This deck presents the Invertible Matrix Theorem as a single phenomenon with many faces, then covers isomorphism of vector spaces and their classification by dimension, the dual space and the dual basis, and the double dual with its natural isomorphism. It targets the traps of supposing that a non-square matrix could be invertible, that isomorphic spaces must share a common set, that the determinant is unrelated to injectivity, and that identifying a space with its dual is canonical.
Subject: Foundations of Higher Mathematics · 117 slides · symbolic lesson
Open the interactive version of this deck · Homework for this lesson
Objectives
By the end of this deck you can:
1. State the Invertible Matrix Theorem and see why its many conditions are one phenomenon.
2. Decide when two vector spaces are isomorphic, and prove it with an explicit map.
3. Build the dual basis of a given basis and check the duality relations.
4. Explain why a space is naturally isomorphic to its double dual, but only unnaturally to its dual.
Warm-up
Discussion prompt
Before we open Invertibility, Isomorphism & Duality: without looking back, what was the main idea of Matrices, Coordinates & Change of Basis, and what could you do by the end of it that you could not do before?
Hint: One sentence for the idea, one for the skill. If the second one is blank, that is the part to revisit.
Answer:
That deck shows how an ordered basis turns abstract vectors into coordinate tuples, how a linear map becomes a matrix, why matrix multiplication is exactly composition, and how a change of basis - that is, similarity - re-coordinates one and the same map. It targets four real errors: confusing a vector with its coordinate tuple, multiplying matrices in the wrong order for a composition, mixing up P and P-inverse in the change-of-basis formula, and thinking that similar matrices are equal.
Concept
A square matrix is invertible when some other matrix undoes it on both sides.
invertible matrix — A square matrix A for which there is a matrix B with AB and BA both equal to the identity. B is called the inverse of A.
\[ AB = BA = I \]
Counterexample
Discussion prompt
A square matrix is invertible when some other matrix undoes it on both sides.
That is stated as though it always holds. Do one of two things: produce a case where it fails, or say precisely what rules such a case out. "It just does" is not on the menu.
Hint: Hunt at the extremes first — zero, one, negative, empty, equal. If every extreme survives, the reason they survive is the proof.
Intuition
Think of the matrix as a machine that transforms an input vector into an output vector.
The inverse is the machine that takes the output and hands you back exactly the original input. Run one, then the other, and you land where you started.
For the round trip to work in both directions, nothing can be lost and nothing can be duplicated along the way. That single demand is the whole story.
Analogy
Discussion prompt
Explain The inverse is the undo button by analogy to something with no Foundations of Higher Mathematics in it at all — a queue, a recipe, a map, a bank balance, whatever fits. Then say where your analogy breaks.
Hint: An analogy that never breaks is not an analogy, it is the same idea wearing a hat. Find the seam — that is the part that is actually new.
Answer:
Think of the matrix as a machine that transforms an input vector into an output vector.
Concept
A matrix cannot have two different inverses. If two candidates both work, they must be equal.
\[ B = B(AC) = (BA)C = C \]
The middle step uses associativity; the outer steps use that each candidate is an inverse. So we may speak of the inverse.
\( A^{-1} \)
Explain it
Discussion prompt
Explain The inverse is unique to a student a year behind you. No notation, no jargon they have not met — and it still has to be true.
Hint: If your explanation needs a symbol they have never seen, you are describing the notation rather than the idea.
Answer:
A matrix cannot have two different inverses. If two candidates both work, they must be equal.
Ranking
Put in order
Put the moves of Computing the inverse of a 2 by 2 into the order they have to happen.
Why: These are the moves of the worked example in the order it makes them, and each one is set up by the one before it. The 2 by 2 inverse formula divides by the determinant, so we need it first.
Worked example
Find the inverse of the matrix below and confirm it works.
\[ A = \begin{pmatrix} 2 & 1 \\ 1 & 1 \end{pmatrix} \]
Compute the determinant
Why: The 2 by 2 inverse formula divides by the determinant, so we need it first.
\[ \det A = 2\cdot 1 - 1\cdot 1 = 1 \]
Apply the swap-and-negate formula
Why: Swap the diagonal entries, negate the off-diagonal ones, then divide by the determinant.
\[ A^{-1} = \frac{1}{\det A}\begin{pmatrix} 1 & -1 \\ -1 & 2 \end{pmatrix} = \begin{pmatrix} 1 & -1 \\ -1 & 2 \end{pmatrix} \]
Verify by multiplying both ways
Why: An inverse must give the identity on both sides, not just one.
\[ A A^{-1} = \begin{pmatrix} 2 & 1 \\ 1 & 1 \end{pmatrix}\begin{pmatrix} 1 & -1 \\ -1 & 2 \end{pmatrix} = \begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix} \]
Picture it
Animation
Shows: Each line of the worked example "Computing the inverse of a 2 by 2", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: An inverse must give the identity on both sides, not just one.
Concept
The inverse of a product reverses the order of the factors.
\[ (AB)^{-1} = B^{-1}A^{-1} \]
You put on socks then shoes; to undo you take off shoes then socks. Last applied is first undone.
Step zero
Discussion prompt
Order matters when inverting a product — before any calculation: what is the plan? Name the moves in order, in plain English, without doing the arithmetic.
Hint: It starts with: Form the product AB
Answer:
Worked example
Take two invertible matrices and test the reversal rule.
\[ A = \begin{pmatrix} 1 & 1 \\ 0 & 1 \end{pmatrix}, \quad B = \begin{pmatrix} 1 & 0 \\ 1 & 1 \end{pmatrix} \]
Form the product AB
Why: We will invert this directly and compare with the formula.
\[ AB = \begin{pmatrix} 2 & 1 \\ 1 & 1 \end{pmatrix}, \quad (AB)^{-1} = \begin{pmatrix} 1 & -1 \\ -1 & 2 \end{pmatrix} \]
Multiply the inverses in reversed order
Why: The rule predicts B inverse on the left, A inverse on the right.
\[ B^{-1}A^{-1} = \begin{pmatrix} 1 & 0 \\ -1 & 1 \end{pmatrix}\begin{pmatrix} 1 & -1 \\ 0 & 1 \end{pmatrix} = \begin{pmatrix} 1 & -1 \\ -1 & 2 \end{pmatrix} \]
Check the wrong order fails
Why: The reversed order matches; the naive same order does not, confirming the rule.
\[ A^{-1}B^{-1} = \begin{pmatrix} 2 & -1 \\ -1 & 1 \end{pmatrix} \neq (AB)^{-1} \]
Picture it
Animation
Shows: Each line of the worked example "Order matters when inverting a product", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: The reversed order matches; the naive same order does not, confirming the rule.
