math.wiki
Linear algebra › Abstract vector spaces › Vector spaces from axioms
Sign in

Vector spaces from axioms

Polynomials, matrices, sequences and functions can all be added and scaled, just like columns of numbers. Which rules must adding and scaling obey so that a theorem of linear algebra, proved once, holds for all of them?

Where you are 5 in Abstract vector spaces

Why this exists

The written course works in \mathbb{R}^n (Adding and scaling vectors in n dimensions), yet its proofs rarely look inside a vector. The proof in Subspaces that a span is a subspace distributes scalars and regroups sums; it never asks for the third entry of a vector. Such a proof works word for word for anything added and scaled under the same rules.

Plenty is. Yuna and Malik run a seismometer station that samples ground motion 100 times a second, so a ten-minute record is 60{,}000 readings. Malik stacks two records by adding them reading by reading; Yuna applies a gain, multiplying every reading by one number. Polynomials, matrices and functions are added and scaled alike. So linear algebra lists the rules that adding and scaling obey in \mathbb{R}^n, calls any set whose operations obey them a vector space, and proves its theorems from the list alone. Each theorem then holds in every vector space at once.

Definition

The scalars come from \mathbb{F}, meaning \mathbb{R} or the Complex numbers \mathbb{C}. Every result in this lesson holds, with the same proof, over any field (Fields), such as \mathbb{Q}: the proofs use only the arithmetic of \mathbb{F} and the fact that a nonzero scalar has a reciprocal.

An addition on a set V is a function assigning to each pair \mathbf{u}, \mathbf{v} \in V an element \mathbf{u} + \mathbf{v} \in V. A scalar multiplication on V assigns to each a \in \mathbb{F} and \mathbf{v} \in V an element a\mathbf{v} \in V. Closure is built in: both results must lie in V.

A vector space over \mathbb{F} is a set V with an addition and a scalar multiplication such that these eight axioms hold for all \mathbf{u}, \mathbf{v}, \mathbf{w} \in V and all a, b \in \mathbb{F}.

AxiomNameStatement
1commutativity\mathbf{u} + \mathbf{v} = \mathbf{v} + \mathbf{u}
2associativity of addition(\mathbf{u} + \mathbf{v}) + \mathbf{w} = \mathbf{u} + (\mathbf{v} + \mathbf{w})
3additive identitysome \mathbf{0} \in V has \mathbf{v} + \mathbf{0} = \mathbf{v} for every \mathbf{v}
4additive inverseeach \mathbf{v} has some \mathbf{w} \in V with \mathbf{v} + \mathbf{w} = \mathbf{0}
5multiplicative identity1\mathbf{v} = \mathbf{v}
6associativity of scaling(ab)\mathbf{v} = a(b\mathbf{v})
7distributivity over vectorsa(\mathbf{u} + \mathbf{v}) = a\mathbf{u} + a\mathbf{v}
8distributivity over scalars(a + b)\mathbf{v} = a\mathbf{v} + b\mathbf{v}

The elements of V are vectors, whatever they are. A vector space over \mathbb{R} is real, one over \mathbb{C} complex. In Axiom 8, a + b adds scalars while the right side adds vectors.

A subspace of V is a subset U \subseteq V that is itself a vector space under the operations of V.

The standard examples, with the expected operations:

  • \mathbb{F}^n, lists added and scaled entry by entry, and the zero space \{\mathbf{0}\}.
  • \mathbb{F}^{S}, for a nonempty set S: all functions f: S \to \mathbb{F}, with (f + g)(x) = f(x) + g(x) and (af)(x) = af(x).
  • \mathbb{F}^{\infty}, sequences (x_1, x_2, \dots), which is \mathbb{F}^{S} for S = \{1, 2, 3, \dots\}. Likewise \mathbb{F}^n is \mathbb{F}^{S} for S = \{1, \dots, n\}, and the matrices \mathbb{F}^{m \times n} are \mathbb{F}^{S} for S = \{1, \dots, m\} \times \{1, \dots, n\} (Cartesian product).
  • \mathcal{P}(\mathbb{F}), the polynomials, and \mathcal{P}_m(\mathbb{F}), those of degree at most m, studied in Polynomials and functions as vectors.

