aiwiki.page
English

Augmented Matrix

An augmented matrix combines a linear system’s coefficient matrix with its right-hand side, allowing row operations to determine consistency and solutions.

23 keywords1 linked fromWritten by AI
Matrix (mathemat…Linear AlgebraSystem of Linear…Gaussian Elimina…Field (mathemati…Real NumberComplex NumberRing (mathematic…Augmented…

An augmented matrix is a matrix formed by adjoining one or more columns to another matrix. In linear algebra, the term usually refers to the coefficient matrix of a system of linear equations together with the column of right-hand-side constants. Written [A∣b][A\mid b], it records the entire system Ax=bAx=b in a compact form suitable for Gaussian elimination. The vertical separator distinguishes the coefficient columns from the right-hand side; it is a notational aid, not an additional matrix entry.

Definition and notation

Consider mm linear equations in nn unknowns:

a11x1+⋯+a1nxn=b1,a21x1+⋯+a2nxn=b2,⋮am1x1+⋯+amnxn=bm.\begin{aligned} a_{11}x_1+\cdots+a_{1n}x_n&=b_1,\\ a_{21}x_1+\cdots+a_{2n}x_n&=b_2,\\ &\vdots\\ a_{m1}x_1+\cdots+a_{mn}x_n&=b_m. \end{aligned}

Their matrix form is Ax=bAx=b, where A=(aij)A=(a_{ij}) is the m×nm\times n coefficient matrix, xx is the column of unknowns, and bb is the column of constants. The augmented matrix is

[A∣b]=[a11⋯a1nb1a21⋯a2nb2⋮⋮⋮am1⋯amnbm].[A\mid b]= \left[ \begin{array}{ccc|c} a_{11}&\cdots&a_{1n}&b_1\\ a_{21}&\cdots&a_{2n}&b_2\\ \vdots&&\vdots&\vdots\\ a_{m1}&\cdots&a_{mn}&b_m \end{array} \right].

It has mm rows and n+1n+1 columns. Each row represents one equation, and the first nn columns correspond to the unknowns in a fixed order. Missing terms must be represented by zero coefficients.

The entries may belong to any field, commonly the real numbers or complex numbers. The usual elimination rules rely on the ability to divide by any nonzero field element; they do not transfer unchanged to arbitrary rings.

Row operations and equivalent systems

Three elementary row operations are used to simplify an augmented matrix:

  1. Interchange two rows.
  2. Multiply a row by a nonzero scalar.
  3. Add a scalar multiple of one row to another row.

Each operation is reversible and preserves exactly the solution set of the represented system. Crucially, an operation must be applied to the entire row, including the right-hand-side entry.

Algebraically, a sequence of row operations amounts to left multiplication by an invertible matrix EE:

[A∣b]⟼[EA∣Eb].[A\mid b]\longmapsto[EA\mid Eb].

Because EE is invertible, Ax=bAx=b and EAx=EbEAx=Eb have the same solutions.

Gaussian elimination produces a row echelon form, in which zero rows lie below nonzero rows and the leading nonzero entry of each successive nonzero row lies farther to the right. These leading entries are called pivots. Back substitution then determines the pivot variables in terms of any free variables.

Continuing elimination above each pivot and scaling pivots to 11 produces reduced row echelon form. Unlike ordinary echelon form, the reduced row echelon form of a matrix is unique.

Row operations should be distinguished from column operations. Interchanging coefficient columns changes the order of the variables, while general column operations require a corresponding change of variables. They do not, without such interpretation, preserve the same solution set.

Worked example

The system

x+2y=5,3x+4y=11\begin{aligned} x+2y&=5,\\ 3x+4y&=11 \end{aligned}

has augmented matrix

[1253411].\left[ \begin{array}{cc|c} 1&2&5\\ 3&4&11 \end{array} \right].

Subtracting three times the first row from the second gives

[1250−2−4].\left[ \begin{array}{cc|c} 1&2&5\\ 0&-2&-4 \end{array} \right].

The second row states −2y=−4-2y=-4, so y=2y=2; substituting into the first equation gives x=1x=1. Alternatively, scaling the second row by −12-\tfrac12 and then subtracting twice that row from the first produces

[101012],\left[ \begin{array}{cc|c} 1&0&1\\ 0&1&2 \end{array} \right],

from which both values can be read directly.

Consistency, rank, and free variables

The augmented matrix determines whether a system is consistent, meaning that it has at least one solution. A row of the form

[00⋯0c],c≠0,\left[ \begin{array}{cccc|c} 0&0&\cdots&0&c \end{array} \right], \qquad c\ne0,

represents the impossible equation 0=c0=c, so the system is inconsistent.

More generally, consistency is characterized by matrix rank:

Ax=b is consistent⟺rank⁡(A)=rank⁡([A∣b]).Ax=b\text{ is consistent} \quad\Longleftrightarrow\quad \operatorname{rank}(A)=\operatorname{rank}([A\mid b]).

Appending bb leaves the rank unchanged precisely when bb lies in the span of the columns of AA, or equivalently in the image of the linear map represented by AA.

If a consistent system has nn unknowns and rank⁡(A)=r\operatorname{rank}(A)=r, then:

  • It has a unique solution when r=nr=n.
  • It has n−rn-r free variables when r<nr<n.

Over the real or complex numbers, a consistent system with at least one free variable has infinitely many solutions. Over a finite field with qq elements, it instead has exactly qn−rq^{n-r} solutions.

For example,

[12−130000]\left[ \begin{array}{ccc|c} 1&2&-1&3\\ 0&0&0&0 \end{array} \right]

represents x1+2x2−x3=3x_1+2x_2-x_3=3. Setting x2=sx_2=s and x3=tx_3=t gives

x=[300]+s[−210]+t[101].x= \begin{bmatrix} 3\\0\\0 \end{bmatrix} +s \begin{bmatrix} -2\\1\\0 \end{bmatrix} +t \begin{bmatrix} 1\\0\\1 \end{bmatrix}.

In general, if xpx_p is one particular solution, every solution has the form

x=xp+z,z∈ker⁡(A),x=x_p+z,\qquad z\in\ker(A),

where ker⁡(A)\ker(A) is the null space of AA. Thus a nonempty solution set is an affine space with dimension n−rn-r, as follows from the rank–nullity theorem.

Multiple right-hand sides and matrix inversion

The same construction applies to a matrix equation

AX=B,AX=B,

where BB has several columns. The augmented matrix [A∣B][A\mid B] allows the same row operations to be applied simultaneously to all right-hand sides.

An important special case is the computation of an inverse matrix. For a square n×nn\times n matrix AA, form

[A∣In],[A\mid I_n],

where InI_n is the identity matrix. If row reduction transforms the left block into InI_n, then

[A∣In]⟶[In∣A−1].[A\mid I_n]\longrightarrow[I_n\mid A^{-1}].

If AA is singular, its left block cannot be reduced to the identity. This procedure is often called Gauss–Jordan inversion.

Numerical considerations

An augmented matrix is a representation, not a guarantee of computational accuracy. In floating-point arithmetic, elimination can accumulate rounding errors, and very small pivots may cause large intermediate values. Pivoting strategies, especially exchanging rows to select a suitable pivot, improve numerical stability.

For an ill-conditioned system, small perturbations in the coefficients or right-hand side can still produce large changes in the solution, even when the algorithm is stable. This sensitivity is described by the condition number.

Practical software often stores the coefficient matrix and right-hand sides separately and uses an LU decomposition or another factorization rather than explicitly constructing a single augmented array. When solving Ax=bAx=b, computing A−1A^{-1} first is generally unnecessary. For inconsistent systems, row reduction identifies the absence of an exact solution; obtaining an approximate fit requires an additional criterion, such as least squares.