Concept
Invertibility is really a statement about linear maps, not just matrices. A bijective linear map has an inverse map, and that inverse is automatically linear.
\[ T : V \to W, \quad T^{-1} : W \to V, \quad T^{-1}\circ T = \mathrm{id}_V, \quad T\circ T^{-1} = \mathrm{id}_W \]
So we never have to assume the inverse respects addition and scaling; it inherits that for free.
Estimation
Predict first
Suppose T is a linear bijection. Show its set-theoretic inverse is also linear.
Commit before you compute: what does The inverse of a linear bijection is linear come out to? A rough magnitude and the right form is enough — the point is to have something concrete to be wrong about.
Correct: Verify both linearity conditions hold
Why: A prediction you can defend turns the computation into a check rather than a leap of faith — and an answer that contradicts it is caught on the spot. The last line is exactly additivity and homogeneity for T inverse, so it is linear.
Worked example
Suppose T is a linear bijection. Show its set-theoretic inverse is also linear.
Take outputs and name their unique preimages
Why: Because T is a bijection, every w has a single v with T of v equal to w.
\[ T(v_1) = w_1, \quad T(v_2) = w_2 \]
Use linearity of T on the combination
Why: T preserves sums and scalars, so it sends the combination of preimages to the combination of outputs.
\[ T(a v_1 + b v_2) = a w_1 + b w_2 \]
Read the equation backwards through T inverse
Why: Applying T inverse to both sides names the unique preimage of the right-hand side.
\[ T^{-1}(a w_1 + b w_2) = a v_1 + b v_2 = a\,T^{-1}(w_1) + b\,T^{-1}(w_2) \]
Verify both linearity conditions hold
Why: The last line is exactly additivity and homogeneity for T inverse, so it is linear.
\[ T^{-1}(a w_1 + b w_2) = a\,T^{-1}(w_1) + b\,T^{-1}(w_2) \]
Picture it
Animation
Shows: Each line of the worked example "The inverse of a linear bijection is linear", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: The last line is exactly additivity and homogeneity for T inverse, so it is linear.
Concept
A two-sided inverse forces the domain and codomain to have the same dimension. For a matrix that means it must be square.
A wide or tall matrix can have a one-sided inverse, but never a two-sided one.
\[ A \in F^{m\times n} \text{ invertible} \;\Rightarrow\; m = n \]
Anomaly
Predict first
A student writes this, and it looks reasonable:
A student finds a matrix that left-cancels a tall matrix and declares it invertible.
It is wrong. Say what breaks — and say it before you turn the page.
Correct: The left product collapses to the 2 by 2 identity, which looks like success.
A genuine inverse must undo the map on both sides, which forces the shapes to match.
Why: The left product collapses to the 2 by 2 identity, which looks like success.
Trap
A student finds a matrix that left-cancels a tall matrix and declares it invertible.
\[ A = \begin{pmatrix} 1 & 0 \\ 0 & 1 \\ 0 & 0 \end{pmatrix}, \quad B = \begin{pmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \end{pmatrix} \]
One side does give the identity
Why: The left product collapses to the 2 by 2 identity, which looks like success.
\[ BA = \begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix} = I_2 \]
But the other side is not the identity
Why: The right product is a 3 by 3 matrix that flattens the third coordinate, so it cannot be the identity.
\[ AB = \begin{pmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 0 \end{pmatrix} \neq I_3 \]
A genuine inverse must undo the map on both sides, which forces the shapes to match.
Demand two-sided cancellation
Why: Only a square matrix can satisfy both products equal the identity, because the identities on each side have the same size.
\[ AB = I_n \;\text{and}\; BA = I_n \;\Rightarrow\; A \text{ is square, } n\times n \]
A non-square matrix may be left- or right-invertible, but never invertible. Injective on one side, not bijective.
Notation
Annotate
From Trap: a one-sided inverse is not an inverse — read this one piece at a time. What is each part doing?
On: \( A = \begin{pmatrix} 1 & 0 \\ 0 & 1 \\ 0 & 0 \end{pmatrix}, \quad B = \begin{pmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \end{pmatrix} \)
Concept
For a square matrix there is a single number that detects invertibility: the determinant.
\[ A \text{ is invertible} \iff \det A \neq 0 \]
A zero determinant is the fingerprint of a matrix that squashes space, so it cannot be undone.
Intuition
The determinant measures how much the map stretches or shrinks volume, with a sign for orientation.
If the determinant is zero, the map has flattened some volume down to nothing: a cube gets pressed into a sheet.
Once volume is crushed to zero you can never inflate it back, so no inverse can exist. That is why a zero determinant means no inverse.
Missing information
Discussion prompt
Test the matrix below and, if it is singular, find a vector it destroys.
What do you need to know — or decide — before the first line can be written? List everything the problem has to hand you.
Hint: Anything you would have to invent to get started is a thing the problem must supply.
Answer:
This decides invertibility in one number.
Worked example
Test the matrix below and, if it is singular, find a vector it destroys.
\[ M = \begin{pmatrix} 1 & 2 \\ 2 & 4 \end{pmatrix} \]
Compute the determinant
Why: This decides invertibility in one number.
\[ \det M = 1\cdot 4 - 2\cdot 2 = 0 \]
Solve for the null vector
Why: A zero determinant means a nonzero vector is sent to zero; find it by solving the homogeneous system.
\[ M\begin{pmatrix} x \\ y \end{pmatrix} = 0 \;\Rightarrow\; x + 2y = 0 \;\Rightarrow\; \begin{pmatrix} -2 \\ 1 \end{pmatrix} \]
Verify the vector is crushed to zero
Why: Plugging the witness back in confirms the kernel is nontrivial, so M cannot be inverted.
\[ M\begin{pmatrix} -2 \\ 1 \end{pmatrix} = \begin{pmatrix} -2 + 2 \\ -4 + 4 \end{pmatrix} = \begin{pmatrix} 0 \\ 0 \end{pmatrix} \]
Picture it
Animation
Shows: Each line of the worked example "A zero determinant exposes the kernel", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: Plugging the witness back in confirms the kernel is nontrivial, so M cannot be inverted.
Concept
A linear map is injective exactly when the only vector it sends to zero is zero itself.
\[ T \text{ injective} \iff \ker T = \{0\} \]
If two inputs had the same output, their difference would be a nonzero vector in the kernel. So an empty kernel and one-to-one are the same condition.
Intuition
Picture the map acting on all of space at once. Injective says distinct arrows stay distinct after the map.
The only way distinctness fails is if some nonzero arrow collapses onto the origin, dragging a whole line of inputs to one output.
So checking injectivity reduces to one question: does anything nonzero die? That is why we only ever look at the kernel.