A seismometer record assigns a reading to each time in the set T of sample times, so the records form \mathbb{R}^{T}; stacking is its addition and a gain its scalar multiplication.

Intuition

The axioms are a contract. A set with operations qualifies by passing eight checks, and in return every theorem proved from them applies to it. A proof about vector spaces may use the eight axioms and the arithmetic of \mathbb{F}, and nothing else: no entries, no list of zeros, no product of vectors, no length. So even 0\mathbf{v} = \mathbf{0} and -\mathbf{v} = (-1)\mathbf{v} must be earned, which the Results do. After that, vector algebra runs exactly as in \mathbb{R}^n.

The zero vector need not be a list of zeros (Example 2), and a familiar set with an unfamiliar operation may break an axiom (Example 3).

Results

Throughout, V is a vector space over \mathbb{F}, with \mathbf{u}, \mathbf{v}, \mathbf{w} \in V and a \in \mathbb{F}. Every result holds over any field.

Result 1 (the zero vector is unique). V has exactly one additive identity.

Result 2 (inverses are unique). Each \mathbf{v} has exactly one additive inverse, written -\mathbf{v}; and \mathbf{w} - \mathbf{v} means \mathbf{w} + (-\mathbf{v}).

Result 3 (cancellation). If \mathbf{u} + \mathbf{w} = \mathbf{v} + \mathbf{w}, then \mathbf{u} = \mathbf{v}. The equation \mathbf{v} + \mathbf{x} = \mathbf{w} has exactly one solution, \mathbf{x} = \mathbf{w} - \mathbf{v}.

Result 4 (the scalar zero). 0\mathbf{v} = \mathbf{0}.

Result 5 (the zero vector scaled). a\mathbf{0} = \mathbf{0}.

Result 6 (minus one). (-1)\mathbf{v} = -\mathbf{v}.

Result 7 (no zero divisors). If a\mathbf{v} = \mathbf{0}, then a = 0 or \mathbf{v} = \mathbf{0}.

Result 8 (function spaces). For every nonempty set S, \mathbb{F}^{S} with pointwise operations is a vector space over \mathbb{F}, with the zero function as zero vector and (-f)(x) = -f(x). In particular \mathbb{F}^n, \mathbb{F}^{\infty} and \mathbb{F}^{m \times n} are vector spaces.

Result 9 (subspace test). A subset U of V is a subspace if and only if \mathbf{0} \in U, \mathbf{u} + \mathbf{w} \in U for all \mathbf{u}, \mathbf{w} \in U, and a\mathbf{u} \in U for all a \in \mathbb{F} and \mathbf{u} \in U. A subspace has the same zero vector as V. This extends the test of Subspaces.

Why it is true

Result 1. If \mathbf{0} and \mathbf{0}' are both additive identities, then by Axiom 3 for \mathbf{0}, Axiom 1, and Axiom 3 for \mathbf{0}',

\mathbf{0}' = \mathbf{0}' + \mathbf{0} = \mathbf{0} + \mathbf{0}' = \mathbf{0}.

Result 2. Let \mathbf{w} and \mathbf{w}' be additive inverses of \mathbf{v}. By Axioms 3, 4, 2, 1, 4, 1 and 3 in turn,

\begin{aligned}\mathbf{w} &= \mathbf{w} + \mathbf{0} = \mathbf{w} + (\mathbf{v} + \mathbf{w}') \\ &= (\mathbf{w} + \mathbf{v}) + \mathbf{w}' = (\mathbf{v} + \mathbf{w}) + \mathbf{w}' \\ &= \mathbf{0} + \mathbf{w}' = \mathbf{w}' + \mathbf{0} = \mathbf{w}'.\end{aligned}

Result 3. Add -\mathbf{w} on the right of both sides. By Axioms 2, 4 and 3 the left side becomes \mathbf{u} + (\mathbf{w} + (-\mathbf{w})), then \mathbf{u} + \mathbf{0}, then \mathbf{u}; the right side becomes \mathbf{v}. For the equation, Axioms 1, 2, 4 and 3 show that \mathbf{w} - \mathbf{v} is a solution:

\begin{aligned}\mathbf{v} + (\mathbf{w} + (-\mathbf{v})) &= (\mathbf{v} + (-\mathbf{v})) + \mathbf{w} \\ &= \mathbf{0} + \mathbf{w} = \mathbf{w}.\end{aligned}

If \mathbf{x} and \mathbf{x}' are solutions, then \mathbf{x} + \mathbf{v} = \mathbf{x}' + \mathbf{v} by Axiom 1, and cancellation gives \mathbf{x} = \mathbf{x}'.

