
Our aim in this paper is to define more concepts that are related to primal topological space. We introduce new operators called γ-diamond and γ∗-diamond and explore their main characterizations. We provide results and examples regarding to these operators. Using these new operators, we create a weaker version of the original topology. Additionally, we present some results related to compatibility.
Citation: Ohud Alghamdi, Ahmad Al-Omari, Mesfer H. Alqahtani. Novel operators in the frame of primal topological spaces[J]. AIMS Mathematics, 2024, 9(9): 25792-25808. doi: 10.3934/math.20241260
[1] | Guohui Zhang, Jinghe Sun, Xing Liu, Guodong Wang, Yangyang Yang . Solving flexible job shop scheduling problems with transportation time based on improved genetic algorithm. Mathematical Biosciences and Engineering, 2019, 16(3): 1334-1347. doi: 10.3934/mbe.2019065 |
[2] | Ruiping Yuan, Jiangtao Dou, Juntao Li, Wei Wang, Yingfan Jiang . Multi-robot task allocation in e-commerce RMFS based on deep reinforcement learning. Mathematical Biosciences and Engineering, 2023, 20(2): 1903-1918. doi: 10.3934/mbe.2023087 |
[3] | Shixuan Yao, Xiaochen Liu, Yinghui Zhang, Ze Cui . An approach to solving optimal control problems of nonlinear systems by introducing detail-reward mechanism in deep reinforcement learning. Mathematical Biosciences and Engineering, 2022, 19(9): 9258-9290. doi: 10.3934/mbe.2022430 |
[4] | Kongfu Hu, Lei Wang, Jingcao Cai, Long Cheng . An improved genetic algorithm with dynamic neighborhood search for job shop scheduling problem. Mathematical Biosciences and Engineering, 2023, 20(9): 17407-17427. doi: 10.3934/mbe.2023774 |
[5] | Jianguo Duan, Mengting Wang, Qinglei Zhang, Jiyun Qin . Distributed shop scheduling: A comprehensive review on classifications, models and algorithms. Mathematical Biosciences and Engineering, 2023, 20(8): 15265-15308. doi: 10.3934/mbe.2023683 |
[6] | Zilong Zhuang, Zhiyao Lu, Zizhao Huang, Chengliang Liu, Wei Qin . A novel complex network based dynamic rule selection approach for open shop scheduling problem with release dates. Mathematical Biosciences and Engineering, 2019, 16(5): 4491-4505. doi: 10.3934/mbe.2019224 |
[7] | Shaofeng Yan, Guohui Zhang, Jinghe Sun, Wenqiang Zhang . An improved ant colony optimization for solving the flexible job shop scheduling problem with multiple time constraints. Mathematical Biosciences and Engineering, 2023, 20(4): 7519-7547. doi: 10.3934/mbe.2023325 |
[8] | Zichen Wang, Xin Wang . Fault-tolerant control for nonlinear systems with a dead zone: Reinforcement learning approach. Mathematical Biosciences and Engineering, 2023, 20(4): 6334-6357. doi: 10.3934/mbe.2023274 |
[9] | Jin Zhang, Nan Ma, Zhixuan Wu, Cheng Wang, Yongqiang Yao . Intelligent control of self-driving vehicles based on adaptive sampling supervised actor-critic and human driving experience. Mathematical Biosciences and Engineering, 2024, 21(5): 6077-6096. doi: 10.3934/mbe.2024267 |
[10] | Lu-Wen Liao . A branch and bound algorithm for optimal television commercial scheduling. Mathematical Biosciences and Engineering, 2022, 19(5): 4933-4945. doi: 10.3934/mbe.2022231 |
Our aim in this paper is to define more concepts that are related to primal topological space. We introduce new operators called γ-diamond and γ∗-diamond and explore their main characterizations. We provide results and examples regarding to these operators. Using these new operators, we create a weaker version of the original topology. Additionally, we present some results related to compatibility.
In this paper, we consider the following diffusion equation on
{−∇⋅(α∇u)=f,inΩ,u=0,on∂Ω. | (1) |
To approximate (1), taking advantage of the adaptive mesh refinement (AMR) to save valuable computational resources, the adaptive finite element method on quadtree mesh is among the most popular ones in the engineering and scientific computing community [20]. Compared with simplicial meshes, quadtree meshes provide preferable performance in the aspects of the accuracy and robustness. There are lots of mature software packages (e.g., [1,2]) on quadtree meshes. To guide the AMR, one possible way is through the a posteriori error estimation to construct computable quantities to indicate the location that the mesh needs to be refined/coarsened, thus to balance the spacial distribution of the error which improves the accuracy per computing power. Residual-based and recovery-based error estimators are among the most popular ones used. In terms of accuracy, the recovery-based error estimator shows more appealing attributes [28,3].
More recently, newer developments on flux recovery have been studied by many researchers on constructing a post-processed flux in a structure-preserving approximation space. Using (1) as an example, given that the data
However, these
More recently, a new class of methods called the virtual element methods (VEM) were introduced in [4,8], which can be viewed as a polytopal generalization of the tensorial/simplicial finite element. Since then, lots of applications of VEM have been studied by many researchers. A usual VEM workflow splits the consistency (approximation) and the stability of the method as well as the finite dimensional approximation space into two parts. It allows flexible constructions of spaces to preserve the structure of the continuous problems such as higher order continuities, exact divergence-free spaces, and many others. The VEM functions are represented by merely the degrees of freedom (DoF) functionals, not the pointwise values. In computation, if an optimal order discontinuous approximation can be computed elementwisely, then adding an appropriate parameter-free stabilization suffices to guarantee the convergence under common assumptions on the geometry of the mesh.
The adoption of the polytopal element brings many distinctive advantages, for example, treating rectangular element with hanging nodes as polygons allows a simple construction of
The major ingredient in our study is an
If
(α∇uT,∇vT)=(f,vT),∀vT∈Qk(T)∩H10(Ω), | (2) |
in which the standard notation is opted.
