Unique, Infinite, or No Solution in Gaussian Elimination

A row of zeros on the left of the augmented matrix does not mean the system has infinitely many solutions. The deciding factor is what sits on the other side of the vertical bar. A row that reads [0 0 0 | 0] is harmless, that equation is redundant. But a row that reads [0 0 0 | c] with any non‑zero c makes the entire system impossible. Spot the difference instantly, name the free parameters when they exist, and write the parametric solution that describes an infinite family of answers. The three cases, unique, infinite, and no solution, are all visible in the final matrix. Learning to read them is faster than re‑solving.

After Gaussian elimination produces a reduced matrix (row‑echelon form or reduced row‑echelon form, as defined in Strang, Introduction to Linear Algebra ch. 2), the shape of the coefficient side tells you which of three outcomes you have. A system with an infinitely many solutions matrix contains at least one column without a pivot; those columns correspond to free parameters. A system with no solution has a contradiction row: zeros on the left, non‑zero on the right. A system with a unique solution has a pivot in every variable column and no contradiction.

Three short rules cover every case you will see in a first linear algebra course:

• Unique solution: Pivots in every column. No row of zeros with a non‑zero constant.
• Infinite solutions: At least one column without a pivot. Every zero‑coefficient row has a matching zero constant.
• No solution: At least one row where the coefficient side is all zeros but the augmented column is non‑zero.

A row of all zeros including the constant is simply a dependent equation, it adds nothing and can be dropped. The count of pivots (Strang calls them the rank) tells you how many independent equations you actually have.

When you see a row with zeros on the coefficient side and a non‑zero number on the constant side, stop solving. That row represents the equation 0 = c where c is not zero. No assignment of the variables can make that true. The system is inconsistent, it has no solution at all.

Worked Example: Inconsistent 3×3 System

Start with this augmented matrix (coefficients on the left, constants on the right):

[ 1 2 -1 | 4 ]
[ 2 4 1 | 9 ]
[ 0 0 0 | 5 ]

The third row is already the contradiction: [0 0 0 | 5]. It says 0 = 5. No further operations can fix it. The original equations conflict, maybe the first two equations were consistent but the third was not, or a data entry error produced a row that cannot be satisfied. In the terminology of OpenStax Linear Algebra, the system is inconsistent. The rank of the coefficient matrix is 2 (two pivots), but the rank of the augmented matrix [A|b] is 3. When those two ranks differ, the system has no solution.

What to do: If you are checking your own work, look for a mistake in the third equation. If the equations came from a real problem, sensor readings, budget constraints, the inconsistency tells you the model is overconstrained; drop or relax one equation.

When the system is consistent and has fewer pivots than variables, you have free parameters. A free parameter corresponds to a column in the coefficient matrix that does not contain a pivot. That parameter can be set to any real number, and the remaining variables (the basic variables, one per pivot) are expressed in terms of it.

In the notation of Strang ch. 2 and OpenStax, a pivot is the first non‑zero entry in a row after the matrix is in row‑echelon form. The number of pivots equals the rank of the matrix. A system with n variables and rank r has n − r free parameters. For a 3×3 system with rank 2, you get one free parameter. For a 3×3 system with rank 1, you get two free parameters.

Worked Example: Dependent 3×3 with One Free Parameter

Reduce this augmented matrix:

[ 1 2 3 | 4 ]
[ 2 4 6 | 8 ]
[ 3 6 9 | 12 ]

After elimination (the second and third rows are multiples of the first) you get:

[ 1 2 3 | 4 ]
[ 0 0 0 | 0 ]
[ 0 0 0 | 0 ]

The coefficient matrix has one pivot (column 1). Columns 2 and 3 have no pivots, so y and z are free. The only remaining equation is x + 2y + 3z = 4. Solving for the basic variable x gives x = 4 − 2y − 3z. To write the parametric solution, set y = t and z = s, where t and s are any real numbers. Then:

• x = 4 − 2t − 3s
• y = t
• z = s

This is the general solution in parametric form. Every choice of (t, s) gives a specific solution. A system with one free parameter produces a line of solutions; two free parameters produce a plane.

Once you have identified the free parameters, the standard procedure is:

1. Label each column without a pivot as a free parameter. Give it a parameter name, t, s, etc.
2. Move all free parameter terms to the right side of the equation for each basic variable.
3. Write the solution as a vector plus a linear combination of parameter vectors.

For the example above, the vector form is:

[x, y, z] = [4, 0, 0] + t[−2, 1, 0] + s[−3, 0, 1]

The constant vector [4, 0, 0] is a particular solution. The vectors [−2, 1, 0] and [−3, 0, 1] span the nullspace of the coefficient matrix. Strang calls this representation the complete solution = particular solution + homogeneous solution.

If the system has no free parameters, every variable is basic and the parametric form collapses to a single numeric vector. That is the unique solution.

The fastest way to classify a system without reading every row is the rank test. Compute the rank of the coefficient matrix A (the number of pivots after elimination). Compute the rank of the augmented matrix [A|b] (the number of pivots when you include the constant column). Then:

• If rank(A) = rank([A|b]) = number of variables → unique solution.
• If rank(A) = rank([A|b]) < number of variables → infinite solutions.
• If rank(A) < rank([A|b]) → no solution.

The third case is the contradiction row: the extra pivot is in the augmented column. For the inconsistent example earlier, rank(A) = 2 but rank([A|b]) = 3, so the system fails the test. This rank condition is covered in every standard linear algebra text, including OpenStax and LibreTexts.

Do not confuse rank with the number of equations. A 3×3 system can have rank 2 even though it has three rows. The extra row is a linear combination of the other two, it contributes nothing to the rank.

The single most common error is misreading a zero row. Students see all zeros on the left and immediately declare infinite solutions, forgetting to check the constant side. The second most common error is forgetting that a row of zeros in the coefficient matrix with a non‑zero constant means no solution, not infinite solutions. A third mistake is treating a system with more variables than equations as automatically having infinite solutions, it does, but only if the system is consistent. A contradictory row can appear even in an underdetermined system, and that kills all solutions.

Before you write the parametric form, always confirm that no row says 0 = c with c ≠ 0. That single check saves more time than any other step.