math.wiki
Linear algebra › Abstract vector spaces › Bases and dimension, proved
Sign in

Bases and dimension, proved

In any vector space, why does every spanning list cut down to a basis and every independent list grow to one, and why do all bases of a finite-dimensional space have the same length?

Where you are 5 in Abstract vector spaces

Why this exists

Amara and Samir build a sudoku app and want to add 3 \times 3 magic squares: grids of real numbers whose three rows, three columns and two diagonals all have the same sum. Each puzzle is stored as a short code. How many numbers must a code hold for the app to rebuild the square, and does the count depend on which numbers they store?

Sums and multiples of magic squares are magic, so the magic squares form a subspace of \mathbb{R}^{3 \times 3}, and the count is its dimension, 3 (Example 2). Basis and dimension proved with row reduction that all bases of a subspace of \mathbb{R}^n have one size. This lesson proves it in every finite-dimensional space from the facts about lists in Span and independence, proved, with the two moves that find bases: cutting a spanning list down and growing an independent list up.

Definition

Let V be a vector space over a field \mathbb{F} (Vector spaces from axioms). A basis of V is a list \mathcal{B} = (\mathbf{b}_1, \dots, \mathbf{b}_n) in V that is linearly independent and spans V; the empty list is the basis of \{\mathbf{0}\}. If \mathbf{v} = a_1\mathbf{b}_1 + \dots + a_n\mathbf{b}_n, the scalars a_j are the coordinates of \mathbf{v} relative to \mathcal{B}, and the column [\mathbf{v}]_{\mathcal{B}} \in \mathbb{F}^n holding them in order is its coordinate vector.

The dimension \dim V of a finite-dimensional V is the length of any basis; Result 2 shows a basis exists and Result 4 that all have one length. When a set carries two fields of scalars, \dim_{\mathbb{F}} V names the field. An infinite-dimensional space has no basis in this sense (Result 2); some books write \dim V = \infty.

E_{jk} is the matrix with 1 in row j, column k and 0 elsewhere, and J is the square matrix of all 1’s.

Intuition

A spanning list can be shortened: walk along it and drop each entry that the earlier entries already build. An independent list can be lengthened: append entries of any spanning list, keeping each one the list so far cannot build. Both walks stop at a basis, long enough to reach every vector and short enough to waste none. The length inequality of Span and independence, proved then pins all bases at one length. Dimension counts free parameters: a vector is fixed by its \dim V coordinates in any basis.

Results

Results 1 to 6 hold over any field. Throughout, V is a vector space over \mathbb{F}.

Result 1 (unique coordinates). A list (\mathbf{b}_1, \dots, \mathbf{b}_n) in V is a basis of V if and only if every \mathbf{v} \in V equals a_1\mathbf{b}_1 + \dots + a_n\mathbf{b}_n for exactly one choice of a_1, \dots, a_n \in \mathbb{F}.

Result 2 (spanning lists shrink to bases). Every list that spans V has a sublist that is a basis of V. Hence V has a basis if and only if V is finite-dimensional.

Result 3 (independent lists extend to bases). Let V be finite-dimensional, (\mathbf{u}_1, \dots, \mathbf{u}_m) independent in V, and (\mathbf{w}_1, \dots, \mathbf{w}_n) a list spanning V. Then appending some of the \mathbf{w}’s to (\mathbf{u}_1, \dots, \mathbf{u}_m) gives a basis of V. In particular every independent list in V extends to a basis.

Result 4 (dimension is well defined). If V is finite-dimensional, any two bases of V have the same length. Moreover \dim \mathbb{F}^n = n, \dim \mathcal{P}_m(\mathbb{F}) = m + 1, \dim \mathbb{F}^{m \times n} = mn and \dim\{\mathbf{0}\} = 0.

Result 5 (the right length is enough). Let \dim V = n. Every independent list of length n in V, and every spanning list of length n, is a basis of V. Every list in V longer than n is dependent, and no list shorter than n spans V.

Result 6 (subspaces). If V is finite-dimensional and U is a subspace of V, then U is finite-dimensional and \dim U \le \dim V, with equality if and only if U = V.

