Solving Linear Systems of Differential Equations

How matrix notation, eigenvectors, and initial conditions turn coupled equations into explicit solutions.

For a homogeneous linear system with constant coefficients, write the equations as x′ = Ax, find exponential solutions from eigenvalues and eigenvectors, and combine enough independent solutions to obtain the general solution. Initial conditions then determine the constants. The worked example below demonstrates the calculation and explains what its exponential terms tell you about the motion. [2]

In this article

1. Start with the matrix, not separate equations
2. Why eigenvalues produce differential-equation solutions
3. Combine solutions—and check that none are missing
4. An original worked example: two coupled, decaying components
5. Read the dynamics without overextending the method

1. Start with the matrix, not separate equations

The key reduction: an exponential trial solution converts the differential equation into an eigenvalue problem. [2]
Completeness matters: finding one solution is not enough; an n-component system needs a fundamental set of n independent solutions. [2]
Separate the tasks: first construct the general solution, then choose its constants to match the initial state. [2], [3]

The central problem here is x′(t) = Ax(t), where x is a column vector of unknown functions and A is a constant square matrix. For two unknowns, equations such as u′ = au + bv and v′ = cu + dv become a single vector equation with matrix rows (a, b) and (c, d). This notation preserves the coupling: each derivative can depend on both unknown functions. It also makes linear algebra available as a systematic way to construct solutions. [2], [3]

The scope is important: this explanation develops the homogeneous, constant-coefficient case. Its right-hand side consists of the matrix acting on the unknown vector, without an added forcing term. The exponential construction below relies on that fixed matrix. It is not a complete procedure for every linear differential system, particularly systems with time-varying coefficients or external forcing. [2]

2. Why eigenvalues produce differential-equation solutions

Begin with the scalar equation z′ = az, whose solutions are z(t) = ce^(at). For a vector system, try the analogous expression x(t) = ve^(λt), where v is a nonzero constant vector. Differentiation gives x′(t) = λve^(λt), while multiplication by A gives Ax(t) = Ave^(λt). Equating these expressions and cancelling the nonzero exponential leaves Av = λv, or equivalently (A − λI)v = 0. This is precisely the eigenvalue–eigenvector condition. [2]

This algebra explains the mechanism. Along an eigenvector direction, multiplication by A changes the vector only by the factor λ. Consequently, a solution can remain in that direction while its magnitude follows an exponential. In a calculation, find candidate eigenvalues by solving det(A − λI) = 0, then solve (A − λI)v = 0 for a nonzero vector for each candidate. Each resulting pair supplies a solution ve^(λt); the next question is whether those solutions are enough to describe every initial state. [2]

3. Combine solutions—and check that none are missing

The principle of superposition says that constant linear combinations of solutions remain solutions of a homogeneous system. Thus, if x₁ and x₂ solve the equation, so does c₁x₁ + c₂x₂. However, two expressions that are constant multiples of each other do not supply two independent choices. To obtain the general solution for an n-component system, the solution vectors must form a fundamental set of n linearly independent solutions. [2]

Place the candidate solution vectors into columns of a matrix X(t). Its determinant W(t) = det X(t) is the Wronskian. A nonzero Wronskian establishes a fundamental set, so the general solution is x(t) = c₁x₁(t) + ⋯ + cₙxₙ(t). For an initial condition x(t₀) = x₀, substitution gives the algebraic system X(t₀)c = x₀. This separates constructing possible trajectories from selecting the particular trajectory required by the initial data. [2]

Object	What it contributes	What to check
Eigenpair (λ, v)	One exponential solution ve^(λt)	v is nonzero and Av = λv [2]
Fundamental solution set	The complete general solution through superposition	The solution-column matrix has nonzero determinant [2]
Initial condition	The constants selecting a particular solution	Evaluate the general solution at the initial time [2], [3]

4. An original worked example: two coupled, decaying components

Consider the illustrative initial-value problem u′ = −2u + v, v′ = u − 2v, with u(0) = 3 and v(0) = 1. Its matrix is A = [[−2, 1], [1, −2]]. The characteristic determinant is (−2 − λ)² − 1 = (λ + 1)(λ + 3), giving eigenvalues λ₁ = −1 and λ₂ = −3. Corresponding eigenvectors are v₁ = (1, 1) and v₂ = (1, −1), as direct multiplication by A verifies. [2]

