WorksMathematics
High-order in-cell discontinuous reconstruction path-conservative methods for conservative systems in primitive variables
Abstract
We present a path-conservative methodology for the numerical approximation of conservative systems formulated exclusively in primitive variables, the most physically relevant set of variables. The approach is designed to preserve the correct weak solution while avoiding the spurious oscillations near contact discontinuities that may arise in conservative formulations. Nonconservative products are treated through the selection of a suitable family of paths combined with an in-cell discontinuous reconstruction strategy, which is locally activated near shocks. To choose the path we introduce a compatibility condition between the Rankine-Hugoniot conditions of the non-conservative form, with the exact ones, assumed to be known. In smooth regions and across contact discontinuities, a standard path-conservative scheme is employed. The method is further extended to achieve high-order accuracy. Numerical results for the Euler multimaterial and shear interface model demonstrates correct shock-capturing property, suppression of spurious oscillations due to contact discontinuities, and convergence without using hybrid formulations.
Key words. In-cell reconstruction; Path-conservative methods; Primitive variables; Nonconservative hyperbolic systems; Multimaterial Euler equations.
MSC codes. 65M08, 76T17, 35L67
Introduction: conservative and non-conservative forms
Let us consider a PDE system of conservation laws of the form:
where is the conservative vector of unknowns that takes values in the convex set of and is the physical flux that goes from to . In this work, we focus on a general family of systems in which the components of the vector are mass densities, momenta, and energies, plus additional variables characterizing the local composition of the flow. This type of structure is encountered very often both in multi-phase or multi-material flows, as well as in many shallow water models.
The main motivation of this work is to design numerical methods able to solve certain fluid models using the so-called primitive variables instead of the conserved ones without altering shock waves. In some cases, the use of primitive variables is mandatory due to the spurious oscillations that appear close to contact discontinuities when solving the problem in conserved variables (see, for example, [16]).
If the variables and are related by a smooth one-to-one transformation , the system in the new variables is written as follows:
where
having denoted with , , the Jacobians of , , and respectively. Therefore the system of conservation laws (1.1) is equivalent to the nonconservative system (1.2) for smooth solutions.
The objective of this work is to solve (1.1) by applying an adequate numerical method to (1.2) so that no oscillations occur at the contact discontinuities and the shock waves are captured correctly. To do this, path-conservative numerical methods combined with the so-called in-cell discontinuous reconstruction technique (see [7, 14, 15]) will be used.
Previous works are based on a hybrid approach (see, for instance, [11, 2, 10]), in which primitive variables are used to handle smooth regions (rarefactions) and contact discontinuities, while conservative variables are employed when shocks appear in the solution. The main advantage of the present work is that the use of conservative variables is no longer required.
A first difficulty to overcome stems from the definition of weak solution of (1.2). Various mathematical frameworks provide ways to define the nonconservative product for discontinuous functions . Here, the approach by Dal Maso, LeFloch, and Murat [8] is applied, introducing a weak solution concept that satisfies (1.2) in the sense of Borel measures. This approach relies on a family of Lipschitz continuous paths, , which must satisfy certain regularity and compatibility conditions:
Let be a function with bounded total variation. According to the definition of the nonconservative product as a Borel measure, the total measure of an interval , which will be denoted by
is given by
where in the integral on the right-hand side represents the derivative at almost every point, and and represent the values to the left and right of the th jump discontinuity. Observe that the family of paths is used to determine the weights of the Dirac masses that accounts for the derivatives of at the jumps.
If nonconservative products are defined in this way, the admissible weak solutions of System (1.2) have to satisfy the generalized Rankine-Hugoniot condition:
where is the speed of propagation of the discontinuity, and and are the values on either side. The notion of weak solutions depends thus on the choice of family of paths. Therefore, this family has to be adequately chosen so that the jump condition (1.3) is equivalent to the Rankine-Hugoniot condition for (1.1):
Although the general question of whether there exists a family of paths that ensures the equivalence of (1.3) and (1.4) is complex and involves the study of viscous profiles (see Appendix A), in some cases, as we shall see in § 3, such a family of paths can be calculated algebraically.