Result 7 (complex spaces over the reals). Let V be a vector space over \mathbb{C} with basis (\mathbf{v}_1, \dots, \mathbf{v}_n). With scalars restricted to \mathbb{R}, V is a vector space over \mathbb{R} with basis (\mathbf{v}_1, i\mathbf{v}_1, \dots, \mathbf{v}_n, i\mathbf{v}_n), so \dim_{\mathbb{R}} V = 2\dim_{\mathbb{C}} V.

Why it is true

“Result k there” means Result k of Span and independence, proved.

Result 1. If the list is a basis, spanning gives an expression for each \mathbf{v}, and if \sum_j a_j\mathbf{b}_j = \sum_j c_j\mathbf{b}_j, then \sum_j (a_j - c_j)\mathbf{b}_j = \mathbf{0} and independence gives a_j = c_j. Conversely, unique expressions make the list span, and since 0\mathbf{b}_1 + \dots + 0\mathbf{b}_n is the only expression for \mathbf{0}, it is independent.

Result 2. Let (\mathbf{v}_1, \dots, \mathbf{v}_m) span V. For j = 1, \dots, m in turn, delete \mathbf{v}_j if it lies in the span of the entries still before it, the kept \mathbf{v}_i with i < j. By the linear dependence lemma (Result 4 there) no deletion changes the span, so the final list spans V. In it, the entries before a kept \mathbf{v}_j are exactly those it was tested against, so no entry lies in the span of the entries before it, and the list is independent by Result 3 there. If V is finite-dimensional, some list spans V and shrinks to a basis. Conversely, a basis is a finite list spanning V.

Result 3. The list L = (\mathbf{u}_1, \dots, \mathbf{u}_m, \mathbf{w}_1, \dots, \mathbf{w}_n) spans V: its span is a subspace containing every \mathbf{w}_k, hence their span V (Result 1 there). Shrink L as in Result 2. When \mathbf{u}_j is tested, the entries before it are \mathbf{u}_1, \dots, \mathbf{u}_{j-1}, and \mathbf{u}_j is outside their span because (\mathbf{u}_1, \dots, \mathbf{u}_j) is independent (Results 2 and 3 there). So no \mathbf{u}_j is deleted, and the basis left is the \mathbf{u}’s followed by the kept \mathbf{w}’s. A finite-dimensional V has a spanning list, so every independent list extends.

Result 4. Let (\mathbf{b}_1, \dots, \mathbf{b}_m) and (\mathbf{c}_1, \dots, \mathbf{c}_n) be bases. The first is independent and the second spans V, so m \le n by the length inequality (Result 5 there); exchanging roles gives n \le m. The standard basis of \mathbb{F}^n has length n, and (1, z, \dots, z^m) is a basis of \mathcal{P}_m(\mathbb{F}) by Result 4 of Polynomials and functions as vectors. In \mathbb{F}^{m \times n}, A = \sum_{j,k} a_{jk}E_{jk}, and this combination has entry a_{jk} in place (j, k), so it is zero only when every a_{jk} = 0: the mn matrices E_{jk} form a basis. The empty list is a basis of \{\mathbf{0}\}.

Result 5. An independent list of length n extends to a basis (Result 3) of length n (Result 4), so nothing was appended. A spanning list of length n shrinks to a basis (Result 2) of length n, so nothing was deleted. A basis spans V, so a list longer than n is dependent by Result 5 there; a basis is independent of length n, so by Result 5 there again no shorter list spans.

Result 6. By Result 6 there, U is spanned by an independent list of length at most n = \dim V, which is a basis of U; so U is finite-dimensional and \dim U \le n. If \dim U = n, a basis of U is an independent list of length n in V, hence a basis of V by Result 5, and V, its span, lies in the subspace U (Result 1 there). So U = V.

Result 7. The axioms of Vector spaces from axioms hold for all complex scalars, so for real ones, and V is a vector space over \mathbb{R}. Each \mathbf{v} = \sum_j c_j\mathbf{v}_j with c_j = a_j + b_ji, a_j, b_j real (Complex numbers), and the distributive and compatibility axioms give

\mathbf{v} = \sum_{j=1}^{n} \big(a_j\mathbf{v}_j + b_j(i\mathbf{v}_j)\big),

so the longer list spans over \mathbb{R}. If such a real combination is \mathbf{0}, the same axioms give \sum_j (a_j + b_ji)\mathbf{v}_j = \mathbf{0}, and independence over \mathbb{C} gives a_j + b_ji = 0, so a_j = b_j = 0.