Qk(T):={v∈H1(Ω):v|K∈Qk(K),∀K∈T}. |
and on
Qk(K):=Pk,k(K)={p(x)q(y),p∈Pk([a,b]),q∈Pk([c,d])}, |
where
On
NH:={z∈N:∃K∈T,z∈∂K∖NK} | (3) |
Otherwise the node
For each edge
\{v\}^{\gamma}_e : = \gamma v^- + (1-\gamma) v^+. |
In this subsection, the quadtree mesh
For the embedded element
Subsequently,
On
\begin{equation} \begin{aligned} \mathcal{V}_{k}(K) : = \Big\{ & \mathit{\boldsymbol{\tau}}\in \mathit{\boldsymbol{H}}(\mathrm{div};K)\cap \mathit{\boldsymbol{H}}(\mathbf{rot};K): \\ & \nabla\cdot \mathit{\boldsymbol{\tau}}\in \mathbb{P}_{k-1}(K), \quad \nabla \times \mathit{\boldsymbol{\tau}} = 0, \\ & \mathit{\boldsymbol{\tau}}\cdot\mathit{\boldsymbol{n}}_e \in \mathbb{P}_{k}(e), \;\forall e\subset \partial K \Big\}. \end{aligned} \end{equation} | (4) |
An
\begin{equation} \mathcal{V}_k : = \bigl\{\mathit{\boldsymbol{\tau}}\in \mathit{\boldsymbol{H}}(\mathrm{div}): \mathit{\boldsymbol{\tau}}|_K \in \mathcal{V}_{k}(K), \;\;{\rm{ on }}\; K\in \mathcal{T}_{\mathrm{poly}} \bigr\}. \end{equation} | (5) |
Next we turn to define the degrees of freedom (DoFs) of this space. To this end, we define the set of scaled monomials
\begin{equation} \mathbb{P}_{k}(e): = \operatorname{span}\left\{1, \frac{s-m_{e}}{h_{e}},\left(\frac{s-m_{e}}{h_{e}}\right)^{2}, \ldots,\left(\frac{s-m_{e}}{h_{e}}\right)^{k}\right\}, \end{equation} | (6) |
where
\begin{equation} \mathbb{P}_{k}({K}): = \operatorname{span}\left\{m_{\alpha}(\mathit{\boldsymbol{x}}): = \left(\frac{\mathit{\boldsymbol{x}}-\mathit{\boldsymbol{x}}_{K}}{h_{K}}\right)^{\mathit{\boldsymbol{\alpha}}}, \quad|\mathit{\boldsymbol{\alpha}}| \leq k\right\}. \end{equation} | (7) |
The degrees of freedom (DoFs) are then set as follows for a
\begin{equation} \begin{aligned} (\mathfrak{e})\; k\geq 1 & \quad \int_e (\mathit{\boldsymbol{\tau}}\cdot \mathit{\boldsymbol{n}}_e) m \,\mathrm{d} s, \quad \forall m \in \mathbb{P}_{k}(e), & \;{\rm{on }}\; \; e\subset \mathcal{E}_{\mathrm{poly}}. \\ (\mathfrak{i})\; k\geq 2 & \quad \int_K \mathit{\boldsymbol{\tau}}\cdot \nabla m\, \mathrm{d} \mathit{\boldsymbol{x}} , \quad \forall m\in \mathbb{P}_{k-1}({K})/\mathbb{R} & \;{\rm{on }}\; \; K\in \mathcal{T}_{\mathrm{poly}}. \end{aligned} \end{equation} | (8) |
Remark 1. We note that in our construction, the degrees of freedom to determine the curl of a VEM function originally in [8] are replaced by a curl-free constraint thanks to the flexibility to virtual element. The reason why we opt for this subspace is that the true flux
As the data
Consider
On each
\begin{equation} \left\{-\alpha \nabla u_{\mathcal{T}} \right\}^{\gamma_e}_e \cdot\mathit{\boldsymbol{n}}_e : = \Big(\gamma_e \left( -\alpha_{K_-} \nabla u_{\mathcal{T}}|_{K_-} \right) + (1-\gamma_e) \left( -\alpha_{K_+} \nabla u_{\mathcal{T}}|_{K_+} \right)\Big)\cdot \mathit{\boldsymbol{n}}_e, \end{equation} | (9) |
where
\begin{equation} \gamma_e : = \frac{\alpha_{K_+}^{1/2}}{\alpha_{K_+}^{1/2} + \alpha_{K_-}^{1/2}}. \end{equation} | (10) |
First for both
\begin{equation} \mathit{\boldsymbol{\sigma}}_{\mathcal{T}}\cdot \mathit{\boldsymbol{n}}_e = \left\{-\alpha \nabla u_{\mathcal{T}} \right\}^{\gamma_e}_e \cdot\mathit{\boldsymbol{n}}_e. \end{equation} | (11) |
In the lowest order case
\begin{equation} |K|\nabla\cdot \mathit{\boldsymbol{\sigma}}_{\mathcal{T}} = \int_K \nabla\cdot\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} \mathrm{d} \mathit{\boldsymbol{x}} = \int_{\partial K} \mathit{\boldsymbol{\sigma}}_{\mathcal{T}} \cdot\mathit{\boldsymbol{n}}_{\partial K}\mathrm{d} s = \sum\limits_{e\subset \partial K} \int_e \mathit{\boldsymbol{\sigma}}_{\mathcal{T}} \cdot\mathit{\boldsymbol{n}}_{\partial K}|_e \mathrm{d} s. \end{equation} | (12) |
If
\begin{equation} \nabla \cdot \mathit{\boldsymbol{\sigma}}_{\mathcal{T}} = \Pi_{k-1} f + c_K. \end{equation} | (13) |
The reason to add
\begin{equation} c_K = \frac{1}{ |K| }\left(-\int_K \Pi_{k-1} f \mathrm{d} \mathit{\boldsymbol{x}} + \sum\limits_{e\subset\partial K} \int_e \left\{-\alpha \nabla u_{\mathcal{T}} \right\}^{\gamma_e}_e \cdot\mathit{\boldsymbol{n}}_{\partial K}|_e \mathrm{d} s\right), \end{equation} | (14) |
Consequently for
\begin{equation} \bigl(\mathit{\boldsymbol{\sigma}}_{\mathcal{T}},\nabla q\bigr)_K = -\left(\Pi_{k-1} f + c_K, q\right)_K + \sum\limits_{e\subset \partial K} \left(\left\{-\alpha \nabla u_{\mathcal{T}} \right\}^{\gamma_e}_e \cdot\mathit{\boldsymbol{n}}_{\partial K}|_e,q\right)_{e}. \end{equation} | (15) |
To the end of constructing a computable local error indicator, inspired by the VEM formulation [8], the recovered flux is projected to a space with a much simpler structure. A local oblique projection
\begin{equation} \bigl({\Pi} \mathit{\boldsymbol{\tau}},\nabla p\bigr)_K = \bigl(\mathit{\boldsymbol{\tau}},\nabla p\bigr)_K,\quad \forall p\in \mathbb{P}_k(K)/\mathbb{R}. \end{equation} | (16) |
Next we are gonna show that this projection operator can be straightforward computed for vector fields in
When
\begin{equation} \bigl(\mathit{\boldsymbol{\tau}},\nabla p\bigr)_K = -\bigl(\nabla \cdot \mathit{\boldsymbol{\tau}}, p\bigr)_K + \bigl(\mathit{\boldsymbol{\tau}}\cdot\mathit{\boldsymbol{n}},p\bigr)_{\partial K}. \end{equation} | (17) |
