1 Introduction
Information on flow in deformable porous media has become of increasing importance in various fields of natural sciences and technology. It offers an abundance of technical, geophysical, environmental and biomedical applications including modern material science polymers and metal foams, gaining significance particularly in lightweight design and aircraft industry, design of batteries or hydrogen fuel cells for green technologies, geothermal energy exploration or reservoir engineering as well as mechanism in the human body and food technology. Consequently, quantitative methods, based on numerical simulations, are desirable in analyzing experimental data and designing theories based on mathematical concepts. Recently, the quasistatic Biot system (cf., e.g., [12, 14]) has attracted researcher’s interest and has been studied as a proper model for the numerical simulation of flow in deformable porous media. The design, analysis and optimization of approximation techniques that are based on an iterative coupling of the subproblems of fluid flow and mechanical deformation were focused strongly. Iterative coupling offers the appreciable advantage over the fully coupled method that existing and highly developed discretizations and algebraic solver technologies can be reused. For the quasistatic Biot system, pioneering work is done in [10, 12]. Further research is presented in, e.g., [2, 4, 7, 8, 9, 13].
In the case of larger contrast coefficients that stand for the ratio between the intrinsical characteristic time and the characteristic domain time scale the fully dynamic hyperbolicparabolic system of poroelasticity has to be considered. In [11], this system (referred to as the Biot–Allard equations) is derived by asymptotic homogenization in the space and time variables. Here, to fix our ideas and carve out the key technique of proof, a simplified form of the system proposed in [11] is studied. However, its mixed hyperbolicparabolic structure is preserved. Our modification of the fully dynamic poroelasticity model in [11] comes through a simplication of the solution’s convolution with the dynamic permeability that is defined as the spatial average of pore system Stokes solutions on the unit cell. The fully dynamic system of poroelasticity to be analyzed here is given by (cf. [14, p. 313])
(1.1a)  
(1.1b)  
(1.1c) 
System (1.1) is equiped with appropriate initial and boundary conditions. In (1.1), the variable is the unkown effective solid phase displacement and is the unkown effective pressure. The quantity
denotes the symmetrized gradient or strain tensor. Further,
is the effective mass density, is Gassmann’s fourth order effective elasticity tensor, is Biot’s pressurestorage coupling tensor and is the specific storage coefficient. In the three field formulation (1.1), the vector field
is Darcy’s velocity and is the permeability tensor. All tensors are assumed to be symmetric, bounded and uniformly positive definite, the constants and are positive. By we denote the Frobenius inner product of and . The functions on the righthand side of (1.1) are supposed to be elements in dual spaces and, therefore, can include body forces and surface data (boundary conditions).So far, the numerical simulation of the system (1.1) has been studied rarely in the literature despite its numerous applications in practice. This might be due to the mixed hyberbolicparabolic character of the system and severe complexities involved in the construction of monolithic solver or iterative coupling schemes with guaranteed stability properties. Spacetime finite element approximations of hyperbolic and parabolic problems and the quasistatic Biot system were recently proposed, analyzed and investigated numerically by the authors in [1, 2, 3]. Here, we propose an iterative coupling scheme for the system (1.1) and prove its convergence. This is done in Banach spaces for the semidiscretization in time of (1.1). An abstract setting is used for the time discretization such that the family of diagonally implicit Runge–Kutta methods becomes applicable. The key ingredient of our proof of convergence is the observation that we can recast the semidiscrete approximation of (1.1) as the minimizer of an energy functional in the displacement and Darcy velocity fields. To solve the minimization problem, the general and abstract framework of alternating minimization (cf. [5, 6]) is applied. The resulting subproblems of this minimization are then reformulated as our final iterative coupling scheme. Thereby, the proof of convergence of the iterative scheme is traced back to the convergence of the alternating minimization approach. This shows that the latter provides an abstract and powerful tool of optimization for the design of iterative coupling schemes.
We use standard notation. In particular, we denote by the standard inner product of and by the norm of .
