Null space, range and the fundamental theorem
Every linear map sends some vectors to zero and reaches some vectors of its target. What are these two sets, and why do their dimensions always add up to the dimension of the domain?
Where you are 5 in Linear maps
Why this exists
For a matrix, Kernel found the vectors sent to \mathbf{0}, Image found the vectors reached, and Rank and nullity proved that their dimensions add up to the number of columns. A linear map between abstract spaces has the same two sets, but no columns to count and no row reduction to run. This lesson defines them for every T \in \mathcal{L}(V, W) and proves the count from bases alone, so it holds for polynomials, matrices and functions, over any field.
Isabel and Malik design the chutes of a log flume ride. A chute’s profile is a cubic p \in \mathcal{P}_3(\mathbb{R}): the boat is p(x) meters above the splash pool when it is x meters along the ground. An inspector records three numbers, the heights at both ends and the slope halfway. Which profiles give the inspector all zeros, which readings can some profile produce, and how far do the readings pin down the chute? Example 2 answers all three.
Definition
Let V and W be vector spaces over \mathbb{F} and T \in \mathcal{L}(V, W). The null space of T is the set of vectors T sends to \mathbf{0}, and the range of T is the set of its outputs:
These are the kernel and image of the written course, with the same symbols; many books write \operatorname{null} T and \operatorname{range} T instead. \ker T lies in the domain V, and \operatorname{im} T in the target W. When \operatorname{im} T is finite-dimensional, \operatorname{rank} T = \dim \operatorname{im} T.
T is injective (one-to-one) if T\mathbf{u} = T\mathbf{v} implies \mathbf{u} = \mathbf{v}, and surjective (onto) if \operatorname{im} T = W (One-to-one, onto and bijective functions). For \mathbf{v}_0 \in V and a subspace U of V, \mathbf{v}_0 + U = \{\mathbf{v}_0 + \mathbf{u} : \mathbf{u} \in U\}.
Intuition
T\mathbf{u} = T\mathbf{v} exactly when T(\mathbf{u} - \mathbf{v}) = \mathbf{0}. So the null space measures what T cannot tell apart: two inputs give one output exactly when they differ by a vector of \ker T, and each output is reached by a whole shifted copy of \ker T.
Dimension makes the trade exact. Extend a basis of \ker T to a basis of V: the vectors inside \ker T are crushed to \mathbf{0}, and the added ones land on a basis of \operatorname{im} T (the proof of Result 5). Every dimension of V is either lost or carried through.
Results
Throughout, V and W are vector spaces over \mathbb{F} and T \in \mathcal{L}(V, W). Every result holds, with the same proof, over any field.
Result 1 (both are subspaces). \ker T is a subspace of V, and \operatorname{im} T is a subspace of W.
Result 2 (the injectivity test). T is injective if and only if \ker T = \{\mathbf{0}\}.
Result 3 (solution sets). Let \mathbf{w} \in W. The equation T\mathbf{v} = \mathbf{w} has a solution if and only if \mathbf{w} \in \operatorname{im} T, and if T\mathbf{v}_0 = \mathbf{w}, its solutions are exactly the vectors of \mathbf{v}_0 + \ker T.
Result 4 (the range from a spanning list). If (\mathbf{v}_1, \dots, \mathbf{v}_m) spans V, then \operatorname{im} T = \operatorname{span}\{T\mathbf{v}_1, \dots, T\mathbf{v}_m\}. In particular, if V is finite-dimensional, so is \operatorname{im} T, and \operatorname{rank} T \le \dim V.
Result 5 (the fundamental theorem of linear maps). Let V be finite-dimensional; W may be any vector space over \mathbb{F}. Then
For T: \mathbb{R}^n \to \mathbb{R}^m this is the rank-nullity theorem of Rank and nullity.
Result 6 (what the dimensions decide). Let V and W be finite-dimensional.
- If \dim V > \dim W, no linear map from V to W is injective.
- If \dim V < \dim W, no linear map from V to W is surjective.
- If \dim V = \dim W, then T is injective if and only if it is surjective.