Worked examples

Example 1

Shrink, then extend

In \mathcal{P}_3(\mathbb{R}) let p_1 = 1 + z^2, p_2 = z - z^3, p_3 = 2 + z + 2z^2 - z^3, p_4 = z^2 + z^3, p_5 = 1 - z^3 and U = \operatorname{span}\{p_1, \dots, p_5\}. (a) Shrink (p_1, \dots, p_5) to a basis of U and give \dim U. (b) Extend that basis to a basis of \mathcal{P}_3(\mathbb{R}) with entries from (1, z, z^2, z^3).

Solution

Polynomials are equal exactly when their coefficients are (Result 2 of Polynomials and functions as vectors).

(a) Walk the list as in the proof of Result 2. p_1 \neq 0 stays. A multiple of p_1 has no z term, so p_2 stays. Next p_3 = 2p_1 + p_2 goes. If ap_1 + bp_2 = p_4, the constant terms give a = 0 and then the z^2 terms give 0 = 1, so p_4 stays. Finally p_5 = p_1 - p_4 goes. The basis is (p_1, p_2, p_4) and \dim U = 3.

(b) Append (1, z, z^2, z^3) and keep walking. If ap_1 + bp_2 + cp_4 = 1, the constant, z and z^2 terms give a = 1, b = 0, c = -1, and then the z^3 term is c - b = -1, not 0. So 1 stays. Now (p_1, p_2, p_4, 1) is independent of length 4 = \dim \mathcal{P}_3(\mathbb{R}), a basis by Result 5, and z, z^2, z^3, already in its span, are deleted. The extension is (p_1, p_2, p_4, 1). Appending (z^3, z^2, z, 1) instead keeps z^3 and gives (p_1, p_2, p_4, z^3): the result depends on the order.

Example 2

The magic squares have dimension three

Find a basis and the dimension of the space \mathcal{M} of real 3 \times 3 magic squares, and the coordinates of Samir’s square

\begin{bmatrix} 9 & 8 & 4 \\ 2 & 7 & 12 \\ 10 & 6 & 5 \end{bmatrix}.
Solution

Write a magic square with line sum s as

M = \begin{bmatrix} x_1 & x_2 & x_3 \\ x_4 & x_5 & x_6 \\ x_7 & x_8 & x_9 \end{bmatrix}.

The middle row, middle column and two diagonals cover every entry once except x_5, which they cover four times, and the three rows add to 3s. So 4s = 3s + 3x_5, and s = 3x_5. Put c = x_5, u = x_1 - c, v = x_3 - c. The top row gives x_2 = 3c - x_1 - x_3 = c - u - v, the diagonals give x_9 = c - u and x_7 = c - v, and the columns then give x_4, x_6, x_8:

M = \begin{bmatrix} c + u & c - u - v & c + v \\ c - u + v & c & c + u - v \\ c - v & c + u + v & c - u \end{bmatrix}.

That is, M = cJ + uA + vB, where

\begin{aligned} A &= \begin{bmatrix} 1 & -1 & 0 \\ -1 & 0 & 1 \\ 0 & 1 & -1 \end{bmatrix}, \\ B &= \begin{bmatrix} 0 & -1 & 1 \\ 1 & 0 & -1 \\ -1 & 1 & 0 \end{bmatrix}. \end{aligned}

Every line of J sums to 3 and every line of A and B to 0, so the three lie in \mathcal{M} and span it. The center and two top corners of cJ + uA + vB are c, c + u, c + v, so it is zero only if c = u = v = 0: the list is independent. So (J, A, B) is a basis and \dim \mathcal{M} = 3. By Result 4 every code needs exactly three numbers, such as the center and two top corners.

For Samir’s square, c = 7, u = 9 - 7 = 2, v = 4 - 7 = -3: it is 7J + 2A - 3B, coordinate vector (7, 2, -3), and each line sums to 21 = 3 \cdot 7.

Example 3

Counting replaces the spanning check

Let U = \{p \in \mathcal{P}_3(\mathbb{R}) : p(1) = p(-1)\}. (a) Find \dim U without solving for the general element. (b) Show that q_1 = z^2 - 1, q_2 = z^3 - z + 2, q_3 = z^3 + z^2 - z form a basis of U, and find the coordinates of p = 4z^3 + z^2 - 4z + 5.

