When a matrix transforms space, we want to know whether we can undo it with an inverse matrix, and if we can, how to find that inverse.
1
Why a zero determinant means no inverse
An inverse matrix A−1 undoes a transformation by sending every point on the transformed grid back to the single point it came from.
To see when that is possible for a 2×2 matrix, remember how the determinant tracks the signed area of the parallelogram formed by its two column arrows: the main diagonal stretches along the axes, while the cross diagonal tilts the arrows inward or outward.
det[acbd]=ad−bc
Watch what happens to the grid, and to our ability to trace points back to where they started, when the two column arrows line up so those two terms cancel to zero.
Why a zero determinant means no inverse
det(A)=6 (invertible)
det(A)=0 (singular)
As long as the column arrows leave non-zero area between them, every point on the new grid comes from one specific starting point, so A−1 exists. Once the arrows line up on a single line (or in 3D, when three column arrows flatten a 1×1×1 cube into a plane or line of volume zero), whole lines of points pile onto the same spot and cannot be pulled back apart:
Invertible⟺det(A)=0singular (no inverse)det(A)=0
Move the column arrow yourself to see the determinant turn negative when one arrow crosses past the other, and hit zero right when the arrows line up and flatten the grid.
Swing one column arrow across the other to see the grid flip over, and stop where the two arrows line up to see the whole grid flatten onto a single line that cannot be undone.
2
Finding values that make a matrix singular
We often want to find which values of an unknown entry k make a matrix singular, meaning it has no inverse:
Find k such that[234k] and [k82k]are singular.
In [234k], you can spot k=6 right away by noticing that the second column [46] is just double the first column [23]. When k shows up in more than one spot, setting the determinant to zero finds every value cleanly.
Finding values that make a matrix singular
[234k][k82k]
singular when k=?
Text solution
Set the determinant of [234k] to zeroThe matrix has no inverse where the main stretch and cross tilt cancel out to give zero area.
det[234k]=2k−12=0
Solve for kIsolate k to find the value where the two columns line up.
2k=12=2(6)⟹k=6
So [234k] is singular when k=6, and invertible for every other number.
Now for [k82k], where k appears in both corners of the main diagonal:
Set the determinant to zeroMultiply the main diagonal and subtract the cross diagonal to get an equation with k2.
det[k82k]=k2−16=0
Take both square rootsSolve for both the positive and negative values of k that flatten the square.
k2=16⟹k=4 or k=−4
So [k82k] has no inverse at k=4 (where the columns point in the same direction) and at k=−4 (where the columns point in opposite directions along the same line).
3
The determinant of an inverse matrix
When det(A)=0, the inverse matrix A−1 exists, and we can find its area or volume scaling factor det(A−1) right away without knowing any of the numbers inside A−1.
Remember that applying one transformation after another multiplies their scaling factors together: det(AB)=det(A)det(B). Applying A and then immediately applying A−1 returns every arrow to the starting identity matrix I=[1001], which leaves area unchanged with a scaling factor of 1.
The determinant of an inverse matrix
Because the two scaling factors must multiply to 1, the determinant of the inverse is the reciprocal of the original determinant:
det(A)det(A−1)⟹det(A−1)=det(I)=1=det(A)1
Notice what the −1 means in each place: the −1 on A−1 names the inverse transformation that undoes A (not 1 divided by a matrix), whereas det(A) is a single number so det(A−1) really is the fraction det(A)1.
Try changing the height of A's second column to see how A−1 scales area by the reciprocal det(A)1: when A shrinks area below 1,A−1 stretches it above 1, and as det(A) gets close to zero, A−1 has to stretch space enormously until the inverse disappears at det(A)=0.
To bring the parallelogram back to the starting square, the inverse scales area by the reciprocal, stretching huge as the shape flattens toward zero and vanishing when it collapses flat.
4
Determinants of products with inverses
We can now find the determinant of a chain of multiplied matrices and inverses even when we only know how each matrix scales space.
Suppose A and B are 3×3 matrices where A scales volumes by det(A)=5 and B scales volumes by det(B)=−2. Find det(A−1),det(A−1B), and det(B−1AB).
Determinants of products with inverses
det(A)det(B)=5=−2
det(A−1)det(A−1B)det(B−1AB)=?=?=?