Why it is true
Result 1. By Result 1 of Linear maps between vector spaces, T\mathbf{0} = \mathbf{0}, so \mathbf{0} lies in both sets. Let a \in \mathbb{F}. If \mathbf{u}, \mathbf{v} \in \ker T, then T(a\mathbf{u} + \mathbf{v}) = aT\mathbf{u} + T\mathbf{v} = \mathbf{0}. If \mathbf{w} = T\mathbf{u} and \mathbf{w}' = T\mathbf{v} are outputs, so is a\mathbf{w} + \mathbf{w}' = T(a\mathbf{u} + \mathbf{v}). Taking a = 1 gives closure under addition, and \mathbf{v} = \mathbf{0} closure under scalar multiplication, so the subspace test (Result 9 of Vector spaces from axioms) applies to both sets.
Result 2. If T is injective and T\mathbf{v} = \mathbf{0} = T\mathbf{0}, then \mathbf{v} = \mathbf{0}, so \ker T = \{\mathbf{0}\}. Conversely, let \ker T = \{\mathbf{0}\}. If T\mathbf{u} = T\mathbf{v}, then T(\mathbf{u} - \mathbf{v}) = T\mathbf{u} - T\mathbf{v} = \mathbf{0}, so \mathbf{u} - \mathbf{v} = \mathbf{0} and \mathbf{u} = \mathbf{v}.
Result 3. A solution exists exactly when \mathbf{w} is an output, which is the definition of \operatorname{im} T. Let T\mathbf{v}_0 = \mathbf{w}. If T\mathbf{v} = \mathbf{w}, then T(\mathbf{v} - \mathbf{v}_0) = \mathbf{w} - \mathbf{w} = \mathbf{0}, so \mathbf{v} = \mathbf{v}_0 + \mathbf{u} with \mathbf{u} = \mathbf{v} - \mathbf{v}_0 \in \ker T. Conversely, T(\mathbf{v}_0 + \mathbf{u}) = \mathbf{w} + \mathbf{0} = \mathbf{w} for every \mathbf{u} \in \ker T.
Result 4. Each T\mathbf{v}_k lies in the subspace \operatorname{im} T (Result 1), so their span does too, because a span is the smallest subspace containing its list (Result 1 of Span and independence, proved). Conversely, each \mathbf{v} \in V is a_1\mathbf{v}_1 + \dots + a_m\mathbf{v}_m for some scalars, and T\mathbf{v} = a_1T\mathbf{v}_1 + \dots + a_mT\mathbf{v}_m by Result 1 of Linear maps between vector spaces. If \dim V = n, apply this to a basis: \operatorname{im} T is spanned by n vectors, and that list shrinks to a basis of \operatorname{im} T (Result 2 of Bases and dimension, proved) of length at most n.
Result 5. \ker T is a subspace of the finite-dimensional V, so it has a basis (\mathbf{u}_1, \dots, \mathbf{u}_m) by Result 6 of Bases and dimension, proved (the empty list if \ker T = \{\mathbf{0}\}). That list is independent in V, so it extends to a basis (\mathbf{u}_1, \dots, \mathbf{u}_m, \mathbf{v}_1, \dots, \mathbf{v}_n) of V (Result 3 there), and \dim V = m + n. It remains to show that (T\mathbf{v}_1, \dots, T\mathbf{v}_n) is a basis of \operatorname{im} T, so that \operatorname{rank} T = n.
It spans. By Result 4, \operatorname{im} T is spanned by T\mathbf{u}_1, \dots, T\mathbf{u}_m, T\mathbf{v}_1, \dots, T\mathbf{v}_n, and each T\mathbf{u}_i is \mathbf{0}, which adds nothing to a span.
It is independent. Suppose c_1T\mathbf{v}_1 + \dots + c_nT\mathbf{v}_n = \mathbf{0}. By linearity T(c_1\mathbf{v}_1 + \dots + c_n\mathbf{v}_n) = \mathbf{0}, so c_1\mathbf{v}_1 + \dots + c_n\mathbf{v}_n lies in \ker T and equals d_1\mathbf{u}_1 + \dots + d_m\mathbf{u}_m for some scalars. Then
and since the basis of V is independent, every coefficient is 0, each c_j included.
This is the second proof in Rank and nullity with the coordinates taken away: it uses only that independent lists extend to bases, so it runs in every finite-dimensional space.
Result 6. A subspace of W has dimension at most \dim W, with equality only for W itself (Result 6 of Bases and dimension, proved). If \dim V > \dim W, Result 5 gives \dim \ker T = \dim V - \operatorname{rank} T \ge \dim V - \dim W > 0, so \ker T \neq \{\mathbf{0}\} and T is not injective by Result 2. If \dim V < \dim W, then \operatorname{rank} T \le \dim V < \dim W by Result 4, so \operatorname{im} T \neq W. If \dim V = \dim W = n, then by Results 2 and 5, T is injective exactly when \dim \ker T = 0, exactly when \operatorname{rank} T = n, exactly when \operatorname{im} T = W.