Result 4. By Axiom 8, 0\mathbf{v} = (0 + 0)\mathbf{v} = 0\mathbf{v} + 0\mathbf{v}, and by Axioms 3 and 1, 0\mathbf{v} = \mathbf{0} + 0\mathbf{v}. Cancelling 0\mathbf{v} (Result 3) in \mathbf{0} + 0\mathbf{v} = 0\mathbf{v} + 0\mathbf{v} gives \mathbf{0} = 0\mathbf{v}.

Result 5. The same argument with Axioms 3 and 7: a\mathbf{0} = a(\mathbf{0} + \mathbf{0}) = a\mathbf{0} + a\mathbf{0}, and a\mathbf{0} = \mathbf{0} + a\mathbf{0}; cancel a\mathbf{0}.

Result 6. By Axioms 5 and 8 and Result 4,

\mathbf{v} + (-1)\mathbf{v} = (1 + (-1))\mathbf{v} = 0\mathbf{v} = \mathbf{0},

so (-1)\mathbf{v} is an additive inverse of \mathbf{v}, and by Result 2 it is -\mathbf{v}.

Result 7. If a\mathbf{v} = \mathbf{0} and a \neq 0, then a has a reciprocal, and by Axioms 5 and 6 and Result 5,

\begin{aligned}\mathbf{v} &= 1\mathbf{v} = (a^{-1}a)\mathbf{v} \\ &= a^{-1}(a\mathbf{v}) = a^{-1}\mathbf{0} = \mathbf{0}.\end{aligned}

Result 8. The sum and multiples of functions S \to \mathbb{F} are functions S \to \mathbb{F}. Two functions are equal when they agree at every x \in S, so evaluate each axiom at a point. Axiom 8 becomes

\begin{aligned}((a + b)f)(x) &= (a + b)f(x) \\ &= af(x) + bf(x) \\ &= (af + bf)(x),\end{aligned}

the distributive law in \mathbb{F}. Likewise Axioms 1, 2, 5, 6 and 7 become, at each point, the commutative and associative laws of addition in \mathbb{F}, the law 1c = c, the associative law of multiplication, and a(c + d) = ac + ad. For Axioms 3 and 4, f(x) + 0 = f(x) and f(x) - f(x) = 0.

Result 9. If U is a subspace, its operations land in U, so it is closed under both. Its identity \mathbf{0}_U satisfies \mathbf{0}_U + \mathbf{0}_U = \mathbf{0}_U = \mathbf{0} + \mathbf{0}_U in V, and cancellation gives \mathbf{0}_U = \mathbf{0}, so \mathbf{0} \in U. Conversely, if U meets the three conditions, the operations of V are operations on U. Axioms 1, 2 and 5 to 8 hold for all vectors of V, so for those in U; Axiom 3 holds since \mathbf{0} \in U; and for \mathbf{u} \in U, -\mathbf{u} = (-1)\mathbf{u} \in U by Result 6 and closure.

Worked examples

Example 1

Which records form a subspace

At Yuna and Malik’s station the records form \mathbb{R}^{T}, where T is the set of 60{,}000 sample times. Which are subspaces: (a) the records reading 0 at the recalibration time t_0; (b) the records with every reading in [-1, 1], the range the digitizer stores; (c) the records f with \sum_{t \in T} f(t) = 0, the output of Malik’s detrending step?

Solution

Run the three checks of Result 9.

(a) The zero record reads 0 at t_0, and if f(t_0) = g(t_0) = 0, then (f + g)(t_0) = f(t_0) + g(t_0) = 0 and (af)(t_0) = af(t_0) = 0. A subspace.