By definition of the space (4) when
When
\begin{equation} \begin{aligned} \bigl(\mathit{\boldsymbol{\tau}},\nabla p\bigr)_K & = -\bigl(\nabla \cdot \mathit{\boldsymbol{\tau}}, \Pi_{k-1} p\bigr)_K + \bigl(\mathit{\boldsymbol{\tau}}\cdot\mathit{\boldsymbol{n}},p\bigr)_{\partial K} \\ & = \bigl(\mathit{\boldsymbol{\tau}}, \nabla \Pi_{k-1} p\bigr)_K + \bigl(\mathit{\boldsymbol{\tau}}\cdot\mathit{\boldsymbol{n}},p - \Pi_{k-1} p\bigr)_{\partial K}, \end{aligned} \end{equation} | (18) |
which can be evaluated using both DoF sets
Given the recovered flux \(\mathit{\boldsymbol{\sigma}}_{\mathcal{T}}\) in Section 3, the recovery-based local error indicator
\begin{equation} \begin{gathered} \eta_{\mathrm{flux},K} : = \big\Vert{\alpha^{-1/2}(\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} + \alpha \nabla u_{\mathcal{T}})} \big\Vert_K, \\ \;{\rm{ and }}\; \; \eta_{\mathrm{res},K} : = \big\Vert{\alpha^{-1/2}(f - \nabla\cdot\mathit{\boldsymbol{\sigma}}_{\mathcal{T}}) } \big\Vert_K, \end{gathered} \end{equation} | (19) |
then
\begin{equation} \eta_K = \left\{ \begin{array}{lc} \eta_{\mathrm{flux},K} & \;{\rm{when }}\; k = 1, \\ \left(\eta_{\mathrm{flux},K}^2 + \eta_{\mathrm{res},K}^2 \right)^{1/2} & \;{\rm{when }}\; k\geq 2. \end{array} \right. \end{equation} | (20) |
A computable
\begin{equation} \widehat{\eta}_{\mathrm{flux},K}: = \big\Vert{\alpha_K^{-1/2}{\Pi}(\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} + \alpha_K \nabla u_{\mathcal{T}})} \big\Vert_K, \end{equation} | (21) |
with the oblique projection
\begin{equation} \widehat{\eta}_{\mathrm{stab},K}: = \big\vert{\alpha_K^{-1/2}(\operatorname{I}-{\Pi})(\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} + \alpha_K \nabla u_{\mathcal{T}})} \big\vert_{S,K}. \end{equation} | (22) |
Here
\begin{equation} S_K(\mathit{\boldsymbol{v}}, \mathit{\boldsymbol{w}}): = \sum\limits_{e\subset \partial K} h_e\big( \mathit{\boldsymbol{v}}\cdot\mathit{\boldsymbol{n}}_e, \mathit{\boldsymbol{w}}\cdot \mathit{\boldsymbol{n}}_e \big)_e + \sum\limits_{\alpha\in \Lambda} (\mathit{\boldsymbol{v}},\nabla m_{\alpha})_K (\mathit{\boldsymbol{w}},\nabla m_{\alpha})_K, \end{equation} | (23) |
where
The computable error estimator
\begin{equation} \widehat{\eta}^2 = \begin{cases} \sum\limits_{K\in \mathcal{T}} \left(\widehat{\eta}_{\mathrm{flux},K}^2 + \widehat{\eta}_{\mathrm{stab},K}^2 \right) = : \sum\limits_{K\in \mathcal{T}} \widehat{\eta}_{K}^2 & \;{\rm{when }}\; k = 1, \\[5pt] \sum\limits_{K\in \mathcal{T}} \left(\widehat{\eta}_{\mathrm{flux},K}^2 + \widehat{\eta}_{\mathrm{stab},K}^2 + \eta_{\mathrm{res},K}^2\right) = : \sum\limits_{K\in \mathcal{T}} \widehat{\eta}_{K}^2 & \;{\rm{when }}\; k\geq 2. \end{cases} \end{equation} | (24) |
In this section, we shall prove the proposed recovery-based estimator
Theorem 4.1. Let
\begin{equation} \widehat{\eta}_{\mathrm{flux},K}^2 \lesssim \mathrm{osc}(f; K)^2 + \eta_{\mathrm{elem},K}^2 + \eta_{\mathrm{edge},K}^2 , \end{equation} | (25) |
where
\begin{align*} \mathrm{osc}(f;K) & = \alpha_K^{-1/2}h_K \big\Vert f- \Pi_{k-1} f \big\Vert_K, \\ \eta_{\mathrm{elem},K} &: = \alpha_K^{-1/2}h_K \big\Vert f + \nabla\cdot(\alpha \nabla u_{\mathcal{T}}) \big\Vert_K, \\ \mathit{\;{\rm{and}}\;}\; \eta_{\mathrm{edge},K} &: = \left(\sum\limits_{e\subset \partial K} \frac{h_e}{\alpha_K + \alpha_{K_e}} \big\Vert \lbrack\lbrack {{\alpha \nabla u_{\mathcal{T}}\cdot\mathit{\boldsymbol{n}}_e}} \rbrack\rbrack_{{ {} }} \big\Vert_e^2\right)^{1/2}. \end{align*} |
In the edge jump term,
Proof. Let
\begin{equation} \begin{aligned} \widehat{\eta}_{\mathrm{flux},K}^2 & = \bigl({\Pi}(\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} + \alpha_K \nabla u_{\mathcal{T}}), \nabla p \bigr)_K = \bigl(\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} + \alpha_K \nabla u_{\mathcal{T}}, \nabla p \bigr)_K \\ & = -\bigl(\nabla \cdot (\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} +\alpha_K \nabla u_{\mathcal{T}}), p \bigr)_K + \sum\limits_{e\subset \partial K}\int_e \big( \mathit{\boldsymbol{\sigma}}_{\mathcal{T}} + \alpha_K \nabla u_{\mathcal{T}}\big)\cdot \mathit{\boldsymbol{n}}_{\partial K}\big|_{e} \, p \, \mathrm{d} s. \end{aligned} \end{equation} | (26) |
By (11), without loss of generality we assume
\begin{equation} \begin{aligned} \big( \mathit{\boldsymbol{\sigma}}_{\mathcal{T}} + \alpha_K \nabla u_{\mathcal{T}}\big)\cdot \mathit{\boldsymbol{n}}_e & = \Big( (1-\gamma_e) \alpha_{K} \nabla u_{\mathcal{T}}|_K - (1-\gamma_e)\alpha_{K_e} \nabla u_{\mathcal{T}}|_{K_e} \Big)\cdot \mathit{\boldsymbol{n}}_e \\ & = \frac{\alpha_{K}^{1/2}}{\alpha_{K}^{1/2} + \alpha_{K_e}^{1/2}}\lbrack\lbrack {{\alpha \nabla u_{\mathcal{T}}\cdot \mathit{\boldsymbol{n}}_e}} \rbrack\rbrack_{{ {e} }}. \end{aligned} \end{equation} | (27) |
The boundary term in (26) can be then rewritten as