Solution

(a) U contains 0 and is closed under sums and multiples, since (p + aq)(\pm 1) = p(\pm 1) + aq(\pm 1). As z \notin U, U \neq \mathcal{P}_3(\mathbb{R}), and Result 6 gives \dim U < 4. The list (1, z^2, z^3 - z) lies in U, and a + bz^2 + c(z^3 - z) = 0 forces c = b = a = 0 from the z^3, z^2 and constant terms. A basis of U spans U, so by the length inequality of Span and independence, proved it has at least 3 entries. So \dim U = 3.

(b) Each q_j takes equal values at \pm 1: 0, 2 and 1. If aq_1 + bq_2 + cq_3 = 0, the z^3 and z^2 terms give b = -c and a = -c, and the constant term -a + 2b = -c must be 0; so a = b = c = 0. Three independent vectors in a space of dimension 3 form a basis by Result 5, with no spanning check.

For p, p(1) = 6 = p(-1). Matching z^3, z^2 and constant terms in aq_1 + bq_2 + cq_3 = p gives b + c = 4, a + c = 1, -a + 2b = 5 (the z terms repeat the first). So -(1 - c) + 2(4 - c) = 7 - c = 5, c = 2, b = 2, a = -1:

p = -q_1 + 2q_2 + 2q_3.
Example 4

Two large subspaces must meet

Prove: if U and W are subspaces of a finite-dimensional V with \dim U + \dim W > \dim V, then U \cap W contains a nonzero vector. Find one for V = \mathcal{P}_3(\mathbb{R}), the U of Example 3, and W = \{p : p(0) = 0\}.

Solution

Let (\mathbf{u}_1, \dots, \mathbf{u}_k) and (\mathbf{w}_1, \dots, \mathbf{w}_l) be bases of U and W (Result 6). The list (\mathbf{u}_1, \dots, \mathbf{u}_k, \mathbf{w}_1, \dots, \mathbf{w}_l) has length k + l > \dim V, so it is dependent by Result 5:

\begin{aligned} &a_1\mathbf{u}_1 + \dots + a_k\mathbf{u}_k \\ &\quad + b_1\mathbf{w}_1 + \dots + b_l\mathbf{w}_l = \mathbf{0} \end{aligned}

with some coefficient nonzero. Then \mathbf{x} = \sum_j a_j\mathbf{u}_j = -\sum_j b_j\mathbf{w}_j lies in U \cap W. If \mathbf{x} = \mathbf{0}, independence of the \mathbf{u}’s gives every a_j = 0 and of the \mathbf{w}’s every b_j = 0, which is false. So \mathbf{x} \neq \mathbf{0}.

For the instance, W is the set of p with constant term 0, so (z, z^2, z^3) spans it and is independent: \dim W = 3. As 3 + 3 > 4, the two meet. One is z^2: its constant term is 0 and its values at \pm 1 are both 1.

Example 5

The field decides the dimension

Show that the 2 \times 2 complex matrices with H^{*} = H (H^{*} = \bar{H}^{T}, the conjugate transpose) form a vector space over \mathbb{R} but not over \mathbb{C}. Find its real dimension and the coordinates of

H = \begin{bmatrix} 3 & 1 - 2i \\ 1 + 2i & -1 \end{bmatrix}.
Solution

Call such matrices Hermitian. For Hermitian H, K and real r, (H + rK)^{*} = H^{*} + \bar{r}K^{*} = H + rK since \bar{r} = r, and 0^{*} = 0, so they form a real subspace of \mathbb{C}^{2 \times 2}, which has real dimension 2 \cdot 4 = 8 by Result 7. H^{*} = H makes each diagonal entry equal its own conjugate, so real; hence E_{11} is Hermitian but iE_{11} is not, and there is no complex subspace.

The condition says the diagonal is real and the corners are conjugates, so every Hermitian matrix is

\begin{aligned} &\begin{bmatrix} a & b + ci \\ b - ci & d \end{bmatrix} \\ &\quad = aE_{11} + dE_{22} + bS + cT \end{aligned}

with a, b, c, d real, S = E_{12} + E_{21} and T = iE_{12} - iE_{21}, both Hermitian. So (E_{11}, E_{22}, S, T) spans, and since the combination shows a, d and b + ci as entries, it is zero only if a = b = c = d = 0. The real dimension is 4. The given H is H = 3E_{11} - E_{22} + S - 2T, coordinates (3, -1, 1, -2).