(b) The constant record f(t) = 0.6 is in the set, but f + f reads 1.2 at every time, outside [-1, 1]. Not closed under addition, so not a subspace. A vector space has no edge that stacking cannot cross.

(c) The zero record sums to 0. For f, g in the set and a \in \mathbb{R}, regrouping the finite sum gives

\begin{aligned}\textstyle\sum_{t}(af + g)(t) &= \textstyle a\sum_{t} f(t) + \sum_{t} g(t) \\ &= a \cdot 0 + 0 = 0.\end{aligned}

Taking a = 1 gives closure under addition, and g = 0 closure under scaling. A subspace.

Example 2

A line whose zero vector is the number 1

On \mathbb{R} define x \oplus y = x + y - 1 and c \odot x = c(x - 1) + 1. Show that this is a real vector space, and find its zero vector and inverses.

Solution

Both formulas give real numbers, so these are operations on \mathbb{R}. Axiom 3 asks for e with x + e - 1 = x for all x, so the zero vector is 1. Axiom 4 asks for w with x + w - 1 = 1, so the inverse of x is 2 - x. For all real x, y, z, a, b:

  • Axioms 1 and 2: x + y - 1 = y + x - 1, and both groupings of x \oplus y \oplus z equal x + y + z - 2.
  • Axioms 5 and 6: 1 \odot x = x and a \odot (b \odot x) = ab(x - 1) + 1 = (ab) \odot x hold.
  • Axiom 7: both sides equal a(x - 1) + a(y - 1) + 1.
  • Axiom 8: both sides equal a(x - 1) + b(x - 1) + 1.

All eight hold, so the Results apply with no new proof. Result 4 says 0 \odot x is the zero vector: 0(x - 1) + 1 = 1. Result 6 says (-1) \odot x is the inverse: -(x - 1) + 1 = 2 - x. Under x \mapsto x - 1 the operations become ordinary addition and scaling, so this is the ordinary line with its numbers moved over by 1.

Example 3

Scaling by the square of the scalar

On \mathbb{R}^2 keep the usual addition and define c \odot (x, y) = (c^2x, c^2y). Which axioms hold?

Solution

Axioms 1 to 4 involve only the usual addition, so they hold. In ordinary terms c \odot \mathbf{v} = c^2\mathbf{v}, and Axioms 5, 6 and 7 hold because 1^2 = 1, (ab)^2 = a^2b^2 and a^2(\mathbf{u} + \mathbf{v}) = a^2\mathbf{u} + a^2\mathbf{v}. Axiom 8 asks for (a + b)^2\mathbf{v} = a^2\mathbf{v} + b^2\mathbf{v}, which fails whenever 2ab\mathbf{v} \neq \mathbf{0}. With a = b = 1 and \mathbf{v} = (1, 0), 2 \odot (1, 0) = (4, 0), while 1 \odot (1, 0) added to itself is (2, 0).

So this is not a vector space, and theorems may fail in it. Result 6 does: (-1) \odot (1, 0) = (1, 0), not the inverse (-1, 0). Since the other seven axioms hold here, Axiom 8 does not follow from them.

Example 4

Solving a vector equation

In a vector space V, solve 4\mathbf{x} - \mathbf{u} = \mathbf{x} + 2\mathbf{w}, naming each step’s reason. Then find \mathbf{x} in \mathbb{R}^3 for \mathbf{u} = (3, -6, 9) and \mathbf{w} = (0, 3, 3).

Solution

Add \mathbf{u} - \mathbf{x} to both sides and regroup by Axioms 1 and 2. On the left, \mathbf{u} - \mathbf{u} = \mathbf{0} by Axiom 4, and 4\mathbf{x} + (-1)\mathbf{x} = 3\mathbf{x} by Result 6 and Axiom 8; the right side becomes \mathbf{u} + 2\mathbf{w}. So 3\mathbf{x} = \mathbf{u} + 2\mathbf{w}. Scaling by \tfrac{1}{3} and using Axioms 5, 6 and 7,

\mathbf{x} = \tfrac{1}{3}\mathbf{u} + \tfrac{2}{3}\mathbf{w}.