\begin{equation} \begin{aligned} & \int_e \big( \mathit{\boldsymbol{\sigma}}_{\mathcal{T}} + \alpha_K \nabla u_{\mathcal{T}}\big)\cdot \mathit{\boldsymbol{n}}_e \,p\, \mathrm{d} s \\ = &\;\int_e \frac{1}{\alpha_{K}^{1/2} + \alpha_{K_e}^{1/2}}\lbrack\lbrack {{\alpha \nabla u_{\mathcal{T}}\cdot \mathit{\boldsymbol{n}}_e}} \rbrack\rbrack_{{ {e} }} \,\alpha_{K}^{1/2}p\, \mathrm{d} s \\ \lesssim & \; \frac{1}{(\alpha_{K}+ \alpha_{K_e})^{1/2}} h_e^{1/2} \big\Vert \lbrack\lbrack {{\alpha \nabla u_{\mathcal{T}}\cdot\mathit{\boldsymbol{n}}_e}} \rbrack\rbrack_{{ {} }} \big\Vert_e \alpha_{K}^{1/2} h_e^{-1/2} \left\Vert{p}\right\Vert_e. \end{aligned} \end{equation} | (28) |
By a trace inequality on an edge of a polygon (Lemma 7.1), and the Poincaré inequality for
h_e^{-1/2}\|p\|_e \lesssim h_K^{-1} \|p\|_K + \|\nabla p\|_K \lesssim \|\nabla p\|_K. |
As a result,
\sum\limits_{e\subset \partial K}\int_e \big( \mathit{\boldsymbol{\sigma}}_{\mathcal{T}} + \alpha_K \nabla u_{\mathcal{T}}\big)\cdot \mathit{\boldsymbol{n}}_e \,p\, \mathrm{d} s \lesssim \eta_{\mathrm{edge},K}\, \alpha_{K}^{1/2} \left\Vert{\nabla p}\right\Vert_e = \eta_{\mathrm{edge},K} \,\widehat{\eta}_{\mathrm{flux},K}. |
For the bulk term on
\begin{aligned} &-\bigl(\nabla \cdot (\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} +\alpha_K \nabla u_{\mathcal{T}}), p \bigr)_K \leq \left|\nabla \cdot (\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} +\alpha_K \nabla u_{\mathcal{T}}) \right| |K|^{1/2} \left\Vert{p}\right\Vert_K \\ \leq & \; \frac{1}{ |K|^{1/2}}\left|\int_K \nabla \cdot (\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} +\alpha_K \nabla u_{\mathcal{T}}) \, \mathrm{d} \mathit{\boldsymbol{x}}\right| \left\Vert{p}\right\Vert_K \\ = & \; \frac{1}{ |K|^{1/2}} \left|\sum\limits_{e\subset \partial K} \int_{e} (\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} +\alpha_K \nabla u_{\mathcal{T}})\cdot \mathit{\boldsymbol{n}}_e \, \mathrm{d} s\right| \left\Vert{p}\right\Vert_K \\ \leq & \; \left(\sum\limits_{e\subset \partial K} \frac{1}{\alpha_{K}^{1/2} + \alpha_{K_e}^{1/2}} \left\Vert{\lbrack\lbrack {{\alpha \nabla u_{\mathcal{T}}\cdot \mathit{\boldsymbol{n}}_e}} \rbrack\rbrack_{{ {} }}}\right\Vert_e \,\alpha_{K}^{1/2}h_e \right) \left\Vert{\nabla p}\right\Vert \\ \lesssim &\; \eta_{\mathrm{edge},K} \, \widehat{\eta}_{\mathrm{flux},K}. \end{aligned} |
When
\begin{equation} \begin{aligned} & -\bigl(\nabla \cdot (\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} +\alpha_K \nabla u_{\mathcal{T}}), p \bigr)_K = -\bigl(\Pi_{k-1} f + c_K + \nabla\cdot(\alpha_K \nabla u_{\mathcal{T}}), p \bigr)_K \\ \leq & \; \left( \big\Vert f- \Pi_{k-1} f \big\Vert_K + \big\Vert f + \nabla\cdot(\alpha \nabla u_{\mathcal{T}}) \big\Vert_K + |c_K| |K|^{1/2}\right)\left\Vert{p}\right\Vert_K. \end{aligned} \end{equation} | (29) |
The first two terms can be handled by combining the weights
\begin{equation} \begin{aligned} c_K |K|^{1/2} & = \frac{1}{ |K|^{1/2} }\Big(-\int_K (\Pi_{k-1} f -f) \mathrm{d} \mathit{\boldsymbol{x}} - \int_K \big(f + \nabla\cdot (\alpha \nabla u_{\mathcal{T}})\big) \mathrm{d} \mathit{\boldsymbol{x}} \\ &\quad + \int_K \nabla\cdot (\alpha \nabla u_{\mathcal{T}}) \mathrm{d} \mathit{\boldsymbol{x}} + \sum\limits_{e\subset\partial K} \int_e \left\{-\alpha \nabla u_{\mathcal{T}} \right\}^{\gamma_e}_e \cdot\mathit{\boldsymbol{n}}_e \mathrm{d} s\Big) \\ &\leq \big\Vert f- \Pi_{k-1} f \big\Vert_K + \big\Vert f + \nabla\cdot(\alpha \nabla u_{\mathcal{T}}) \big\Vert_K \\ & \quad + \frac{1}{ |K|^{1/2}}\sum\limits_{e\subset\partial K} \int_e (\alpha_K \nabla u_{\mathcal{T}} - \left\{\alpha \nabla u_{\mathcal{T}} \right\}^{\gamma_e}_e) \cdot\mathit{\boldsymbol{n}}_e \mathrm{d} s \\ & \leq \big\Vert f- \Pi_{k-1} f \big\Vert_K + \big\Vert f + \nabla\cdot(\alpha \nabla u_{\mathcal{T}}) \big\Vert_K \\ & \quad + \sum\limits_{e\subset\partial K} \frac{\alpha_{K}^{1/2}}{\alpha_{K}^{1/2} + \alpha_{K_e}^{1/2}} \left\Vert{\lbrack\lbrack {{\alpha \nabla u_{\mathcal{T}}\cdot \mathit{\boldsymbol{n}}_e}} \rbrack\rbrack_{{ {} }}}\right\Vert_e . \end{aligned} \end{equation} | (30) |
The two terms on
-\bigl(\nabla \cdot (\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} +\alpha_K \nabla u_{\mathcal{T}}), p \bigr)_K \lesssim \Big( \mathrm{osc}(f; K) + \eta_{\mathrm{elem},K} + \eta_{\mathrm{edge},K} \Big) \alpha_K^{1/2}\left\Vert{\nabla p}\right\Vert |
and the theorem follows.
Theorem 4.2. Under the same setting with Theorem 4.1, let
\begin{equation} \widehat{\eta}_{\mathrm{stab},K}^2 \lesssim \mathrm{osc}(f; K)^2 + \eta_{\mathrm{elem},K}^2 + \eta_{\mathrm{edge},K}^2 , \end{equation} | (31) |
The constant depends on
Proof. This theorem follows directly from the norm equivalence Lemma 7.3:
\big\vert{\alpha_K^{-1/2}(\operatorname{I}-{\Pi})(\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} + \alpha_K \nabla u_{\mathcal{T}})} \big\vert_{S,K} \lesssim \big\vert{\alpha_K^{-1/2}(\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} + \alpha_K \nabla u_{\mathcal{T}})} \big\vert_{S,K}, |
while evaluating the DoFs
Theorem 4.3. Under the same setting with Theorem 4.1, on any
\begin{equation} \widehat{\eta}_{K} \lesssim \mathrm{osc}(f;K) + \big\Vert \alpha^{1/2}\nabla (u-u_{\mathcal{T}})\big\Vert_{\omega_K}, \end{equation} | (32) |
with a constant independent of
Proof. This is a direct consequence of Theorem 4.1 and 4.2 and the fact that the residual-based error indicator is efficient by a common bubble function argument.
