We want a way to solve a system of linear equations directly from determinants, without having to row-reduce the matrix.
1
Why Cramer's Rule Works
Suppose we want to solve the system
3x+y=7
x+2y=4
Instead of combining rows algebraically, we can read the system by its columns, c1=[31] and c2=[12], and ask what stretch factors x and y make xc1+yc2 land on the target vector b=[74].
Remember that when we put two column vectors into a 2×2 matrix A=[acbd], the determinant det(A)=ad−bc gives the signed area of the parallelogram formed by those columns.
Why Cramer's Rule Works
Text solution
Original parallelogram areaPut the system's columns into the matrix A=[3112] and find the signed area of their parallelogram.
det(A)=det[3112]=(3)(2)−(1)(1)=5
Replace column 1 with b to find xBecause sliding xc1 by yc2 parallel to c2 does not change the parallelogram's area, we replace the first column of A with b=[74] and divide by det(A).
x=det(A)det[7412]=5(7)(2)−(1)(4)=510=2
Replace column 2 with b to find yBy the same parallel slide along c1, we replace the second column of A with b and divide by det(A).
y=det(A)det[3174]=5(3)(4)−(7)(1)=55=1
Stretching the first column by x=2 and the second column by y=1 makes the two vectors add tip-to-tail to reach [74].
In the parallelogram below, sliding the target vector b parallel to c2 changes the stretch factor y while keeping the base c2, the perpendicular height, and Area(b,c2) unchanged, so dividing Area(b,c2) by Area(c1,c2) always gives x.
Sliding the target vector along one column vector changes that column vector's stretch factor without changing the area between the target vector and that column vector, so dividing that area by the original area between the two column vectors always gives the other column vector's stretch factor.
Whenever det(A)=0, a system A[xy]=b with columns c1 and c2 has one unique solution given by Cramer's Rule:
In 3D, determinants measure volume instead of area, and the exact same column-replacement rule works.
If det(A)=0, the denominator is zero and the original parallelogram is squished flat onto a single line, so there is no unique solution and Cramer's Rule cannot be used.
det(A)=0
no solution
infinitely many solutions
2
Systems with an Unknown Parameter
When a matrix is full of plain numbers, row reduction is fast. Cramer's Rule becomes much easier than row reduction when an unknown parameter k is in the matrix, where row-reducing would immediately force us to divide by expressions with k.
Let's find which values of k give the system
kx+2y=3
2x+ky=1
a unique solution, and then solve for x and y in terms of k.
Systems with an Unknown Parameter
kx+2y2x+ky=3=1
one solution for which k?
x=?y=?
Text solution
Set det(A)=0 for a unique solutionThe matrix A=[k22k] is invertible and gives a unique solution when its determinant is not zero.
det(A)=(k)(k)−(2)(2)=k2−4=0⟹k=2 and k=−2
Replace column 1 with [31] to find xFor k=±2, we put the target vector into the first column and divide by det(A).
x=k2−4det[312k]=k2−43k−2
Replace column 2 with [31] to find yPutting the target vector into the second column and dividing by det(A) gives y.
y=k2−4det[k231]=k2−4k−6
For every value of k except 2 and −2, the system has the unique solution x=k2−43k−2 and y=k2−4k−6.
3
From Cramer's Rule to the Adjugate
Cramer's Rule solves Ax=b for one target vector b at a time. If we want a single inverse matrix A−1 that takes any target vector b=b1b2b3 and immediately gives x=x1x2x3, we can leave b1,b2,b3 as unknown inputs inside Cramer's Rule and factor out the matrix multiplying b. That transposed grid of cofactors is the adjugate of A, written adj(A).
Remember that to expand a 3×3 determinant down a column, we multiply each entry by its cofactorCij: cross out row i and column j, take the 2×2 determinant left over, and apply the sign from the checkerboard grid +−+−+−+−+.
From Cramer's Rule to the Adjugate
Expanding the Cramer's Rule numerator for x1 down the first column separates the inputs b1,b2,b3 from the first column's cofactors, which depend only on A:
Because matrix multiplication multiplies rows by columns, we have to turn each column of cofactors sideways into a row so that it multiplies b1b2b3:
On the main diagonal, row 1 of adj(A) multiplies column 1 of A, pairing the first column's cofactors with the first column's own entries to give the cofactor expansion of det(A):
C11a11+C21a21+C31a31=det(A)
Off the main diagonal, row 1 of adj(A) multiplies column 2 of A, pairing column 1's cofactors with column 2's entries. That is the determinant of a matrix whose first two columns are identical, which forms a 3D box flattened into a plane of volume 0:
Putting the diagonal and off-diagonal entries together gives an equation that holds for every matrix A, even when det(A)=0:
adj(A)⋅A=det(A)000det(A)000det(A)=det(A)I
In the product adj(A)⋅A below, pairing row i of adj(A) with the matching column i of A keeps the full 3D box of volume det(A), while pairing row i with a different column j replaces column i with a copy of column j and flattens the box into a plane of volume zero.