Both steps can be undone (add \mathbf{x} - \mathbf{u}, scale by 3), so this is the only solution: the algebra of \mathbb{R}^n, now justified in every vector space. In \mathbb{R}^3, \mathbf{x} = (1, -2, 3) + (0, 2, 2) = (1, 0, 5) and, as a check, 4\mathbf{x} - \mathbf{u} = (4, 0, 20) - (3, -6, 9) = (1, 6, 11) and \mathbf{x} + 2\mathbf{w} = (1, 0, 5) + (0, 6, 6) = (1, 6, 11) agree.

Example 5

A nonzero vector has infinitely many multiples

Let V be a vector space over \mathbb{R} or \mathbb{C} and \mathbf{v} \neq \mathbf{0}. Prove that a\mathbf{v} \neq b\mathbf{v} whenever a \neq b, and conclude that the only finite real or complex vector space is \{\mathbf{0}\}.

Solution

By Proof by contraposition, show that a\mathbf{v} = b\mathbf{v} forces a = b. Add -(b\mathbf{v}) to both sides. Result 6 and Axiom 6 give -(b\mathbf{v}) = (-1)(b\mathbf{v}) = (-b)\mathbf{v}, so Axioms 8 and 4 turn the equation into (a - b)\mathbf{v} = \mathbf{0}. Since \mathbf{v} \neq \mathbf{0}, Result 7 gives a - b = 0.

So a \mapsto a\mathbf{v} is an injective function from the infinite set \mathbb{F} into V (One-to-one, onto and bijective functions), and a space with a nonzero vector is infinite. Only the last step used that \mathbb{F} is infinite (see Boundaries).

Example 6

Real scalars or complex scalars

In \mathbb{C}^2 let U = \{(z, \bar{z}) : z \in \mathbb{C}\}. Is U a subspace of the complex vector space \mathbb{C}^2? Of \mathbb{C}^2 with the same addition but only real scalars?

Solution

With real scalars, \mathbb{C}^2 is a real vector space: each axiom for real a, b is a case of the axiom for complex a, b.

Over \mathbb{C}: (1, 1) \in U, but i(1, 1) = (i, i) is not, because \bar{i} = -i \neq i. Not a subspace.

Over \mathbb{R}: (0, 0) \in U; the conjugate of a sum is the sum of the conjugates, so (z + w, \bar{z} + \bar{w}) \in U; and a real c has \bar{c} = c, so (cz, c\bar{z}) = (cz, \overline{cz}) \in U. By Result 9, U is a subspace. With z = s + ti,

(s + ti, s - ti) = s(1, 1) + t(i, -i),

the real combinations of (1, 1) and (i, -i). The same set and addition have different subspaces over different scalars.

Boundaries

The scalars must form a field. Result 7 divides by a nonzero scalar. The integer pairs \mathbb{Z}^2 obey all eight rules with integer scalars, but over \mathbb{R} they are not closed: \tfrac{1}{2}(1, 0) \notin \mathbb{Z}^2. The multiples of 3, closed under the integer combinations of Proving statements about divisibility, fail the same way under scaling by \tfrac{1}{3}. Such structures (modules) belong to abstract algebra.

A finite field gives finite spaces. Over the integers mod 2 (Modular arithmetic), where 1 + 1 = 0, the lists of length n form a vector space with 2^n vectors, and \mathbf{v} + \mathbf{v} = \mathbf{0} for every \mathbf{v}. Results 1 to 9 still hold; Example 5’s conclusion and Exercise 11 do not.

The empty set is not a vector space. Axiom 3 demands an element. That is why the subspace test asks for \mathbf{0} \in U and not closure alone (Exercise 13).

No length, angle or product. In \mathbb{R}^n these come from the dot product, extra structure that inner products (Inner products and norms) generalize.

Where this goes next

Polynomials and functions as vectors works through polynomials, functions and sequences. Span and independence, proved and Bases and dimension, proved prove in any vector space what the written course found in \mathbb{R}^n, and Sums and direct sums of subspaces combines subspaces. A function between vector spaces that respects both operations is a linear map (Linear maps between vector spaces), as x \mapsto x - 1 is in Example 2. In abstract algebra, Axioms 1 to 4 say a vector space under addition is an abelian group (Groups), and Field extensions and degree treats a larger field as a vector space over a smaller one.