The resulting solutions are e^(−t)(1, 1) and e^(−3t)(1, −1). Their solution-column determinant is −2e^(−4t), which never vanishes, so their combination is the general solution: u(t) = c₁e^(−t) + c₂e^(−3t) and v(t) = c₁e^(−t) − c₂e^(−3t). At t = 0, the conditions require c₁ + c₂ = 3 and c₁ − c₂ = 1. Therefore c₁ = 2 and c₂ = 1. [2]

The particular solution is u(t) = 2e^(−t) + e^(−3t) and v(t) = 2e^(−t) − e^(−3t). Check it directly: differentiation gives u′ = −2e^(−t) − 3e^(−3t) and v′ = −2e^(−t) + 3e^(−3t). Substituting u and v into the original right-hand sides gives exactly these expressions. Evaluating at zero gives 3 and 1. Both the equations and the initial conditions are therefore satisfied. [2]

5. Read the dynamics without overextending the method

The example reveals more than two formulas. Adding and subtracting them gives u + v = 4e^(−t) and u − v = 2e^(−3t). The shared component decays with exponent −1, while the difference decays with exponent −3. Both variables approach zero, but their difference disappears faster than their shared component. These combinations expose the two independent modes hidden inside the coupled equations: the eigenvectors identify the combinations, and the eigenvalues specify their exponential rates. [2]

Do not assume that every matrix yields the same simple collection of real exponential modes. The general treatment distinguishes real distinct, complex, and repeated eigenvalues. The worked example covers the real-distinct case. For other cases, the essential checkpoint remains completeness: have you actually obtained enough independent solutions? Merely counting eigenvalues, including repetitions, does not perform the Wronskian check or construct the missing solutions. The detailed procedures for those other cases are beyond the method developed here. [2]

Frequently asked questions

Why can’t I solve each equation separately?

In a coupled system, one component’s derivative depends on other components, so treating each equation as an isolated scalar equation ignores that dependence. Matrix notation retains all those relationships. The eigenvector construction then identifies directions in which an exponential solution satisfies every equation simultaneously. [3], [2]

Does finding one eigenvector solve the whole problem?

Usually it supplies only one solution mode. The general solution of an n-component homogeneous system requires a fundamental set of n independent solutions. Constant multiples of one mode do not create additional independence; assembling the solution vectors into a matrix and checking its determinant tests completeness. [2]

Does rescaling an eigenvector change the answer?

Multiplying an eigenvector by a nonzero constant rescales its exponential solution, but that factor can be absorbed into the arbitrary coefficient in the general solution. It changes the representation, not the family of solutions. Keep the rescaling consistent when solving for constants from initial conditions. [2]

What does a negative real eigenvalue tell me?

For its associated mode ve^(λt), a negative real λ makes the exponential tend to zero as time increases. That describes this mode, not automatically every solution of the system: the other independent modes and their coefficients also matter. In the worked example, both exponents are negative, so both modes decay. [2]

What is the most useful final verification?

Differentiate the proposed vector solution and check that it equals A times the solution. Then substitute the initial time and verify the prescribed initial vector. These are separate checks: satisfying the differential equation alone does not confirm that the constants match the requested initial-value problem. [2], [3]

Conclusion

The central strategy is to replace a coupled differential problem with an algebraic one: eigenpairs generate exponential modes, independence makes their combination complete, and initial conditions select the constants. The worked example also shows why the method is informative: its modes distinguish shared motion from a faster-decaying difference. Keep the scope explicit, verify the proposed solution in the original equations, and treat complex or repeated eigenvalues as cases requiring further construction rather than forcing them into the real-distinct template. [2]

Source notes

Detailed constructions for complex eigenvalues, repeated eigenvalues, nonhomogeneous forcing, and time-varying coefficient matrices are not established by the passages used here.

Explore next

Sources

Open a source or expand its cited passages.

Relevance scores are model estimates of how well each source addresses this question, not probabilities or fact-check ratings. Reasoned guidance is identified on hover.