Text solution
Take the reciprocal to find det(A−1)Since A stretches volume by 5, undoing A shrinks volume by 51.
det(A−1)=det(A)1=51
Multiply the scaling factors for det(A−1B)Track what happens to a box by multiplying the volume scaling factor of A−1 by the volume scaling factor of B.
det(A−1B)=det(A−1)det(B)=51×(−2)=−52
Multiply and cancel for det(B−1AB)Multiply all three scaling factors in a row, where B−1 scales by −21 and cancels the −2 scaling of B.
det(B−1AB)=det(B−1)det(A)det(B)=(−21)×5×(−2)=5
So A−1 scales volumes by 51,A−1B scales volumes by −52, and B−1AB scales volumes by 5 because B and B−1 completely cancel each other's scaling out.
5
Building the inverse of a 2 by 2 matrix
When we have the entries of an invertible 2×2 matrix A=[acbd] in front of us, the determinant shows us how to build the entire inverse matrix A−1 from scratch.
Watch how we undo both the axis stretches and the sideways tilts of A=[3214] to return the transformed grid to the starting square [1001].
Building the inverse of a 2 by 2 matrix
Text solution
Find the determinantMultiply the main diagonal and subtract the cross diagonal to see how much A scales area.
det(A)=3(4)−1(2)=12−2=10
Swap the main diagonal and negate the off-diagonal tiltsSwap 3 and 4 to balance the X and Y stretches, and flip the signs on 1 and 2 to cancel the tilts into det(A) times the identity matrix.
[3214][4−2−13]=[100010]=10[1001]
Divide by the determinantDivide every entry by 10 to scale the square back down to 1×1.
A−1=101[4−2−13]=[0.4−0.2−0.10.3]
So A−1=[0.4−0.2−0.10.3] un-tilts the parallelogram and shrinks its area by 101 back to the starting square.
For any invertible 2×2 matrix, multiplying by the swapped-and-negated matrix cancels the tilts to leave det(A)I, and dividing by det(A) gives the inverse:
⟹A−1[acbd]swap main diagonal,flip off-diagonal signs[d−c−ba]=[det(A)00det(A)]=det(A)1[d−c−ba]
Drag and let go below to see [d−c−ba] un-tilt A's parallelogram into an upright square of side det(A), and det(A)1 scale that square back down to 1×1.
Swapping the main diagonal and flipping the signs on the other diagonal un-tilts the parallelogram into an upright square whose side equals the determinant, so dividing by the determinant scales it right back to the unit square.
6
Checking invertibility and finding a 2 by 2 inverse
Now let's put the determinant check and the 2×2 inverse formula together on a new matrix:
For A=[5723], is A invertible, what is det(A−1), and what is A−1?
Checking invertibility and finding a 2 by 2 inverse
A=[5723]
A invertible?
det(A−1)A−1=?=?
Text solution
Check the determinant and its reciprocalCompute det(A) first to make sure the grid does not collapse to area zero.
det(A)⟹det(A−1)=5(3)−2(7)=15−14=1=0=11=1
Swap the main diagonal, negate the tilts, and divide by det(A)Swap 5 and 3, flip the signs on 2 and 7, and divide by 1.
A−1=11[3−7−25]=[3−7−25]
So A is invertible, det(A−1)=1, and A−1=[3−7−25] returns every point on the transformed grid to where it started while leaving areas unchanged.
You can check a 2×2 inverse in a few seconds by multiplying A by A−1 and making sure you get the identity matrix. For example, the top row of A times the second column of A−1 is 5(−2)+2(5)=0, so that sign flip actually canceled the tilt.
Key concepts
Everything here comes down to tracking what happens to area or volume and undoing it:
Checking whether an inverse exists (singular vs. invertible)
A matrix has an inverse if and only if it does not flatten space to zero area or volume (det(A)=0). To find what value of k makes a matrix singular or non-invertible, set det(A)=0 and solve. For [k82k],k2−16=0 gives k=±4. For a 3×3 matrix, do not try to use the 2×2 rule ad−bc; expand its 3×3 determinant and set it equal to zero.