After prescribing the family of paths, defining a concept of entropy solution becomes necessary: the notion of entropy induced by the one of the conservative system (that can be given by an entropy pair or by Lax’s condition, for instance) will be assumed here, i.e. a weak solution of (1.2) will be considered an entropy solution if is an entropy solution for (1.1).
Once the family of paths and the notion of entropy have been prescribed so that the weak solutions of (1.1) and (1.2) are the same, defining numerical methods for the latter that converge to the correct weak solutions is not a simple task. The framework of path-conservative methods introduced in [13] makes easy the design of numerical methods that are consistent with the definition of weak-solutions but, if the numerical dissipation is not controlled, the limits of the numerical solutions can be weak solutions satisfying a different jump condition: see [5], [3]. The in-cell reconstruction technique will be used here to control the numerical dissipation so that isolated shock waves are correctly captured and (1.4) is satisfied: see [7, 14, 15].
The paper is organized as follows. In Section 2 we present the general methodology developed in this work. We first introduce a path-conservative semi-discrete formulation of the method, detailing the incorporation of the in-cell reconstruction operator, to finally obtain the fully discrete method. In § 3 we focus on the Euler equations with an interface and shear, providing all the necessary ingredients to apply the proposed method. § 4 demonstrates the numerical performance of the method, comparing it with existing approaches to highlight the improvements achieved. Finally, § 5 includes some final conclusions and remarks.
High-order in-cell discontinuous reconstruction path-conservative methods for conservative systems in primitive variable
Following [13], high-order path-conservative numerical schemes in semi-discrete form:
will be used to solve (1.2), where:
and are the space and the time steps respectively. They are assumed to be constant in the description of the numerical method for simplicity.
are the computational cells with length .
,
is the approximation to the cell average of the solution at the -th cell in time ;
is a space-time reconstruction operator, which may be discontinuous, to be defined in Subsection 2.1. As it will be seen, it is designed so that the method is high-order accurate in smooth regions while suppressing viscosity near shocks, thereby ensuring the correct jump conditions are recovered;
, where and are two Lipschitz continuous functions from to that satisfy
and
for every set . See [6] for details.
Once the numerical approximations at time have been computed, the ODE system (2.1) has to be solved with an adequate time stepping to update the numerical solution.
Definition of reconstruction operators
The first step is to identify the cells where a shock wave is present or may be generated. To do this, the Riemann problem with initial data
is used as a detector: if its solution involves at least one shock wave, the cell is marked. If the only wave detected is a shock, it will be said that the shock wave is isolated. Let us denote by the set of indices corresponding to the marked cells, namely
The second step is to select a space-time reconstruction operator that, given the cell averages at the cells of a smooth function at time , provides smooth approximations of at the cells
where is the stencil, i.e. the set of indices of the cells whose averages are used to compute . In the case of first order methods, the piecewise constant reconstruction operator will be considered. For higher order methods, Taylor expansions and the Cauchy-Kowaleskaya procedure will be followed to compute these reconstruction operators as in the ADER approach [17, 9].
On the basis of the marked cells and the standard reconstruction operator, we define as follows:
If then the reconstruction operator is defined by
where – , are two states and a scalar to be chosen so that if and can be linked by an admissible shock with speed , then
– is chosen so that
where is a predetermined index (in the case of fluids, the one corresponding to the mass conservation equation). Observe that this in-cell discontinuous reconstruction can only be done if , i.e. if
otherwise the index is removed from the set .
Remark 2.1. The idea behind this reconstruction operator is the following (see Figure 1):
– Two intermediate states of the solution of the Riemann problem (8) (or approximations of them), , , linked by an admissible shock with speed are selected.
– At time , the constant state is replaced in the cell by a discontinuity placed at linking and , where is chosen so that the mass is conserved.
– The reconstruction operator represents the displacement at velocity of this discontinuity in the cell .
The states , are such that, if and can be linked by an admissible shock traveling at speed , the reconstruction coincides with the exact solution of the Riemann problem.