In this section, we shall prove that the computable error estimator
Assumption 1 (
By Assumption 1, we denote the father
Assumption 2 (Quasi-monotonicity of
Denote
\begin{equation} \pi_{z} v = \left\{\begin{array}{ll} \frac{\int_{\omega_{z} \cap \omega_{m(z)}} v \phi_z }{\int_{\omega_{z} \cap \omega_{m(z)}} \phi_z }& {\rm { if }}\; \mathit{\boldsymbol{z}} \in \Omega, \\ 0 & {\rm { if }}\; \mathit{\boldsymbol{z}} \in \partial \Omega. \end{array}\right. \end{equation} | (33) |
We note that if
\begin{equation} \mathcal{I} v : = \sum\limits_{z\in \mathcal{N}_1} (\pi_z v)\phi_z. \end{equation} | (34) |
Lemma 4.4 (Estimates for
\begin{equation} \alpha_K^{1/2} h_K^{-1} \left\Vert{v - \mathcal{I}v}\right\Vert_{K} + \alpha_K^{1/2} \left\Vert{\nabla \mathcal{I}v}\right\Vert_{K} \lesssim \big\Vert \alpha^{1/2} \nabla v\big\Vert_{\omega_K}, \end{equation} | (35) |
and for
\begin{equation} \sum\limits_{K\subset\omega_z} h_{z}^{-2} \|\alpha^{1/2}(v-\pi_{z} v)\phi_z\|_{K}^2 \lesssim \big\Vert \alpha^{1/2} \nabla v\big\Vert_{\omega_z}^2, \end{equation} | (36) |
in which
Proof. The estimate for
Denotes the subset of nodes
\begin{equation} \begin{aligned} \mathrm{osc}(f;\mathcal{T})^2 : = & \sum\limits_{z \in \mathcal{N}_1\cap( \mathcal{N}_{\partial \Omega} \cup \mathcal{N}_I)} h_{z}^{2} \big\|\alpha^{-1/2} f\big\|_{\omega_{z}}^2 \\ +& \sum\limits_{z \in \mathcal{N}_1 \backslash ( \mathcal{N}_{\partial \Omega} \cup \mathcal{N}_I)} h_{z}^{2} \big\|\alpha^{-1/2} (f - f_z)\big\|_{\omega_{z}}^2, \end{aligned} \end{equation} | (37) |
with
Theorem 4.5. Let
\begin{equation} \big\Vert{\alpha^{1/2}\nabla (u - u_{\mathcal{T}})}\big\Vert \lesssim \left(\widehat{\eta}^2 +\mathrm{osc}(f;\mathcal{T})^2 \right)^{1/2}. \end{equation} | (38) |
For
\begin{equation} \big\Vert{\alpha^{1/2}\nabla (u - u_{\mathcal{T}})}\big\Vert \lesssim \widehat{\eta}, \end{equation} | (39) |
where the constant depends on
Proof. Let
\begin{align*} &\big\Vert{\alpha^{1/2}\nabla \varepsilon}\big\Vert^2 = \big(\alpha\nabla (u - u_{\mathcal{T}}), \nabla(\varepsilon -\mathcal{I}\varepsilon)\big) \\ = & \big(\alpha \nabla u + \mathit{\boldsymbol{\sigma}}_{\mathcal{T}}, \nabla(\varepsilon -\mathcal{I}\varepsilon)\big) - \big(\alpha \nabla u_{\mathcal{T}} + \mathit{\boldsymbol{\sigma}}_{\mathcal{T}}, \nabla(\varepsilon -\mathcal{I}\varepsilon)\big) \\ = & \big(f -\nabla\cdot \mathit{\boldsymbol{\sigma}}_{\mathcal{T}}, \varepsilon -\mathcal{I}\varepsilon\big) - \big(\alpha \nabla u_{\mathcal{T}} + \mathit{\boldsymbol{\sigma}}_{\mathcal{T}}, \nabla(\varepsilon -\mathcal{I}\varepsilon)\big) \\ \leq& \left(\sum\limits_{K\in\mathcal{T}} \alpha_K^{-1}h_K^2\left\Vert{f - \nabla\cdot\mathit{\boldsymbol{\sigma}}_{\mathcal{T}}}\right\Vert_{K}^2 \right)^{1/2} \left(\sum\limits_{K\in\mathcal{T}} \alpha_K h_K^{-2}\left\Vert{\varepsilon - \mathcal{I}\varepsilon}\right\Vert_{K}^2 \right)^{1/2} \\ & \left(\sum\limits_{K\in\mathcal{T}} \alpha_K^{-1}\big\Vert{\alpha \nabla u_{\mathcal{T}} + \mathit{\boldsymbol{\sigma}}_{\mathcal{T}}} \big\Vert_{K}^2 \right)^{1/2} \left(\sum\limits_{K\in\mathcal{T}} \alpha_K \left\Vert{\nabla(\varepsilon - \mathcal{I}\varepsilon)}\right\Vert_{K}^2 \right)^{1/2}. \\ & \lesssim \left(\sum\limits_{K\in\mathcal{T}} (\eta_{\mathrm{res},K}^2+\eta_{\mathrm{flux},K}^2) \right)^{1/2} \left(\sum\limits_{K\in\mathcal{T}} \big\Vert{\alpha^{1/2}\nabla \varepsilon}\big\Vert_{\omega_K} \right)^{1/2}. \end{align*} |
Applying the norm equivalence of
When
\begin{equation} (f, \varepsilon-\mathcal{I}\varepsilon) = \sum\limits_{z \in \mathcal{N}_1} \sum\limits_{K \subset \omega_{z}} \big(f,\left(\varepsilon-\pi_{z} \varepsilon\right) \phi_{z}\big)_{K}, \end{equation} | (40) |
in which a patch-wise constant
\begin{align*} & \big(f -\nabla\cdot \mathit{\boldsymbol{\sigma}}_{\mathcal{T}}, \varepsilon - \mathcal{I}\varepsilon\big) = \big(f, \varepsilon - \mathcal{I}\varepsilon\big) - \big(\nabla\cdot (\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} + \alpha_K\nabla u_{\mathcal{T}}), \varepsilon - \mathcal{I}\varepsilon\big) \\ = & \sum\limits_{z \in \mathcal{N}} \sum\limits_{K \subset \omega_{z}} \big(f,\left(\varepsilon-\pi_{z} \varepsilon\right) \phi_{z}\big)_{K} - \big(\nabla\cdot (\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} + \alpha_K\nabla u_{\mathcal{T}}), \varepsilon - \mathcal{I}\varepsilon\big) \\ \leq & \left(\mathrm{osc}(f;\mathcal{T})^2 \right)^{1/2} \left(\sum\limits_{z \in \mathcal{N}_1} \sum\limits_{K\subset\omega_z} h_{z}^{-2} \|\alpha^{1/2}(\varepsilon-\pi_{z} \varepsilon )\phi_z\|_{K}^2\right)^{1/2} \\ & \; + \left(\sum\limits_{K\in\mathcal{T}} \alpha_K^{-1} h_K^{2} \big\Vert \nabla\cdot (\mathit{\boldsymbol{\sigma}}_{\mathcal{T}} + \alpha_K\nabla u_{\mathcal{T}}) \big\Vert_{K}^2 \right)^{1/2} \left(\sum\limits_{K\in\mathcal{T}} \alpha_K h_K^{-2}\left\Vert{\varepsilon - \mathcal{I}\varepsilon}\right\Vert_{K}^2 \right)^{1/2}. \end{align*} |
Applied an inverse inequality in Lemma 7.2 on
The numerics is prepared using the bilinear element for common AMR benchmark problems. The codes for this paper are publicly available on https://github.com/lyc102/ifem implemented using
The adaptive finite element (AFEM) iterative procedure is following the standard
\texttt{SOLVE}\longrightarrow \texttt{ESTIMATE} \longrightarrow \texttt{MARK} \longrightarrow \texttt{REFINE}. |
The linear system is solved by MATLAB
\sum\limits_{K \in \mathcal{M}} \widehat{\eta}^{2}_{K} \geq \theta \sum\limits_{K \in \mathcal{T}} \widehat{\eta}^{2}_{K}, \quad \rm { for } \theta \in(0,1). |
