Difficulty: Introductory | Prerequisites: Augmented matrices, REF, RREF, Gaussian elimination (see Part 1 of Lecture 03 notes)
Once you can row-reduce a matrix to RREF, the next question is: what does the solution set look like? This section gives you a systematic method for reading solutions directly from RREF, introduces the concept of rank, and establishes one of the most important results in the course: every linear system has either no solutions, exactly one solution, or infinitely many solutions. Nothing else is possible. Understanding this classification is essential for nearly every topic that follows.
Variables corresponding to columns without pivots become free parameters. The solution set is written in terms of those parameters. The rank of a matrix equals the number of pivots, and it controls how many free parameters (and therefore how many degrees of freedom) the solution has.
Leading entry (pivot)
The first nonzero entry in a row of a matrix in REF or RREF. Pivot columns are the columns that contain a leading entry.
Think of pivots as the "anchored" variables, the ones you can solve for directly.
Free variable
A variable whose corresponding column in the RREF has no leading entry. Free variables are assigned parameters (t1, t2, ...) and can take any real value.
In simple terms, free variables are the ones you get to choose freely; everything else is determined by them.
Rank
The number of leading entries (pivots) in the REF or RREF of a matrix. If a matrix has n variable columns and k columns without leading entries, the rank equals n − k.
Think of rank as a measure of how many independent constraints the system actually imposes.
k-parameter solution
A solution set described in terms of k free parameters, where k equals the number of columns (excluding the augmented column) without leading entries.
In simple terms, k tells you how many dimensions of freedom the solution set has.
The solution set of a system whose augmented matrix is in RREF can be described in a standard way:
Identify which columns have leading entries (pivot columns) and which do not (free columns).
Assign a free parameter to each free variable.
Use each equation (row) to express the corresponding pivot variable in terms of the free parameters.
Write the full solution as a set of tuples parametrised by the free variables.
Augmented matrix in RREF:
( 1 2 0 1 0 | 2 )
( 0 0 1 3 0 | 3 )
( 0 0 0 0 1 | 4 )
This corresponds to:
E1: x1 + 2x2 + x4 = 2
E2: x3 + 3x4 = 3
E3: x5 = 4
Columns 2 and 4 have no leading entries, so x2 and x4 are free variables. Set x2 = t1 and x4 = t2.
Solving for the pivot variables:
From E1: x1 = 2 − 2t1 − t2
From E2: x3 = 3 − 3t2
From E3: x5 = 4
Solution set: {(2 − 2t1 − t2, t1, 3 − 3t2, t2, 4) | t1, t2 ∈ ℝ}
This is a 2-parameter family of solutions.
For a system with n variables (columns excluding the augmented column):
k = number of variable columns without leading entries (free variables)
Rank = n − k (number of pivots)
Corollary: Any linear system falls into exactly one of three cases:
No solutions: A row of the form (0 0 ... 0 | c) with c ≠ 0 appears. The system is inconsistent.
Exactly one solution: k = 0, meaning every variable column has a pivot. No free parameters.
Infinitely many solutions: k > 0 and the system is consistent. The solution set is a k-parameter family.
No other outcome is possible.
System:
x1 + 2x2 − x3 = 1
x1 + x2 + 2x3 = 0
5x1 + 8x2 + x3 = 1
Augmented matrix:
( 1 2 −1 1 )
( 1 1 2 0 )
( 5 8 1 1 )
After row reduction:
( 1 2 −1 0 )
( 0 −1 3 −1 )
( 0 0 0 −2 )
The last row says 0 = −2, which is a contradiction. This system has no solutions.
Change the third equation's right-hand side to 3:
x1 + 2x2 − x3 = 1
x1 + x2 + 2x3 = 0
5x1 + 8x2 + x3 = 3
After row reduction to RREF:
( 1 0 5 −2 )
( 0 1 −3 1 )
( 0 0 0 0 )
Column 3 has no pivot, so x3 is free. Set x3 = t.
x1 = −2 − 5t
x2 = 1 + 3t
Solution set: {(−2 − 5t, 1 + 3t, t) | t ∈ ℝ}
Rank = 2, free parameters = 1. This is a line in ℝ³.
Rank tells you the effective number of independent constraints in a system. In engineering, if you are modelling a structure with more unknowns than independent equations (rank < number of unknowns), the system is under-determined and you have design freedom. If the system is inconsistent (no solutions), your constraints are contradictory and the design is infeasible.
Students sometimes forget to check for inconsistency before writing the solution set. Always scan for a row of the form (0 0 ... 0 | c) with c ≠ 0 first.
Confusing "zero row" with "inconsistent row." A row (0 0 ... 0 | 0) is consistent (it says 0 = 0, which is fine). Only a nonzero constant on the right makes it inconsistent.
Thinking that more equations always means fewer solutions. Redundant equations (like the third equation in the consistent example above) can reduce to a zero row and add no new constraints.
Forgetting that free parameters range over all of ℝ. A 1-parameter solution is not "one solution," it is infinitely many.
⚠️ Being able to read off the solution set from RREF is a core exam skill. Practice until you can do it quickly and reliably.
⚠️ The three-cases classification (no solution, unique solution, infinitely many) appears frequently in both conceptual and computational questions.
⚠️ Watch for the difference between "the system is inconsistent" and "the system has a free variable." These are independent properties. A system can be inconsistent regardless of whether it would have had free variables.
⚠️ Rank questions often ask you to determine the rank without fully reducing the matrix. Counting pivots in REF (you do not need to go all the way to RREF) is sufficient.
True or false: If a system has more unknowns than equations, it must have infinitely many solutions.
Fill in the blank: The rank of a matrix equals the number of ______ in its REF.
True or false: A row of all zeros in the augmented matrix means the system is inconsistent.
Fill in the blank: Free variables correspond to columns in RREF that have no ______.
True or false: A system with rank equal to the number of variables and no inconsistent rows has exactly one solution.
Answers: 1. False (it could be inconsistent). 2. leading entries (pivots). 3. False (a zero row is consistent; 0 = 0 is true). 4. leading entry (pivot). 5. True.
Q: A 3×5 augmented matrix (3 equations, 4 variables) is row-reduced to RREF with 3 pivots and no inconsistent rows. How many free parameters does the solution have?
A: k = 4 − 3 = 1 free parameter. The solution is a 1-parameter family.
Q: You row-reduce a system and find the last nonzero row is (0 0 0 0 | 5). What can you conclude?
A: The system is inconsistent. That row represents the equation 0 = 5, which is impossible.
Q: In the RREF example with the solution set {(2 − 2t1 − t2, t1, 3 − 3t2, t2, 4)}, what is the rank of the coefficient matrix?
A: There are 5 variables and 2 free parameters, so the rank is 5 − 2 = 3.
Q: Can a homogeneous system (all right-hand sides equal to zero) be inconsistent?
A: No. The trivial solution (all variables equal to zero) always satisfies a homogeneous system. A homogeneous system is always consistent.
Rank reappears throughout the course. It connects to the dimension of the column space and the row space of a matrix, and later to the rank-nullity theorem. The classification of solution sets (no solution, unique, infinitely many) is a direct consequence of the structure of RREF and will be revisited when you study linear transformations and their kernels. The parametric form of solution sets is the first step toward understanding the null space of a matrix.
Related Terms / Search Tags: solution set, free variable, pivot, leading entry, rank, k-parameter solution, consistent system, inconsistent system, augmented matrix, RREF, Gaussian elimination, null space, homogeneous system, MATH 416, abstract linear algebra