If and , then
Remark 2.2. The idea behind this reconstruction operator is the following (see Figure 2): suppose that and can be linked by an admissible shock. Then, will be a marked cell and the reconstructions at , will be the constant states and respectively: no wave interaction occurs at the intercells then and the reconstructions at the three cells coincide with the solution of the Riemann problem.
Otherwise (i.e., if )
i.e. the standard reconstruction operator is considered in the cells where no shocks are detected.

Figure 1. In-cell discontinuous reconstruction operator.

Figure 2. Constant reconstruction operator.
Fully discrete form
is computed by integrating (2.1) in the interval :
where will be defined in Subsection 2.4. Two quadrature formulas
are used to approximate the integrals in space and time, respectively. Taking into account the expression of the reconstruction operators we obtain:
Remark 2.3. Observe that this method is basically a standard high-order path-conservative method in the smooth region and the standard in-cell discontinuous reconstruction developed in [7, 14, 15] in the presence of shocks for pure nonconservative systems.
How to mark a cell? Choices of , and
Let us assume that a path-conservative Roe linearization for the nonconservative system is available (see [18]), i.e. a matrix-valued function satisfying the following properties:
(R1) for each , has distinct real eigenvalues ;
(R2) , for every ;
(R3) for any ,
Then, for every the coordinates of in the basis of eigenvectors of , are computed, i.e.
and the intermediate states
are defined. The cell is then marked if there exists an index of those corresponding to genuinely nonlinear fields such that
where is a parameter to be selected. Observe that, if (2.14) is satisfied with , then the solution of the linear Riemann problem essentially consists of only one shock wave that, due to the Roe property (R3), is also an admissible discontinuity of the system. In practice we use . Moreover, the second condition in (2.14) allows one to discard non-entropy shocks. In this case, the in-cell discontinuous reconstruction is used (2.6) with the following choices:
Summing up, the set is given by
Time step
Following [14, 15] the time step is chosen as follows:
Here
where is the stability parameter, represent the eigenvalues of , and
where and are given in (2.8) and (2.15), respectively, only for marked cells. Observe that, besides the stability requirement, this choice of time step ensures that no discontinuous reconstruction leaves a marked cell. Note that this time restriction could be avoided as in [7] but, for simplicity, we keep it.
Path-conservative fluctuations
The fluctuations corresponding to Roe method will be considered here
where is the Roe linearization used to mark cells. As usual represent the matrices whose eigenvalues are the absolute value of with same eigenvectors.
First-order method
In this case of first-order methods, the space-time reconstruction operator is the piecewise constant one, i.e.
so that the first-order in-cell discontinuous reconstruction path-conservative method (2.12) method is reduced to:
Second-order method
The MUSCL-Hancock space-time reconstruction operator (see [19, 20]) is considered in this case:
and the stencil is . Here, is the minmod approximation of the first order spatial derivative of at at time , whose th component is given by
where represents the th component of and
The mid-point rule is chosen as quadrature formula both in time and space, so that the second-order in-cell discontinuous reconstruction path-conservative method (2.12) is given by:
where so
and
Shock-capturing and Rankine-Hugoniot compatibility
We recall the theorem in [7, 14, 15] that states that the scheme (2.12) captures exactly isolated shocks in the following way:
Theorem 2.4. Assume that and can be joined by an entropy shock of speed satisfying (1.4). Then, the numerical method (2.12) provides an exact numerical solution of the Riemann problem with initial conditions
in the sense that
where is the exact solution.
Proof. Since (1.4) and (1.3) are equivalents for the given family of paths, the proof reduces to the one in [14].
Extensions
In this work only first- and second-order methods are considered, but this strategy can be applied to higher order methods as it has been done in [15]. Moreover in the mentioned paper the methods are combined with the MOOD approach. The method can also be extended to problem with source terms in a well-balanced manner as it was done in [14].
Euler equations with an interface and shear
As an example of a nonlinear system we consider the Euler equations of gas dynamics for two immiscible materials, plus a shear wave modeling the presence of transverse flow. The resulting system of PDEs is a good representative example for many other applications in compressible flows as well as shallow water flows. It is well known (see e.g. [16, 1] and may others) that the use of conservative variables produce spurious oscillations close to contact discontinuities related to material or velocity interfaces.