Throughout all examples, we fix
\eta_{\;{\rm{Residual}}\;,K}^2 : = \alpha_K^{-1}h_K^2 \big\Vert f + \nabla\cdot(\alpha \nabla u_{\mathcal{T}}) \big\Vert_K^2 + \frac{1}{2}\sum\limits_{e\subset \partial K} \frac{h_e}{\alpha_K + \alpha_{K_e}} \big\Vert \lbrack\lbrack {{\alpha \nabla u_{\mathcal{T}}\cdot\mathit{\boldsymbol{n}}_e}} \rbrack\rbrack_{{ {} }} \big\Vert_e^2, |
Let
\;{\rm{effectivity index}}\; : = {\eta}/{\big\Vert{\alpha^{1/2}\nabla \varepsilon }\big\Vert}, \quad \;{\rm{ where }}\;\; \varepsilon: = u - u_{\mathcal{T}}, \; \eta = \eta_{\;{\rm{Residual}}\;} \;{\rm{ or }}\; \widehat{\eta}, |
i.e., the closer to 1 the effectivity index is, the more accurate this estimator is to measure the error of interest. We use an order
\ln \eta_n \sim -r_{\eta} \ln N_n + c_1,\quad\;{\rm{and}}\;\quad \ln \Vert{\alpha^{1/2} \nabla (u - u_{\mathcal{T}}) } \Vert \sim -r_{\;{\rm{err}}\;} \ln N_n + c_2, |
where the subscript
In this example, a standard AMR benchmark on the L-shaped domain is tested. The true solution
The solution
This example is a common benchmark test problem introduced in [9], see also [17,12]) for elliptic interface problems. The true solution
\mu(\theta) = \left\{\begin{array}{ll} \cos ((\pi / 2-\delta) \gamma) \cdot \cos ((\theta-\pi / 2+\rho) \gamma) & {\rm { if }}\; 0 \leq \theta \leq \pi / 2 \\ \cos (\rho \gamma) \cdot \cos ((\theta-\pi+\delta) \gamma) & {\rm { if }}\; \pi / 2 \leq \theta \leq \pi \\ \cos (\delta \gamma) \cdot \cos ((\theta-\pi-\rho) \gamma) & {\rm { if }}\; \pi \leq \theta < 3 \pi / 2 \\ \cos ((\pi / 2-\rho) \gamma) \cdot \cos ((\theta-3 \pi / 2-\delta) \gamma) & {\rm { if }}\; 3 \pi / 2 \leq \theta \leq 2 \pi \end{array}\right. |
While
\gamma = 0.1,\;\; R \approx 161.4476387975881, \;\; \rho = \pi / 4,\;\; \delta \approx -14.92256510455152, |
By this choice, this function is very singular near the origin as the maximum regularity it has is
The AFEM procedure for this problem stops when the relative error reaches
A postprocessed flux with the minimum
However, we do acknowledge that the technical tool involving interpolation is essentially limited to
The author is grateful for the constructive advice from the anonymous reviewers. This work was supported in part by the National Science Foundation under grants DMS-1913080 and DMS-2136075, and no additional revenues are related to this work.
Unlike the identity matrix stabilization commonly used in most of the VEM literature, for
\begin{equation} (\!(\mathit{\boldsymbol{\sigma}}, {\mathit{\boldsymbol{\tau}}})\!)_{K} : = \big({\Pi} \mathit{\boldsymbol{\sigma}}, {\Pi} \mathit{\boldsymbol{\tau}} \big)_K + {S}_K\big(({\rm I}-{\Pi} )\mathit{\boldsymbol{\sigma}}, ({\rm I}-{\Pi} )\mathit{\boldsymbol{\tau}}\big), \end{equation} | (41) |
where
To show the inverse inequality and the norm equivalence used in the reliability bound, on each element, we need to introduce some geometric measures. Consider a polygonal element
Proposition 1. Under Assumption 1,
Lemma 7.1 (Trace inequality on small edges [13]). If Proposition 1 holds, for
\begin{equation} h_e^{-1/2}\left\Vert{v}\right\Vert_{e} \lesssim h_K^{-1} \left\Vert{v}\right\Vert_{K} + \left\Vert{\nabla v}\right\Vert_{K}, \quad \mathit{on} \;e\subset K. \end{equation} | (42) |
Proof. The proof follows essentially equation (3.9) in [13,Lemma 3.3] as a standard scaled trace inequality on
h_e^{-1/2}\left\Vert{v}\right\Vert_{e} \lesssim h_e^{-1} \left\Vert{v}\right\Vert_{T_e} + \left\Vert{\nabla v}\right\Vert_{T_e} \lesssim h_K^{-1} \left\Vert{v}\right\Vert_{K} + \left\Vert{\nabla v}\right\Vert_{K}. |
Lemma 7.2 (Inverse inequalities). Under Assumption 1, we have the following inverse estimates for
\begin{equation} \|\nabla \cdot \mathit{\boldsymbol{\tau}}\|_K \lesssim h_K^{-1} \|\mathit{\boldsymbol{\tau}}\|_K, \quad \mathit{and} \quad \|\nabla \cdot \mathit{\boldsymbol{\tau}}\|_K \lesssim h_K^{-1} S_K\big(\mathit{\boldsymbol{\tau}},\mathit{\boldsymbol{\tau}}\big)^{1/2}. \end{equation} | (43) |
Proof. The first inequality in (43) can be shown using a bubble function trick. Choose
\|\nabla \cdot \mathit{\boldsymbol{\tau}}\|_K^2 \lesssim (\nabla \cdot \mathit{\boldsymbol{\tau}}, p b_K) = -(\mathit{\boldsymbol{\tau}}, \nabla (p b_K)) \leq \left\Vert{\mathit{\boldsymbol{\tau}}}\right\Vert_K \left\Vert{ \nabla (p b_K)}\right\Vert_K, |
and then
\left\Vert{ \nabla (p b_K)}\right\Vert \leq \left\Vert{ b_K \nabla p }\right\Vert_K + \left\Vert{p\nabla b_K}\right\Vert_K \leq \left\Vert{ b_K }\right\Vert_{\infty,\Omega} \left\Vert{\nabla p }\right\Vert_K + \left\Vert{p}\right\Vert_K \left\Vert{\nabla b_K}\right\Vert_{\infty,K}. |
Consequently, the first inequality in (43) follows above by the standard inverse estimate for polynomials
To prove the second inequality in (43), by integration by parts we have
\begin{equation} \left\Vert{\nabla\cdot\mathit{\boldsymbol{\tau}}}\right\Vert^2 = (\nabla\cdot\mathit{\boldsymbol{\tau}}, p) = -(\mathit{\boldsymbol{\tau}},\nabla p) + \sum\limits_{e\subset\partial K} (\mathit{\boldsymbol{\tau}}\cdot \mathit{\boldsymbol{n}}_e, p). \end{equation} | (44) |
Expand
\begin{equation} \left\Vert{p}\right\Vert_K^2 = \mathbf{p}^{\top} \mathbf{M} \mathbf{p} \geq \mathbf{p}^{\top} \operatorname{diag}(\mathbf{M}) \mathbf{p} \geq \min\limits_j \mathbf{M}_{jj}\left\Vert{\mathbf{p}}\right\Vert_{\ell^2}^2 \simeq h_K^2 \left\Vert{\mathbf{p}}\right\Vert_{\ell^2}^2, \end{equation} | (45) |
since
