定义曲面 χ Γ \chi_\Gamma χΓ,定义 u Γ ∈ H # 1 ( Γ ) u_{\Gamma} \in H_{\#}^{1}(\Gamma) uΓ∈H#1(Γ),它是如下弱形式的解:
( 33 ) ∫ Γ ∇ Γ u Γ ⋅ ∇ Γ v = ∫ Γ f Γ v ∀ v ∈ H # 1 ( Γ ) (33) \quad \int_{\Gamma} \nabla_{\Gamma} u_{\Gamma} \cdot \nabla_{\Gamma} v=\int_{\Gamma} f_{\Gamma} v \quad \forall v \in H_{\#}^{1}(\Gamma) (33)∫Γ∇ΓuΓ⋅∇Γv=∫ΓfΓv∀v∈H#1(Γ)
定义误差矩阵 E , E Γ ∈ R ( n + 1 ) × ( n + 1 ) \mathbf{E}, \mathbf{E}_{\Gamma} \in \mathbb{R}^{(n+1) \times(n+1)} E,EΓ∈R(n+1)×(n+1),
( 34 ) ∫ Γ ∇ Γ v ⋅ ∇ Γ w − ∫ γ ∇ γ v ~ ⋅ ∇ γ w ~ = ∫ γ ∇ γ v ~ ⋅ E ∇ γ w ~ = ∫ Γ ∇ Γ v ⋅ E Γ ∇ Γ w (34) \quad \int_{\Gamma} \nabla_{\Gamma} v \cdot \nabla_{\Gamma} w-\int_{\gamma} \nabla_{\gamma} \widetilde{v} \cdot \nabla_{\gamma} \widetilde{w}=\int_{\gamma} \nabla_{\gamma} \widetilde{v} \cdot \mathbf{E} \nabla_{\gamma} \widetilde{w}=\int_{\Gamma} \nabla_{\Gamma} v \cdot \mathbf{E}_{\Gamma} \nabla_{\Gamma} w (34)∫Γ∇Γv⋅∇Γw−∫γ∇γv ⋅∇γw =∫γ∇γv ⋅E∇γw =∫Γ∇Γv⋅EΓ∇Γw
对于所有的 v , w ∈ H 1 ( Γ ) v, w \in H^{1}(\Gamma) v,w∈H1(Γ)和所有的lift v ~ , w ~ ∈ H 1 ( γ ) \widetilde{v}, \widetilde{w} \in H^{1}(\gamma) v ,w ∈H1(γ)都成立。所谓的lift,指的是定义曲线 Γ \Gamma Γ上的函数值 v v v,它和 γ \gamma γ上的函数 v ~ \tilde{v} v~满足如下的关系,
v = v ~ ∘ χ ∘ χ Γ − 1 v=\tilde{v} \circ \chi \circ \chi_{\Gamma}^{-1} v=v~∘χ∘χΓ−1
对于不同正则性的曲面,我们有不同的对于 E \mathbf{E} E和 E Γ \mathbf{E_\Gamma} EΓ的估计。
首先,我们
那么我们有如下的引理,
Lemma 14 (relation between tangential gradients). If v ~ : γ → R is of class H 1 , then the tangential gradients ∇ γ v ~ and ∇ Γ v satisfy (35) ∇ Γ v = D χ Γ g Γ − 1 D χ t ∇ γ v ~ , ∇ γ v ~ = D χ g − 1 D χ Γ t ∇ Γ v \begin{array}{l}{\text { Lemma } 14 \text { (relation between tangential gradients). If } \tilde{v} : \gamma \rightarrow \mathbb{R} \text { is of class } H^{1} \text { , }} \\ {\text { then the tangential gradients } \nabla_{\gamma} \widetilde{v} \text { and } \nabla_{\Gamma} v \text { satisfy }} \\ {\begin{array}{llll}{\text { (35) }} & {\nabla_{\Gamma} v=D \chi_{\Gamma} \mathbf{g}_{\Gamma}^{-1} D \chi^{t} \nabla_{\gamma} \widetilde{v},} & {\nabla_{\gamma} \widetilde{v}=D \chi \mathbf{g}^{-1} D \chi_{\Gamma}^{t} \nabla_{\Gamma} v}\end{array}}\end{array} Lemma 14 (relation between tangential gradients). If v~:γ→R is of class H1 , then the tangential gradients ∇γv and ∇Γv satisfy (35) ∇Γv=DχΓgΓ−1Dχt∇γv ,∇γv =Dχg−1DχΓt∇Γv
