Why Small Row-Reduction Errors Matter
Row reduction is a sequence of simple operations, but a single incorrect step can change the system being solved. A sign error may produce a plausible-looking solution that fails the original equations. An omitted right-hand-side calculation can silently replace an equation with a different one. Even a correct reduced matrix can be read incorrectly if a free variable is mistaken for a pivot variable or a contradiction row is overlooked.
The safest way to diagnose a reduction is to ask three questions at each stage: Was the operation allowed? Was it applied to every entry in the row? And what do the coefficient columns say about pivots, free variables, and consistency? In this tutorial, an augmented matrix represents a system of linear equations, and row operations are applied to the entire augmented row, including its rightmost column.
Multiplying by zero is not an allowed row operation: it can erase an equation. Adding a multiple of one column to another is also not a row operation. In particular, column operations on an augmented matrix can change which coefficients belong to which variables. The reason for being strict about these rules is that the purpose of elimination is to produce an equivalent system, one with exactly the same solution set.
Proof. A row interchange only changes the order of two equations, so a vector satisfies the system before the interchange exactly when it satisfies the system afterward. If an equation \(L=b\) is multiplied by a nonzero scalar \(c\), then \(L=b\) implies \(cL=cb\); conversely, \(cL=cb\) implies \(L=b\) after division by \(c\). Finally, suppose the equation \(L_i=b_i\) is replaced by \(L_i+cL_j=b_i+cb_j\), while \(L_j=b_j\) remains. Any solution of the original equations satisfies the new one by addition. Conversely, a solution of the new equation and the unchanged equation \(L_j=b_j\) satisfies \(L_i=b_i\), by subtracting \(cL_j=cb_j\). Thus each operation preserves precisely the same solutions. \(\square\)
Applying an Operation to the Whole Augmented Row
Worked Example: The Right-Hand Side Was Left Behind
Solve the system represented by the augmented matrix below. A common mistake is to subtract \(2R_1\) from \(R_2\) in the coefficient entries but forget to do so in the right-hand-side entry.
The intended operation is \(R_2\to R_2-2R_1\). It must be applied to all three entries of row 2:
The second row now gives \(-3y=-6\), so \(y=2\). Dividing row 2 by \(-3\), then using \(R_1\to R_1-R_2\), gives
The solution is \(x=1,\ y=2\). Check: the original equations are \(x+y=3\) and \(2x-y=0\). Substitution gives \(1+2=3\) and \(2(1)-2=0\).
The incorrect version leaves the right-hand side of row 2 at \(0\), producing \([0\ \ -3\mid 0]\). That row says \(y=0\), and the first row then gives \(x=3\). But \(2x-y=2(3)-0=6\ne0\), so \((3,0)\) does not solve the original system. The correction is not to adjust the answer afterward; it is to perform the row operation on the full augmented row.
Arithmetic Errors: Recompute the Entire New Row
A legal operation can still be carried out incorrectly. When a row is replaced by a difference, calculate each entry as “old entry minus the specified multiple,” with its sign intact. Do not infer the sign of the new entry from the sign of the old one alone.
Worked Example: A Sign Error Changes the Answer
Consider the system with augmented matrix shown below. The first equation is \(x+2y=5\), and the second is \(3x+y=7\).
Entry by entry, the new second row is \((3-3\cdot1,\ 1-3\cdot2,\ 7-3\cdot5)=(0,-5,-8)\). In particular, \(7-15=-8\), not \(8\). Scaling row 2 by \(-\tfrac15\) gives \([0\ \ 1\mid\tfrac85]\), and then \(R_1\to R_1-2R_2\) gives
Thus \(x=\tfrac95,\ y=\tfrac85\). Check: \(x+2y=\tfrac95+\tfrac{16}{5}=\tfrac{25}{5}=5\), and \(3x+y=\tfrac{27}{5}+\tfrac85=\tfrac{35}{5}=7\).
If the sign error changes \(-8\) to \(8\), the second equation in the reduced system becomes \(-5y=8\), leading to \(y=-\tfrac85\) and \(x=\tfrac{41}{5}\). The first original equation happens to hold for those values, since \(\tfrac{41}{5}+2(-\tfrac85)=5\), but the second does not: \(3(\tfrac{41}{5})-\tfrac85=\tfrac{115}{5}=23\ne7\). This is why checking every original equation matters: satisfying one equation does not establish that a proposed vector solves the system.
Do Not Miss a Contradiction Row
A row with zero coefficients and a nonzero right-hand side represents an impossible equation, such as \(0=1\). It is not a row of zeros that can be ignored. Earlier material on the existence criterion for solutions establishes that such a row means the system is inconsistent. Here the practical error to watch for is reading only the coefficient entries and overlooking the augmented entry.
