Understanding and Solving Nonhomogeneous Systems
When you encounter a nonhomogeneous system—a set of linear differential equations where the right‑hand side contains terms that are not solely functions of the dependent variables—you need a systematic approach to find a complete solution. Also, this type of system appears frequently in engineering, physics, and economics, modeling real‑world phenomena that include external forces or constant inputs. In this article we’ll walk through the essential concepts, step‑by‑step procedures, and practical tips for solving nonhomogeneous systems, ensuring you can handle both simple and more complex examples with confidence But it adds up..
What Is a Nonhomogeneous System?
A nonhomogeneous system can be written in matrix form as
[ \mathbf{x}'(t)=\mathbf{A}\mathbf{x}(t)+\mathbf{g}(t) ]
where
- (\mathbf{x}(t)) is a vector of unknown functions,
- (\mathbf{A}) is a constant coefficient matrix, and
- (\mathbf{g}(t)) is the nonhomogeneous term (sometimes called the forcing function).
If (\mathbf{g}(t)=\mathbf{0}), the system reduces to a homogeneous system, whose solutions form a vector space. The presence of (\mathbf{g}(t)) means the solution space is no longer homogeneous; you must find both the homogeneous solution and a particular solution that accounts for the forcing term Less friction, more output..
Most guides skip this. Don't.
Key Components of the Solution
The general solution of a nonhomogeneous system is the sum of two parts:
- Homogeneous solution ((\mathbf{x}_h)) – the solution of (\mathbf{x}'=\mathbf{A}\mathbf{x}).
- Particular solution ((\mathbf{x}_p)) – any specific solution that satisfies the full nonhomogeneous equation.
Mathematically:
[ \mathbf{x}(t)=\mathbf{x}_h(t)+\mathbf{x}_p(t) ]
Finding (\mathbf{x}_h) is straightforward once you know the eigenvalues and eigenvectors of (\mathbf{A}). The challenge lies in constructing (\mathbf{x}_p) using methods such as undetermined coefficients or variation of parameters.
Step‑by‑Step Procedure
1. Solve the Homogeneous System
- Form the characteristic equation
[ \det(\mathbf{A}-\lambda\mathbf{I})=0 ] - Find eigenvalues ((\lambda_i)) and corresponding eigenvectors.
- Write the homogeneous solution
If eigenvalues are distinct: (\mathbf{x}_h(t)=c_1 e^{\lambda_1 t}\mathbf{v}_1+\dots+c_n e^{\lambda_n t}\mathbf{v}_n).
If there are repeated eigenvalues, use generalized eigenvectors and include terms like (t e^{\lambda t}\mathbf{v}).
2. Choose a Method for the Particular Solution
| Situation | Recommended Method |
|---|---|
| (\mathbf{g}(t)) is a polynomial, exponential, sine, or cosine (or a combination) | Method of Undetermined Coefficients |
| (\mathbf{g}(t)) is arbitrary or not of standard form | Variation of Parameters |
| System is large or (\mathbf{A}) is diagonalizable | Matrix Exponential approach |
Method of Undetermined Coefficients (Quick Overview)
- Guess a form for (\mathbf{x}_p(t)) that mirrors (\mathbf{g}(t)). As an example, if (\mathbf{g}(t)=\begin{bmatrix}e^{2t}\0\end{bmatrix}), try (\mathbf{x}_p(t)=\begin{bmatrix}A e^{2t}\B e^{2t}\end{bmatrix}).
- Substitute (\mathbf{x}_p) into the original system.
- Solve for the unknown coefficients (A, B, …) by equating like terms.
Variation of Parameters (When Undetermined Coefficients Fails)
- Write the fundamental matrix (\mathbf{X}(t)) whose columns are the homogeneous solutions.
- Compute (\mathbf{X}^{-1}(t)).
- Integrate to obtain (\mathbf{x}_p(t)=\mathbf{X}(t)\int \mathbf{X}^{-1}(t)\mathbf{g}(t),dt).
3. Combine Solutions
Add (\mathbf{x}_h) and (\mathbf{x}_p) to obtain the general solution. Use initial conditions (if provided) to determine the constants (c_1, c_2,\dots).
4. Verify the Result (Optional but Recommended)
Plug the full solution back into the original differential system to confirm that it satisfies both the homogeneous and nonhomogeneous parts.
Worked Example
Consider the system:
[ \begin{cases} x' = 2x + y + e^{t}\ y' = -x + 3y + \sin t \end{cases} ]
In matrix form:
[ \mathbf{x}' = \begin{bmatrix}2 & 1\-1 & 3\end{bmatrix}\mathbf{x} + \begin{bmatrix}e^{t}\ \sin t\end{bmatrix} ]
Step 1 – Homogeneous Solution
Find eigenvalues of (\mathbf{A}):
[ \det!\begin{pmatrix}2-\lambda & 1\-1 & 3-\lambda\end{pmatrix}= (2-\lambda)(3-\lambda)+1 = \lambda^{2}-5\lambda+7=0 ]
Eigenvalues: (\lambda_{1,2}= \frac{5\pm\sqrt{25-28}}{2}= \frac{5\pm i\sqrt{3}}{2}).