Concept
For a square matrix, being invertible is the same as its columns being independent, which for the right count is the same as their forming a basis.
\[ \operatorname{rank} A = n \iff \text{the } n \text{ columns are a basis of } F^n \]
Full rank means the columns span the whole space with no redundancy, so the map reaches everything and loses nothing.
Concept
For a square matrix, a long list of conditions all say the same thing. Any one of them holds if and only if all of them do.
\[ A \text{ invertible} \iff \det A \neq 0 \iff \ker A = \{0\} \]
\[ \iff A \text{ injective} \iff A \text{ surjective} \iff \operatorname{rank} A = n \iff \text{columns are a basis} \]
Intuition
The theorem looks like a pile of separate facts, but it is really a single idea seen from different angles.
The map either preserves all the information in a vector or it does not. If it does, every description of health holds at once: nonzero determinant, empty kernel, onto, full rank.
If it loses any information, every single condition fails together. There is no in-between for a square matrix.
Fill the middle
Fill in the blanks
From Every condition holds for an invertible matrix — finish the line. Write what belongs on the right of the equals sign before you look.
\det A = 1 \neq 0
Why: Producing the right-hand side unprompted is the difference between recognising this line and being able to use it. This is the fastest single test and we computed it earlier.
Worked example
Take a matrix we already inverted and confirm the whole list of conditions at once.
\[ A = \begin{pmatrix} 2 & 1 \\ 1 & 1 \end{pmatrix} \]
Determinant is nonzero
Why: This is the fastest single test and we computed it earlier.
\[ \det A = 1 \neq 0 \]
Kernel is trivial
Why: Solving the homogeneous system forces both coordinates to zero, so nothing but zero is destroyed.
\[ 2x + y = 0, \; x + y = 0 \;\Rightarrow\; x = 0,\, y = 0 \]
Columns are a basis of the plane
Why: Two independent vectors in a two dimensional space automatically span it, giving full rank.
\[ \left\{\begin{pmatrix}2\\1\end{pmatrix},\begin{pmatrix}1\\1\end{pmatrix}\right\} \text{ independent}, \quad \operatorname{rank} A = 2 \]
Verify the conditions agree
Why: Nonzero determinant, trivial kernel, and full rank all fired together, exactly as the theorem promises.
\[ \det A \neq 0 \;\wedge\; \ker A = \{0\} \;\wedge\; \operatorname{rank} A = 2 \]
Picture it
Animation
Shows: Each line of the worked example "Every condition holds for an invertible matrix", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: Nonzero determinant, trivial kernel, and full rank all fired together, exactly as the theorem promises.
Step zero
Discussion prompt
A singular matrix fails every condition — before any calculation: what is the plan? Name the moves in order, in plain English, without doing the arithmetic.
Hint: It starts with: Determinant is zero
Answer:
Worked example
Now run the same checklist on a singular matrix and watch every item fail.
\[ M = \begin{pmatrix} 1 & 2 \\ 2 & 4 \end{pmatrix} \]
Determinant is zero
Why: The one-number test already flags trouble.
\[ \det M = 0 \]
Kernel is nontrivial
Why: We found a nonzero witness earlier, so the map is not injective.
\[ \begin{pmatrix} -2 \\ 1 \end{pmatrix} \in \ker M \]
Rank drops and the image is a line
Why: The second column is twice the first, so the columns span only a line, not the plane; the map is not surjective.
\[ \begin{pmatrix} 2 \\ 4 \end{pmatrix} = 2\begin{pmatrix} 1 \\ 2 \end{pmatrix}, \quad \operatorname{rank} M = 1 \]
Verify all conditions collapse together
Why: Zero determinant, nonzero kernel, deficient rank, and non-surjectivity arrived as a package, confirming the all-or-nothing law.
\[ \det M = 0 \;\wedge\; \ker M \neq \{0\} \;\wedge\; \operatorname{rank} M < 2 \]
Picture it
Animation
Shows: Each line of the worked example "A singular matrix fails every condition", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: Zero determinant, nonzero kernel, deficient rank, and non-surjectivity arrived as a package, confirming the all-or-nothing law.
Anomaly
Predict first
A student writes this, and it looks reasonable:
A student claims the determinant only measures area and says nothing about whether the map is one-to-one.
It is wrong. Say what breaks — and say it before you turn the page.
Correct: The belief treats a zero determinant as a purely geometric fact with no bearing on inputs mapping to the same output.
A zero determinant means volume collapses, and collapse is exactly non-injectivity: a whole line of inputs is crushed to one point.
Why: The belief treats a zero determinant as a purely geometric fact with no bearing on inputs mapping to the same output.
Trap
A student claims the determinant only measures area and says nothing about whether the map is one-to-one.
The false split
Why: The belief treats a zero determinant as a purely geometric fact with no bearing on inputs mapping to the same output.
\[ \det A = 0 \;\not\Rightarrow\; A \text{ non-injective} \quad (\text{claimed}) \]
A zero determinant means volume collapses, and collapse is exactly non-injectivity: a whole line of inputs is crushed to one point.
Chain the equivalences
Why: The theorem links the determinant directly to the kernel, so a zero determinant guarantees a nonzero vector goes to zero.
\[ \det A = 0 \iff \ker A \neq \{0\} \iff A \text{ not injective} \]
The determinant and injectivity are two readings of the same instrument, not separate concerns.
Concept
For a linear map from a finite-dimensional space to itself, injective and surjective are the same condition. Winning either one wins both.
\[ T : V \to V,\; \dim V < \infty: \quad T \text{ injective} \iff T \text{ surjective} \iff T \text{ bijective} \]
This is a pigeonhole effect for vector spaces, and it fails in infinite dimensions, where a shift can be injective without being onto.
Worked example
Prove that an injective linear map on a finite-dimensional space is automatically invertible.
Read injectivity as zero nullity
Why: Injective means the kernel is trivial, so its dimension is zero.
\[ \ker T = \{0\} \;\Rightarrow\; \operatorname{nullity} T = 0 \]
Apply rank-nullity
Why: The dimensions of kernel and image add to the dimension of the domain.
\[ \operatorname{rank} T + \operatorname{nullity} T = \dim V \;\Rightarrow\; \operatorname{rank} T = \dim V \]
Conclude the image is everything
Why: The image is a subspace of V with the full dimension of V, so it must equal V; the map is onto.
\[ \operatorname{im} T \subseteq V,\; \dim(\operatorname{im} T) = \dim V \;\Rightarrow\; \operatorname{im} T = V \]
Verify bijectivity, hence invertibility
Why: Injective plus surjective is bijective, and a bijective linear map has a linear inverse.
\[ T \text{ injective} + T \text{ surjective} \;\Rightarrow\; T \text{ invertible} \]
Picture it
Animation
Shows: Each line of the worked example "Rank-nullity forces invertibility", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: Injective plus surjective is bijective, and a bijective linear map has a linear inverse.