\begin{aligned} \left\Vert{\nabla\cdot\mathit{\boldsymbol{\tau}}}\right\Vert^2 & \leq \left(\sum\limits_{\alpha\in \Lambda} (\mathit{\boldsymbol{\tau}}, m_{\alpha})_K^2 \right)^{1/2} \left(\sum\limits_{\alpha\in \Lambda} p_{\alpha}^2 \right)^{1/2} \\ & \quad + \left(\sum\limits_{e\subset \partial K} h_e \left\Vert{\mathit{\boldsymbol{\tau}}\cdot \mathit{\boldsymbol{n}}_e}\right\Vert_e^2 \right)^{1/2} \left(\sum\limits_{e\subset \partial K} h_e^{-1} \left\Vert{p}\right\Vert_e^2 \right)^{1/2} \\ & \lesssim S_K(\mathit{\boldsymbol{\tau}},\mathit{\boldsymbol{\tau}})^{1/2} \left(\left\Vert{\mathbf{p}}\right\Vert_{\ell^2} + h_K^{-1} \left\Vert{p}\right\Vert_K + \left\Vert{\nabla p}\right\Vert_K \right). \end{aligned} |
As a result, the second inequality in (43) is proved when apply an inverse inequality for
Remark 2. While the proof in Lemma 7.2 relies on
Lemma 7.3 (Norm equivalence). Under Assumption 1, let
\begin{equation} \gamma_* \Vert{\mathit{\boldsymbol{\tau}}}\Vert_K \leq \Vert{ \mathit{\boldsymbol{\tau}}}\Vert_{h,K} \leq \gamma^*\Vert{\mathit{\boldsymbol{\tau}}}\Vert_K, \end{equation} | (46) |
where both
Proof. First we consider the lower bound, by triangle inequality,
\Vert{\mathit{\boldsymbol{\tau}}}\Vert_{K}\leq \big\Vert{{\Pi}\mathit{\boldsymbol{\tau}}}\big\Vert_{K} + \big\Vert{(\mathit{\boldsymbol{\tau}} - {\Pi}\mathit{\boldsymbol{\tau}}) }\big\Vert_{K}. |
Since
\begin{equation} \Vert{\mathit{\boldsymbol{\tau}} }\Vert_{K}^2 \leq S_K\big(\mathit{\boldsymbol{\tau}},\mathit{\boldsymbol{\tau}}\big), \quad \;{\rm{ for }}\; \mathit{\boldsymbol{\tau}}\in \mathcal{V}_k(K). \end{equation} | (47) |
To this end, we consider the weak solution to the following auxiliary boundary value problem on
\begin{equation} \left\{ \begin{aligned} \Delta \psi & = \nabla\cdot \mathit{\boldsymbol{\tau}}&\;{\rm{ in }}\; K, \\ \frac{\partial \psi}{\partial n} & = \mathit{\boldsymbol{\tau}} \cdot\mathit{\boldsymbol{n}}_{\partial K} &\;{\rm{ on }}\;\partial K. \end{aligned} \right. \end{equation} | (48) |
By a standard Helmholtz decomposition result (e.g. Proposition 3.1, Chapter 1[23]), we have
\left\Vert{\mathit{\boldsymbol{\tau}} - \nabla \psi}\right\Vert_K^2 = (\mathit{\boldsymbol{\tau}} - \nabla \psi, \nabla^{\perp} \phi) = 0. |
Consequently, we proved essentially the unisolvency of the modified VEM space (4) and
\begin{equation} \begin{aligned} & \big\Vert{\mathit{\boldsymbol{\tau}} }\big\Vert_{K}^2 = (\mathit{\boldsymbol{\tau}}, \nabla \psi)_K = \big(\mathit{\boldsymbol{\tau}}, \nabla \psi \big)_K \\ = & \; -\big(\nabla\cdot\mathit{\boldsymbol{\tau}}, \psi \big)_K+ (\mathit{\boldsymbol{\tau}}\cdot\mathit{\boldsymbol{n}}_{\partial K} ,\psi )_{\partial K} \\ \leq & \;\|\nabla \cdot \mathit{\boldsymbol{\tau}}\|_K \| \psi\|_K + \sum\limits_{e\subset \partial K} \|\mathit{\boldsymbol{\tau}}\cdot\mathit{\boldsymbol{n}}_e\|_e\| \psi \|_e \\ \leq &\; \|\nabla \cdot \mathit{\boldsymbol{\tau}}\|_K \| \psi\|_K + \left(\sum\limits_{e\subset \partial K} h_e\|\mathit{\boldsymbol{\tau}}\cdot\mathit{\boldsymbol{n}}_e\|_e^2\right)^{1/2} \left(\sum\limits_{e\subset \partial K} h_e^{-1}\|\psi\|_e^2\right)^{1/2} \end{aligned} \end{equation} | (49) |
Proposition 1 allows us to apply an isotropic trace inequality on an edge of a polygon (Lemma 7.1), combining with the Poincaré inequality for
h_e^{-1/2}\|\psi\|_e \lesssim h_K^{-1} \|\psi\|_K + \|\nabla \psi\|_K \lesssim \|\nabla \psi\|_K. |
Furthermore applying the inverse estimate in Lemma 7.2 on the bulk term above, we have
\big\Vert{\mathit{\boldsymbol{\tau}} }\big\Vert_{K}^2 \lesssim S_K\big(\mathit{\boldsymbol{\tau}},\mathit{\boldsymbol{\tau}}\big)^{1/2} \|\nabla \psi\|_K, |
which proves the validity of (47), thus yield the lower bound.
To prove the upper bound, by
\begin{equation} h_e\|\mathit{\boldsymbol{\tau}}\cdot\mathit{\boldsymbol{n}}_e\|_e^2 \lesssim \left\Vert{\mathit{\boldsymbol{\tau}}}\right\Vert_K,\quad \;{\rm{ and }}\; \quad |(\mathit{\boldsymbol{\tau}}, \nabla m_{\alpha})_K| \leq \left\Vert{\mathit{\boldsymbol{\tau}}}\right\Vert_K. \end{equation} | (50) |
To prove the first inequality, by Proposition 1 again, consider the edge bubble function
\begin{equation} \left\Vert{\nabla b_e}\right\Vert_{\infty,K} = O(1/h_e), \;{\rm{ and }}\; \left\Vert{b_e}\right\Vert_{\infty, K} = O(1). \end{equation} | (51) |
Denote
\begin{aligned} \|\mathit{\boldsymbol{\tau}}\cdot\mathit{\boldsymbol{n}}_e\|_e^2 & \lesssim \big(\mathit{\boldsymbol{\tau}}\cdot\mathit{\boldsymbol{n}}_e, b_e q_e \big)_e = x\big(\mathit{\boldsymbol{\tau}}\cdot\mathit{\boldsymbol{n}}_e, b_e q_e \big)_{\partial K} \\ & = \big(\mathit{\boldsymbol{\tau}}, q_e\nabla b_e \big)_K + \big(\nabla\cdot\mathit{\boldsymbol{\tau}}, b_e q_e\big)_K \\ & \leq \left\Vert{\mathit{\boldsymbol{\tau}}}\right\Vert_K \left\Vert{q_e\nabla b_e}\right\Vert_{T_e} + \left\Vert{\nabla\cdot\mathit{\boldsymbol{\tau}}}\right\Vert_K \left\Vert{q_e b_e}\right\Vert_{T_e}, \\ & \leq \left\Vert{\mathit{\boldsymbol{\tau}}}\right\Vert_K \left\Vert{q_e}\right\Vert_{T_e} \left\Vert{\nabla b_e}\right\Vert_{\infty,K} + \left\Vert{\nabla\cdot\mathit{\boldsymbol{\tau}}}\right\Vert_K \left\Vert{q_e}\right\Vert_{T_e} \left\Vert{b_e}\right\Vert_{\infty,K}. \end{aligned} |
Now by the fact that
The second inequality in (50) can be estimated straightforward by the scaling of the monomials (7)
\begin{equation} \left|(\mathit{\boldsymbol{\tau}}, \nabla m_{\alpha})_K\right| \leq \left\Vert{\mathit{\boldsymbol{\tau}}}\right\Vert_K \left\Vert{\nabla m_{\alpha}}\right\Vert_K \leq \left\Vert{\mathit{\boldsymbol{\tau}}}\right\Vert_K . \end{equation} | (52) |
Hence, (46) is proved.