证明:
由曲面梯度的第一个定义,
∇ Γ v = D χ Γ g Γ − 1 ∇ ( v ∘ χ Γ ) = D χ Γ g Γ − 1 ∇ ( v ~ ∘ χ ) = D χ Γ g Γ − 1 D χ t ∇ γ v ~ \nabla_{\Gamma} v=D \chi_{\Gamma} \mathrm{g}_{\Gamma}^{-1} \nabla\left(v \circ \chi_{\Gamma}\right)=D \chi_{\Gamma} \mathrm{g}_{\Gamma}^{-1} \nabla(\tilde{v} \circ \chi)=D \chi_{\Gamma} \mathrm{g}_{\Gamma}^{-1} D \chi^{t} \nabla_{\gamma} \tilde{v} ∇Γv=DχΓgΓ−1∇(v∘χΓ)=DχΓgΓ−1∇(v~∘χ)=DχΓgΓ−1Dχt∇γv~
因为 C 1 C^1 C1在 H 1 H^1 H1是稠密的,得证。第二个证明类似。
Lemma 15 (geometric consistency). The error matrices E and E Γ read on V (36) E = D χ ( q Γ q g Γ − 1 − g − 1 ) D χ t (37) E Γ = D χ Γ ( g Γ − 1 − q q Γ g − 1 ) D χ Γ t \begin{array}{l}{\text { Lemma } 15 \text { (geometric consistency). The error matrices } \mathrm{E} \text { and } \mathrm{E}_{\Gamma} \text { read on } \mathcal{V}} \\ {\text { (36) } \quad \mathrm{E}=D \chi\left(\frac{q_{\Gamma}}{q} \mathrm{g}_{\Gamma}^{-1}-\mathrm{g}^{-1}\right) D \chi^{t}} \\ {\text { (37) } \quad \mathrm{E}_{\Gamma}=D \chi_{\Gamma}\left(\mathrm{g}_{\Gamma}^{-1}-\frac{q}{q_{\Gamma}} \mathrm{g}^{-1}\right) D \chi_{\Gamma}^{t}}\end{array} Lemma 15 (geometric consistency). The error matrices E and EΓ read on V (36) E=Dχ(qqΓgΓ−1−g−1)Dχt (37) EΓ=DχΓ(gΓ−1−qΓqg−1)DχΓt
证明:
使用(35),以及投影算子的参数化定义,我们立得:
∫ Γ ∇ Γ v ⋅ ∇ Γ w = ∫ γ ∇ γ v ~ ⋅ q Γ q ( D χ g Γ − 1 D χ t ) ∇ γ w ~ \int_{\Gamma} \nabla_{\Gamma} v \cdot \nabla_{\Gamma} w=\int_{\gamma} \nabla_{\gamma} \widetilde{v} \cdot \frac{q_{\Gamma}}{q}\left(D \chi \mathbf{g}_{\Gamma}^{-1} D \chi^{t}\right) \nabla_{\gamma} \widetilde{w} ∫Γ∇Γv⋅∇Γw=∫γ∇γv ⋅qqΓ(DχgΓ−1Dχt)∇γw
下面,我们定义两个重要的常量,用 ∣ D χ ( y ) ∣ |D \chi(\mathrm{y})| ∣Dχ(y)∣和 ∣ D − χ ( y ) ∣ \left|D^{-} \chi(\mathrm{y})\right| ∣D−χ(y)∣来表示 D χ ( y ) D \chi(\mathbf{y}) Dχ(y)最大奇异值和最小奇异值。
定义稳定化常数如下,