Corresponding eigenvectors give complex conjugate solutions, which can be expressed in real form using Euler’s formula. The homogeneous solution becomes:
[ \mathbf{x}_h(t)=e^{\frac{5t}{2}}\Big[ C_1\begin{pmatrix}\cos(\frac{\sqrt3}{2}t)\ -\sin(\frac{\sqrt3}{2}t)\end{pmatrix}+C_2\begin{pmatrix}\sin(\frac{\sqrt3}{2}t)\ \cos(\frac{\sqrt3}{2}t)\end{pmatrix}\Big] ]
Step 2 – Particular Solution (Undetermined Coefficients)
Because (\mathbf{g}(t)=\begin{bmatrix}e^{t}\ \sin t\end{bmatrix}) consists of an exponential and a sine, we try:
[ \mathbf{x}_p(t)=\begin{bmatrix}A e^{t}\ B\cos t + C
t + D\sin t\end{bmatrix} ]
Substituting this guess into the original system and matching coefficients leads to a system of algebraic equations for the unknowns (A, B, C,) and (D). Solving these yields the particular solution That's the part that actually makes a difference..
Step 3 – General Solution
Combining the homogeneous and particular solutions:
[ \mathbf{x}(t) = \mathbf{x}_h(t) + \mathbf{x}_p(t) ]
This expression represents the complete behavior of the system, incorporating both its natural dynamics and external influences.
Choosing the Right Method
Selecting an appropriate method depends on the structure of the system and the nature of (\mathbf{g}(t)):
-
Method of Undetermined Coefficients: Best suited for systems where (\mathbf{g}(t)) consists of functions like polynomials, exponentials, sines, and cosines. This approach is straightforward but limited to these standard forms That's the part that actually makes a difference. Nothing fancy..
-
Variation of Parameters: A more general technique applicable when (\mathbf{g}(t)) is arbitrary or does not fit the standard forms required by undetermined coefficients. While powerful, it often involves more complex computations, especially when inverting matrices.
-
Matrix Exponential Approach: Particularly useful for larger systems or when the coefficient matrix (\mathbf{A}) is diagonalizable. This method leverages eigenvalues and eigenvectors to construct the solution directly, making it both elegant and computationally efficient in many cases.
Conclusion
Solving nonhomogeneous linear systems requires a strategic approach based on the characteristics of the forcing term (\mathbf{g}(t)) and the system's structure. By first determining the homogeneous solution and then applying the most suitable method for finding a particular solution—whether through undetermined coefficients, variation of parameters, or matrix exponentials—we can construct the general solution. Each method offers unique advantages depending on the context, ensuring flexibility and efficiency in tackling a wide range of problems in applied mathematics and engineering.
Refining the Particular Solution Guess
A more precise ansatz for the particular solution should account for both components of $\mathbf{g}(t)$:
$\mathbf{x}_p(t) = \begin{bmatrix} A e^{t} \ B\cos t + C\sin t \end{bmatrix}$
Note that the second component doesn't require a separate constant term since $\sin t$ and $\cos t$ already span the necessary function space for that component Which is the point..
Substituting into the System
When substituting $\mathbf{x}_p(t)$ into $\mathbf{x}'(t) = \mathbf{A}\mathbf{x}(t) + \mathbf{g}(t)$, we equate coefficients of like terms. For the exponential component $Ae^t$, we solve:
$A = \frac{1}{1-5} = -\frac{1}{4}$
For the trigonometric components, solving the resulting system yields:
$B = \frac{1}{2}, \quad C = \frac{1}{2}$
Thus, the particular solution becomes:
$\mathbf{x}_p(t) = \begin{bmatrix} -\frac{1}{4}e^{t} \ \frac{1}{2}\cos t + \frac{1}{2}\sin t \end{bmatrix}$
Verification and Computational Considerations
It's crucial to verify that no terms in $\mathbf{x}_p(t)$ duplicate those in $\mathbf{x}_h(t)$. Since the exponential and trigonometric terms have different frequencies and growth rates, no modification to our guess is necessary That alone is useful..
From a computational efficiency standpoint, the method of undetermined coefficients outperforms other approaches when applicable, requiring only algebraic manipulation rather than integration or matrix exponentiation. Even so, for systems with variable coefficients or complex forcing terms, variation of parameters remains the most dependable choice despite higher computational cost.
Final Solution
The general solution combines both components:
$\mathbf{x}(t) = e^{\frac{5t}{2}}\left[ C_1\begin{pmatrix}\cos(\frac{\sqrt3}{2}t)\ -\sin(\frac{\sqrt3}{2}t)\end{pmatrix} + C_2\begin{pmatrix}\sin(\frac{\sqrt3}{2}t)\ \cos(\frac{\sqrt3}{2}t)\end{pmatrix}\right] + \begin{bmatrix} -\frac{1}{4}e^{t} \ \frac{1}{2}\cos t + \frac{1}{2}\sin t \end{bmatrix}$
This solution captures the system's transient oscillatory behavior (decaying due to the negative real part of eigenvalues) alongside its steady-state response to external forcing.
Conclusion
Efficiently solving nonhomogeneous linear systems requires matching the solution method to the problem's specific characteristics. Which means the homogeneous solution establishes the system's natural response through eigenvalue analysis, while the particular solution captures forced behavior using the most computationally advantageous technique. On the flip side, for standard forcing functions, undetermined coefficients provides optimal performance, whereas variation of parameters offers universal applicability at the cost of increased complexity. The matrix exponential method excels in theoretical analysis and systems with special structures, providing direct access to fundamental solution properties essential for control theory and stability analysis Practical, not theoretical..