Boundaries

A basis here is a finite list. \mathcal{P}(\mathbb{F}), \mathbb{F}^{\infty} and C[0, 1] have none (Result 2). Allowing infinite sets, with every vector a finite combination of members in exactly one way, makes 1, z, z^2, \dots a basis of \mathcal{P}(\mathbb{F}); that every space has such a basis rests on the axiom of choice.

Dimension depends on the field. \mathbb{C}^n has dimension n over \mathbb{C} and 2n over \mathbb{R}, and \mathbb{R} is infinite-dimensional over \mathbb{Q} (Span and independence, proved). The Hermitian matrices of Example 5 have a real dimension and no complex one.

Results 5 and 6 need the right space. The q_j of Example 3 are three independent vectors of \mathcal{P}_3(\mathbb{R}), a basis of U but not of \mathcal{P}_3(\mathbb{R}). Result 6 needs U \subseteq V: \mathcal{P}_2(\mathbb{R}) and that U both have dimension 3 inside \mathcal{P}_3(\mathbb{R}), yet z^3 - z \in U and z^3 - z \notin \mathcal{P}_2(\mathbb{R}).

Where this goes next

Sums and direct sums of subspaces counts the dimension of a sum of subspaces with these results. Null space, range and the fundamental theorem proves its fundamental theorem by extending a basis of the null space (Result 3). Orthonormal bases and Gram-Schmidt, proved uses Result 5 to see that Gram-Schmidt outputs a basis. Field extensions and degree reads a larger field as a vector space over a smaller one, as Result 7 reads \mathbb{C} over \mathbb{R}. Coordinates in a basis and change of basis did [\mathbf{v}]_{\mathcal{B}} in \mathbb{R}^n.

Exercises

Shrinking and extending

  1. In \mathcal{P}_2(\mathbb{R}), shrink (1 - z, 2 - 2z, z + z^2, 1 + z^2, z^2) to a basis of its span. What is the span?
  2. In \mathbb{R}^{2 \times 2} let M_1 = \begin{bmatrix} 1 & 1 \\ 0 & 1 \end{bmatrix} and M_2 = \begin{bmatrix} 0 & 1 \\ 1 & 0 \end{bmatrix}. Extend (M_1, M_2) to a basis with entries from (E_{11}, E_{12}, E_{21}, E_{22}), following the proof of Result 3.
  3. Find a basis and the dimension of \{p \in \mathcal{P}_3(\mathbb{R}) : p(1) = p(2) = 0\}.
  4. Find a basis and the dimension of the space of matrices in \mathbb{R}^{2 \times 3} whose rows and columns all sum to 0.
  5. (a) Find the coordinates of \begin{bmatrix} 4 & 9 & 2 \\ 3 & 5 & 7 \\ 8 & 1 & 6 \end{bmatrix} relative to the basis (J, A, B) of Example 2. (b) Find a basis and the dimension of the space of magic squares whose lines all sum to 0.

Counting instead of checking

  1. Show that (1 + 2z, z - z^2, 3 + z^2) is a basis of \mathcal{P}_2(\mathbb{R}) by checking only independence, and find the coordinates of 4 + z.
  2. Find a basis and the dimension of the space of real 3 \times 3 matrices A with A^{T} = -A.

Changing the field

  1. Find a basis of \mathbb{C}^2 as a vector space over \mathbb{R}, and the coordinates of (1 + 2i, 3 - i) in it.
  2. Regard \mathbb{R} as a vector space over \mathbb{Q}, and let \displaystyle \mathbb{Q}(\sqrt{2}) = \{a + b\sqrt{2} : a, b \in \mathbb{Q}\}. Show that it is a subspace with basis \displaystyle (1, \sqrt{2}), that \displaystyle (1 + \sqrt{2}, 1 - \sqrt{2}) is also a basis, and find the coordinates of \displaystyle 5 + \sqrt{2} in the second basis.
  3. Find the dimension of the space of 2 \times 2 complex matrices with trace 0, over \mathbb{C} and over \mathbb{R}, with a basis for each.