(38) S χ : = sup y ∈ V max { ∣ D χ ( y ) ∣ , ∣ D χ Γ ( y ) ∣ } min { ∣ D − χ ( y ) ∣ , ∣ D − χ Γ ( y ) ∣ } \begin{array}{ll}{\text { (38) }} & {S_{\chi} :=\sup _{\mathbf{y} \in \mathcal{V}} \frac{\max \left\{|D \chi(y)|,\left|D \chi_{\Gamma}(y)\right|\right\}}{\min \left\{\left|D^{-} \chi(y)\right|,\left|D^{-} \chi_{\Gamma}(y)\right|\right\}}}\end{array} (38) Sχ:=supy∈Vmin{∣D−χ(y)∣,∣D−χΓ(y)∣}max{∣Dχ(y)∣,∣DχΓ(y)∣}
定义几何精度的相对度量如下,
( 39 ) λ ∞ : = sup y ∈ V ∣ D ( χ − χ Γ ) ( y ) ∣ min { ∣ D − χ ( y ) ∣ , ∣ D − χ Γ ( y ) ∣ } (39) \quad \lambda_{\infty} :=\sup _{\mathbf{y} \in \mathcal{V}} \frac{\left|D\left(\chi-\chi_{\Gamma}\right)(\mathbf{y})\right|}{\min \left\{\left|D^{-} \chi(\mathbf{y})\right|,\left|D^{-} \chi_{\Gamma}(\mathbf{y})\right|\right\}} (39)λ∞:=y∈Vsupmin{∣D−χ(y)∣,∣D−χΓ(y)∣}∣D(χ−χΓ)(y)∣
那么我们有,
Lemma 16 (error estimates for g and q ) . The following error estimates are valid (40) ∥ I − g Γ g − 1 ∥ L ∞ ( V ) , ∥ I − g Γ − 1 g ∥ L ∞ ( V ) ≲ S X λ ∞ (41) ∥ 1 − q − 1 q I ∥ L ∞ ( V ) , ∥ 1 − q Γ − 1 q ∥ L ∞ ( V ) ≲ S X n λ ∞ \begin{array}{l}{\text { Lemma } 16 \text { (error estimates for } \mathrm{g} \text { and } q) . \text { The following error estimates are valid}} \\ {\begin{array}{ll}{\text { (40) }} & {\left\|\mathbf{I}-\mathrm{g}_{\Gamma} \mathrm{g}^{-1}\right\|_{L_{\infty}(\mathcal{V})},\left\|\mathbf{I}-\mathrm{g}_{\Gamma}^{-1} \mathrm{g}\right\|_{L_{\infty}(\mathcal{V})} \lesssim S_{\mathcal{X}} \lambda_{\infty}} \\ {\text { (41) }} & {\left\|1-q^{-1} q_{\mathrm{I}}\right\|_{L_{\infty}(\mathcal{V})},\left\|1-q_{\Gamma}^{-1} q\right\|_{L_{\infty}(\mathcal{V})} \lesssim S_{\mathcal{X}}^{n} \lambda_{\infty}}\end{array}}\end{array} Lemma 16 (error estimates for g and q). The following error estimates are valid (40) (41) ∥∥I−gΓg−1∥∥L∞(V),∥∥I−gΓ−1g∥∥L∞(V)≲SXλ∞∥∥1−q−1qI∥∥L∞(V),∥∥1−qΓ−1q∥∥L∞(V)≲SXnλ∞
这里,如何理解矩阵的L无穷范数,好的,让我们来看看。
证明:
