pdsolve  find exact solutions for systems of partial differential equations (PDEs)

Calling Sequence


pdsolve(PDE_sys, optional_1, optional_2,...)


Parameters


PDE_sys



system of PDEs; it can contain inequations

optional_i



(optional) arguments that can be given in any order and are described as follows

funcs



set or list with indeterminate functions or names  can also be a ranking

HINT = `+`



indicate that the PDE system is to be solved by separating the variables by sum

HINT = `*`



indicate that the PDE system is to be solved by separating the variables by product

HINT = TWS



indicate that the PDE system is to be solved by computing a traveling wave solution as a power series in tanh(x), where xi represents a linear combination of the independent variables

HINT = TWS(foo)



indicate that the PDE system is to be solved by computing a traveling wave solution as a power series in foo, where foo is the name of a mathematical function

singsol=false



avoid the computation of the singular solutions when the system is nonlinear

mindim=N



avoid the computation of solutions if the dimension of the solution space is less than N

parameters=P



P is a list or set of names or functions which are solving variables with priority lower than any other variable

diffalg



request that the differential elimination step for all PDE systems be performed by using the DifferentialAlgebra package





Description


•

Given a PDE system, possibly including ODEs, algebraic constraints, and inequations, pdsolve returns an exact solution if: 1) the uncoupling of the system does not require more resources than those available in the computer and 2) the pdsolve routines for solving a PDE subsystem in a single unknown as well as the dsolve routines for solving related ODE systems succeed in solving the PDE and ODE subsystems arising in the uncoupling process. Depending on the complexity of the system, step 2) may not be fully successful. However, step 1) is systematic in that, generally speaking, whenever the system is consistent, the uncoupling is feasible. pdsolve also works with anticommutative variables set using the Physics package using the approach explained in PerformOnAnticommutativeSystem.


The solving process


1.

The system is first uncoupled by using differential algebra techniques for polynomial systems. When the PDE system, rational in the unknowns and their derivatives, contains nonpolynomial coefficients, the uncoupling is performed by rewriting the system in polynomial form by using a differential extension approach (see PDEtools[dpolyform]).


Due to the intrinsic nature of the differential algebra elimination process, this first step always produces (possibly many) PDE subsystems, such that:

–

The equations inside each subsystem satisfy all the integrability conditions;

–

The union of the nonsingular solutions of each subsystem is equal to the general and singular solutions of the original system;

–

Provided that the original system is not subdetermined, one of the PDE subsystems depends on a single unknown.

2.

The PDE subsystem for a single unknown is solved, and the unknown removed from all other subsystems. Due to nature of the differential elimination process, after removing this single unknown, there exists a PDE subsystem that depends on a single unknown.


Step 2 is repeated until all the PDE subsystems obtained in step 1 are solved, thus arriving at the solution to the original input PDE system.

•

When solving each of these PDE subsystems involving a single unknown, two situations may arise: the subsystem consists of only one PDE, or it consists of many PDEs. In the former case, pdsolve solves the problem by using older subroutines. In the latter case, pdsolve uses a set of routines that, at each step, change variables introducing differential invariants as new variables. This approach permits solving the problem when the PDE subsystem is essentially nonlinear.



The output


•

When successful, pdsolve returns a sequence of solution sets, including the singular solutions of the given PDE system. These singular solutions exist only when the system is nonlinear and can be identified by observing that the dimension of the solution space (basically the number of integration constants appearing) is less than the dimension of the general solution.

•

When pdsolve is partially successful, the output is a sequence of sets and lists. The sets contain solutions while the lists also contain unsolved differential equations. It is valuable to have these partial solutions because they may be easier to solve. For example, the unsolved equations may be ODEs instead of PDEs, less in number, or of lower differential order.

•

pdsolve can fail to solve any of the PDE subsystems obtained in step 1 of "The solving process" described above. In such a case, the output consists of a sequences of lists, where each list contains a PDE subsystem. The value of this output is that these PDE subsystems represent the uncoupling (also known as triangularization) of the original system. Depending on the PDE system, this triangularization may represent a relevant step towards solving the problem.