Conservative form
We consider the Euler equations enhanced with a marker function to detect moving interfaces and two velocities:
with
The fluid is assumed to be composed of two immiscible ideal gases with equations of states
Typical values are for example for air, and for water. The marker function takes values in , denoting one material and the other one. The equation of state of the fluid is then given by
Two cases will be considered here
two materials, one velocity (),
one material (), two velocities.
Rankine-Hugoniot conditions for the variables , , and writes as follows:
From the first one, we derive:
Therefore, from the Rankine-Hugoniot condition for :
we derive
Since for shock waves we obtain that is constant through shocks, so that the RH condition for is the same that the one for . The same happens for the variable . Therefore, (30) are the RH conditions to be considered for shocks. The variables and can only be discontinuous at contact discontinuities that travel at speed .
Nonconservative form with
We can consider the nonconservative form of the system reading
The matrix has 5 eigenvalues given by
with
The corresponding eigenvectors of are given by
Correct choice of the family of paths
Given a family of paths
the generalized Rankine-Hugoniot conditions (3) for the variables , , are given by:
Again, (and thus ) and are constant though shocks, so that their corresponding RH conditions are equal to the one of .
Theorem 3.1. Given and , if the family of paths
is chosen, then the Rankine-Hugoniot conditions (30) and (32) are equivalent.
Proof. The proof is given in Appendix B.
Nonconservative Roe matrix
Given the family of paths (33), a Roe matrix must verify:
It can be easily checked that a Roe matrix verifying (34) is given by
where
The eigenvalues of (35) are given by:
where , with the eigenvectors
Note that, close to an isolated shock wave, the Roe matrix is only used to mark the cells. The convergence to the right weak solution is guaranteed by the use of the DR method with the proper shock speed. We have now all the ingredients to apply the in-cell discontinuous reconstruction method (16).
Numerical tests
Several tests will be considered for validation and application for each model. The following notation will be used in the simulations:
O1_Primitive+DR: first-order path-conservative in-cell discontinuous reconstruction method (2.20).
O2_Primitive+DR: second-order path-conservative in-cell discontinuous reconstruction method (2.22).
O1_Conservative: standard first-order fully conservative Roe method.
O1_Primitive: standard first-order path-conservative Roe method without discontinuous in-cell reconstruction.
O1_Hybrid: first-order primitive-conservative method without discontinuous in-cell reconstruction. This approach is similar to the one in [10, 4]: once the cells have been marked, O1_Conservative is used in marked cells and O1_Primitive in the rest. The marking is relaxed in comparison with the methods based on discontinuous reconstruction: is chosen in (2.14).
Reference_Solution: it is computed with the O1_Hybrid scheme using a mesh of 5000 points.
In all tests the CFL number is fixed to 0.5.
Euler equations with an interface
All tests are performed in the spatial domain using a mesh of points. In all test cases and . We remember refers to and to . In all these tests, , and we evaluate the ability of the schemes to handle a material interface.
Stationary contact discontinuity. The initial condition is given by
Figure 3 shows the numerical results: we observe how all numerical schemes capture well the isolated contact discontinuity except the fully conservative one that introduce spurious oscillations. This is what was pointed out in [16]. Since no shock wave is detected, in this case the numerical results provided by O1_Primitive and O1_Hybrid coincide. In Table 1, we report the numerical errors obtained with the O1_Primitive+DR, O2_Primitive+DR and O1_Conservative schemes for different meshes. The first two schemes achieve errors at machine precision, whereas the conservative formulation exhibits nonzero errors that decrease with mesh refinement, as expected.
