F(u) = 0, u \in {W_0} \end{aligned} $ (3.3)

求Frechet导数, 于是对一切$u, \phi \in {W_0}(D)$, 有

$ \begin{aligned} F'(u)(\phi ) = L\phi - { h'_u}(x, u)\phi, u, \phi \in {W_0}. \end{aligned} $ (3.4)

定理3.1  假设

(ⅰ)$\mathop {{\rm{inf}}}\limits_D {h'_u} > - \lambda $;

(ⅱ)对任意$u \in {W_0}$, $\displaystyle\int_1^{ + \infty } {\frac{{ds}}{{1 + R(s) + T(s)R(s)}}} = \infty, $

这里$T(s) = \mathop {{\rm{sup}}}\limits_D {h'_u}(x, u)$, $R(s) = \sup \left\| {{{[L-{{h'}_u}(x, u)]}^{ - 1}}} \right\|$, 则方程F(u)=0在${W_0}$中有唯一的解.

  条件$\mathop {{\rm{inf}}}\limits_D {h'_u} > - \lambda $表明, 零不是$L\phi - {\dot h'_u}(x, u)\phi = 0$的特征值, 所以对所有$u \in {W_0}(D)$, 算子$F'(u) = L - {h'_u}(x, u)I$是可逆的, 其中$I$${W_0}(D)$上的恒同算子, 于是$F$是一个从${W_0}(D)$${L_2}(D)$局部同胚.

任取$u \in {W_0}(D)$, 记${[L + {h'_u}(x, u)I]^{ - 1}}\phi = \xi$.由定理2.5, 有

$ \begin{aligned} \left\| \xi \right\| &\le C({\left\| \xi \right\|_0} + {\left\| {L\xi } \right\|_0}) = C(\left\| {{{[L-{{h'}_u}(x, u)I]}^{ - 1}}\phi } \right\| + \left\| {L{{[L-{{h'}_u}(x, u)I]}^{ - 1}}\phi } \right\|\\ &\le C(R(\left\| u \right\|)\left\| \phi \right\| + \left\| \phi \right\| + \left\| {{{h'}_u}(x, u)} \right\|\left\| {{{[L-{h_u}^\prime (x, u)I]}^{ - 1}}\phi } \right\|)\\ &\le C(R(\left\| u \right\|)\left\| \phi \right\| + \left\| \phi \right\| + \mathop {\sup }\limits_{\Omega \times R} \left\| {{{h'}_u}(x, u)} \right\|R(\left\| u \right\|)\left\| \phi \right\|)\\ &\le C(R(\left\| u \right\|)\left\| \phi \right\| + \left\| \phi \right\| + T(\left\| u \right\|)R(\left\| u \right\|)\left\| \phi \right\|), \end{aligned} $

所以得到

$ \begin{aligned} \left\| {F'{{(u)}^{ - 1}}} \right\| = \left\| {{{[L-{{h'}_u}(x, u)I]}^{ - 1}}} \right\| \le C(1 + R(\left\| u \right\|) + T(\left\| u \right\|)R(\left\| u \right\|)). \end{aligned} $

考虑映射F的路径提升问题

$ \begin{aligned} \left\{ \begin{aligned} &{F({p_u}(t)) - F({p_u}(0)) = {e^{ - t}}(F(u) - F({u_0}))}, \\ &{{p_u}(0) = {u_0}}, \end{aligned} \right. \end{aligned} $

$ \begin{aligned} {p'_u}(t) = - {[F'({p_u}(t))]^{ - 1}}{e^{ - t}}(F(u) - F({u_0})), \end{aligned} $

于是

$ \begin{aligned} \left\| {{p_u}(t) - {p_u}(0)} \right\| &= \left\| {\displaystyle\int_0^t {{{p'}_u}(t)dt} } \right\| \le \displaystyle\int_0^t {\left\| {{{p'}_u}(t)} \right\|} dt\\ &= \displaystyle\int_0^t {\left\| {{{[F'({p_u}(t))]}^{ -1}}{e^{ -t}}(F(u) -F({u_0}))} \right\|} dt\\ &\le C\displaystyle\int_0^t {\left| {1 + R({p_u}(t)) + T({p_u}(t))R({p_u}(t))} \right|} {e^{ - t}}\left\| {F(u) - F({u_0})} \right\|dt, \\ \left\| {{p_u}(t)} \right\| &\le \left\| {{p_u}(0)} \right\| + C\displaystyle\int_0^t {\left| {1 + R({p_u}(t)) + T({p_u}(t))R({p_u}(t))} \right|} {e^{ - t}}\left\| {F(u) - F({u_0})} \right\|dt. \end{aligned} $