The unknowns of the problem and rankings


•

By default, when only the PDE system is given, pdsolve considers all the unknown functions present in the system as the unknowns of the problem  herein also called the solving variables. This default can be changed by calling pdsolve with an extra argument, specifying the solving variables either as a set or as a list containing the names of these functions or the functions themselves.

•

When the solving variables are indicated as a set, the ordering chosen by pdsolve for uncoupling the system is chosen by the PDEtools[casesplit] command, which performs the triangularization . When the solving variables are indicated as a list, for instance, , the uncoupling ordering attempted is: to obtain a PDE subsystem depending on only , another PDE subsystem depending on only , and another one depending on . If such a triangularization is successful, pdsolve first solves the subsystem for , then the one for by substituting the solution found for and solving the resulting system for , and finally the one for by substituting the solutions found for and solving the resulting system for .



Other optional arguments


•

When the PDE system is nonlinear, by default, pdsolve's output contains, in separate solution sets, the singular solutions of the system. This involves both computing the singular cases (when uncoupling the system inside PDEtools[casesplit]) and solving them. In many applications, however, the computation of the singular cases is not relevant. Also their computation is sometimes more time consuming than the computation of the general solution. For these reasons, an optional argument is provided to override this behavior, so that by giving singsol=false these singular solutions are not computed.

•

When a PDE system is solved, the dimension of the solution space (typically indicated by the number of integration constants present) depends on various factors, not only the "apparent differential order" of the system. In some applications, however, it is the case that the solutions of interest are only those that have at least a minimum number of dimensions. In such a case, to discard solutions with dimensions less than (some positive integer), use the optional argument. For nonlinear PDE systems, this option may also speed up the computational process since there is no time consumed in computing and solving the lowerdimension cases. For linear PDE systems, where there is only one solution, pdsolve returns NULL if N is greater than the dimension of the solution.

•

It is possible to ask pdsolve to solve a PDE system by trying to separate the variables by sum or product. For this purpose, use the optional argument or , respectively.

•

It is possible to ask pdsolve to solve a PDE system by computing a Traveling Wave Solution, as a power series expansion in tanh (default) or one of the following functions: exp, ln, sin, cos, tan, their multiplicative inverses csc, sec and cot, the corresponding six hyperbolic functions, arcsinh, the twelve elliptic Jacobi functions, the WeierstrassP function and also the identity function x > x, useful to compute purely polynomial solutions when they exist.

•

By default, pdsolve performs all differential elimination processes for PDE systems by using the DEtools[Rif] package. This default can be changed by using the optional argument diffalg, anywhere in the calling sequence, which causes the computation to use the DifferentialAlgebra package.

•

For further mathematical conventions related to the output by pdsolve, see the conventions used in solve, dsolve, and int.




Examples


The computations below were performed on a Pentium II400 laptop PC running Windows. Maple solved all five examples in 36 seconds.
Example 1.
Compute the point symmetries of an ordinary differential equation (ODE), that is, solve the determining PDE system for the infinitesimals of the symmetry generator. Consider example 11 from Kamke's book:
>


 (1) 
>


 (2) 
>


 (3) 
>


 (4) 
The PDE system satisfied by the symmetries, that is, infinitesimals of the symmetry generator, of the ODE above is given by
>


 (5) 
>


>


 (6) 
This is a second order linear PDE system, with two unknowns and four equations. Its general solution is given by
>


 (7) 
Since symmetries are defined up to a multiplicative constant, we can drop . Hence, for arbitrary , the ODE has only one point symmetry.
Solutions to PDE systems can be tested by using pdetest (the command tests whether the solution satisfies each PDE in the system).
>


 (8) 
Now, an interesting question is: Are there other solutions related to particular values of the parameters n and r?
To answer that question one must solve a more difficult, now nonlinear, PDE system by indicating to pdsolve that and are parameters.
Also, in this particular problem, from the form of the ODE , the case is of no interest since the ODE would become linear, and so, computing its symmetries is the same as computing its solution. We are interested in a solution for different values of and but also for . The first step is then to add this inequation to the PDE system.
>


Next we indicate to pdsolve that and are parameters of the problem, thus arriving at the desired solutions.
>