Figure 3. Numerical results for Test: Stationary contact discontinuity.
| O1 Primitive+DR | ||||
| N | ||||
| 100 | 4.01e-16 | 0.00e+00 | 9.54e-16 | 3.47e-16 |
| 200 | 3.04e-16 | 0.00e+00 | 2.60e-16 | 2.60e-16 |
| 400 | 2.17e-17 | 0.00e+00 | 2.60e-16 | 0.00e+00 |
| 800 | 9.76e-17 | 0.00e+00 | 1.73e-16 | 8.67e-17 |
| O2 Primitive+DR | ||||
| N | ||||
| 100 | 4.01e-16 | 0.00e+00 | 9.54e-16 | 3.47e-16 |
| 200 | 3.04e-16 | 0.00e+00 | 2.60e-16 | 2.60e-16 |
| 400 | 2.17e-17 | 0.00e+00 | 2.60e-16 | 0.00e+00 |
| 800 | 9.76e-17 | 0.00e+00 | 1.73e-16 | 8.67e-17 |
| O1 Conservative | ||||
| N | Order | Order | ||
| Order | Order | |||
| 100 | 7.27e-03 | 0.00 | 2.68e-02 | 0.00 |
| 1.40e-02 | 0.00 | 1.45e-02 | 0.00 | |
| 200 | 5.06e-03 | 0.52 | 1.36e-02 | 0.98 |
| 6.67e-03 | 1.07 | 6.42e-03 | 1.18 | |
| 400 | 3.14e-03 | 0.69 | 6.52e-03 | 1.06 |
| 3.11e-03 | 1.10 | 3.93e-03 | 0.71 | |
| 800 | 1.98e-03 | 0.67 | 3.18e-03 | 1.04 |
| 1.51e-03 | 1.05 | 2.71e-03 | 0.54 |
Table 1. Test Stationary contact discontinuity: order of accuracy for O1 Primitive+DR, O2 Primitive+DR and O1 Conservative: L1 errors at time .
Isolated shock wave. The initial condition is given by
where
Only one material (air ) is considered in this test. See Figure 4 for numerical results. In this case all the schemes capture the correct shock wave but O1_Primitive converge as expected. In this case, the numerical results corresponding to O1_Conservative and O1_Hybrid coincide. Observe that the use of the discontinuous reconstruction makes the numerical method to capture the isolated shock wave much more sharply.

Figure 4. Numerical results for Test: Isolated shock.
Quirk and Karni shock helium-bubble interaction. This test is taken from Quirk and Karni in [16]. The initial condition is given by
This test corresponds to the interaction between a shock and a helium bubble. Figure 5 shoes the numerical results. We observe that the poorest results are produced by the fully conservative scheme, due to spurious oscillations arising in the vicinity of contact discontinuities. The standard path-conservative method, when used without the in-cell reconstruction operator, fails to correctly capture the plateau near the left and shocks. In the remaining cases, all schemes appear to converge to the correct solution; however, those that incorporate the in-cell discontinuous reconstruction operator exhibit superior performance in shock regions. Moreover, as expected, the second-order scheme gives better results in rarefaction waves. In 2 we observe the errors obtained with O1 Primitive+DR, O2 Primitive+DR, O1 Conservative. All seem to converge to the correct solution, but together with Figure 6, the convergence is much faster in the case of the in-cell discontinuous reconstruction path-conservative methods.

Figure 5. Numerical results for Test: Quirk and Karni shock helium-bubble interaction.
Figure 6. Convergence for Test Quirk and Karni shock helium-bubble interaction: Numerical errors obtained in log scale for O1_Primitive+DR, O2_Primitive+DR, O1_Conservative at time .