Worked examples
A map on cubics
Let T \in \mathcal{L}(\mathcal{P}_3(\mathbb{R})) be (Tp)(x) = xp'(x) - 2p(x). Find \ker T and \operatorname{im} T, confirm Result 5, and decide whether T is injective or surjective.
SolutionT is linear because differentiation and multiplication by x are. On the basis (1, x, x^2, x^3), T(x^k) = kx^k - 2x^k = (k - 2)x^k, so T(1) = -2, T(x) = -x, T(x^2) = 0 and T(x^3) = x^3. By Result 1 of Linear maps between vector spaces,
This is 0 exactly when a = b = d = 0, so \ker T = \operatorname{span}\{x^2\}, of dimension 1. By Result 4, \operatorname{im} T = \operatorname{span}\{1, x, x^3\}, of dimension 3, and 1 + 3 = 4 = \dim \mathcal{P}_3(\mathbb{R}). T is not injective, since T(x^2) = 0, and not surjective, since no output has an x^2 term. Result 6 predicted that the two failures come together: the domain and the target have the same dimension.
Readings on a log flume chute
The inspector records Sp = (p(0),\ p(4),\ p'(2)) for a chute profile p \in \mathcal{P}_3(\mathbb{R}), with x running from 0 to 4 meters. Find \ker S and \operatorname{im} S. Isabel wants every profile with readings (2, 0, -\tfrac12), and Malik the one among them that lands flat, with p'(4) = 0.
SolutionFor p = a + bx + cx^2 + dx^3,
Sp = \mathbf{0} gives a = 0, then b + 4c + 16d = 0 and b + 4c + 12d = 0. Subtracting, 4d = 0, so d = 0 and b = -4c. Hence \ker S = \operatorname{span}\{x^2 - 4x\}: the profiles cx(x - 4), level at both ends and flat at the middle, are the ones the inspector cannot see. By Result 5, \operatorname{rank} S = 4 - 1 = 3, so \operatorname{im} S is a 3-dimensional subspace of \mathbb{R}^3, which is all of \mathbb{R}^3 (Result 6 of Bases and dimension, proved). Every triple of readings comes from some chute.
For the readings (2, 0, -\tfrac12), the straight chute p_0 = 2 - \tfrac12 x works: p_0(0) = 2, p_0(4) = 0 and p_0' = -\tfrac12 everywhere. By Result 3 the profiles with these readings are exactly
Then p'(x) = -\tfrac12 + c(2x - 4), so p'(4) = -\tfrac12 + 4c, which is 0 for c = \tfrac18. Malik’s chute is p = 2 - x + \tfrac18 x^2.
A commutator on 2 by 2 matrices
Let N = \begin{bmatrix} 0 & 1 \\ 0 & 0 \end{bmatrix} and define T \in \mathcal{L}(\mathbb{F}^{2 \times 2}) by TA = AN - NA. Find \ker T and \operatorname{im} T, and show that they share a nonzero matrix.
SolutionT is linear because matrix multiplication distributes over sums and commutes with scalars. For A = \begin{bmatrix} a & b \\ c & d \end{bmatrix},
TA = 0 exactly when c = 0 and a = d, so \ker T = \operatorname{span}\{I, N\}, the matrices that commute with N, of dimension 2. With H = \begin{bmatrix} -1 & 0 \\ 0 & 1 \end{bmatrix}, the outputs are cH + (a - d)N, with c and a - d free, so \operatorname{im} T = \operatorname{span}\{H, N\}, also of dimension 2, and 2 + 2 = 4.
The matrix N lies in both: TN = N^2 - N^2 = 0, and T\begin{bmatrix} 1 & 0 \\ 0 & 0 \end{bmatrix} = N. So \ker T + \operatorname{im} T = \operatorname{span}\{I, N, H\} has dimension only 3.
Differentiation and integration on all polynomials
On \mathcal{P}(\mathbb{R}) let Dp = p' and (Jp)(x) = \int_0^x p(t)\,dt. Find the null space and range of D, J and JD, and check JD on p = 3 - 4x + 6x^2.
