1. Introduction
Fictitious domain methods have a long history, dating back to the pioneering work of Peskin [22] and are currently enjoying great popularity, having been successfully applied to a variety of problems. Several variants include such methods as the ghostcell finite difference method [24], cut–cell volume method [21], immersed interface [17], ghost fluid [6], shifted boundary methods [2, 13, 20], –FEM [9], and CutFEM [3, 4, 5, 18, 11], among others. For a comprehensive overview of this research area, the interested reader is referred to the review paper [19]. Considerable impetus has been provided in the contexts of fluid–structure interaction and reduced order modeling for parametrically–dependent domains [12, 14].
Such cases pose severe challenges in the discretization and even result to simulations of diminished quality. For instance, the generation of a suitable conforming mesh is a challenging and computationally intensive task. As a means to bypass such complications, it is instructive to consider the actual computational domain of interest as being embedded in an unfitted background mesh. More precisely, this can be achieved via a geometric parameterization of its boundary via level–set geometries, using a fixed Cartesian background and its associated mesh for each new domain configuration. This approach avoids the need to remesh, as well as the need to develop a reference domain formulation. In such cases, immersed and embedded methods compare favorably to fitted mesh FEMs, providing simple and efficient schemes for the numerical approximation of PDEs in both cases of static and evolving geometries.
The overall objective of this note is to extend the a–priori analysis of cutFEM beyond the realm of linear problems. To this end, we propose an unfitted framework for the numerical solution of a semilinear elliptic boundary value problem with a polynomial nonlinearity. We start by introducing the model problem and the necessary notation in Section 2. Then, Section 3 focuses on the derivation of the a–priori error estimates and a numerical experiment is reported in Section 4, verifying the theoretical convergence rates and showcasing the accuracy of the method. The paper concludes with a brief discussion of our contributions and suggestions for future work in Section 5.
2. The model problem and preliminaries
As a model problem, we consider a semilinear elliptic boundary value problem of the form
(2.1)  
where is a simply connected open domain with boundary . The nonlinearity is assumed to be of polynomial type . Such equations have been studied previously in the context of problems with critical exponents [8] and are referred to in the theory of boundary layers of viscous fluids [23] as Emden–Fowler equations. It is straightforward to verify that the weak formulation
(2.2) 
of (2.1) admits a weak solution . Following a standard energy argument and assuming the force , the a–priori error bound
readily follows, indicating a continuous dependence of the solution on the data.
Implementation of an unfitted FEM for the discretization of (2.2) requires a fixed background domain which contains ; let its corresponding shape–regular mesh. The active mesh
is the minimal submesh of which covers and is in general unfitted to its boundary . As usual, the subscript indicates the global mesh size. The finite element space for discrete solutions will in fact be built upon the extended domain which corresponds to . Fictitious domain methods require boundary conditions at to be weakly satisfied through a variant of Nitsche’s method. On the other hand, coercivity over the whole computational domain is ensured by means of additional ghost penalty terms which act on the gradient jumps in the boundary zone; see, for instance, [5]. Therefore, a more detailed analysis of the interface grid is required; the submesh consisting of all cut elements is denoted
and the relevant set of faces upon which ghost penalty will be applied is given by
Considering the finite element space
for approximate solutions, we define discrete counterparts to the continuous bilinear and linear forms in (2.2), setting
(2.3)  
(2.4) 
for . Here,
denotes the outward pointing unit normal vector on the boundary
. The cutFEM discretization scheme reads as follows: find a discrete state , such that(2.5) 
for all , where the stabilization term
(2.6) 
acts on the gradient jumps of over element faces in the interface zone and is included in the bilinear form to extend its coercivity from the physical domain to . The quantities and in (2.3) and (2.6) are positive penalty parameters; see Lemma 3.2 below.
3. Norms, approximation properties and a–priori analysis
The convergence analysis of the method (2.5) is based on the following mesh–dependent norms:
The trace inequality for and implies in particular:
A necessary approximation result is stated next:
Lemma 3.1 ([5], Lemma 5).
Let a linear –extension operator on , such that , , and
the Clémenttype extended interpolation operator defined by
, where is the standard Clément interpolant. Then, the estimate(3.1) 
holds for every .
Regarding stability, the coercivity and continuity properties of the augmented bilinear form now read as follows:
Lemma 3.2 ([5], Lemmata 6 and 7).
Defining the method (2.5) with sufficiently large parameter and , then
(3.2) 
for every , and
(3.3) 
independently of and of the way in which the boundary intersects the background mesh.