Worked Example: A Zero Coefficient Row Is Not Always Harmless
Reduce the augmented matrix for \(x+y=2\) and \(2x+2y=5\). Subtract twice row 1 from row 2, applying the operation to the right-hand side as well:
The last row says \(0x+0y=1\), or \(0=1\), which is false. Therefore the system has no solution. This conclusion can also be checked against the original equations: doubling \(x+y=2\) would give \(2x+2y=4\), but the second equation requires \(2x+2y=5\).
A mistaken reading is to see the two zero coefficient entries and call the row “all zeros,” then assign a variable freely. The row is not all zeros: its right-hand side is \(1\). Free variables arise from nonpivot coefficient columns in a consistent system; a contradiction row makes the entire system inconsistent before any assignment of free variables can produce a solution.
Read Pivot and Free Variables from Coefficient Columns
Proof. A pivot is the leading nonzero entry of a nonzero row in echelon form. A coefficient column containing a pivot identifies the variable solved for by that row; a coefficient column without a pivot does not lead any row, so its variable is free. To show that every free-variable assignment determines pivot variables, start at the bottommost nonzero row. Because there is no contradiction row, that row has a pivot in a coefficient column, and its equation determines its pivot variable once all variables to the right have been assigned. Proceed upward one row at a time: at each pivot row, the variables to its right have already been assigned or determined, so its equation determines the pivot variable. This gives values for every pivot variable. The zero rows impose only \(0=0\), because the stated condition excludes a nonzero right-hand side in such a row. Finally, each pivot variable is uniquely determined by its row once the variables to its right are fixed, so no second assignment of pivot-variable values can accompany the same free-variable assignment. \(\square\)
This result depends on the matrix being in echelon form and the system being consistent. It does not say that every variable in a row-echelon system is free, nor that a variable is free merely because its coefficient is zero in one row. Look for pivots in the coefficient columns of the entire matrix. Do not count a leading entry in the augmented column as a pivot variable.
Worked Example: The Middle Variable Is Free
Read the solutions from this echelon augmented matrix for variables \(x_1,x_2,x_3\):
The pivots are in coefficient columns 1 and 3. Column 2 has no pivot, so \(x_2\) is free. Let \(x_2=t\), where \(t\in\mathbb{R}\). The first row gives \(x_1+2t=3\), hence \(x_1=3-2t\); the second row gives \(x_3=4\). Thus the solutions are
Check: substituting into the first row gives \((3-2t)+2t=3\), and substituting into the second gives \(4=4\), for every real \(t\). A common incorrect assignment is to call \(x_3\) free because it appears to the right of the missing pivot in column 2. But column 3 contains a pivot, and its row fixes \(x_3=4\). The variable’s position in the list does not determine whether it is free; the pivot columns do.
A Practical Error-Diagnosis Routine
When a worked reduction looks suspicious, inspect it in a fixed order. This helps distinguish an illegal operation from an arithmetic slip, and both from an error in interpreting the final form.
Check that each step is a row interchange, a nonzero row scaling, or the addition of a multiple of one row to another. Reject division by zero and operations on columns.
Include the augmented entry and keep parentheses around negative numbers while multiplying or subtracting. For example, \(7-3(5)=-8\).
A row with zero coefficient entries and a nonzero right-hand side is a contradiction, not a free-variable equation.
Nonpivot coefficient columns identify free variables, provided no contradiction row is present.
Substitute into every original equation. A failure pinpoints that either the reduction or the reading of its result needs correction.
This routine is not a substitute for doing the reduction carefully. It is a way to locate the source of a disagreement: an operation must preserve the solution set, arithmetic must produce the claimed row, and the final interpretation must respect both the coefficient columns and the augmented column.
Common Mistakes at a Glance
| Mistake | Why it fails | Correction |
|---|---|---|
| Changing coefficients but not the right-hand side | The resulting row represents a different equation. | Apply the operation to every entry of the augmented row. |
| Scaling a row by zero | It can erase an equation and change the solution set. | Scale only by a nonzero scalar. |
| Dropping a minus sign in elimination | The new system may have different solutions. | Recalculate each entry and substitute the answer into the original equations. |
| Ignoring a nonzero augmented entry in a zero coefficient row | The system contains a false equation \(0=b\), with \(b\ne0\). | Record that the system is inconsistent; do not assign free variables. |
| Calling a nonpivot variable a pivot variable | It leads to an incorrect parametrization of the solution set. | Locate pivots in the coefficient columns, then assign parameters to the nonpivot columns. |
Check Your Understanding
For each question, identify the relevant error or rule and give a reason for your conclusion.
- In a row replacement \(R_2\to R_2-4R_1\), which entries of an augmented matrix must be recalculated?
- Why is \(R_1\to 0R_1\) not an allowed elementary row operation?
- What does a row \([0\ \ 0\mid -3]\) say about the system, and why?
- In an echelon augmented matrix for four variables, the coefficient pivots are in columns 1, 2, and 4, and there is no contradiction row. Which variable is free?
- A proposed solution satisfies the first two equations of a three-equation system but not the third. Is it a solution of the system? Explain.