Proofs

  1. Prove that the symmetric n \times n matrices over \mathbb{F} form a subspace of \mathbb{F}^{n \times n} of dimension \displaystyle \frac{n(n + 1)}{2}.
  2. Let \dim V = n. Prove that for each k = 0, 1, \dots, n, V has a subspace of dimension k.
  3. Let U be a subspace of \mathcal{P}(\mathbb{F}) with \dim U = k \ge 1. Prove that U contains a polynomial of degree at least k - 1.
  4. Let V be finite-dimensional, U a subspace with \dim U = \dim V - 1, and \mathbf{v} \in V with \mathbf{v} \notin U. Prove that every \mathbf{x} \in V is \mathbf{u} + c\mathbf{v} for exactly one \mathbf{u} \in U and one c \in \mathbb{F}.
  5. Let (\mathbf{v}_1, \dots, \mathbf{v}_n) be a basis of V and (\mathbf{w}_1, \dots, \mathbf{w}_n) a list in V whose span contains every \mathbf{v}_j. Prove that (\mathbf{w}_1, \dots, \mathbf{w}_n) is a basis of V.
Answers
  1. Basis (1 - z, z + z^2, z^2); the span is \mathcal{P}_2(\mathbb{R}). 2 - 2z = 2(1 - z) and 1 + z^2 = (1 - z) + (z + z^2) go; z + z^2 has a z^2 term, so it stays; a(1 - z) + b(z + z^2) = z^2 forces a = 0, then b = 0, so z^2 stays. Three independent vectors in \mathcal{P}_2(\mathbb{R}) span it by Result 5.
  2. (M_1, M_2, E_{11}, E_{12}). aM_1 + bM_2 + cE_{11} = \begin{bmatrix} a + c & a + b \\ b & a \end{bmatrix}. With c = 0 it equals E_{11} only if a = 1 and a = 0, so E_{11} stays; it equals E_{12} only if a = b = 0 and a + b = 1, so E_{12} stays. Length 4 = \dim \mathbb{R}^{2 \times 2} makes it a basis (Result 5), and E_{21}, E_{22} are deleted.
  3. Basis (z^2 - 3z + 2, z^3 - 3z^2 + 2z); dimension 2. By the The remainder and factor theorems, p(1) = 0 gives p = (z - 1)q, and q(2) = 0 gives q = (z - 2)(a + bz). So p = a(z - 1)(z - 2) + bz(z - 1)(z - 2), and the two, of degrees 2 and 3, are independent.
  4. Basis \begin{bmatrix} 1 & 0 & -1 \\ -1 & 0 & 1 \end{bmatrix}, \begin{bmatrix} 0 & 1 & -1 \\ 0 & -1 & 1 \end{bmatrix}; dimension 2. Row 1 is (a, b, -a - b), and zero column sums make row 2 its negative. So the matrix is a times the first plus b times the second, and entries (1, 1), (1, 2) show independence.
  5. (a) (5, -1, -3): c = 5, u = 4 - 5, v = 2 - 5, and 5J - A - 3B reproduces every entry. (b) Basis (A, B), dimension 2. Since s = 3c, the line sum is 0 exactly when c = 0, leaving uA + vB; A, B are independent as a sublist of a basis.
  6. Coordinates \displaystyle (-\frac{1}{5}, \frac{7}{5}, \frac{7}{5}). If a(1 + 2z) + b(z - z^2) + c(3 + z^2) = 0, then c = b, \displaystyle a = -\frac{b}{2} and \displaystyle a + 3c = \frac{5b}{2} = 0, so all are 0, and Result 5 gives a basis. For 4 + z: a + 3c = 4, 2a + b = 1, c = b give \displaystyle b = c = \frac{7}{5}, \displaystyle a = -\frac{1}{5}.
  7. Basis (E_{12} - E_{21}, E_{13} - E_{31}, E_{23} - E_{32}); dimension 3. A^{T} = -A forces a_{jj} = -a_{jj}, so a_{jj} = 0, and a_{kj} = -a_{jk}. So A is a_{12}(E_{12} - E_{21}) plus a_{13}(E_{13} - E_{31}) plus a_{23}(E_{23} - E_{32}), and the entries above the diagonal show independence.
  8. Basis (\mathbf{e}_1, i\mathbf{e}_1, \mathbf{e}_2, i\mathbf{e}_2) by Result 7; coordinates (1, 2, 3, -1), since (1 + 2i, 3 - i) = \mathbf{e}_1 + 2(i\mathbf{e}_1) + 3\mathbf{e}_2 - (i\mathbf{e}_2).