2 Variational formulation of a semidiscrete approximation of the system of dynamic poroelasticity
Firstly, we discretize the continuous system of dynamic poroelasticity (1.1) in time by using arbitrary (diagonally implicit) Runge–Kutta methods and formulate the semidiscrete approximation as solution to a minimization problem, following the approach in [5]. For this, we consider an equidistant partition of the time interval of interest with time step size . In the sequel, we use the following function spaces for displacement, pressure, and flux, respectively,
Further, let , , and denote the corresponding natural test spaces, and , , and their dual spaces.
Applying any diagonally implicit Runge–Kutta method for the temporal discretization of (1.1), eventually involves solving systems of the following structure.
Problem 2.1
In the th time step, find the displacement , pressure , and flux , satisfying for all the equations
(2.1a)  
(2.1b)  
(2.1c) 
In (2.1), the quantities are discretization parameters, and the righthand side functions , , include information on external volume and surface terms, as well as previous time steps depending on the choice of the implicit Runge–Kutta discretization.
Assuming positive compressibility, i.e., for the specific storage coefficient, the semidiscrete approximation satisfies equivalently the following variational problem; cf. [5] for the derivation of a similar equivalence in the framework of the quasistatic Biot system.
Problem 2.2
Find , satisfying
(2.2) 
where the energy at time is defined by ()
(2.3)  
The semidiscrete pressure may then be recovered by the postprocessing step
(2.4) 
3 Iterative coupling for the system of dynamic poroelasticity
Following the philosophy of [5], we propose an iterative coupling of the semidiscrete equations (2.1) of dynamic poroelasticity by firstly applying the fundamental alternating minimization to the variational formulation (2.2); cf. Alg. 1.
Secondly, the resulting scheme is equivalently reformated in terms of a stabilized splitting scheme applied to the threefield formulation (2.1). For this, a pressure iterate , , is introduced, consistent with (2.4), and the optimality conditions corresponding to the two steps of Alg. 1 are reformulated. The calculations are skipped here. We immediately present the resulting scheme, which in the end is closely related to the undrained split for the quasistatic Biot system [10].
Problem 3.1
Let be given and .
1. Step (Update of mechanical deformation): For given , find satisfying for all ,
(3.1)  
where denotes the standard tensor product.
2. Step (Update of Darcy velocity and pressure): For given find satisfying for all ,
(3.2a)  
(3.2b) 
4 Convergence of the iterative coupling scheme
The identification of the undrained split approach (3.1), (3.2) as the application of the alternating minimization, cf. Alg. 1, to the variational problem (2.2) yields the basis for a simple convergence analysis. For this, we utilize the following abstract convergence result, that is rewritten here in terms of the specific formulation of Alg. 1.
Theorem 4.1 (Convergence of the alternating minimization [6])
Let , , and denote seminorms on , , and , respectively. Let satisfy the inequalities
for all . Furthermore, assume that the energy functional of (2.3) satisfies the following conditions:

The energy is Frechét differentiable with denoting its derivative.

The energy is strongly convex wrt. with modulus , i.e., for all and it holds that

The partial functional derivatives and are uniformly Lipschitz continuous wrt. and with Lipschitz constants and , respectively, i.e., for all and it holds that
A simple application of Theorem 4.1 now yields the main result of the work, namely the global linear convergence of the undrained split (3.1), (3.2).
Corollary 4.2 (Linear convergence of the undrained split)
Proof.
We first examine convexity and smoothness properties of defined in (2.3) by analyzing the second functional derivative of . For this, let and be arbitrary. Then, for the second functional derivative of it holds that
(4.2) 
Next, we define a norm on by considering the partial second functional derivative of with respect to the displacement field,
Similarly, we define a norm on by means of
It directly follows that is strongly convex wrt. with modulus , and the partial functional derivatives and are uniformly Lipschitz continuous wrt. and with Lipschitz constants and , respectively.