( g − g Γ ) ( y ) = D χ ( y ) t D ( χ − χ T ) ( y ) + D ( χ − χ T ) ( y ) t D χ Γ ( y ) ∀ y ∈ V \left(\mathrm{g}-\mathrm{g}_{\mathrm{\Gamma}}\right)(\mathrm{y})=D \chi(\mathrm{y})^{t} D\left(\chi-\chi_{\mathrm{T}}\right)(\mathrm{y})+D\left(\chi-\chi_{\mathrm{T}}\right)(\mathrm{y})^{t} D \chi_{\Gamma}(\mathrm{y}) \quad \forall \mathrm{y} \in \mathcal{V} (g−gΓ)(y)=Dχ(y)tD(χ−χT)(y)+D(χ−χT)(y)tDχΓ(y)∀y∈V
(40)的一个断言成立,第二个类似。下面来证明(41),
q ( y ) − q Γ ( y ) = det g ( y ) − det g Γ ( y ) q ( y ) + q Γ ( y ) ∀ y ∈ V q(\mathbf{y})-q_{\Gamma}(\mathbf{y})=\frac{\operatorname{det} \mathbf{g}(\mathbf{y})-\operatorname{det} \mathbf{g}_{\Gamma}(\mathbf{y})}{q(\mathbf{y})+q_{\Gamma}(\mathbf{y})} \quad \forall \mathbf{y} \in \mathcal{V} q(y)−qΓ(y)=q(y)+qΓ(y)detg(y)−detgΓ(y)∀y∈V
有平方差公式,我们知道这是对的。因为 q = det g = ∏ i = 1 n λ i ( g ) q=\sqrt{\operatorname{det} \mathbf{g}}=\sqrt{\prod_{i=1}^{n} \lambda_{i}(\mathbf{g})} q=detg=∏i=1nλi(g),我们有
( 42 ) ∣ D − χ ( y ) ∣ n ≤ q ( y ) ≤ ∣ D χ ( y ) ∣ n ∀ y ∈ V (42) \quad\left|D^{-} \chi(\mathrm{y})\right|^{n} \leq q(\mathrm{y}) \leq|D \chi(\mathrm{y})|^{n} \quad \forall \mathrm{y} \in \mathcal{V} (42)∣∣D−χ(y)∣∣n≤q(y)≤∣Dχ(y)∣n∀y∈V
所以呢,
∣ q ( y ) − 1 ( q − q Γ ) ( y ) ∣ ≲ ∣ D − χ ( y ) ∣ − n ∣ D ( χ − χ Γ ) ( y ) ∣ ∣ D χ ( y ) ∣ n − 1 ∀ y ∈ V \left|q(\mathbf{y})^{-1}\left(q-q_{\Gamma}\right)(\mathbf{y})\right| \lesssim\left|D^{-} \chi(\mathbf{y})\right|^{-n}\left|D\left(\chi-\chi_{\Gamma}\right)(\mathbf{y})\right||D \chi(\mathbf{y})|^{n-1} \quad \forall \mathbf{y} \in \mathcal{V} ∣∣q(y)−1(q−qΓ)(y)∣∣≲∣∣D−χ(y)∣∣−n∣D(χ−χΓ)(y)∣∣Dχ(y)∣n−1∀y∈V
这个第一个不等号怎么理解?我也很迷惑。由此,(41)第一部分立可证,第二部分类似证法。
下面介绍一个模等价定理,
给定一个利普希茨的曲面 γ \gamma γ,让 S e q \mathcal{S}_{e q} Seq是利普希茨曲面集合,使得引理17成立于一个一致等价的常数 C e q C_{e q} Ceq。
Lemma 18 (uniform Poincar e ˊ -Friedrichs constant). Given a Lipschitz surface γ for every v ∈ H # 1 ( Γ ) with Γ ∈ S eq there holds that (45) ∥ v ∥ L 2 ( Γ ) ≲ ∥ ∇ Γ u ∥ L 2 ( Γ ) with the constant hidden in ≲ depending only on γ and C e q \begin{array}{l}{\text { Lemma 18 (uniform Poincaré-Friedrichs constant). Given a Lipschitz surface } \gamma} \\ {\text { for every } v \in H_{\#}^{1}(\Gamma) \text { with } \Gamma \in \mathcal{S}_{\text {eq }} \text { there holds that }} \\ {\begin{array}{ll}{\text { (45) }} & {\|v\|_{L_{2}(\Gamma)} \lesssim\left\|\nabla_{\Gamma} u\right\|_{L_{2}(\Gamma)}} \\ {\text { with the constant hidden in }} & {\lesssim \text { depending only on } \gamma \text { and } C_{e q}}\end{array}}\end{array} Lemma 18 (uniform Poincareˊ-Friedrichs constant). Given a Lipschitz surface γ for every v∈H#1(Γ) with Γ∈Seq there holds that (45) with the constant hidden in ∥v∥L2(Γ)≲∥∇Γu∥L2(Γ)≲ depending only on γ and Ceq
证明:
只需要证明
∥ v ∥ L 2 ( Γ ) ≤ C ∥ ∇ Γ v ∥ L 2 ( Γ ) ∀ v ∈ H 1 ( Γ ) \|v\|_{L_{2}(\Gamma)} \leq C\left\|\nabla_{\Gamma} v\right\|_{L_{2}(\Gamma)} \quad \forall v \in H^{1}(\Gamma) ∥v∥L2(Γ)≤C∥∇Γv∥L2(Γ)∀v∈H1(Γ)
我们用反证法,假设 v k ∈ H # 1 ( Γ k ) v_{k} \in H_{\#}^{1}\left(\Gamma_{k}\right) vk∈H#1(Γk),使得
∥ v k ∥ L 2 ( Γ k ) = 1 ∥ ∇ Γ k v k ∥ L 2 ( Γ k ) → 0 \left\|v_{k}\right\|_{L_{2}\left(\Gamma_{k}\right)}=1 \quad\left\|\nabla_{\Gamma_{k}} v_{k}\right\|_{L_{2}\left(\Gamma_{k}\right)} \rightarrow 0 ∥vk∥L2(Γk)=1∥∇Γkvk∥L2(Γk)→0
根据引理17,我们有
∥ v ~ k ∥ L 2 ( γ ) ≃ 1 , ∥ ∇ γ v ~ k ∥ L 2 ( γ ) → 0 \left\|\tilde{v}_{k}\right\|_{L_{2}(\gamma)} \simeq 1, \quad\left\|\nabla_{\gamma} \tilde{v}_{k}\right\|_{L_{2}(\gamma)} \rightarrow 0 ∥v~k∥L2(γ)≃1,∥∇γv~k∥L2(γ)→0
因为 L 2 L_2 L2嵌入到 H 1 H_1 H1,根据嵌入定理,我们知道,在 H 1 H^1 H1中,存在 { v ~ k } k \left\{\tilde{v}_{k}\right\}_{k} {v~k}k收敛到 v ~ ∈ H 1 ( γ ) \tilde{v} \in H^{1}(\gamma) v~∈H1(γ)。易知,因 ∇ γ v ~ = 0 \nabla_{\gamma} \tilde{v}=0 ∇γv~=0,所以 v ~ \widetilde{v} v 是一个常数。下面只要证明 v ~ = 0 \tilde{v}=0 v~=0,即可推出矛盾。
对 ∀ ϵ > 0 \forall \epsilon>0 ∀ϵ>0, ∃ k \exists k ∃k足够大,使得 ∥ v ~ k − v ~ ∥ L 2 ( γ ) ≤ ϵ \left\|\tilde{v}_{k}-\tilde{v}\right\|_{L_{2}(\gamma)} \leq \epsilon ∥v~k−v~∥L2(γ)≤ϵ,那么
∣ v ~ ∣ = ∣ Γ k ∣ − 1 ∣ ∫ Γ k v ~ ∣ = ∣ Γ k ∣ − 1 ∣ ∫ Γ k v ~ − v k ∣ ≤ ∣ Γ k ∣ − 1 / 2 ∥ v ~ − v k ∥ L 2 ( Γ k ) ≤ C e q ∣ Γ k ∣ − 1 / 2 ∥ v ~ − v ~ k ∥ L 2 ( γ ) ≤ C e q ∣ Γ k ∣ − 1 / 2 ϵ \begin{aligned}|\widetilde{v}| &=\left|\Gamma_{k}\right|^{-1}\left|\int_{\Gamma_{k}} \tilde{v}\right|=\left|\Gamma_{k}\right|^{-1}\left|\int_{\Gamma_{k}} \tilde{v}-v_{k}\right| \\ & \leq\left|\Gamma_{k}\right|^{-1 / 2}\left\|\tilde{v}-v_{k}\right\|_{L_{2}\left(\Gamma_{k}\right)} \leq C_{e q}\left|\Gamma_{k}\right|^{-1 / 2}\left\|\tilde{v}-\tilde{v}_{k}\right\|_{L_{2}(\gamma)} \leq C_{e q}\left|\Gamma_{k}\right|^{-1 / 2} \epsilon \end{aligned} ∣v ∣=∣Γk∣−1∣∣∣∣∫Γkv~∣∣∣∣=∣Γk∣−1∣∣∣∣∫Γkv~−vk∣∣∣∣≤∣Γk∣−1/2∥v~−vk∥L2(Γk)≤Ceq∣Γk∣−1/2∥v~−v~k∥L2(γ)≤Ceq∣Γk∣−1/2ϵ
对常数1使用引理17,我们有 ∣ Γ k ∣ ≃ ∣ Γ ∣ \left|\Gamma_{k}\right| \simeq|\Gamma| ∣Γk∣≃∣Γ∣,结果立得。
Lemma 19 (perturbation error estimate for C 1 , α surfaces). Let u ~ ∈ H # 1 ( γ ) solve (19) and u Γ ∈ H # 1 ( Γ ) solve (33). Then, the following error estimate for u − u Γ holds (46) ∥ ∇ γ ( u ~ − u ~ Γ ) ∥ L 2 ( γ ) ≲ λ ∞ ∥ f Γ ∥ H − 1 ( Γ ) + ∥ f q q Γ − 1 − f Γ ∥ H # − 1 ( Γ ) where the hidden constant depends on S χ defined in (38). \begin{array}{l}{\text { Lemma } 19 \text { (perturbation error estimate for } C^{1, \alpha} \text { surfaces). Let } \widetilde{u} \in H_{\#}^{1}(\gamma) \text { solve }} \\ {\text { (19) and } u_{\Gamma} \in H_{\#}^{1}(\Gamma) \text { solve (33). Then, the following error estimate for } u-u_{\Gamma}} \\ {\text { holds }} \\ {\begin{array}{ll}{\text { (46) }} & {\left\|\nabla_{\gamma}\left(\widetilde{u}-\widetilde{u}_{\Gamma}\right)\right\|_{L_{2}(\gamma)} \lesssim \lambda_{\infty}\left\|f_{\Gamma}\right\|_{H^{-1}(\Gamma)}+\left\|f q q_{\Gamma}^{-1}-f_{\Gamma}\right\|_{H_{\#}^{-1}(\Gamma)}} \\ {\text { where the hidden constant depends on } S_{\chi} \text { defined in (38). }}\end{array}}\end{array} Lemma 19 (perturbation error estimate for C1,α surfaces). Let u ∈H#1(γ) solve (19) and uΓ∈H#1(Γ) solve (33). Then, the following error estimate for u−uΓ holds (46) where the hidden constant depends on Sχ defined in (38). ∥∇γ(u −u Γ)∥L2(γ)≲λ∞∥fΓ∥H−1(Γ)+∥∥fqqΓ−1−fΓ∥∥H#−1(Γ)
证明:我们分三步证明。
1、误差表示
使用误差矩阵的定义,我们可以写下