SolutionJ is linear by the linearity of the integral. A polynomial with zero derivative is constant, so \ker D = \operatorname{span}\{1\}. By the The fundamental theorem of calculus, D(Jq) = q for every q, that is, DJ = I; so every polynomial is a derivative and \operatorname{im} D = \mathcal{P}(\mathbb{R}). D is surjective but not injective.
If Jp = 0, then p = DJp = 0, so \ker J = \{0\} and J is injective. Every Jp vanishes at 0; conversely, if q(0) = 0, then J(Dq)(x) = q(x) - q(0) = q(x). So \operatorname{im} J = \{q : q(0) = 0\}, which misses the constant 1: J is injective but not surjective.
The same computation gives (JDp)(x) = p(x) - p(0), so \ker JD is the constants and \operatorname{im} JD = \operatorname{im} J. For p = 3 - 4x + 6x^2, Dp = -4 + 12x and JDp = -4x + 6x^2 = p - 3.
Here the domain equals the target, yet injective and surjective come apart: Result 6 needs finite dimension.
Count first, then compute
Let T: \mathcal{P}_4(\mathbb{R}) \to \mathbb{R}^4 be
Show without computing that some nonzero p has Tp = \mathbf{0}. Then find \ker T and decide whether T is surjective.
SolutionEach entry is linear in p, so T is linear. Since \dim \mathcal{P}_4(\mathbb{R}) = 5 > 4 = \dim \mathbb{R}^4, Result 6 says T is not injective, and by Result 2 its null space contains a nonzero p.
To find it: p(1) = p(-1) = 0 makes x - 1 and x + 1 factors of p (The remainder and factor theorems), so p = (1 - x^2)(a + bx + cx^2). Over [-1, 1] the integral of an odd polynomial is 0, and
So the third entry is \tfrac43 a + \tfrac{4}{15}c and the fourth is \tfrac{4}{15}b. Both vanish exactly when b = 0 and c = -5a, and (1 - x^2)(1 - 5x^2) = 1 - 6x^2 + 5x^4, so
By Result 5, \operatorname{rank} T = 5 - 1 = 4, so T is surjective: any two end values and any two integrals are achieved by some p \in \mathcal{P}_4(\mathbb{R}), and two such p differ by a multiple of 1 - 6x^2 + 5x^4 (Result 3).
Prescribing the null space and the range
Let V and W be finite-dimensional, U a subspace of V and X a subspace of W. Prove that some T \in \mathcal{L}(V, W) has \ker T = U and \operatorname{im} T = X if and only if \dim U + \dim X = \dim V. Then give such a map on \mathcal{P}_3(\mathbb{R}) with U = X = \mathcal{P}_1(\mathbb{R}).
SolutionIf such a T exists, Result 5 reads \dim V = \dim U + \dim X.
Conversely, let \dim U + \dim X = \dim V. Take a basis (\mathbf{u}_1, \dots, \mathbf{u}_m) of U and extend it to a basis (\mathbf{u}_1, \dots, \mathbf{u}_m, \mathbf{v}_1, \dots, \mathbf{v}_n) of V (Result 3 of Bases and dimension, proved); then n = \dim V - \dim U = \dim X. Let (\mathbf{x}_1, \dots, \mathbf{x}_n) be a basis of X. By the linear map lemma (Result 5 of Linear maps between vector spaces) some T \in \mathcal{L}(V, W) has T\mathbf{u}_i = \mathbf{0} and T\mathbf{v}_j = \mathbf{x}_j for all i and j. By Result 4, \operatorname{im} T is the span of \mathbf{0} and the \mathbf{x}_j, which is X. Every \mathbf{u}_i lies in the subspace \ker T, so U \subseteq \ker T, and by Result 5, \dim \ker T = \dim V - \dim X = \dim U. A subspace of \ker T of the same dimension is all of it (Result 6 of Bases and dimension, proved), so \ker T = U.
For U = X = \mathcal{P}_1(\mathbb{R}) inside \mathcal{P}_3(\mathbb{R}), 2 + 2 = 4. The construction with the bases (1, x) of U, (1, x, x^2, x^3) of V and (1, x) of X gives T(a + bx + cx^2 + dx^3) = c + dx. The second derivative is another such map, since D^2(a + bx + cx^2 + dx^3) = 2c + 6dx. For each of them the null space equals the range.
Boundaries
- Result 5 needs V finite-dimensional, and Result 6 needs both spaces to be. In Example 4, D and J act on the infinite-dimensional \mathcal{P}(\mathbb{R}); one is surjective and not injective, the other the reverse. The target in Result 5 may be infinite-dimensional: the range is finite-dimensional anyway, by Result 4.