When multiplying the adjugate by the matrix, pairing a row of cofactors with its matching column keeps the full 3D box volume, while pairing it with any other column repeats an edge and flattens the box to zero volume.
5
Finding the Adjugate and Inverse
Let's cross out rows and columns to find the cofactor grid, adj(A),det(A), and A−1 for the matrix
A=101210021
Finding the Adjugate and Inverse
A=101210021
adj(A)A−1=?=?
Text solution
Find the nine 2×2 leftover determinantsCover each row and column of A one spot at a time and take the 2×2 determinant of the four entries left uncovered.
Flip the four minus spots to get the cofactor gridApply the checkerboard signs +−+−+−+−+ by flipping the signs of the four edge entries.
Cofactors=1−2421−2−121
Multiply row 1 of A by row 1 of the cofactors to get det(A)Before transposing, multiply the top row of A,[120], by the top row of the cofactor grid, [12−1], to get the determinant right away.
det(A)=(1)(1)+(2)(2)+(0)(−1)=5
Transpose the cofactor grid to get adj(A)Turn each column of the cofactor grid into a row.
adj(A)=12−1−2124−21
Divide by det(A) to get A−1Since det(A)=5=0, divide the adjugate matrix by 5.
A−1=5112−1−2124−21
We now have the complete inverse matrix A−1, and applying these same steps (covering a row and column, applying the sign, and transposing) to a 2×2 matrix [acbd] gives the familiar 2×2 inverse formula A−1=ad−bc1[d−c−ba].
Across the nine entries of A below, covering row i and column j leaves a 2×2 determinant whose signed cofactor always lands in the transposed position (row j, column i) of adj(A) and A−1.
Covering a row and column of the starting matrix gives a signed smaller determinant that goes into the swapped row and column of the adjugate and inverse.
6
Finding One Entry of an Inverse
What if a question asks for just one entry of A−1 instead of the whole inverse? Row reduction would force us to reduce the entire matrix, whereas the adjugate formula needs only det(A) and a single cofactor.
Let's find the entry in row 2, column 3 of A−1 for the same matrix
A=101210021
using det(A)=5.
Finding One Entry of an Inverse
(A−1)2,3=?
A=101210021
?
A−1
Text solution
Flip the indices to pick the cofactorBecause adj(A) is the transpose of the cofactor grid, the entry in row 2, column 3 of A−1 comes from row 3, column 2 of the cofactors.
(A−1)2,3=det(A)C3,2
Cover row 3 and column 2 of A and apply the signCrossing out row 3 and column 2 of A leaves a 2×2 determinant in a minus spot of the sign grid.
C3,2=−det[1002]=−((1)(2)−(0)(0))=−2
Divide by det(A)Divide C3,2=−2 by det(A)=5.
(A−1)2,3=−52
The entry in row 2, column 3 of A−1 is −52, which matches row 2, column 3 of our full inverse above while needing only one cofactor instead of nine.
Key concepts
Both Cramer's Rule and the adjugate inverse formula divide by det(A), so they work only when det(A)=0. If det(A)=0, the columns flatten onto a line or plane and there is no unique solution or inverse; and if you need to solve a larger system made of plain numbers, ordinary row reduction is faster.
When a 2×2 system gives you parallelogram areas directly or has an unknown parameter k (like [k22k][xy]=[31], where row reduction would divide by messy expressions with k), use Cramer's Rule by checking det(A)=k2−4=0 and replacing one column at a time with b:
Expanding Cramer's Rule down each column puts the signed leftover determinants Cij (with signs +−+−+−+−+) into the transposed cofactor matrix adj(A), giving the inverse formula
A−1=det(A)1adj(A)
If a problem asks for only one entry of A−1, skip the other cofactors and flip the indices right away: to find row 2, column 3 of A−1, cover row 3 and column 2 of A to get (A−1)2,3=det(A)C3,2. And whenever cofactors multiply matrix entries directly, use adj(A)⋅A=det(A)I: multiplying a column's cofactors by its own entries gives det(A) (such as a11C11+a21C21+a31C31=det(A)), while multiplying them by a different column's entries gives 0.
7
Practice
1
Use Cramer's Rule to solve the system 4x−y=5,2x+3y=13.
Which column of [42−13] gets replaced by [513] when you solve for x?