Estimation
Predict first
Use the inverse to solve a linear system in one shot.
Commit before you compute: what does Solving a system with the inverse come out to? A rough magnitude and the right form is enough — the point is to have something concrete to be wrong about.
Correct: Verify by substituting back
Why: A prediction you can defend turns the computation into a check rather than a leap of faith — and an answer that contradicts it is caught on the spot. Plugging the solution into the original system must reproduce b exactly.
Worked example
Use the inverse to solve a linear system in one shot.
\[ A\mathbf{x} = \mathbf{b}, \quad A = \begin{pmatrix} 2 & 1 \\ 1 & 1 \end{pmatrix}, \quad \mathbf{b} = \begin{pmatrix} 3 \\ 2 \end{pmatrix} \]
Multiply both sides by the inverse
Why: Since A is invertible the solution is unique and given by the inverse times b.
\[ \mathbf{x} = A^{-1}\mathbf{b} = \begin{pmatrix} 1 & -1 \\ -1 & 2 \end{pmatrix}\begin{pmatrix} 3 \\ 2 \end{pmatrix} \]
Carry out the product
Why: Row by row multiplication gives the coordinate values.
\[ \mathbf{x} = \begin{pmatrix} 3 - 2 \\ -3 + 4 \end{pmatrix} = \begin{pmatrix} 1 \\ 1 \end{pmatrix} \]
Verify by substituting back
Why: Plugging the solution into the original system must reproduce b exactly.
\[ A\begin{pmatrix} 1 \\ 1 \end{pmatrix} = \begin{pmatrix} 2 + 1 \\ 1 + 1 \end{pmatrix} = \begin{pmatrix} 3 \\ 2 \end{pmatrix} = \mathbf{b} \]
Translation
\( \mathbf{x} = \begin{pmatrix} 3 - 2 \\ -3 + 4 \end{pmatrix} = \begin{pmatrix} 1 \\ 1 \end{pmatrix} \)
Draw it
Translate both ways. First write the expression above as a sentence with no symbols in it at all. Then cover it, and write your sentence back as notation. If the two versions disagree, the disagreement is the thing to fix.
Ranking
Put in order
Put the moves of The matrix of the inverse map into the order they have to happen.
Why: These are the moves of the worked example in the order it makes them, and each one is set up by the one before it. Apply the matrix A to a test vector to see where the map takes it.
Worked example
If a linear map has a matrix in some basis, the inverse map has the inverse matrix.
\[ [T]_B = A = \begin{pmatrix} 2 & 1 \\ 1 & 1 \end{pmatrix} \;\Rightarrow\; [T^{-1}]_B = A^{-1} \]
Send a vector forward through T
Why: Apply the matrix A to a test vector to see where the map takes it.
\[ T\begin{pmatrix} 1 \\ 1 \end{pmatrix} = A\begin{pmatrix} 1 \\ 1 \end{pmatrix} = \begin{pmatrix} 3 \\ 2 \end{pmatrix} \]
Send the output back through T inverse
Why: Apply the inverse matrix to the output; it should return the original input.
\[ T^{-1}\begin{pmatrix} 3 \\ 2 \end{pmatrix} = A^{-1}\begin{pmatrix} 3 \\ 2 \end{pmatrix} = \begin{pmatrix} 1 \\ 1 \end{pmatrix} \]
Verify the round trip is the identity
Why: Starting vector, forward, then back returns the starting vector, confirming the inverse matrix represents the inverse map.
\[ T^{-1}\big(T(\mathbf{v})\big) = \mathbf{v} \]
Reverse engineer
Discussion prompt
Work backwards. The example finished here:
Verify the round trip is the identity
What was it asked to do, and what must it have been given? Reconstruct the problem from its answer.
Hint: Every quantity in the result had to enter somewhere. Account for each one.
Answer:
If a linear map has a matrix in some basis, the inverse map has the inverse matrix.
Pattern
1. Confirm the matrix is square
Why: Only equal dimensions permit a two-sided inverse; a non-square matrix is out immediately.
2. Pick the cheapest equivalent condition
Why: For small matrices compute the determinant; for larger ones row reduce and read the rank, or check the kernel.
3. Any one condition settles all of them
Why: By the Invertible Matrix Theorem a single pass or fail decides invertibility, so you never need the whole list.
4. For an endomorphism, injective alone suffices
Why: In finite dimensions injective forces surjective, so a trivial kernel already proves invertibility.
Elimination
Eliminate the wrong options
Is M invertible, and what is the reason?
3 of these 4 are wrong. Strike them one at a time, and say what rules each one out before you strike the next. The survivor is the answer.
Survives elimination: B
Why: The determinant is 1 times 6 minus 2 times 3, which equals zero. The second column is three times the first, so the columns are dependent, the rank is 1, and the matrix is singular.
Check
Look at the matrix below and decide.
\[ M = \begin{pmatrix} 1 & 2 \\ 3 & 6 \end{pmatrix} \]
Check your understanding
Is M invertible, and what is the reason?
Answer: B
Why: The determinant is 1 times 6 minus 2 times 3, which equals zero. The second column is three times the first, so the columns are dependent, the rank is 1, and the matrix is singular.
Concept
An isomorphism is a linear map that is also a bijection. When one exists, the two spaces are called isomorphic.
isomorphism — A bijective linear map between vector spaces. Its inverse is again a linear map, so structure is preserved in both directions.
\[ V \cong W \iff \exists\, T : V \to W \text{ linear and bijective} \]
Definition probe
Sort into buckets
Every line below is part of the definition of invertible matrix or of isomorphism — one or the other, never both. Put each where it belongs.
Intuition
Two isomorphic spaces are the same object wearing different names. The map is a dictionary translating every vector, sum, and scalar multiple faithfully.
Anything you can say with addition and scaling in one space has an exact translation in the other. No algebraic fact can tell them apart.
This is the linear-algebra echo of a bijection of sets and of an isomorphism of groups: structure-preserving, invertible, and therefore an equivalence of objects.
Concept
Being isomorphic behaves like equality between spaces: reflexive, symmetric, and transitive.
\[ V \cong V, \quad V \cong W \Rightarrow W \cong V, \quad U \cong V \wedge V \cong W \Rightarrow U \cong W \]
The identity witnesses reflexivity, the inverse map witnesses symmetry, and composition of isomorphisms witnesses transitivity.
Fill the middle
Fill in the blanks
From The composite of isomorphisms is an isomorphism — finish the line. Write what belongs on the right of the equals sign before you look.
(T \circ S)^} S^{-1} \circ T^{-1} \overset______}
Why: Producing the right-hand side unprompted is the difference between recognising this line and being able to use it. A composition of linear maps is linear, so the candidate map respects addition and scaling.