- The dimensions add, but the subspaces need not split V. For an operator, \ker T and \operatorname{im} T both lie in V and their dimensions add to \dim V, but they can overlap (Example 3) or coincide (Example 6), so V = \ker T \oplus \operatorname{im} T can fail (Sums and direct sums of subspaces).
- Dimensions do not name the subspaces. Result 5 fixes \dim \ker T but not which subspace it is; by Example 6 every subspace of the right dimension occurs.
- Surjectivity depends on the target. Example 1’s formula is not onto \mathcal{P}_3(\mathbb{R}), but it is onto \operatorname{span}\{1, x, x^3\}.
Where this goes next
The matrix of a linear map records T as a matrix relative to chosen bases and shows that \operatorname{rank} T is the rank of that matrix. Invertibility and isomorphic spaces uses Results 2 and 6 to show that between spaces of the same finite dimension, one-to-one alone or onto alone makes a map invertible. Invariant subspaces studies \ker(T - \lambda I) and \operatorname{im}(T - \lambda I) for an operator T, and Dual spaces applies Result 5 to linear functionals, as in Exercise 14.
Exercises
Null spaces and ranges
- Let T: \mathcal{P}_3(\mathbb{R}) \to \mathbb{R}^2, Tp = (p(1), p'(1)). Find a basis of \ker T and of \operatorname{im} T, and confirm Result 5.
- Let T: \mathbb{R}^{2 \times 2} \to \mathcal{P}_2(\mathbb{R}), T\begin{bmatrix} a & b \\ c & d \end{bmatrix} = (a - d) + (b + c)x + (a - d)x^2. Find bases of \ker T and \operatorname{im} T.
- Let T: \mathcal{P}_2(\mathbb{R}) \to \mathcal{P}_3(\mathbb{R}), (Tp)(x) = (x - 2)p(x). Show that T is injective, and find a basis of \operatorname{im} T.
- Let T \in \mathcal{L}(\mathbb{C}^3), T(z_1, z_2, z_3) = (z_1 + iz_2,\ iz_1 - z_2,\ z_3 - iz_1). Find \ker T, and describe \operatorname{im} T by one equation.
- Let S \in \mathcal{L}(\mathbb{F}^{\infty}), S(x_1, x_2, x_3, \dots) is (x_1 - x_2,\ x_2 - x_3,\ x_3 - x_4, \dots). Find \ker S and show that S is surjective.
- Let T \in \mathcal{L}(\mathcal{P}_2(\mathbb{R})), (Tp)(x) = p(x + 1) - p(x). Find \ker T and \operatorname{im} T.
Counting with the fundamental theorem
- (a) A linear T: \mathcal{P}_5(\mathbb{R}) \to \mathbb{R}^{2 \times 2} has \dim \ker T = 3. Find \operatorname{rank} T. Is T surjective? (b) A linear T: \mathbb{R}^{2 \times 3} \to \mathcal{P}_3(\mathbb{R}) is surjective. Find \dim \ker T.
- Is there a linear T: \mathbb{R}^5 \to \mathbb{R}^2 with \ker T = \{\mathbf{x} : x_1 = x_2,\ x_3 = x_4 = x_5\}?
- Show without computing that some nonzero p \in \mathcal{P}_3(\mathbb{R}) has \int_0^1 x^k p(x)\,dx = 0 for k = 0, 1, 2. Then find all such p.
- For which k is T: \mathcal{P}_2(\mathbb{R}) \to \mathbb{R}^3, Tp = (p(0), p(1), p(k)), injective? Explain why it is then surjective, and for k = 1 find \ker T and \operatorname{im} T.
- Let \dim V = 7 and \dim W = 3. Which values can \dim \ker T take for T \in \mathcal{L}(V, W)?
Proofs
- Let T \in \mathcal{L}(V, W) be injective and (\mathbf{v}_1, \dots, \mathbf{v}_m) independent in V. Prove that (T\mathbf{v}_1, \dots, T\mathbf{v}_m) is independent.
- Let V be finite-dimensional and S, T \in \mathcal{L}(V) with ST = 0. Prove that \operatorname{rank} S + \operatorname{rank} T \le \dim V. Deduce that T^2 = 0 forces \operatorname{rank} T \le \tfrac12 \dim V, and give T \in \mathcal{L}(\mathbb{F}^4) with T^2 = 0 and rank 2.