  9. Coordinates (3, 2). \displaystyle \mathbb{Q}(\sqrt{2}) = \operatorname{span}\{1, \sqrt{2}\} over \mathbb{Q}, a subspace (Result 1 of Span and independence, proved). If \displaystyle a + b\sqrt{2} = 0 with b \neq 0, then \displaystyle \sqrt{2} = -\frac{a}{b} is rational, which is false; so b = a = 0. If \displaystyle a(1 + \sqrt{2}) + b(1 - \sqrt{2}), which is \displaystyle (a + b) + (a - b)\sqrt{2}, is 0, then a + b = a - b = 0, so a = b = 0, and Result 5 gives a basis. Matching a + b = 5, a - b = 1 gives a = 3, b = 2.
  10. Over \mathbb{C}: basis (E_{11} - E_{22}, E_{12}, E_{21}), dimension 3, since \begin{bmatrix} a & b \\ c & -a \end{bmatrix} = a(E_{11} - E_{22}) + bE_{12} + cE_{21} shows a, b, c as entries. Over \mathbb{R}, by Result 7: basis (E_{11} - E_{22}, i(E_{11} - E_{22}), E_{12}, iE_{12}, E_{21}, iE_{21}), dimension 6.
  11. If A^{T} = A and B^{T} = B, then (A + cB)^{T} = A + cB, and 0^{T} = 0: a subspace. For j < k let S_{jk} = E_{jk} + E_{kj}, and let S_{jj} = E_{jj}; each is symmetric. A symmetric A has a_{kj} = a_{jk}, so A = \sum_{j \le k} a_{jk}S_{jk}: the list spans. In \sum_{j \le k} c_{jk}S_{jk} the entry in place (j, k), j \le k, is c_{jk}, since only S_{jk} is nonzero there; so the sum is 0 only if every c_{jk} = 0. The pairs j \le k number \displaystyle n + \frac{n(n - 1)}{2} = \frac{n(n + 1)}{2}.
  12. Let (\mathbf{v}_1, \dots, \mathbf{v}_n) be a basis (Result 2) and U_k = \operatorname{span}\{\mathbf{v}_1, \dots, \mathbf{v}_k\}, with U_0 = \{\mathbf{0}\}, a subspace (Result 1 of Span and independence, proved). The list (\mathbf{v}_1, \dots, \mathbf{v}_k) spans U_k and is independent as a sublist of an independent list (Result 2 there), so \dim U_k = k.
  13. Let (p_1, \dots, p_k) be a basis of U. No p_j is 0 (Result 2 of Span and independence, proved), so m = \max_j \deg p_j \ge 0. Each p_j lies in the subspace \mathcal{P}_m(\mathbb{F}), so U \subseteq \mathcal{P}_m(\mathbb{F}) (Result 1 there), and Results 6 and 4 give k \le m + 1. So m \ge k - 1, and the p_j of degree m is the polynomial required.
  14. Let n = \dim V and (\mathbf{u}_1, \dots, \mathbf{u}_{n-1}) a basis of U. As \mathbf{v} is outside its span U, the list (\mathbf{u}_1, \dots, \mathbf{u}_{n-1}, \mathbf{v}) is independent (Result 3 of Span and independence, proved), of length n, so a basis of V (Result 5). Existence: \mathbf{x} = \sum_j a_j\mathbf{u}_j + c\mathbf{v}, and \mathbf{u} = \sum_j a_j\mathbf{u}_j \in U. Uniqueness: if \mathbf{u} + c\mathbf{v} = \mathbf{u}' + c'\mathbf{v} with \mathbf{u} = \sum_j a_j\mathbf{u}_j and \mathbf{u}' = \sum_j a'_j\mathbf{u}_j, both sides are combinations of the basis, so Result 1 gives c = c' and every a_j = a'_j, hence \mathbf{u} = \mathbf{u}'.
  15. \operatorname{span}\{\mathbf{w}_1, \dots, \mathbf{w}_n\} is a subspace containing every \mathbf{v}_j, so it contains their span V (Result 1 of Span and independence, proved). So the \mathbf{w}’s span V, and their number is n = \dim V (Result 4), so they form a basis by Result 5.
math.wiki · CC-BY-SA-4.0