Finding the determinant of an inverse or product
You do not need to compute the entries of A−1 to find det(A−1) or det(A−1B). Just replace each inverse with the reciprocal scaling factor det(A−1)=det(A)1 and multiply the scaling factors together: if det(A)=5 and det(B)=−2, then det(A−1B)=51(−2)=−52 and det(B−1AB)=det(A)=5.
Building a 2×2 inverse
When det(A)=ad−bc=0, swap the main diagonal to balance the stretches, negate the cross diagonal to cancel the tilts (which gives A[d−c−ba]=det(A)I), and divide by det(A):
A−1[5723]−1=ad−bc1[d−c−ba], so =11[3−7−25]
7
Practice
1
Let A=[42−3−1]. Is A invertible? If so, find det(A−1) and A−1.
Watch the double negative when subtracting (−3)(2): dropping a minus sign gives −4−6=−10, which throws off both det(A−1)=−101 and every entry of A−1.
First check whether the parallelogram has non-zero area: what does the main stretch minus the cross tilt give?
Compute the determinant of A:
det(A)=4(−1)−(−3)(2)=−4+6=2
(Watch the double negative when subtracting (−3)(2): dropping a minus sign gives −4−6=−10, which throws off both det(A−1)=−101 and every entry of A−1.)
Since det(A)=2=0,yes, A is invertible, and its inverse scales area by the reciprocal:
det(A−1)=21
Now swap the main diagonal entries 4 and −1, flip the signs on the off-diagonal entries −3 and 2, and divide by 2(be careful to flip −3 to +3; leaving it negative gives [−0.5−1−1.52]):
A−1=21[−1−234]=[−0.5−11.52]
2
For what values of k is the matrix [k−12−3k] singular (non-invertible), and how are its two column arrows related at each value?
What equals zero when the two column arrows line up along a single line?
Set the determinant equal to zero:
k(k)−(−3)(−12)=k2−36=0⟹k2=36
(Remember to take both square roots here, not just k=6.)
k=6 or k=−6
Now plug each value back into the columns [k−12] and [−3k](don't assume positive k means "same direction" without checking the signs of the coordinates):
At k=6, the columns are [6−12] and [−36], which point in opposite directions along the same line.
At k=−6, the columns are [−6−12] and [−3−6], which point in the exact same direction.
3
Suppose A and B are 3×3 matrices where A scales volumes by −4 and B scales volumes by 31. Find det(B−1),det(A−1B), and det(B−1AB).
Take the reciprocal 311=3, not the opposite sign −31.
If a matrix scales volume by a factor, what factor must its inverse multiply volume by to get back to 1?
Take the reciprocal of det(B)=31(take the reciprocal 311=3, not the opposite sign −31):
det(B−1)=311=3
For det(A−1B), multiply det(A−1)=−41 by det(B)=31:
det(A−1B)=(−41)(31)=−121
For det(B−1AB), multiply all three scaling factors in order (make sure to use det(B−1)=3 for the first factor rather than 31, which would give −94):
det(B−1AB)=3×(−4)×31=−4
4
For what value of k does the 3×3 matrix 10223104k have no inverse?
Remember the minus sign on the middle cofactor term −2(0⋅k−4⋅2); forgetting to flip that sign gives 3k−20=0 and k=320.
A 3×3 matrix has no inverse when its 3×3 determinant is zero; expand along the top row.
Expand the determinant along the top row (remember the minus sign on the middle cofactor term −2(0⋅k−4⋅2); forgetting to flip that sign gives 3k−20=0 and k=320):
Multiplying a 2×2 matrix A=[acbd] by its swapped-and-negated matrix gives [acbd][2−3−52]=[−1100−11]. Find det(A),det(A−1), the original matrix A, and A−1.
Do not take the determinant of the right-hand side to get (−11)(−11)=121.
Compare the right-hand side [−1100−11] directly with det(A)[1001].
Because multiplying A by [d−c−ba] always gives det(A) times the identity matrix [1001](do not take the determinant of the right-hand side to get (−11)(−11)=121):
det(A)=−11⟹det(A−1)=−111
Undo the diagonal swap and sign flips on [d−c−ba]=[2−3−52] to get A:
A=[2352]
Finally, divide the swapped-and-negated matrix by det(A)=−11(remember to divide by −11, not +11, or every entry of A−1 will have the wrong sign):