Find the determinant of the coefficient matrix A=[42−13], being careful to subtract (−1)(2)=−2 rather than dropping the minus sign on −1 (which would give 12−2=10 instead of 14):
det(A)=det[42−13]=(4)(3)−(−1)(2)=14
Replace column 1 with [513] to find x, and replace column 2 with [513] to find y:
x=14det[513−13]=14(5)(3)−(−1)(13)=1428=2
y=14det[42513]=14(4)(13)−(5)(2)=1442=3
2
Two column vectors c1 and c2 form a parallelogram with signed area Area(c1,c2)=6. A target vector b=xc1+yc2 forms a parallelogram with c2 of signed area Area(b,c2)=−18, and a parallelogram with c1 of signed area Area(c1,b)=15. Find x and y.
In Area(b,c2), the target b replaced the first column c1, so do not pair Area(b,c2) with y just because c2 is still in the parentheses (which would swap the answers to x=25 and y=−3).
Look at which original column was replaced by b inside Area(b,c2) and Area(c1,b).
In Area(b,c2), the target b replaced the first column c1, so do not pair Area(b,c2) with y just because c2 is still in the parentheses (which would swap the answers to x=25 and y=−3):
x=Area(c1,c2)Area(b,c2)=6−18=−3
y=Area(c1,c2)Area(c1,b)=615=25
3
Consider the system kx+3y=6,3x+ky=2. (a) For what values of k does this system have a unique solution? (b) For those values of k, use Cramer's Rule to find x and y in terms of k.
Set the determinant of the coefficient matrix not equal to zero, remembering both 3 and −3 when solving k2=9 so you don't miss k=−3 and write only k=3.
(a) What must be true about det[k33k] for the system to have a unique solution? (b) Replace one column at a time with [62].
(a) Set the determinant of the coefficient matrix not equal to zero, remembering both 3 and −3 when solving k2=9 so you don't miss k=−3 and write only k=3:
det(A)=det[k33k]=k2−9=0⟹k=3 and k=−3
(b) Replace the first and second columns with [62] and divide by k2−9:
x=k2−9det[623k]=k2−96k−6
y=k2−9det[k362]=k2−92k−18
4
Let A=210011121. Find the cofactor matrix, adj(A),det(A), and A−1.
After finding the nine 2×2 leftover determinants, apply the checkerboard signs +−+−+−+−+ before transposing.
Covering each row and column gives the 2×2 leftover determinants −1−1−1123122. Flip the signs on the four minus spots of +−+−+−+−+ to get the cofactor matrix:
Cofactors=−11−1−12−31−22
Transpose the cofactor matrix by turning its columns into rows (forgetting to transpose here would leave rows and columns swapped and give the wrong inverse 1−111−23−12−2):
adj(A)=−1−1112−2−1−32
Multiply row 1 of A by row 1 of the cofactor matrix to find det(A), and divide adj(A) by det(A):
det(A)=(2)(−1)+(0)(−1)+(1)(1)=−1
A−1=−11−1−1112−2−1−32=11−1−1−2213−2
5
The matrix A=301124213 has det(A)=3. Without finding the whole inverse, find the entry in row 1, column 2 of A−1.
Flip the indices first so you cover row 2, column 1 of A rather than row 1, column 2 (covering row 1, column 2 without flipping would give C1,2=1 and the wrong answer 31).Covering row 2 and column 1 leaves [1423]; apply the minus sign from the checkerboard grid at row 2, column 1 (don't leave it off and get −35).
Because adj(A) is transposed, which row and column of A should you cover to get row 1, column 2 of A−1?
Flip the indices first so you cover row 2, column 1 of A rather than row 1, column 2 (covering row 1, column 2 without flipping would give C1,2=1 and the wrong answer 31):
(A−1)1,2=det(A)C2,1
Covering row 2 and column 1 leaves [1423], and applying the minus sign from the checkerboard grid at row 2, column 1 (don't leave it off and get −35) gives:
C2,1=−det[1423]=−((1)(3)−(2)(4))=5
(A−1)1,2=35
6
Let A=2−1453−2106. The cofactors of the first column of A are C11=18,C21=−32, and C31=−3. Without finding any other cofactors, find: (a) det(A) (b) the value of 5C11+3C21−2C31 (c) the matrix product adj(A)⋅A
Expand down column 1 by multiplying column 1's entries (2,−1,4) by column 1's cofactors (be careful not to multiply column 1's cofactors by row 1's entries 2,5,1, which would give −127).
Compare the numbers multiplying the cofactors in (a) and (b) to the columns of A, and remember what happens when column 1's cofactors multiply column 1 versus column 2.
(a) Expand down column 1 by multiplying column 1's entries (2,−1,4) by column 1's cofactors (be careful not to multiply column 1's cofactors by row 1's entries 2,5,1, which would give −127):
det(A)=2(18)+(−1)(−32)+4(−3)=56
(b) The numbers 5,3,−2 are the entries of column 2, and multiplying column 1's cofactors by column 2's entries gives the determinant of a matrix whose first two columns are identical:
5C11+3C21−2C31=0
(c) Multiply det(A)=56 by the 3×3 identity matrix I so the result is a 3×3 matrix rather than the single number 56: