Setting Up Linear Systems

From Word Problems to Matrix Form

Engineering is full of problems that boil down to finding unknown quantities. A circuit has unknown branch currents. A structural frame has unknown joint forces. A data-fitting task has unknown polynomial coefficients. In each case, the constraints that the unknowns must satisfy can be written as a collection of linear equations — and any collection of linear equations can be written compactly in matrix form.

The central object is the equation Ax = b. The matrix A encodes the relationships between unknowns, the vector x holds the unknowns themselves, and the vector b holds the known right-hand-side values. Recognizing this structure is the gateway to solving — and understanding — virtually any linear problem.

Matrix Form

Ax = b

A is an m×n matrix (m equations, n unknowns), x is an n×1 unknown vector, and b is an m×1 known vector. Together they encode every linear system.

Building the Coefficient Matrix

The first practical skill is translating a set of equations into matrix form. The rule is straightforward: each row of A corresponds to one equation, and each column of A corresponds to one unknown. The entry A_ij is the coefficient of the j-th unknown in the i-th equation.

Consider the simple 2×2 system:

2x₁ + 3x₂ = 5
x₁ − x₂ = 1

Reading off the coefficients row by row gives the matrix A = [[2, 3], [1, −1]], the unknown vector x = [x₁, x₂]ᵀ, and the right-hand-side vector b = [5, 1]ᵀ. In matrix form:

Example System

\begin{pmatrix}2&3\\1&-1\end{pmatrix}\begin{pmatrix}x_1\\x_2\end{pmatrix}=\begin{pmatrix}5\\1\end{pmatrix}

The first row encodes 2x₁ + 3x₂ = 5; the second encodes x₁ − x₂ = 1. Every column corresponds to one unknown: column 1 for x₁, column 2 for x₂.

Geometric Interpretation

Beyond the algebra, each linear equation has a clean geometric meaning. In two dimensions, the equation ax₁ + bx₂ = c defines a straight line in the (x₁, x₂) plane. Solving a 2×2 system means finding the point where two lines intersect.

In three dimensions, each equation ax₁ + bx₂ + cx₃ = d defines a plane. Solving a 3×3 system means finding where three planes meet. The same logic extends to higher dimensions: each equation defines a hyperplane, and the solution is the intersection of all the hyperplanes.

This geometric picture immediately predicts the three types of outcome a system can have — and why each arises.

Existence and Uniqueness

The geometry reveals three fundamentally different cases:

Unique solution: The hyperplanes meet at exactly one point. This happens when the system is "full rank" — no equation is redundant and no contradiction exists.
No solution: The hyperplanes are parallel (or otherwise fail to share a common point). The system is inconsistent — the equations contradict each other.
Infinitely many solutions: Two or more hyperplanes coincide (or all pass through a common line, plane, etc.). The equations are dependent — some contain redundant information.

Which case applies is determined by the rank of A compared to the rank of the augmented matrix [A|b]. This will be made precise when we study Gaussian elimination.

The Augmented Matrix

A convenient way to represent an entire system is the augmented matrix [A|b], formed by appending b as a new column to the right of A. This single object contains all the information about the system — the coefficient structure and the right-hand side — and is the direct input to Gaussian elimination in the next lesson.

Augmented Matrix

\left[\begin{array}{cc|c}2&3&5\\1&-1&1\end{array}\right]

The vertical bar separates A from b. Row operations on [A|b] simultaneously transform the left and right halves, preserving the solution set while simplifying the system.

Types of Linear Systems

The relative sizes of m (equations) and n (unknowns) shape the expected behavior of a system before we even look at specific values:

Square System

m = n

Equal equations and unknowns — typically yields a unique solution if A is invertible. This is the "nice" case that arises most often in engineering.

Overdetermined

m > n

More equations than unknowns — usually no exact solution exists. The best approximate solution is found via least squares. Common in data fitting.

Underdetermined

m < n

Fewer equations than unknowns — infinitely many solutions if the system is consistent. Additional constraints (e.g., minimum norm) are needed to pick one.

Real-World Examples

Linear systems appear throughout engineering and applied science. Recognizing the Ax = b structure is the first step to solving any of these problems:

Electrical circuits: Kirchhoff's voltage and current laws generate linear equations in the unknown branch currents and node voltages. Every circuit analysis reduces to solving a linear system.
Structural analysis: Force and moment balance at each joint of a truss or frame yields a linear system in the unknown member forces or displacements.
Computer graphics: Finding the intersection of a ray with a surface (plane, triangle, or quadric) requires solving a linear system for the ray parameter and surface coordinates.
Data fitting: Fitting a polynomial of degree d through n data points leads to a Vandermonde system — a specific, structured linear system whose solution gives the polynomial coefficients.
Traffic flow: Conservation of vehicles at each intersection node produces linear equations in the unknown traffic flows on each road segment.

Key Takeaways

Any system of linear equations can be written in matrix form Ax = b, where A captures the coefficients, x holds the unknowns, and b holds the right-hand side. The geometry of the system — intersecting hyperplanes — determines whether solutions exist: a unique solution when the system has full rank, no solution when it is inconsistent, and infinitely many when equations are dependent. The augmented matrix [A|b] encodes everything needed to solve the system. Whether m equals, exceeds, or falls short of n shapes what kind of solution to expect. Next up: Gaussian elimination transforms [A|b] into row echelon form, systematically revealing the solution.