Exercises

Checking the axioms

  1. On \mathbb{R}^2 let (x_1, y_1) \oplus (x_2, y_2) be (x_1 + x_2 - 2, y_1 + y_2 + 3) and let c \odot (x, y) be (c(x - 2) + 2, c(y + 3) - 3), a vector space by the method of Example 2. Find the zero vector and the inverse of (5, 1), check that (-1) \odot (5, 1) is that inverse, and compute (3 \odot (5, 1)) \oplus ((-1) \odot (4, 0)).
  2. On \mathbb{R}^2 keep the usual addition and define c \odot (x, y) = (cx, -cy). Which axioms fail? Give numbers that break each.
  3. On \mathbb{R}^2 keep the usual scaling and define \mathbf{u} \oplus \mathbf{v} = \mathbf{u} + 2\mathbf{v}. Which axioms fail? Give numbers that break each.
  4. On the real 2 \times 2 matrices define A \oplus B = AB, with the usual scaling. Show that Axiom 1 fails for A = \begin{bmatrix} 1 & 1 \\ 0 & 1 \end{bmatrix} and B = \begin{bmatrix} 1 & 0 \\ 1 & 1 \end{bmatrix}, find the only possible zero vector, and show that Axiom 4 fails.

Subspaces

  1. Which are subspaces of \mathbb{R}^{2 \times 2}? (a) the matrices with trace 0; (b) those with determinant 0; (c) the symmetric ones, A^{T} = A.
  2. Which are subspaces of the complex vector space \mathbb{C}^3? (a) z_1 + 2iz_2 - z_3 = 0; (b) z_1z_2 = 0; (c) \lvert z_1 \rvert = \lvert z_2 \rvert.
  3. At the station, let t_1 \neq t_2 be sample times. Which sets of records are subspaces of \mathbb{R}^{T}? (a) f(t_1) = 2f(t_2); (b) f(t_1) \geq 0; (c) largest reading minus smallest at most 0.5.
  4. Show that the sequences in \mathbb{R}^{\infty} with x_{k+2} = x_{k+1} + x_k for all k \geq 1 form a subspace. Write the first six terms of the one with x_1 = 2, x_2 = -1, add the one with x_1 = x_2 = 1, and check the rule on the sum.
  5. In a vector space, solve 3(\mathbf{x} - \mathbf{u}) + 2\mathbf{w} = \mathbf{x} + \mathbf{u}, naming each step’s reason. Then find \mathbf{x} in \mathbb{C}^2 for \mathbf{u} = (1 + i, 2) and \mathbf{w} = (3, -i).