Hence, due to the gradient penalty in the boundary zone, control of the –norm of the gradient may be extended over the whole active mesh .
We next quantify how the additional term affects the Galerkin orthogonality and consistency of the variational formulation (2.5).
Lemma 3.3 (Galerkin orthogonality).
Proof.
The following preliminary result investigating optimality with respect to interpolation is a key ingredient of our approach.
Proposition 3.4.
Proof.
Adapting for our purposes the procedure in the proof of [7, Thm. 5.3.3, p. 319] for the –Laplacian, a first observation is that there exists , such that
(3.6) 
Then, denoting , we successively apply the coercivity estimate (3.2), (3.6) and the Galerkin orthogonality (3.4) to estimate
A bound for the leading two terms is readily implied by the continuity estimate (3.3), the Cauchy–Schwarz inequality and (3.1):
while the third term is estimated by
Hence, the assertion (3.5) already follows for . ∎
Under some additional regularity requirements for the solution , we are now in a position to derive error estimates for our finite element approximations:
Theorem 3.5 (Optimal convergence).
Proof.
We first decompose the total error into its discrete–error and projection–error components; i.e.,
The desired estimate for the first term is already provided by (3.1). Hence, it suffices to prove the assertion for the latter, which is in turn bounded by Proposition 3.4. Indeed, by (3.1) and the properties of the Clément interpolant [10, p.69], estimate (3.5) yields
for . Recalling is the conjugate index of , clearly and the bound is optimal. ∎
4. Numerical validation
In order to verify the validity of the apriori error estimate in Theorem 3.5
, numerical simulations have been implemented in a python environment, using the open–source Netgen/NGSolvengsxfem finite element software. We consider a two–dimensional test case of (
2.1) for with manufactured exact solution and right–hand side force defined byin ; i.e., the unit disc centered at the origin. As in Section 2, the original domain is immersed in the background domain . To investigate orders of convergence, we consider a sequence of successively refined tessellations for with mesh parameters (). The stabilization constants , in (2.3) and (2.6) are taken to be equal to and respectively. By the theoretical error estimate stated in Theorem 3.5, we should expect first–order convergence rates with respect to the –norm and additionally second order for the –norm.
EOC  EOC  

7.74620e2  2.47468e3  
3.90601e2  0.988  5.83351e4  2.085  
1.93383e2  1.014  1.33451e4  2.128  
9.63082e3  1.006  3.34143e5  1.999  
4.80627e3  1.003  8.12293e6  2.040  
2.40450e3  0.999  2.01406e6  2.012  
Mean  1.002  2.049 
As illustrated in Table 1, the numerical findings validate the theoretically predicted rates of convergence and verify the effectiveness of the proposed framework.
5. Conclusions
The present note concentrated on the derivation of an a–priori error estimate for a cut finite element approximation of a semilinear model problem. To the authors’ best knowledge, this is one of the few instances in the literature that such an analysis has been carried out beyond a linear context. Our approach is based on classical arguments for the –Laplacian [7] and on key results from [5] for a stabilized unfitted method for the Poisson problem. Future work will delve more deeply in the analysis of unfitted FEMs for general time–dependent problems with nonlinearities. From a computational point of view, the effect of preconditioning on the performance of the method will be assessed in the spirit of [1, 16]. Finally, the method seems promising for controlling nonlinear PDEs with uncertainties, involving large deformations and/or topological changes [12, 14, 15].
Acknowledgments
This project has received funding from the Hellenic Foundation for Research and Innovation (HFRI) and the General Secretariat for Research and Technology (GSRT), under grant agreement No[1115] (PI: E. Karatzas), and the support of the National Infrastructures for Research and Technology S.A. (GRNET S.A.) in the National HPC facility  ARIS  under project ID pa190902.
References
 [1] Aik. Aretaki, E.N. Karatzas, Random geometries, preconditioned optimal control PDE problems discretized by a FEM with cut elements and a Quasi Monte Carlo method simulation, submitted for publication. arXiv preprint:2003.00352, 2020.
 [2] N.M. Atallah, C. Canuto, G. Scovazzi, Analysis of the shifted boundary method for the Stokes problem, Computer Methods in Appl. Mech. Engrg. 358 (2020),112609.

[3]
E. Burman, S. Claus, P. Hansbo, M.G. Larson, A. Massing, CutFEM: discretizing geometry and partial differential equations,
Int. J. Numer. Meth. Engrg. 104 (2014), 472501.  [4] E. Burman, P. Hansbo, Fictitious domain finite element methods using cut elements: I. A stabilized Lagrange multiplier method, Comput. Methods Appl. Mech. Engrg. 199 (4144) (2010), 26802686.