Worked example
Show that composing two isomorphisms gives another one, with the reversed composite as inverse.
\[ S : U \to V, \quad T : V \to W \text{ both isomorphisms} \]
The composite is linear
Why: A composition of linear maps is linear, so the candidate map respects addition and scaling.
\[ T \circ S : U \to W \text{ is linear} \]
Propose the reversed composite as inverse
Why: By the socks-and-shoes principle the inverse should undo T first, then S.
\[ (T \circ S)^{-1} \overset{?}{=} S^{-1} \circ T^{-1} \]
Verify the candidate cancels on both sides
Why: Composing in both orders collapses to the identity, so the composite is bijective and hence an isomorphism.
\[ (S^{-1}\!\circ T^{-1})\circ(T\circ S) = S^{-1}\circ S = \mathrm{id}_U \]
Picture it
Animation
Shows: Each line of the worked example "The composite of isomorphisms is an isomorphism", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: Composing in both orders collapses to the identity, so the composite is bijective and hence an isomorphism.
Step zero
Discussion prompt
Polynomials of degree at most two look like triples — before any calculation: what is the plan? Name the moves in order, in plain English, without doing the arithmetic.
Hint: It starts with: Confirm the map is linear
Answer:
Worked example
Build an explicit isomorphism between the degree at most two polynomials and ordered triples.
\[ \Phi : P_2 \to \mathbb{R}^3, \quad \Phi(a + bx + cx^2) = (a, b, c) \]
Confirm the map is linear
Why: Reading off coefficients is additive and scales correctly, so the map is linear.
\[ \Phi\big(p + \lambda q\big) = \Phi(p) + \lambda\,\Phi(q) \]
Exhibit the inverse
Why: Every triple comes from exactly one polynomial, so the map is a bijection with an explicit inverse.
\[ \Phi^{-1}(a, b, c) = a + bx + cx^2 \]
Verify on a sample polynomial
Why: Sending a polynomial forward then back returns the original, confirming an isomorphism.
\[ 1 + 2x + 3x^2 \;\xrightarrow{\Phi}\; (1,2,3) \;\xrightarrow{\Phi^{-1}}\; 1 + 2x + 3x^2 \]
Picture it
Animation
Shows: Each line of the worked example "Polynomials of degree at most two look like triples", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: Sending a polynomial forward then back returns the original, confirming an isomorphism.
Check
Consider the linear map below.
\[ T : \mathbb{R}^2 \to \mathbb{R}^2, \quad T(x, y) = (x + y,\; 2x + 2y) \]
Check your understanding
Is T an isomorphism?
Answer: B
Why: The matrix has rows (1,1) and (2,2), whose determinant is zero. The vector (1,-1) maps to (0,0), so the kernel is nontrivial and T is neither injective nor bijective.
Concept
Choosing an ordered basis pins every vector to a unique tuple of coordinates, and this correspondence is itself an isomorphism.
\[ [\,\cdot\,]_B : V \to F^n, \quad v \mapsto [v]_B \]
This is why coordinates are trustworthy: the tuple carries exactly the same linear structure as the abstract vector it names.
Concept
Two finite-dimensional spaces over the same field are isomorphic if and only if they have the same dimension.
\[ V \cong W \iff \dim V = \dim W \]
So over a fixed field there is essentially one space of each dimension, and the standard tuple space is its representative.
\[ \dim V = n \;\Rightarrow\; V \cong F^n \]
Intuition
Everything else about a finite-dimensional space, its favorite basis, whether its vectors are polynomials or matrices or arrows, is cosmetic.
The single number that survives relabeling is the dimension. Match the dimension and you can translate one space into the other perfectly.
This is a rare and clean classification: one integer completely determines the object up to isomorphism.
Hypothesis
Predict first
Every n-dimensional space is a copy of the tuple space is about to be worked. State your hypothesis first: which rule or definition decides this one, and what is the first move it forces? Then watch whether the example agrees with you.
Correct: Define the coordinate map
Why: Each vector has a unique expansion in the basis; send it to the tuple of its coefficients.
A hypothesis you wrote down is falsifiable; a vague sense of how it will go is not. If the example opens somewhere else, that gap is the thing worth chasing.
Worked example
Prove that any n-dimensional space over F is isomorphic to the standard tuple space.
\[ \dim V = n, \quad B = \{v_1, \dots, v_n\} \text{ a basis} \]
Define the coordinate map
Why: Each vector has a unique expansion in the basis; send it to the tuple of its coefficients.
\[ \Phi\Big(\textstyle\sum_i c_i v_i\Big) = (c_1, \dots, c_n) \]
Injective and surjective from uniqueness
Why: Unique coefficients mean no two vectors share a tuple and every tuple is realized, so the map is a bijection.
\[ \ker \Phi = \{0\}, \quad \operatorname{im}\Phi = F^n \]
Verify linearity closes the argument
Why: Coordinates add and scale coordinate-wise, so the bijection is linear, giving the isomorphism.
\[ \Phi(u + \lambda w) = \Phi(u) + \lambda\,\Phi(w) \;\Rightarrow\; V \cong F^n \]
Picture it
Animation
Shows: Each line of the worked example "Every n-dimensional space is a copy of the tuple space", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: Coordinates add and scale coordinate-wise, so the bijection is linear, giving the isomorphism.
Missing information
Discussion prompt
Treat the complex numbers as a vector space over the reals and identify its dimension.
What do you need to know — or decide — before the first line can be written? List everything the problem has to hand you.
Hint: Anything you would have to invent to get started is a thing the problem must supply.
Answer:
Adding complex numbers adds the real pairs, and scaling by a real scalar scales the pair, giving a linear bijection.
Worked example
Treat the complex numbers as a vector space over the reals and identify its dimension.
\[ \Psi : \mathbb{C} \to \mathbb{R}^2, \quad \Psi(a + bi) = (a, b) \]
Check the map is real-linear and bijective
Why: Adding complex numbers adds the real pairs, and scaling by a real scalar scales the pair, giving a linear bijection.
\[ \Psi\big((a+bi) + t(c+di)\big) = (a + tc,\; b + td) \]
Read off the real dimension
Why: A basis over the reals is one and the imaginary unit, so the real dimension is two.
\[ \dim_{\mathbb{R}} \mathbb{C} = 2 \]
Verify the field changes the answer
Why: Over the complex numbers a single element already spans, so the dimension drops; the scalar field is decisive.
\[ \dim_{\mathbb{C}} \mathbb{C} = 1 \neq 2 = \dim_{\mathbb{R}} \mathbb{C} \]
Picture it
Animation
Shows: Each line of the worked example "The complex numbers as a real plane", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: Adding complex numbers adds the real pairs, and scaling by a real scalar scales the pair, giving a linear bijection.