Proofs

  1. Prove that (-a)\mathbf{v} = -(a\mathbf{v}) = a(-\mathbf{v}).
  2. Let V be a vector space over \mathbb{R} or \mathbb{C}. Prove that \mathbf{v} + \mathbf{v} = \mathbf{0} implies \mathbf{v} = \mathbf{0}.
  3. Prove that the intersection of any nonempty collection of subspaces of V is a subspace.
  4. Prove that a nonempty subset of V closed under addition and scalar multiplication is a subspace, and show that “nonempty” cannot be dropped.
  5. Let U be a subspace of V and \mathbf{v} \in V. Prove that \mathbf{v} + U = \{\mathbf{v} + \mathbf{u} : \mathbf{u} \in U\} is a subspace if and only if \mathbf{v} \in U.
  6. Prove that Axioms 2 to 8 imply Axiom 1. (First show \mathbf{w} + \mathbf{v} = \mathbf{0} and \mathbf{0} + \mathbf{v} = \mathbf{v} from Axioms 3 and 4 as stated; then expand (1 + 1)(\mathbf{u} + \mathbf{v}) two ways.)
Answers
  1. Zero (2, -3); inverse (-1, -7); combination (9, 6). The zero needs x + e_1 - 2 = x and y + e_2 + 3 = y; the inverse needs 5 + w_1 - 2 = 2 and 1 + w_2 + 3 = -3. Also (-1) \odot (5, 1) = (-3 + 2, -4 - 3) = (-1, -7) as Result 6 says. Then 3 \odot (5, 1) = (11, 9) and (-1) \odot (4, 0) = (0, -6), so the combination is (11 + 0 - 2, 9 - 6 + 3) = (9, 6).
  2. Axioms 5 and 6 fail. 1 \odot (1, 1) = (1, -1); and with \mathbf{v} = (1, 1), 6 \odot \mathbf{v} = (6, -6) but 2 \odot (3 \odot \mathbf{v}) = 2 \odot (3, -3) = (6, 6). Axioms 1 to 4 concern the usual addition, and 7 and 8 hold since (ax, -ay) is linear in both a and (x, y).
  3. Axioms 1, 2 and 8 fail. (1, 0) \oplus (0, 0) = (1, 0) but (0, 0) \oplus (1, 0) = (2, 0). With \mathbf{w} = (1, 0), (\mathbf{0} \oplus \mathbf{0}) \oplus \mathbf{w} = (2, 0) but \mathbf{0} \oplus (\mathbf{0} \oplus \mathbf{w}) = (0, 0) \oplus (2, 0) = (4, 0) for the other grouping. With a = b = 1, 2\mathbf{w} = (2, 0) but \mathbf{w} \oplus \mathbf{w} = (3, 0). Axiom 3 holds with zero (0, 0), Axiom 4 with inverse -\tfrac{1}{2}\mathbf{v}, 5 and 6 by usual scaling, and 7 since a(\mathbf{u} + 2\mathbf{v}) = a\mathbf{u} + 2a\mathbf{v}.
  4. AB = \begin{bmatrix} 2 & 1 \\ 1 & 1 \end{bmatrix} but BA = \begin{bmatrix} 1 & 1 \\ 1 & 2 \end{bmatrix}. A zero E needs AE = A for all A; A = I_2 forces E = I_2, which works. The zero matrix has OW = O \neq I_2 for every W, so no inverse.
  5. (a) Yes: \operatorname{tr}(cA + B) = c\operatorname{tr} A + \operatorname{tr} B, and \operatorname{tr} O = 0. (b) No: \begin{bmatrix} 1 & 0 \\ 0 & 0 \end{bmatrix} and \begin{bmatrix} 0 & 0 \\ 0 & 1 \end{bmatrix} have determinant 0, but their sum I_2 has determinant 1. (c) Yes: (cA + B)^{T} = cA^{T} + B^{T} (The transpose and symmetric matrices).
  6. (a) Yes: the left side is linear in \mathbf{z}, so solutions are kept by sums and complex multiples, and \mathbf{0} is one. (b) No: (1, 0, 0) and (0, 1, 0) are in it, but (1, 1, 0) has z_1z_2 = 1. (c) No: (1, 1, 0) and (1, -1, 0) are in it, but their sum is (2, 0, 0).
  7. (a) Yes: the condition is kept by sums and multiples, since (af + g)(t_1) = 2(af + g)(t_2), and holds at 0. (b) No: if g(t_1) = 1, then (-1)g reads -1 there. (c) No: the record reading 0.4 at t_1 and 0 elsewhere has spread 0.4, and twice it has spread 0.8.