| O1 Primitive+DR | ||||||||
| N | Order | Order | Order | $\lVert\Delta c\rVert_1 | Order | ||||
| 100 | 5.05e-02 | 0.00 | 1.15e-02 | 0.00 | 1.86e-02 | 0.00 | 3.32e-02 | 0.00 |
| 200 | 3.41e-02 | 0.57 | 6.88e-03 | 0.74 | 1.08e-02 | 0.79 | 2.33e-02 | 0.52 |
| 400 | 2.22e-02 | 0.62 | 4.37e-03 | 0.66 | 6.55e-03 | 0.72 | 1.54e-02 | 0.60 |
| 800 | 1.35e-02 | 0.72 | 2.61e-03 | 0.75 | 3.85e-03 | 0.77 | 9.42e-03 | 0.71 |
| O2 Primitive+DR | ||||||||
| N | Order | Order | Order | $\lVert\Delta c\rVert_1 | Order | ||||
| 100 | 1.79e-02 | 0.00 | 3.97e-03 | 0.00 | 5.74e-03 | 0.00 | 1.19e-02 | 0.00 |
| 200 | 9.77e-03 | 0.87 | 1.98e-03 | 1.00 | 2.87e-03 | 1.00 | 6.63e-03 | 0.85 |
| 400 | 3.47e-03 | 1.49 | 8.28e-04 | 1.26 | 1.23e-03 | 1.22 | 2.21e-03 | 1.58 |
| 800 | 1.90e-03 | 0.87 | 6.21e-04 | 0.41 | 9.74e-04 | 0.34 | 1.06e-03 | 1.05 |
| O1 Conservative | ||||||||
| N | Order | Order | Order | Order | ||||
| 100 | 6.87e-02 | 0.00 | 3.32e-02 | 0.00 | 5.45e-02 | 0.00 | 6.69e-02 | 0.00 |
| 200 | 4.73e-02 | 0.54 | 2.20e-02 | 0.60 | 3.66e-02 | 0.57 | 4.77e-02 | 0.49 |
| 400 | 3.22e-02 | 0.55 | 1.45e-02 | 0.60 | 2.45e-02 | 0.58 | 3.37e-02 | 0.50 |
| 800 | 2.07e-02 | 0.64 | 9.18e-03 | 0.66 | 1.60e-02 | 0.62 | 2.36e-02 | 0.51 |
Table 2. Test Quirk and Karni shock helium-bubble interaction: convergence of O1 Primitive+DR, O2 Primitive+DR, O1 Conservative: L1 errors at time .
Gaburro et. al. shock helium-bubble interaction. This test is taken from [10]. The initial condition is given by
This test corresponds to the interaction between a shock and a helium bubble too. Figure 7 show the numerical results. The same conclusions apply as in the previous test.

Figure 7. Numerical results for Test: Shock helium-bubble interaction
Euler equations with shear
All tests are performed in the spatial domain using a mesh of points. In all test cases we take and the variable is set to 1, so that there is only one material (air) and we evaluate the ability of the schemes to handle a shear interface.
Stationary contact discontinuity. The initial condition is given by
Figure 8 shows the numerical results: as in the two-material interface, we observe how all numerical schemes capture well the isolated contact discontinuity except O1_Conservative that introduce spurious oscillations. Again, the numerical results given by O1_Primitive and O1_Hybrid coincide.
Figure 8. Numerical results for Test: Stationary contact discontinuity.
Modified Quirk and Karni small shock-shear interaction. This test is a modification of the one taken from Quirk and Karni [16]. The initial condition is given by
This test corresponds to the interaction between a shock and a small shear wave. Figure 9 shows the numerical results. In this case, the only one numerical method that is not able to capture the correct solution is the standard path-conservative scheme without in-cell discontinuous reconstruction. It seems that the contact discontinuity related to the shear interface is not big enough to produce spurious oscillations in the fully conservative method. Again, the schemes with the in-cell discontinuous reconstruction operator perform better close to shock waves. The second-order scheme captures better the rarefactions as expected.

Figure 9. Numerical results for Test: modified Quirk and Karni small shock-shear interaction.
Modified Quirk and Karni strong shock-shear interaction. This test is a modification of the previous one but now the shear wave in greater. The initial condition is given by
Figure 10 shows the numerical results. In this case, the strong contact discontinuities associated with the shear interface generate spurious oscillations in the fully conservative scheme, which completely ruins the numerical solution. For the remaining methods, similar conclusions can be drawn: standard path-conservative schemes fail to capture the correct shock waves, whereas the hybrid and in-cell discontinuous methods do. Among these, the schemes incorporating in-cell discontinuous reconstruction provide a more accurate resolution of shock waves. Once again, the second-order scheme exhibits superior performance in rarefaction regions.

Figure 10. Numerical results for Test: modified Quirk and Karni strong shock-shear interaction.