Anomaly
Predict first
A student writes this, and it looks reasonable:
A student refuses to call the polynomials and triples isomorphic because their elements are visibly different objects.
It is wrong. Say what breaks — and say it before you turn the page.
Correct: The mistake demands equality of the underlying sets, which is far stronger than isomorphism.
Isomorphism only asks for a structure-preserving bijection, not identical elements.
Why: The mistake demands equality of the underlying sets, which is far stronger than isomorphism.
Trap
A student refuses to call the polynomials and triples isomorphic because their elements are visibly different objects.
The false requirement
Why: The mistake demands equality of the underlying sets, which is far stronger than isomorphism.
\[ P_2 \neq \mathbb{R}^3 \text{ as sets} \;\Rightarrow\; P_2 \not\cong \mathbb{R}^3 \quad (\text{claimed}) \]
Isomorphism only asks for a structure-preserving bijection, not identical elements.
Compare dimensions, not elements
Why: Both spaces are three dimensional over the reals, so the classification theorem makes them isomorphic despite different-looking members.
\[ \dim P_2 = 3 = \dim \mathbb{R}^3 \;\Rightarrow\; P_2 \cong \mathbb{R}^3 \]
Same structure, different costumes. That is exactly what isomorphic means.
Break the constraint
Discussion prompt
The rule this trap just fixed:
Both spaces are three dimensional over the reals, so the classification theorem makes them isomorphic despite different-looking members.
Now break it on purpose. Build a case that violates it and follow the consequences until something visibly fails. Where does the failure first show up — and would you have noticed it if you had not been looking?
Hint: The dangerous rules are the ones whose violation still produces an answer. If yours fails loudly, try to find one that fails quietly.
Answer:
The mistake demands equality of the underlying sets, which is far stronger than isomorphism.
Prediction
Predict first
Which of these real vector spaces is NOT isomorphic to the standard four-dimensional tuple space?
Answer it in your own words, now, with nothing to choose from. The options are on the next slide — and picking the right one off a list is an easier skill than producing it.
Correct: The polynomials of degree at most four.
Why: Isomorphism of finite-dimensional spaces is decided by dimension. Degree at most four polynomials have dimension five, while the others all have dimension four, so it is the exception.
Check
All but one of these real vector spaces are isomorphic to each other.
Check your understanding
Which of these real vector spaces is NOT isomorphic to the standard four-dimensional tuple space?
Answer: C
Why: Isomorphism of finite-dimensional spaces is decided by dimension. Degree at most four polynomials have dimension five, while the others all have dimension four, so it is the exception.
Concept
The dual space collects every linear map from a space to its field of scalars. Such a map is called a linear functional.
dual space — The space of all linear maps from V to its scalar field F, written V star. Its elements are linear functionals.
\[ V^{*} = L(V, F) = \{\, \varphi : V \to F \mid \varphi \text{ linear} \,\} \]
Matching
Match the pairs
Match each term to the definition this lesson gave it — not the one you would guess from the word.
Why: These are the working definitions of invertible matrix, isomorphism, dual space as Invertibility, Isomorphism & Duality uses them. Pairing them correctly is the test of whether you could state each one with the slide switched off.
Intuition
A linear functional is a measuring device: feed it a vector, and it returns a single number in a way that respects addition and scaling.
Reading off one coordinate, taking a weighted total, integrating a function against a fixed weight, all are functionals.
The dual space is the space of all such measurements. Studying a vector through every possible linear reading is the dual point of view.
Concept
Functionals can be added and scaled pointwise, and the result is again a functional, so the dual is a vector space in its own right.
\[ (\varphi + \psi)(v) = \varphi(v) + \psi(v), \quad (\lambda\varphi)(v) = \lambda\,\varphi(v) \]
The zero functional sends every vector to zero and plays the role of the additive identity.
Estimation
Predict first
Take the functional that reads the first coordinate and confirm it is linear.
Commit before you compute: what does A coordinate functional in action come out to? A rough magnitude and the right form is enough — the point is to have something concrete to be wrong about.
Correct: Verify it is a genuine element of the dual
Why: A prediction you can defend turns the computation into a check rather than a leap of faith — and an answer that contradicts it is caught on the spot. Linearity plus scalar output is exactly the definition of a functional, so it lives in the dual of the tuple space.
Worked example
Take the functional that reads the first coordinate and confirm it is linear.
\[ \varphi : \mathbb{R}^3 \to \mathbb{R}, \quad \varphi(x, y, z) = x \]
Evaluate on a sample vector
Why: Applying the functional just returns the chosen coordinate.
\[ \varphi(5, 7, 9) = 5 \]
Test additivity and scaling
Why: The first coordinate of a sum is the sum of first coordinates, and scaling scales it, so the map is linear.
\[ \varphi\big((x,y,z) + \lambda(x',y',z')\big) = x + \lambda x' = \varphi(x,y,z) + \lambda\,\varphi(x',y',z') \]
Verify it is a genuine element of the dual
Why: Linearity plus scalar output is exactly the definition of a functional, so it lives in the dual of the tuple space.
\[ \varphi \in (\mathbb{R}^3)^{*} \]
Picture it
Animation
Shows: Each line of the worked example "A coordinate functional in action", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: Linearity plus scalar output is exactly the definition of a functional, so it lives in the dual of the tuple space.
Concept
Given a basis, there is a matched family of functionals, each returning the coordinate along one basis vector and ignoring the others.
\[ \varphi_i(v_j) = \delta_{ij} = \begin{cases} 1 & i = j \\ 0 & i \neq j \end{cases} \]
These functionals form a basis of the dual space, called the dual basis of the chosen basis.
Step zero
Discussion prompt
Building a dual basis — before any calculation: what is the plan? Name the moves in order, in plain English, without doing the arithmetic.
Hint: It starts with: Read the functionals from the inverse matrix
Answer:
Worked example
Construct the dual basis for the basis of the plane given below.
\[ v_1 = \begin{pmatrix} 2 \\ 1 \end{pmatrix}, \quad v_2 = \begin{pmatrix} 1 \\ 1 \end{pmatrix} \]
Read the functionals from the inverse matrix
Why: The dual basis functionals are the rows of the inverse of the matrix whose columns are the basis vectors.
\[ \begin{pmatrix} 2 & 1 \\ 1 & 1 \end{pmatrix}^{-1} = \begin{pmatrix} 1 & -1 \\ -1 & 2 \end{pmatrix} \]
Write down the two functionals
Why: Each row becomes a functional acting on coordinates.
\[ \varphi_1(x, y) = x - y, \quad \varphi_2(x, y) = -x + 2y \]
Verify the duality relations
Why: Each functional must return one on its own basis vector and zero on the other; all four checks succeed.
\[ \varphi_1(v_1) = 1,\; \varphi_1(v_2) = 0,\; \varphi_2(v_1) = 0,\; \varphi_2(v_2) = 1 \]
Picture it
Animation
Shows: Each line of the worked example "Building a dual basis", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: Each functional must return one on its own basis vector and zero on the other; all four checks succeed.