 (9) 
>


 (10) 
So there exist particular values of and for which the system has additional solutions. The solution set with and integers and with xi linear in is in fact a particular case of the general solution computed previously, but the other solution sets are not.
Example 2.
Compute the general solution of the following (linear) overdetermined system involving two PDEs, three unknown functions, one of which depends on 2 variables and the other two depend on only 1 variable.
>


 (11) 
>


 (12) 
The solution for the unknowns G, H, is given by
>


 (13) 
and depends on 3 arbitrary functions, .
>


 (14) 
Example 3.
Compute the solution of the following overdetermined first order linear system of 3 PDEs in a single unknown .
>


>


 (15) 
The solving technique makes use of differential invariants and the solution is given by
>


 (16) 
>


 (17) 
Example 4.
Compute the solution of the following nonlinear system, consisting of Burger's equation and a possible potential.
>


 (18) 
>


 (19) 
Apart from the solution to the general or main case, nonlinear systems usually have singular solutions, which solve the subsystems appearing during the splitting into cases of the differential elimination process. Depending on the case, the computation of the singular cases may be quite time consuming. To compute only the solution to the general case, one can give the optional argument singsol = false.
>


 (20) 
>


 (21) 
Without giving singsol = false, pdsolve attempts to compute the solutions to both the general and singular cases, in this example leading to
>


 (22) 
In the singular case above, , only Burger's equation remains in the system and the solution computed is a Traveling Wave Solution. The introduction of a potential in the system leads to the general solution only when is different from zero.
To see the cases, pass to PDEtools[casesplit] the same arguments passed to pdsolve.
>


 (23) 
Example 5.
Compute the solutions of the following overdetermined linear system of 38 PDEs that arises when computing point symmetries for the wave equation (more like an exercise concerning the solving of a system with many equations). The system is given by
>


 (24) 
>


 (25) 
>


 (26) 
The solution is given by
>


 (27) 
>


 (28) 
Example 6.
The following is a nonlinear PDE system example in five unknowns, containing four PDEs, extracted from Landau and Lifshitz's "Fluid Mechanics" book. To obtain enhanced display and simplify the input
>


 (29) 
>


 (30) 
>


>


The PDE system is:
>


 (31) 
>


 (32) 
>


 (33) 
>


 (34) 
>


This system represents a 2D (coordinates x, z) fluid layer, open to air, subjected to a vertical temperature gradient (Nabla of T(x,z,t)) and embedded in a gravitational field. The constants sigma and rho are related to the viscosity, thermal diffusivity, the gravitational constant, and the coefficient of thermal expansion. A set of traveling wave solutions can be computed for this system in 2 seconds (typical PC 2 GigaHertz, 2004). The nonsingular case after removing redundant constant solutions:
>


 (35) 
For p = constant, this system also admits solutions in terms of the WeierstrassP elliptic function; this is last in the sequence of nonsingular solutions (operand [1]).
>


 (36) 
Both dsolve and pdsolve can now solve PDEs that involve anticommutative variables set using the Physics package using the approach explained in PerformOnAnticommutativeSystem.
>


 (37) 
Set first and as suffixes for variables of type/anticommutative (see Setup)
>


 (38) 
A PDE system example with two unknown anticommutative functions of four variables, two commutative and two anticommutative; to avoid redundant typing in the input that follows and redundant display of information on the screen let's use PDEtools:diff_table PDEtools:declare
>


 (39) 
>


Now we can enter derivatives directly as the function's name indexed by the differentiation variables and see the display the same way; two PDEs
>


 (40) 
>


 (41) 
The solution to this system:
>


 (42) 
This solution involves an anticommutative constant , analogous to the commutative constants where n is an integer. The arbitrary functions introduced are all commutative, as usual and the Grasmannian parity of the left and righthandsides is preserved##
>


 (43) 
>




See Also


casesplit, dchange, declare, DEtools, diff, DifferentialAlgebra, dpolyform, dsolve, dsolve,system, gensys, int, odetest, pdetest, PDEtools, pdsolve, PerformOnAnticommutativeSystem, Physics, Rif, Setup, solve, TWSolutions


References



ChebTerrab, E.S. "A Computational Approach for the Exact Solving of Systems of Partial Differential Equations". Submitted to Computer Physics Communications, 2001.