[1] | G. Choquet, Sur les notions de filter et grille, C. R. Acad. Sci., 224 (1947), 171–173. |
[2] | K. Kuratowski, Topologie I, Warszawa: Academic Press Inc., 1966. |
[3] |
A. Al-Omari, T. Noiri, On \Psi_{\mathcal{G}}-sets in grill topological spaces, Filomat, 25 (2011), 187–196. https://doi.org/10.2298/FIL1102187A doi: 10.2298/FIL1102187A
![]() |
[4] | K. K. Azad, Fuzzy grills and a characterization of fuzzy proximity, J. Math. Anal. Appl., 79 (1981), 13–17. |
[5] | R. Vaidyanathaswamy, The Localisation theory in set topology, P. Ind. AS-Math. Sci., 20 (1944), 51–61. |
[6] | R. Vaidyanathaswamy, Set Theory, 2 Eds., New York: Chelsea Publishing Company, 1960. |
[7] | D. Janković, T. R. Hamlett, New Topologies from old via Ideals, Am. Math. Mon., 97 (1990), 295–310. |
[8] |
D. Sarkar, Fuzzy ideal theory fuzzy local function and generated fuzzy topology, Fuzzy Sets Sys., 87 (1977), 117–123. https://doi.org/10.1016/S0165-0114(96)00032-2 doi: 10.1016/S0165-0114(96)00032-2
![]() |
[9] |
A. Kandil, O. A. E. Tantawy, S. A. El-Sheikh, A. M. Abd El-latif, Soft ideal theory soft local function and generated soft topological spaces, Appl. Math. Inf. Sci., 8 (2014), 1595–1603. http://dx.doi.org/10.12785/amis/080413 doi: 10.12785/amis/080413
![]() |
[10] |
Z. Ameen, S. Al Ghour, Cluster soft sets and cluster soft topologies, Comput. Appl. Math., 42 (2023), 337. https://doi.org/10.1007/s40314-023-02476-7 doi: 10.1007/s40314-023-02476-7
![]() |
[11] | S. Acharjee, M. Özkoç, F. Y. Issaka, Primal topological spaces, preprint paper, 2022. https://doi.org/10.48550/arXiv.2209.12676 |
[12] |
T. M. Al-shami, Z. A. Ameen, R. Abu-Gdairi, A. Mhemdi, On primal soft topology, Mathematics, 11 (2023), 2329. https://doi.org/10.3390/math11102329 doi: 10.3390/math11102329
![]() |
[13] |
A. Al-Omari, M. H. Alqahtani, Some operators in soft primal spaces, AIMS Math., 9 (2024), 10756–10774. http://dx.doi.org/10.3934/math.2024525 doi: 10.3934/math.2024525
![]() |
[14] | Z. A. Ameen, R. A. Mohammed, T. M. Al-shami, B. A. Asaad, Novel fuzzy topologies formed by fuzzy primal frameworks, J. Intell. Fuzzy Sys., pre-press, 2024, 1–10. https://doi.org/10.3233/JIFS-238408 |
[15] | A. Al-Omari, M. Özkoç, S. Acharjee, Primal-proximity spaces, preprint paper, 2023. https://doi.org/10.48550/arXiv.2306.07977 |
[16] | A. Al-Omari, S. Acharjee, M. Özkoç, A new operator of primal topological spaces, Mathematica, 65 (2023), 175–183. |
[17] |
A. Al-Omari, M. H. Alqahtani, Primal structure with closure operators and their applications, Mathematics, 11 (2023), 4946. https://doi.org/10.3390/math11244946 doi: 10.3390/math11244946
![]() |
[18] |
A. Al-Omari, O. Alghamdi, Regularity and normality on primal spaces, AIMS Math., 9 (2024), 7662–7672. https://doi.org/10.3934/math.2024372 doi: 10.3934/math.2024372
![]() |
[19] |
H. Al-Saadi, H. Al-Malki, Categories of open sets in generalized primal topological spaces, Mathematics, 12 (2024), 207. https://doi.org/10.3390/math12020207 doi: 10.3390/math12020207
![]() |
[20] |
H. Al-Saadi, H. Al-Malki, Generalized primal topological spaces, AIMS Math., 8 (2023), 24162–24175. http://dx.doi.org/10.3934/math.20231232 doi: 10.3934/math.20231232
![]() |
[21] | N. V. Veličko, H-closed topological spaces, Am. Math. Soc. Transl., 78 (1968), 103–118. |
[22] | T. Noiri, On \alpha-continuous functions, Casopis Pest. Mat., 109 (1984), 118–126. |
1. | Hengliang Tang, Jinda Dong, Solving Flexible Job-Shop Scheduling Problem with Heterogeneous Graph Neural Network Based on Relation and Deep Reinforcement Learning, 2024, 12, 2075-1702, 584, 10.3390/machines12080584 | |
2. | Chen Han, Xuanyin Wang, TPN:Triple network algorithm for deep reinforcement learning, 2024, 591, 09252312, 127755, 10.1016/j.neucom.2024.127755 | |
3. | Miguel S. E. Martins, João M. C. Sousa, Susana Vieira, A Systematic Review on Reinforcement Learning for Industrial Combinatorial Optimization Problems, 2025, 15, 2076-3417, 1211, 10.3390/app15031211 | |
4. | Tianhua Jiang, Lu Liu, A Bi-Population Competition Adaptive Interior Search Algorithm Based on Reinforcement Learning for Flexible Job Shop Scheduling Problem, 2025, 24, 1469-0268, 10.1142/S1469026824500251 | |
5. | Tianyuan Mao, A Review of Scheduling Methods for Multi-AGV Material Handling Systems in Mixed-Model Assembly Workshops, 2025, 5, 2710-0723, 227, 10.54691/p4x5a536 | |
6. | Peng Zhao, You Zhou, Di Wang, Zhiguang Cao, Yubin Xiao, Xuan Wu, Yuanshu Li, Hongjia Liu, Wei Du, Yuan Jiang, Liupu Wang, 2025, Dual Operation Aggregation Graph Neural Networks for Solving Flexible Job-Shop Scheduling Problem with Reinforcement Learning, 9798400712746, 4089, 10.1145/3696410.3714616 | |
7. | Yuxin Peng, Youlong Lyu, Jie Zhang, Ying Chu, Heterogeneous Graph Neural-Network-Based Scheduling Optimization for Multi-Product and Variable-Batch Production in Flexible Job Shops, 2025, 15, 2076-3417, 5648, 10.3390/app15105648 |