Detecting Inconsistent Systems from Row Reduction

What Inconsistency Looks Like

A system of linear equations is consistent if at least one choice of its variables satisfies every equation. It is inconsistent if no such choice exists. Row reduction helps detect inconsistency by revealing a scalar equation that cannot be true. In particular, a row of zeros in the coefficient positions paired with a nonzero right-hand entry represents an equation such as \(0=1\).

For a system with \(m\) equations and \(n\) variables, its augmented matrix has \(n\) coefficient columns and one additional column for the right-hand sides. Each row represents one equation. A row of the form \([\,0\ \cdots\ 0\mid c\,]\), where \(c\ne0\), therefore says that a sum of zero multiples of the variables must equal \(c\). The left side is zero for every assignment of the variables, so that equation cannot hold.

The location of the nonzero entry matters. A row \([\,0\ \cdots\ 0\mid 0\,]\) represents \(0=0\), which is true for every assignment; it does not create a contradiction. The diagnostic pattern is specifically zeroes in all coefficient positions and a nonzero entry in the right-hand position.

Why Row Reduction Preserves Consistency

Row reduction applies elementary row operations to the augmented matrix. These operations are swapping two rows, multiplying a row by a nonzero scalar, and adding a multiple of one row to another. To use a row in the reduced matrix as evidence about the original system, we need to know that these operations preserve whether a solution exists.

Proof. Swapping two rows only changes the order in which the equations are listed, so a vector of variable values satisfies the system before the swap exactly when it satisfies the system after the swap. Multiplying a row by a nonzero scalar replaces an equation \(L=c\) by \(dL=dc\), where \(d\ne0\). The equations have the same solutions because \(L=c\) implies \(dL=dc\), and \(dL=dc\) implies \(L=c\) after division by \(d\). Finally, adding a multiple \(d\) of one equation \(L_2=c_2\) to another \(L_1=c_1\) replaces the latter by \(L_1+dL_2=c_1+dc_2\). Any solution of the original equations satisfies this new equation. Conversely, if the unchanged second equation \(L_2=c_2\) and the new equation hold, subtracting \(dL_2=dc_2\) recovers \(L_1=c_1\). Thus each operation preserves precisely the same solutions. \(\square\)

Consequently, a contradiction found after row reduction is a contradiction for the original system as well. For example, if an original system had a solution, the proposition says that the same variable values would satisfy every system obtained by row operations. They could not satisfy a row that says \(0=1\).

Proof. First suppose the reduced matrix contains \([\,0\ \cdots\ 0\mid c\,]\) with \(c\ne0\). Its equation is \(0=c\), which no variable values satisfy. By the proposition, the row operations preserve the solution set, so the original system is inconsistent.

For the converse, suppose row echelon form has no such row. Every nonzero row must then have a leading coefficient in one of the variable columns: if its coefficient entries were all zero, its right-hand entry would also have to be zero. Assign zero to each variable in a non-pivot column. Starting from the last nonzero row and moving upwards, solve each row for its leading variable. That variable has a nonzero coefficient, and all variables to its right already have assigned values, so the row determines a value for it. This gives values for every pivot variable. Any rows below the nonzero rows are zero rows and impose only \(0=0\). Thus the reduced system has a solution. The proposition then gives a solution to the original system, so it is consistent. Reducing to RREF is also a sequence of elementary row operations and has the same criterion. \(\square\)

The theorem gives a complete test, not just a one-way warning: a contradictory row guarantees inconsistency, and if row echelon form has no contradictory row, back-substitution can produce a solution. The test concerns the augmented matrix, so always distinguish its coefficient columns from its final right-hand column.

Worked Examples

Common Mistake: Treating Every Zero Row as a Contradiction

A row of zeroes in the coefficient positions is not enough to conclude that a system is inconsistent. The right-hand entry must be checked too. For example, consider \(x+y=2\) and \(2x+2y=4\). Subtracting twice the first row from the second gives

A wrong conclusion would be: “The bottom row has only zeroes, so the system is inconsistent.” The correction is to translate that row into its equation: \(0x+0y=0\), or \(0=0\). This equation is true, not contradictory. For instance, \(x=2,y=0\) satisfies both original equations: \(2+0=2\) and \(2(2)+2(0)=4\).

In short, read the entire augmented row. A row with zero coefficients and a nonzero right-hand side is a contradiction; a row with zeroes in every position is a redundant equation that imposes no restriction. A nonzero coefficient row, meanwhile, expresses a genuine equation and should not be mistaken for either case.

A Reliable Detection Procedure

When checking a system for inconsistency, keep the interpretation of the augmented matrix visible throughout the calculation. Each row operation changes the equations but preserves their solution set, so the final rows can be read as an equivalent system. The following procedure isolates the relevant test:

This test is useful because it turns a question about all possible variable values into a visible check of a finite matrix. It does not require finding every solution: a single contradictory row rules out all solutions. Conversely, the theorem shows that if the reduced system contains no contradictory row, back-substitution can produce at least one solution. The essential habit is to translate the row back into its scalar equation before drawing a conclusion.