- Let V be finite-dimensional and \varphi \in \mathcal{L}(V, \mathbb{F}) nonzero. Prove that \varphi is surjective and \dim \ker \varphi = \dim V - 1.
- Let V be finite-dimensional and T \in \mathcal{L}(V). Prove that \ker T \subseteq \ker T^2 and \operatorname{im} T^2 \subseteq \operatorname{im} T, and that if \ker T = \ker T^2, then \operatorname{im} T = \operatorname{im} T^2.
- Let V be finite-dimensional, U a subspace of V and T \in \mathcal{L}(V, W). Prove that \dim T(U) \ge \dim U - \dim \ker T, where T(U) = \{T\mathbf{u} : \mathbf{u} \in U\}.
- Prove that no T \in \mathcal{L}(\mathbb{R}^5) has \ker T = \operatorname{im} T, and give a T \in \mathcal{L}(\mathbb{R}^4) that does.
Answers
- If p(1) = 0, then p = (x - 1)q (The remainder and factor theorems) and p'(1) = q(1), so p'(1) = 0 as well means q = (x - 1)r. Hence \ker T = \{(x - 1)^2(a + bx)\}, with basis ((x - 1)^2,\ x(x - 1)^2). Since T(1) = (1, 0) and T(x - 1) = (0, 1), \operatorname{im} T = \mathbb{R}^2 with the standard basis, and 2 + 2 = 4.
- TA = 0 exactly when a = d and c = -b, so \ker T has basis I, \begin{bmatrix} 0 & 1 \\ -1 & 0 \end{bmatrix}. The outputs are (a - d)(1 + x^2) + (b + c)x, so \operatorname{im} T has basis (1 + x^2, x); 2 + 2 = 4.
- A nonzero p gives (x - 2)p of degree \deg p + 1, not the zero polynomial, so \ker T = \{0\} and T is injective (Result 2). By Result 4, \operatorname{im} T is spanned by x - 2, x(x - 2), x^2(x - 2), which have degrees 1, 2, 3 and so are independent: a basis. It is the set of q \in \mathcal{P}_3(\mathbb{R}) with q(2) = 0, of dimension 3 = 4 - 1.
- The second entry is i(z_1 + iz_2), so T\mathbf{z} = \mathbf{0} exactly when z_1 = -iz_2 and z_3 = iz_1 = z_2: \ker T = \operatorname{span}\{(-i, 1, 1)\}. By Result 5, \operatorname{rank} T = 2. Every output has w_2 = iw_1, and that subspace has dimension 2, so (Result 6 of Bases and dimension, proved) \operatorname{im} T = \{\mathbf{w} \in \mathbb{C}^3 : w_2 = iw_1\}.
- S\mathbf{x} = \mathbf{0} means x_k = x_{k+1} for every k, so \ker S = \operatorname{span}\{(1, 1, 1, \dots)\}. Given \mathbf{y}, set x_1 = 0 and x_{k+1} = x_k - y_k; then x_k - x_{k+1} = y_k, so S\mathbf{x} = \mathbf{y} and S is surjective, though not injective.
- T(1) = 0, T(x) = 1, T(x^2) = 2x + 1, so T(a + bx + cx^2) = (b + c) + 2cx. This is 0 exactly when b = c = 0: \ker T = \operatorname{span}\{1\}. \operatorname{im} T = \operatorname{span}\{1, 2x + 1\} = \mathcal{P}_1(\mathbb{R}), and 1 + 2 = 3.
- (a) \operatorname{rank} T = 6 - 3 = 3 < 4, so T is not surjective. (b) \dim \ker T = 6 - 4 = 2.
- No. The proposed null space has basis ((1, 1, 0, 0, 0),\ (0, 0, 1, 1, 1)), so Result 5 would force \operatorname{rank} T = 5 - 2 = 3 > 2 = \dim \mathbb{R}^2.
- The map p \mapsto \big(\int_0^1 p,\ \int_0^1 xp,\ \int_0^1 x^2p\big) is linear from a space of dimension 4 to \mathbb{R}^3, so it is not injective (Result 6). For p = a + bx + cx^2 + dx^3, \int_0^1 x^kp\,dx = \tfrac{a}{k + 1} + \tfrac{b}{k + 2} + \tfrac{c}{k + 3} + \tfrac{d}{k + 4}. Solving the three equations gives a = -\tfrac{1}{20}d, b = \tfrac35 d, c = -\tfrac32 d, so the solutions are p = t(20x^3 - 30x^2 + 12x - 1), t \in \mathbb{R}.