 [5] E. Burman, P. Hansbo, Fictitious domain finite element methods using cut elements II. A stabilized Nitsche method, Appl. Num. Math. 2(4) (2012), 328341.
 [6] W. Bo, J.W. Grove, A volume of fluid method based ghost fluid method for compressible multifluid flows, Computers & Fluids 90 (2014), 113122.
 [7] P.G. Ciarlet, The Finite Element Method for Elliptic Problems. North–Holland Publishing Co., 7th edition, 1978.
 [8] P. Clément, D. Guedes de Figueiredo, E. Mitidieri, Quasilinear elliptic equations with critical exponents, Topological Methods in Nonlinear Analysis 7(1) (1996), 133170.
 [9] M. Duprez, A. Lozinski, –FEM: a finite element method on domains defined by level–sets, arXiv preprint arXiv: 1901.03966v3, 2019.
 [10] PA. Ern, J.–L. Guermond, Theory and Practice of Finite Elements. Applied Mathematical Sciences, Vol. 159, Springer Verlag, 2004.
 [11] A. Hansbo, P. Hansbo, Nitsche’s method for interface problems in computational mechanics, Comput. Methods Appl. Mech. Eng. 191 (2002), 55375552.
 [12] E.N. Karatzas, F. Ballarin, G. Rozza, Projection–based reduced order models for a cut finite element method in parametrized domains, Computers & Mathematics with Applications 79(3) (2020), 833851.
 [13] E.N. Karatzas, G. Stabile, N. Attalah, G. Scovazzi, G. Rozza, A reduced order approach for the embedded shifted boundary FEM and a heat exchange system on parametrized geometries, In: Fehr J., Haasdonk B. (eds) IUTAM Symposium on Model Order Reduction of Coupled Systems, Stuttgart, Germany, May 22–25, 2018. IUTAM Bookseries, Vol. 36. Springer, Cham (2020).
 [14] E.N. Karatzas, G. Stabile, L. Nouveau, G. Scovazzi, G. Rozza, A reduced basis approach for PDEs on parametrized geometries based on the shifted boundary finite element method and application to a Stokes flow, Comput. Methods Appl. Mech. Engrg. 347 (2019), 568587.
 [15] E.N. Karatzas, G. Stabile, L. Nouveau, G. Scovazzi, G. Rozza, A reduced–order shifted boundary finite element method for parametrized incompressible Navier–Stokes equations, submitted for publication, arXiv preprint: 1907.10549, 2019.
 [16] G. Katsouleas, E.N. Karatzas, Aik. Aretaki, An unfitted discontinuous Galerkin method for the Stokes system, in preparation, 2020.
 [17] E.M. Kolahdouz, A.P.S. Bhalla, B.A. Craven, B.E. Griffith, An Immersed Interface Method for Faceted Surfaces, J. Comput. Physics, 400 (2020), 1008854.
 [18] A. Lozinski, CutFEM without cutting the mesh cells: a new way to impose Dirichlet and Neumann boundary conditions on unfitted meshes, arXiv preprint: 1901.03966, 2019.
 [19] R. Mittal, and G. Iaccarino, Immersed boundary methods, Annual Review of Fluid Mechanics 37 (1) (2005), 239261.
 [20] A. Main, G. Scovazzi, The shifted boundary method for embedded domain computations. Part I: Poisson and Stokes problems, Journal of Computational Physics 372 (2018), 972995.
 [21] V. Pasquariello, G. Hammerl, F. ’́Orley, S. Hickel, C. Danowski, A. Popp, W. A. Wall, N. A. Adams, A cutcell finite volume – finite element coupling approach for fluid–structure interaction in compressible flow, Journal of Computational Physics 307 (2016), 670695.
 [22] C. S. Peskin, Flow patterns around heart valves: A numerical method, Journal of Computational Physics 10 (1972), 252271.
 [23] J. Wong, On the generalized Emden–Fowler equation, SIAM Review 17 (2) (1975), 339360.
 [24] C. H. Wu, O. M. Faltinsen, B. F. Chen, TimeIndependent Finite Difference and Ghost Cell Method to Study Sloshing Liquid in 2D and 3D Tanks with Internal Structures, Communications in Computational Physics 13 (3) (2013), 780800.
Comments
There are no comments yet.