Conclusions
In this work, we propose a methodology based on [7, 14, 15], for the numerical treatment of conservative systems formulated exclusively in primitive variables. The convergence difficulties associated with nonconservative products are addressed by identifying an appropriate family of paths and by incorporating in-cell discontinuous reconstructions within a path-conservative framework. In smooth regions or in the presence of contact discontinuities, a standard path-conservative method is used, while in regions where shocks appear, the in-cell discontinuous reconstruction is activated. The proposed approach is further extended to achieve high-order accuracy. Several numerical tests for the Euler multimaterial and shear interface demonstrate the improvement in the numerical solution: when using primitive variables exclusively together with this methodology, the scheme converges to the correct solution even in the presence of shocks, and spurious oscillations are effectively suppressed when dealing with contact discontinuities. No hybrid approaches are necessary to deal with these problems.
Appendix A. Paths and viscous profiles
Although in general the choice of a family of paths for that ensures that the jump condition is consistent with the physics of the problem may be difficult, when comes from a conservative system through a change of variables, there is a clear criterion for deciding which is the correct family of paths: it has to be such that a triple satisfies if and only if satisfies . The interesting question is whether there is always a family of paths with this property. Let us assume that a triple is an admissible discontinuity for if and can be linked by a traveling wave of the viscous problem
i.e., if there exists a smooth function satisfying
such that
is a solution of (A.1) for every . The function , which is called the viscous profile, has to solve the following boundary value problem:
In this case, the path given by
is a compatible path, in the sense that it links and and satisfies (3). Indeed, one has:
Integrating this equality in we obtain:
as we wanted to prove. In order to obtain a family of paths with the sought property, first the paths found for states that can be linked by an admissible discontinuity have to be re-parametrized in the interval , and then the family of paths have to be extended to arbitrary pair of states: this can be done, at least locally, on the basis of the solutions of Riemann problems as in [12].
Choice of the family of paths
Proof of Theorem 3.1. Let us see which choice of family of paths we should take to obtain equivalence of (3.3) and (3.5). We will focus on the second and third equations since the first one is the same in both cases. Let us suppose that (3.5) is equivalent to (3.3) for some family of paths . From (3.3) we have:
Now dividing the second equation of (3.3) by the first one, after some straightforward calculations, we obtain the equality
And using this equality, we finally obtain that the family of paths has to satisfy
A family of paths that satisfies this equality is given by
If we define
the equivalence of the third equations of (3.3) and (3.5) is proved using the same strategy of adding and subtracting .
Acknowledgments
Ernesto Pimentel-García and Carlos Parés have been supported by projects PID2022-137637NB-C21 and PCI2024-155061-2 funded by MCIN/AEI/10.13039/501100011033/ and ERDF A way of making Europe. Both have also been supported by the Grant PPRO-FQM216-2023 funded by Consejería de Universidad, Investigación e Innovación and by ERDF Andalusia Program 2021-2027. MR is member of the CARDAMOM team at the INRIA research center at University of Bordeaux.
References
References
- [1]R. Abgrall, How to prevent pressure oscillations in multicomponent flow calculations: A quasi conservative approach, Journal of Computational Physics, 125 (1996), pp. 150–160, https://doi.org/10.1006/jcph.1996.0085, https://www.sciencedirect.com/science/article/pii/S0021999196900856.
- [2]R. Abgrall and S. Karni, Computations of compressible multifluids, Journal of computational physics, 169 (2001), pp. 594–623.DOI
- [3]R. Abgrall and S. Karni, A comment on the computation of non-conservative products, Journal of Computational Physics, 229 (2010), pp. 2759–2763, http://dx.doi.org/10.1016/j.jcp.2009.12.015, http://www.sciencedirect.com/science/article/pii/S0021999109006949.
- [4]B. Audebert and F. Coquel, Hybrid Godunov-Glimm method for a nonconservative hyperbolic system with kinetic relations, in Numerical Mathematics and Advanced Applications, Springer Berlin Heidelberg, 2006, pp. 646–653, https://doi.org/10.1007/978-3-540-34288-5_62.
- [5]M. J. Castro, P. G. LeFloch, M. L. Muñoz-Ruiz, and C. Parés, Why many theories of shock waves are necessary: Convergence error in formally path-consistent schemes, Journal of Computational Physics, 227 (2008), pp. 8107–8129, http://dx.doi.org/10.1016/j.jcp.2008.05.012, http://www.sciencedirect.com/science/article/pii/S0021999108002842.