Concept
In finite dimensions the dual basis has exactly as many functionals as the basis has vectors, so a space and its dual share a dimension.
\[ \dim V^{*} = \dim V \;\Rightarrow\; V \cong V^{*} \]
By the classification theorem, matching dimensions already forces an isomorphism between a space and its dual.
Anomaly
Predict first
A student writes this, and it looks reasonable:
A student concludes that since a space and its dual are isomorphic, there is a god-given identification of vectors with functionals.
It is wrong. Say what breaks — and say it before you turn the page.
Correct: The isomorphism from V to its dual was built by sending each basis vector to its dual functional, which secretly depends on the basis.
The isomorphism exists but is not canonical: change the basis and the identification changes.
Why: The isomorphism from V to its dual was built by sending each basis vector to its dual functional, which secretly depends on the basis.
Trap
A student concludes that since a space and its dual are isomorphic, there is a god-given identification of vectors with functionals.
The hidden choice
Why: The isomorphism from V to its dual was built by sending each basis vector to its dual functional, which secretly depends on the basis.
\[ v_i \mapsto \varphi_i \quad (\text{depends on the chosen basis}) \]
The isomorphism exists but is not canonical: change the basis and the identification changes.
No basis-free rule sends V to its dual
Why: There is no way to turn an arbitrary vector into a functional without first picking extra structure, so the isomorphism is unnatural.
\[ V \cong V^{*} \text{ (unnatural)}, \quad \text{no canonical } v \mapsto \varphi_v \]
Same dimension guarantees an isomorphism, but not a preferred one. The dual is a different space that merely happens to be the same size.
Notation
Annotate
From Trap: the dual is canonically the same as V — read this one piece at a time. What is each part doing?
On: \( V \cong V^{*} \text{ (unnatural)}, \quad \text{no canonical } v \mapsto \varphi_v \)
Commit first
Predict first
Which functional is the dual vector for v1, that is, the one giving 1 on v1 and 0 on v2?
Commit to an answer, then rate it — certain, fairly sure, or guessing — and write the rating down before you turn the page.
Correct: The functional sending (x, y) to x minus y.
Why: We need a functional equal to 1 on (1,0) and 0 on (1,1). The map (x,y) to x minus y gives 1 minus 0 equal to 1 on v1 and 1 minus 1 equal to 0 on v2, exactly the duality relations.
The rating matters as much as the answer: confident-and-wrong is the combination that survives revision, because nothing about it feels like it needs revisiting.
Check
Use the basis below and the defining duality relations.
\[ v_1 = (1, 0), \quad v_2 = (1, 1) \]
Check your understanding
Which functional is the dual vector for v1, that is, the one giving 1 on v1 and 0 on v2?
Answer: A
Why: We need a functional equal to 1 on (1,0) and 0 on (1,1). The map (x,y) to x minus y gives 1 minus 0 equal to 1 on v1 and 1 minus 1 equal to 0 on v2, exactly the duality relations.
Concept
The dual of the dual space is the double dual. There is a map into it that needs no choices at all.
Given a vector, define the functional that eats a functional and returns its value at that vector. This is the evaluation map.
\[ \mathrm{ev} : V \to V^{**}, \quad \mathrm{ev}(v)(\varphi) = \varphi(v) \]
Intuition
A vector can play two roles. Usually it is the thing being measured; now let it be the measurement instead.
Fix a vector and let functionals pass through it. Each functional yields a number, so the vector becomes a rule that acts on functionals: a functional on functionals.
Crucially this rule is written using nothing but the vector itself. No basis, no extra data. That is what makes it natural.
Explain it
Discussion prompt
Explain Evaluate at v to a student a year behind you. No notation, no jargon they have not met — and it still has to be true.
Hint: If your explanation needs a symbol they have never seen, you are describing the notation rather than the idea.
Answer:
A vector can play two roles. Usually it is the thing being measured; now let it be the measurement instead.
Ranking
Put in order
Put the moves of The evaluation map is injective into the order they have to happen.
Why: These are the moves of the worked example in the order it makes them, and each one is set up by the one before it. To find the kernel, suppose the evaluation of a vector is the zero functional on the dual.
Worked example
Show that the evaluation map has trivial kernel, so it embeds a space into its double dual.
Assume a vector lands at zero
Why: To find the kernel, suppose the evaluation of a vector is the zero functional on the dual.
\[ \mathrm{ev}(v) = 0 \;\Rightarrow\; \varphi(v) = 0 \text{ for every } \varphi \in V^{*} \]
Contrapositive: a nonzero vector is seen by some functional
Why: If the vector were nonzero, extend it to a basis and use its coordinate functional to read a nonzero value.
\[ v \neq 0 \;\Rightarrow\; \exists\, \varphi \text{ with } \varphi(v) = 1 \]
Force the vector to be zero
Why: Since every functional vanishes on v, no such witness exists, so v cannot be nonzero.
\[ \ker(\mathrm{ev}) = \{0\} \]
Verify injectivity via the trivial kernel
Why: A linear map with trivial kernel is injective, so evaluation embeds V into its double dual.
\[ \ker(\mathrm{ev}) = \{0\} \;\Rightarrow\; \mathrm{ev} \text{ injective} \]
Picture it
Animation
Shows: Each line of the worked example "The evaluation map is injective", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: A linear map with trivial kernel is injective, so evaluation embeds V into its double dual.
Concept
In finite dimensions the evaluation map is not just injective; it is an isomorphism, and it required no choice of basis.
\[ \dim V^{**} = \dim V^{*} = \dim V \;\Rightarrow\; \mathrm{ev} \text{ is onto, hence an isomorphism} \]
Because it is basis-free, this is called a natural or canonical isomorphism, and we routinely identify a space with its double dual.
\[ V \cong V^{**} \text{ naturally} \]
Analogy
Discussion prompt
Explain V is naturally isomorphic to its double dual by analogy to something with no Foundations of Higher Mathematics in it at all — a queue, a recipe, a map, a bank balance, whatever fits. Then say where your analogy breaks.
Hint: An analogy that never breaks is not an analogy, it is the same idea wearing a hat. Find the seam — that is the part that is actually new.
Answer:
In finite dimensions the evaluation map is not just injective; it is an isomorphism, and it required no choice of basis.
Anomaly
Predict first
A student writes this, and it looks reasonable:
A student reasons that if the double dual is naturally the same as V, then the dual must be too, since all three have equal dimension.
It is wrong. Say what breaks — and say it before you turn the page.