By the Hölder inequality we deduce that
(4.3)  
Hence, it follows that
On the other hand, applying the triangle inequality and Young’s inequality, and balancing the arising constants properly yields that
Together with (4.3), we also conclude that
Thereby, the assumptions of Theorem 4.1 are fulfilled and (4.1) is ensured with constants and . Finally, the assertion follows directly, since is quadratic and is a local minimum of and relates to the second functional derivative of via (4.2). Therefore, we have that for all .
Remark 4.3 (Convergence of )
The convergence of the sequence of pressures follows now immediately by a standard infsup argument.
Remark 4.4 (Comparison with quasistatic case)
The final convergence rate in Corollary 4.2 coincides with the one for the undrained split applied to the quasistatic Biot equations for an homogeneous and isotropic bulk; cf. [12]. In that case, the Biot tensor reduces to for some constant , and is defined by the Lamé parameters, such that , where is the drained bulk modulus.
Acknowledgement
This work was supported by the German Academic Exchange Service (DAAD) under the grant ID 57458510 and by the Research Council of Norway (RCN) under the grant ID 294716.
References
 [1] M. Bause, U. Köcher, F. A. Radu, F. Schieweck, Postprocessed Galerkin approximation of improved order for wave equations, Math. Comp., 2019, DOI: 10.1090/mcom/3464, pp. 1–33; arXiv:1803.03005.
 [2] M. Bause, F. A. Radu, U. Köcher, Spacetime finite element approximation of the Biot poroelasticity system with iterative coupling, Comput. Methods Appl. Mech. Engrg., 320 (2017), pp. 745–768.
 [3] M. Bause, F. A. Radu, U. Köcher, Error analysis for discretizations of parabolic problems using continuous finite elements in time and mixed finite elements in space, Numer. Math., 137 (2017), pp. 773–818.
 [4] J. W. Both, M. Borregales, J. M. Nordbotten, K. Kundan, F. A. Radu, Robust fixed stress splitting for Biot’s equations in heterogeneous media, Appl. Math. Lett., 68, pp. 101–108.
 [5] J. W. Both, K. Kundan, J. M. Nordbotten, F. A. Radu, The gradient flow structures of thermoporoviscoelasticity, arXiv:1907.03134.
 [6] J. W. Both, On the rate of convergence of alternating minimization for nonsmooth nonstrongly convex optimization in Banach spaces, arXiv:1911.00404.
 [7] N. Castelletto, J. A. White, H. A. Tchelepi, Accuracy and convergence properties of the fixedstress iterative solution of twoway coupled poromechanics, Int. J. Num. Anal. Meth. Geomechanics, 39 (2015), pp. 1593–1618.
 [8] N. Castelletto, J. A. White, M. Ferronato, Scalable algorithms for threefield mixed finite element coupled poromechanics, J. Comp. Phys., 327 (2016), pp. 894–918.
 [9] Q. Hong, J. Kraus, M. Lymbery, F. Philo, Conservative discretizations and parameterrobust preconditioners for Biot and multiplenetwork fluxbased poroelasticity models, Numer. Linear Algebra Appl., 26 (2019), e2242, pp. 1–25.
 [10] J. Kim, H. A. Tchelepi, R. Juanes, Stability and convergence of sequential methods for coupled flow and geomechanics: Drained and undrained splits, Comput. Methods Appl. Mech. Engrg., 200 (2011), pp. 2094–2116.
 [11] A. Mikelić, M. F. Wheeler, Theory of the dynamic Biot–Allard equations and their link to the quasistatic Biot system, J. Math. Phys., 53 (2012), pp. 123702:1–15.
 [12] A. Mikelić, M. F. Wheeler, Convergence of iterative coupling for coupled flow and geomechanics, Comput. Geosci., 17 (2013), pp. 479–496.
 [13] A. Mikelić, B. Wang, M. F. Wheeler, Numerical convergence study of iterative coupling for coupled flow and geomechanics, Comput. Geosci., 18 (2014), pp. 325–341.
 [14] R. Showalter, Diffusion in poroelastic media, J. Math. Anal. Appl., 251 (2000), pp. 310–340.
Comments
There are no comments yet.