- Injective exactly when k \ne 0 and k \ne 1. Then a p in the null space has three distinct roots 0, 1, k and degree at most 2, so p = 0; and \dim \mathcal{P}_2(\mathbb{R}) = 3 = \dim \mathbb{R}^3, so T is surjective by Result 6. For k \in \{0, 1\}, x^2 - x is a nonzero vector of the null space. For k = 1: \ker T = \operatorname{span}\{x^2 - x\} and \operatorname{im} T = \{(s, t, t)\} = \operatorname{span}\{(1, 0, 0), (0, 1, 1)\}, and 1 + 2 = 3.
- 4, 5, 6 or 7. Since \operatorname{rank} T \le 3, \dim \ker T = 7 - \operatorname{rank} T \ge 4. Each value occurs: with bases (\mathbf{v}_1, \dots, \mathbf{v}_7) of V and (\mathbf{w}_1, \mathbf{w}_2, \mathbf{w}_3) of W, the map with T\mathbf{v}_j = \mathbf{w}_j for j \le r and T\mathbf{v}_j = \mathbf{0} otherwise has rank r, for r = 0, 1, 2, 3.
- Suppose a_1T\mathbf{v}_1 + \dots + a_mT\mathbf{v}_m = \mathbf{0}. By linearity T(a_1\mathbf{v}_1 + \dots + a_m\mathbf{v}_m) = \mathbf{0}, so a_1\mathbf{v}_1 + \dots + a_m\mathbf{v}_m \in \ker T = \{\mathbf{0}\} (Result 2). The \mathbf{v}_k are independent, so every a_k = 0.
- ST = 0 means S(T\mathbf{v}) = \mathbf{0} for every \mathbf{v}, so \operatorname{im} T \subseteq \ker S. A subspace has dimension at most that of the space containing it (Result 6 of Bases and dimension, proved), so by Result 5, \operatorname{rank} T \le \dim \ker S = \dim V - \operatorname{rank} S. With S = T this reads 2\operatorname{rank} T \le \dim V. Example: T(x_1, x_2, x_3, x_4) = (x_2, 0, x_4, 0), for which T^2 = 0 and \operatorname{im} T = \operatorname{span}\{\mathbf{e}_1, \mathbf{e}_3\}.
- \operatorname{im} \varphi is a subspace of \mathbb{F}, which has dimension 1, and it is not \{0\} because \varphi \ne 0. So it has dimension 1 and equals \mathbb{F} (Result 6 of Bases and dimension, proved): \varphi is surjective. By Result 5, \dim \ker \varphi = \dim V - 1.
- If T\mathbf{v} = \mathbf{0}, then T^2\mathbf{v} = T\mathbf{0} = \mathbf{0}. If \mathbf{w} = T^2\mathbf{v}, then \mathbf{w} = T(T\mathbf{v}) \in \operatorname{im} T. If \ker T = \ker T^2, Result 5 for T and for T^2 gives \operatorname{rank} T = \dim V - \dim \ker T, which is \dim V - \dim \ker T^2 = \operatorname{rank} T^2. So \operatorname{im} T^2 is a subspace of \operatorname{im} T of the same finite dimension, and the two are equal (Result 6 of Bases and dimension, proved).
- U is finite-dimensional (Result 6 of Bases and dimension, proved). The restriction R \in \mathcal{L}(U, W), R\mathbf{u} = T\mathbf{u}, is linear because T is; its range is T(U) and its null space is U \cap \ker T. By Result 5 for R, \dim T(U) = \dim U - \dim(U \cap \ker T), and U \cap \ker T is a subspace of \ker T, so \dim(U \cap \ker T) \le \dim \ker T. Hence \dim T(U) \ge \dim U - \dim \ker T.
- If \ker T = \operatorname{im} T, Result 5 gives 5 = \dim \ker T + \operatorname{rank} T = 2\dim \ker T, impossible for an integer. On \mathbb{R}^4, T(x_1, x_2, x_3, x_4) = (x_3, x_4, 0, 0) has \ker T = \{\mathbf{x} : x_3 = x_4 = 0\}, which is \operatorname{span}\{\mathbf{e}_1, \mathbf{e}_2\} = \operatorname{im} T.