∥ ∇ γ ( u ~ − u ~ Γ ) ∥ L 2 ( γ ) 2 = ∫ γ ∇ γ u ~ ⋅ ∇ γ v ~ − ∫ Γ ∇ Γ u Γ ⋅ ∇ Γ v + ∫ γ ∇ γ u ~ Γ ⋅ E ∇ γ v ~ \left\|\nabla_{\gamma}\left(\widetilde{u}-\widetilde{u}_{\Gamma}\right)\right\|_{L_{2}(\gamma)}^{2}=\int_{\gamma} \nabla_{\gamma} \widetilde{u} \cdot \nabla_{\gamma} \widetilde{v}-\int_{\Gamma} \nabla_{\Gamma} u_{\Gamma} \cdot \nabla_{\Gamma} v+\int_{\gamma} \nabla_{\gamma} \widetilde{u}_{\Gamma} \cdot \mathbf{E} \nabla_{\gamma} \widetilde{v} ∥∇γ(u −u Γ)∥L2(γ)2=∫γ∇γu ⋅∇γv −∫Γ∇ΓuΓ⋅∇Γv+∫γ∇γu Γ⋅E∇γv
进而有,
∥ ∇ γ ( u ~ − u ~ Γ ) ∥ L 2 ( γ ) 2 = ∫ Γ ( f q q Γ − f Γ ) v + ∫ γ ∇ γ u ~ Γ ⋅ E ∇ γ v ~ \left\|\nabla_{\gamma}\left(\widetilde{u}-\widetilde{u}_{\Gamma}\right)\right\|_{L_{2}(\gamma)}^{2}=\int_{\Gamma}\left(f \frac{q}{q_{\Gamma}}-f_{\Gamma}\right) v+\int_{\gamma} \nabla_{\gamma} \widetilde{u}_{\Gamma} \cdot \mathbf{E} \nabla_{\gamma} \widetilde{v} ∥∇γ(u −u Γ)∥L2(γ)2=∫Γ(fqΓq−fΓ)v+∫γ∇γu Γ⋅E∇γv
2、几何误差矩阵
重写 E \mathbf{E} E为
E = D χ ( ( q − 1 q Γ − 1 ) g Γ − 1 − g − 1 ( I − g g Γ − 1 ) ) D χ t \mathbf{E}=D \chi\left(\left(q^{-1} q_{\Gamma}-1\right) \mathrm{g}_{\Gamma}^{-1}-\mathbf{g}^{-1}\left(\mathbf{I}-\mathbf{g g}_{\Gamma}^{-1}\right)\right) D \chi^{t} E=Dχ((q−1qΓ−1)gΓ−1−g−1(I−ggΓ−1))Dχt
使用 g g g和 q q q的误差估计式,我们可以得到
∥ E ∥ L ∞ ( γ ) ≲ λ ∞ \|\mathbf{E}\|_{L_{\infty}(\gamma)} \lesssim \lambda_{\infty} ∥E∥L∞(γ)≲λ∞
3、最后的估计
由holder不等式,我们有
∫ γ ∇ u ~ Γ ⋅ E ∇ γ v ~ ≤ ∥ ∇ γ v ~ ∥ L 2 ( γ ) ∥ ∇ γ u ~ Γ ∥ L 2 ( γ ) ∥ E ∥ L ∞ ( γ ) \int_{\gamma} \nabla \widetilde{u}_{\Gamma} \cdot \mathbf{E} \nabla_{\gamma} \widetilde{v} \leq\left\|\nabla_{\gamma} \widetilde{v}\right\|_{L_{2}(\gamma)}\left\|\nabla_{\gamma} \widetilde{u}_{\Gamma}\right\|_{L_{2}(\gamma)}\|\mathbf{E}\|_{L_{\infty}(\gamma)} ∫γ∇u Γ⋅E∇γv ≤∥∇γv ∥L2(γ)∥∇γu Γ∥L2(γ)∥E∥L∞(γ)
使用模等价定理,我们可以得到
∫ Γ ( f q q Γ − f Γ ) v = ∫ Γ ( f q q Γ − f Γ ) ( v − v ‾ ) ≤ ∥ f q q Γ − 1 − f Γ ∥ H # − 1 ( Γ ) ∥ ∇ Γ v ∥ L 2 ( Γ ) \int_{\Gamma}\left(f \frac{q}{q_{\Gamma}}-f_{\Gamma}\right) v=\int_{\Gamma}\left(f \frac{q}{q_{\Gamma}}-f_{\Gamma}\right)(v-\overline{v}) \leq\left\|f q q_{\Gamma}^{-1}-f_{\Gamma}\right\|_{H_{\#}^{-1}(\Gamma)}\left\|\nabla_{\Gamma} v\right\|_{L_{2}(\Gamma)} ∫Γ(fqΓq−fΓ)v=∫Γ(fqΓq−fΓ)(v−v)≤∥∥fqqΓ−1−fΓ∥∥H#−1(Γ)∥∇Γv∥L2(Γ)
最后利用引理17即可。