Correct: The mistake treats equal dimension as the whole story and ignores whether the isomorphism uses a choice.
Equal dimension gives an isomorphism, but naturality is about whether a canonical basis-free rule exists.
Why: The mistake treats equal dimension as the whole story and ignores whether the isomorphism uses a choice.
Trap
A student reasons that if the double dual is naturally the same as V, then the dual must be too, since all three have equal dimension.
The dimension-only argument
Why: The mistake treats equal dimension as the whole story and ignores whether the isomorphism uses a choice.
\[ \dim V = \dim V^{*} = \dim V^{**} \;\Rightarrow\; \text{all naturally the same} \quad (\text{claimed}) \]
Equal dimension gives an isomorphism, but naturality is about whether a canonical basis-free rule exists.
Separate the two cases
Why: The evaluation map into the double dual is written with the vector alone; any map to the dual must first choose a basis, so only one is natural.
\[ V \cong V^{**} \text{ naturally}, \qquad V \cong V^{*} \text{ only after a choice} \]
So V and its double dual can be identified once and for all, while V and its dual cannot be identified canonically.
Two truths and a lie
Sort into buckets
Some of these hold up and some are the exact mistakes this lesson is built to prevent. Sort them.
Concept
Vector spaces and linear maps form a category, and the isomorphisms are the invertible linear maps we have been studying.
This is one instance of a recurring pattern: a bijection is the isomorphism of sets, a group isomorphism is the invertible homomorphism, and a linear isomorphism is the invertible linear map.
Each time, the structure-preserving invertible map is the right notion of sameness, and the interesting classifications, dimension here, are exactly the invariants those maps preserve.
Counterexample
Discussion prompt
Vector spaces and linear maps form a category, and the isomorphisms are the invertible linear maps we have been studying.
That is stated as though it always holds. Do one of two things: produce a case where it fails, or say precisely what rules such a case out. "It just does" is not on the menu.
Hint: Hunt at the extremes first — zero, one, negative, empty, equal. If every extreme survives, the reason they survive is the proof.
Answer:
Each time, the structure-preserving invertible map is the right notion of sameness, and the interesting classifications, dimension here, are exactly the invariants those maps preserve.
Pattern
1. For a quick verdict, compare dimensions
Why: Over the same field, finite-dimensional spaces are isomorphic exactly when their dimensions match, so counting a basis often settles it.
2. To prove it constructively, exhibit a map
Why: Define a linear map on a basis, then show it is bijective, most cheaply by proving the kernel is trivial in equal dimensions.
3. To disprove, find a preserved quantity that differs
Why: Any isomorphism invariant, above all dimension, that disagrees rules out an isomorphism.
4. Ask whether the isomorphism is natural
Why: If your map needs a basis it is merely an isomorphism; if it is written with the given data alone, like evaluation, it is canonical.
Real world
Discussion prompt
Outside this lesson: where does Invertibility, Isomorphism & Duality actually turn up? Name one concrete situation — a job, a piece of software someone ships, a decision somebody has to make — and say which part of How to prove two spaces are isomorphic is doing the work in it.
Hint: Vague is the failure mode here. "Engineering" is not a situation; "deciding whether this build is fast enough to ship" is.
Answer:
That deck presents the Invertible Matrix Theorem as a single phenomenon with many faces, then covers isomorphism of vector spaces and their classification by dimension, the dual space and the dual basis, and the double dual with its natural isomorphism. It targets the traps of supposing that a non-square matrix could be invertible, that isomorphic spaces must share a common set, that the determinant is unrelated to injectivity, and that identifying a space with its dual is canonical.
Prediction
Predict first
Why is the isomorphism from V to its double dual considered natural, while a chosen isomorphism from V to its dual is not?
Answer it in your own words, now, with nothing to choose from. The options are on the next slide — and picking the right one off a list is an easier skill than producing it.
Correct: The evaluation map is defined using only the vector, with no choice of basis, whereas identifying V with its dual requires choosing one.
Why: Naturality means basis-free. The evaluation map sends a vector to evaluation at that vector using nothing else, while every isomorphism to the dual depends on a chosen basis, so only the double-dual map is canonical.
Check
Consider a finite-dimensional space, its dual, and its double dual.
Check your understanding
Why is the isomorphism from V to its double dual considered natural, while a chosen isomorphism from V to its dual is not?
Answer: A
Why: Naturality means basis-free. The evaluation map sends a vector to evaluation at that vector using nothing else, while every isomorphism to the dual depends on a chosen basis, so only the double-dual map is canonical.
Check
Let T be a linear map from a finite-dimensional space to itself with trivial kernel.
\[ T : V \to V, \quad \dim V < \infty, \quad \ker T = \{0\} \]
Check your understanding
What can you conclude about T?
Answer: B
Why: Trivial kernel means nullity zero, so by rank-nullity the rank equals the dimension of V and the image is all of V. Injective and surjective together make T bijective, hence invertible.
Elimination
Eliminate the wrong options
What is the inverse of the product A times B?
3 of these 4 are wrong. Strike them one at a time, and say what rules each one out before you strike the next. The survivor is the answer.
Survives elimination: B
Why: By socks and shoes the inverse reverses the order: multiply the inverse of B on the left by the inverse of A on the right. Then A B times that collapses to the identity, since the inner B and its inverse cancel first.
Check
Let A and B be invertible square matrices of the same size.
\[ A, B \in F^{n\times n} \text{ invertible} \]
Check your understanding
What is the inverse of the product A times B?
Answer: B
Why: By socks and shoes the inverse reverses the order: multiply the inverse of B on the left by the inverse of A on the right. Then A B times that collapses to the identity, since the inner B and its inverse cancel first.
Connect it up
Draw it
One page, no notation unless you need it: draw how these connect — How to test invertibility · How to prove two spaces are isomorphic · What invertible means · The inverse is the undo button · The inverse is unique. Put an arrow wherever one of them is what makes another possible, and label the arrow with why.
Recap
Invertibility is one phenomenon: for a square matrix, nonzero determinant, trivial kernel, injective, surjective, full rank, and columns-a-basis all hold together or fail together.
On a finite-dimensional space, an injective endomorphism is automatically surjective, so a trivial kernel already proves invertibility.
Two finite-dimensional spaces over the same field are isomorphic exactly when they share a dimension; every space of dimension n is a copy of the standard tuple space.
A space and its dual have equal dimension and so are isomorphic, but only after a choice of basis; a space and its double dual are naturally isomorphic through the basis-free evaluation map.
The unifying thread: the invertible structure-preserving map is the right notion of sameness, echoing bijections of sets and isomorphisms of groups, and dimension is the invariant it preserves.
Want this taught 1-on-1? Alexander tutors Foundations of Higher Mathematics — $55/session, free consultation.