- [6]M. J. Castro, T. Morales de Luna, and C. Parés, Well-Balanced Schemes and Path-Conservative Numerical Methods, in Handbook of Numerical Analysis, vol. 18 of Handbook of Numerical Methods for Hyperbolic Problems Applied and Modern Issues, Elsevier, 2017, pp. 131–175, http://www.sciencedirect.com/science/article/pii/S1570865916300333 (accessed 2017-07-24).DOI
- [7]C. Chalons, Path-conservative in-cell discontinuous reconstruction schemes for non conservative hyperbolic systems, Communications in Mathematical Sciences, 18 (2020), pp. 1–30, https://doi.org/10.4310/cms.2020.v18.n1.a1.
- [8]G. Dal Maso, P. G. LeFloch, and F. Murat, Definition and weak stability of nonconservative products, Journal de mathématiques pures et appliquées, 74 (1995), pp. 483–548.
- [9]M. Dumbser, M. Castro, C. Parés, and E. F. Toro, Ader schemes on unstructured meshes for nonconservative hyperbolic systems: Applications to geophysical flows, Computers & Fluids, 38 (2009), pp. 1731–1748.DOI
- [10]E. Gaburro, W. Boscherri, S. Chiocchetti, and M. Ricchiuto, Discontinuous galerkin schemes for hyperbolic systems in non-conservative variables: Quasi-conservative formulation with subcell finite volume corrections, Computer Methods in Applied Mechanics and Engineering, 431 (2024), p. 117311, https://doi.org/10.1016/j.cma.2024.117311, https://www.sciencedirect.com/science/article/pii/S004578252400567X.
- [11]S. Karni, Hybrid multifluid algorithms, SIAM Journal on Scientific Computing, 17 (1996), pp. 1019–1039.DOI
- [12]M. L. Muñoz-Ruiz and C. Parés, Godunov method for nonconservative hyperbolic systems, ESAIM: Mathematical Modelling and Numerical Analysis, 41 (2007), p. 169–185, https://doi.org/10.1051/m2an:2007011.
- [13]C. Parés, Numerical methods for nonconservative hyperbolic systems: a theoretical framework., SIAM Journal on Numerical Analysis, 44 (2006), pp. 300–321.DOI
- [14]E. Pimentel-García, M. J. Castro, C. Chalons, T. Morales de Luna, and C. Parés, In-cell discontinuous reconstruction path-conservative methods for non conservative hyperbolic systems - second-order extension, Journal of Computational Physics, 459 (2022), p. 111152, https://doi.org/10.1016/j.jcp.2022.111152, https://www.sciencedirect.com/science/article/pii/S0021999122002145.
- [15]E. Pimentel-García, M. J. Castro, C. Chalons, and C. Parés, High-order in-cell discontinuous reconstruction path-conservative methods for nonconservative hyperbolic systems–dr.mood method, Numerical Methods for Partial Differential Equations, 40 (2024), p. e23133, https://doi.org/10.1002/num.23133, https://onlinelibrary.wiley.com/doi/abs/10.1002/num.23133, https://arxiv.org/abs/https://onlinelibrary.wiley.com/doi/pdf/10.1002/num.23133.
- [16]J. J. Quirk and S. Karni, On the dynamics of a shock-bubble interaction, Journal of Fluid Mechanics, 318 (1996), p. 129 – 163, https://doi.org/10.1017/S0022112096007069, https://www.scopus.com/inward/record.uri?eid=2-s2.0-0030184154&doi=10.1017%2fS0022112096007069&partnerID=40&md5=03387c7b519a7b621f2b8daef4a7b245. Cited by: 411; All Open Access, Green Open Access.
- [17]V. A. Titarev and E. F. Toro, Ader schemes for three-dimensional non-linear hyperbolic systems, Journal of Computational Physics, 204 (2005), pp. 715–736.DOI
- [18]I. Toumi, A weak formulation of Roe’s approximate Riemann solver, Journal of Computational Physics, 102 (1992), pp. 360–373.DOI