利用Gronwall不等式, 有

$ \begin{aligned} \displaystyle\int_{\left\| {{p_u}(0)} \right\|}^{\left\| {{p_u}(t)} \right\|} {\frac{{ds}}{{1 + R(s) + T(s)R(s)}}} \le C\displaystyle\int_0^t {{e^{ - t}}\left\| {F(u) - F({u_0})} \right\|} dt \le C\left\| {F(u) - F({u_0})} \right\|. \end{aligned} $

结合式条件(ⅱ), 得$\exists M > 0$, $\begin{aligned} \left\| {{p_u}(t)} \right\| \le M, t > 0. \end{aligned}$由定理1.1, 现在只需要证明${p_u}(t)$可以向$ - \infty $延伸即可.

$g( - t) = {p_u}(t), t \in (b, 0], b < 0$.那么对于${t_1}, {t_2} \in (b, 0]$,

$ \begin{aligned} \left\| {{p_u}({t_1}) -{p_u}({t_2})} \right\| &= \left\| {g( -{t_1}) -g( - {t_2})} \right\| = \displaystyle\int_{ - {t_2}}^{ - {t_1}} {\left\| {g'(s)} \right\|} ds\\ &= \displaystyle\int_{ - {t_2}}^{ - {t_1}} {\left\| {F'({p_u}{{(t)}^{ - 1}}} \right\|} {e^s}\left\| {F(u) - F({u_0})} \right\|ds\\ &\le C\displaystyle\int_{ - {t_2}}^{ - {t_1}} {\left| {1 + R(M) + T(M)R(M)} \right|} {e^s}\left\| {F(u) - F({u_0})} \right\|ds\\ &\le C(1 + R(M) + T(M)R(M)){e^b}\left\| {F(u) - F({u_0})} \right\|\left| {{t_1} - {t_2}} \right|, \end{aligned} $

所以${p_u}(t)$$( - b, 0]$上Lipschitz连续, ${p_u}(t)$可以向$- \infty$延伸, 也就是说, $F$${W_0}(D)$${L_2}(D)$上的全局同胚, 并且在${W_0}$中方程$F(u) = 0$有唯一的解, 命题得证.

备注  在定理3.1中的条件(ii)可以替换为$\left\| {{{h'}_u}} \right\| \le \omega (\left\| u \right\|)$, 其中$\omega $连续的满足$\displaystyle\int_a^\infty {\frac{{dt}}{{\omega (t)}}} = \infty $.证明与定理3.1相类似.

参考文献
[1] Elcrat A R, Sigillito V G. An explicit a priori estimate for parabolic equations with applications to semilinear equations[J]. J. Math. Anal., 1976, 7(3): 746–753.
[2] Chen P J, Gurtin M E. On a theory of heat conduction involving two temperatures[J]. Z. Angew. Math. Phys., 1968, 19(4): 614–627. DOI:10.1007/BF01594969
[3] Sigillito V G. Pointwise bounds for solutions of semilinear parabolic equations[J]. SIAM J. Appl. Math., 1967, 9(3): 581–285.
[4] Sigillito V G. On a continuous method of approximating solutions of the heat equation[J]. J. Assoc. Comp. Mach., 1967, 14(4): 732–741. DOI:10.1145/321420.321432
[5] Feng Y Q, Wang Z Y. The application of the basin of attraction to the existence and uniqueness of solutions for the second order parabolic boundary value problem[J]. J. Math., 2016, 36(5): 949–954.
[6] Gorni G. Acriterion of invertibility in the large for local diffeomorphisms between Banach spaces[J]. Nonl. Anal., 1993, 21(1): 43–47. DOI:10.1016/0362-546X(93)90176-S
[7] Plastock R. Homeomorphism between Banach space[J]. Trans. Am. Math. Soc., 1974, 200: 169–183. DOI:10.1090/S0002-9947-1974-0356122-6
[8] Wang W X, Shen Z H. The basin of attraction in Banach spaces and its applications[J]. Acta Math. Sinica Chin. Ser., 2006, 49(5): 1013–1020.
[9] Ladyzhenskaya O A, Uraltseva N N. Linear and quasiliear elliptic equation[M]. New York: Academic Press, 1968.
[10] Elcrat A R. Constructive existence for semilinear eliptic equations with discontinuous coefficients[J]. SIAM J. Math. Anal., 1974, 5(4): 663–672. DOI:10.1137/0505066
[11] Feng Y Q, Wang Z Y, Wen C J. Global homeomorphism and applications tothe existence and uniqueness of solutions ofsome differential equations[J]. Adv. Diff. Equ., 2014: 52. DOI:10.1186/1687-1847-2014-52