  8. The zero sequence obeys the rule, and (cx_{k+2} + y_{k+2}) = (cx_{k+1} + y_{k+1}) + (cx_k + y_k) when \mathbf{x} and \mathbf{y} do. Terms 2, -1, 1, 0, 1, 1 plus 1, 1, 2, 3, 5, 8 give 3, 0, 3, 3, 6, 9, and 3 + 0 = 3, 0 + 3 = 3, 3 + 3 = 6, 3 + 6 = 9.
  9. \mathbf{x} = 2\mathbf{u} - \mathbf{w}, so in \mathbb{C}^2, \mathbf{x} = (-1 + 2i, 4 + i). By Axiom 7 and Exercise 10, 3(\mathbf{x} - \mathbf{u}) = 3\mathbf{x} - 3\mathbf{u}. Adding 3\mathbf{u} - 2\mathbf{w} - \mathbf{x} and using Axioms 1, 2, 4 and 8 and Result 6 gives 2\mathbf{x} = 4\mathbf{u} - 2\mathbf{w}; scale by \tfrac{1}{2}. Check: 3(\mathbf{x} - \mathbf{u}) + 2\mathbf{w} = (-6 + 3i, 6 + 3i) + (6, -2i) = (3i, 6 + i) and \mathbf{x} + \mathbf{u} = (3i, 6 + i) agree.
  10. By Axiom 8 and Result 4, a\mathbf{v} + (-a)\mathbf{v} = 0\mathbf{v} = \mathbf{0}, so (-a)\mathbf{v} = -(a\mathbf{v}) by Result 2. By Axioms 7 and 4 and Result 5, a\mathbf{v} + a(-\mathbf{v}) = a\mathbf{0} = \mathbf{0}, so a(-\mathbf{v}) = -(a\mathbf{v}) too.
  11. By Axioms 5 and 8, \mathbf{v} + \mathbf{v} = (1 + 1)\mathbf{v} = 2\mathbf{v}. So 2\mathbf{v} = \mathbf{0} with 2 \neq 0, and Result 7 gives \mathbf{v} = \mathbf{0}.
  12. Let U be the intersection. Every subspace in the collection contains \mathbf{0}, so U does. If \mathbf{u}, \mathbf{w} \in U, they lie in each subspace, so \mathbf{u} + \mathbf{w} and a\mathbf{u} do too, by Result 9; hence both lie in U, and Result 9 makes U a subspace.
  13. Pick \mathbf{u} in the set. Closure puts 0\mathbf{u} in it, and 0\mathbf{u} = \mathbf{0} by Result 4, so Result 9 applies. The empty set is closed under both operations, but it is not a subspace: nothing in it satisfies Axiom 3.
  14. If \mathbf{v} \in U, each \mathbf{v} + \mathbf{u} is in U by closure, and each \mathbf{u}' \in U is \mathbf{v} + (\mathbf{u}' - \mathbf{v}) with \mathbf{u}' - \mathbf{v} \in U; so \mathbf{v} + U = U. Conversely, if \mathbf{v} + U is a subspace, it contains \mathbf{0} (Result 9), so \mathbf{v} + \mathbf{u} = \mathbf{0} for some \mathbf{u} \in U. Then \mathbf{v} is an additive inverse of \mathbf{u} (Axiom 1), so \mathbf{v} = -\mathbf{u} = (-1)\mathbf{u} \in U by Results 2 and 6.
  15. Let \mathbf{v} + \mathbf{w} = \mathbf{0} and \mathbf{w} + \mathbf{x} = \mathbf{0} (Axiom 4). By Axioms 3 and 4, \mathbf{w} + \mathbf{v} equals (\mathbf{w} + \mathbf{v}) + (\mathbf{w} + \mathbf{x}), which regroups by Axiom 2 to \mathbf{w} + ((\mathbf{v} + \mathbf{w}) + \mathbf{x}), which is (\mathbf{w} + \mathbf{0}) + \mathbf{x} by Axioms 4 and 2, which is \mathbf{w} + \mathbf{x} = \mathbf{0} by Axioms 3 and 4. Then by Axioms 4, 2 and 3, \mathbf{0} + \mathbf{v} equals (\mathbf{v} + \mathbf{w}) + \mathbf{v}, then \mathbf{v} + (\mathbf{w} + \mathbf{v}), then \mathbf{v} + \mathbf{0} = \mathbf{v}. So if \mathbf{y} + \mathbf{p} = \mathbf{y} + \mathbf{q}, adding an inverse of \mathbf{y} on the left gives \mathbf{p} = \mathbf{q} by Axiom 2 and these two facts; if \mathbf{p} + \mathbf{y} = \mathbf{q} + \mathbf{y}, adding it on the right gives \mathbf{p} = \mathbf{q} by Axioms 2, 4 and 3. By Axioms 8 and 5, (1 + 1)(\mathbf{u} + \mathbf{v}) = \mathbf{u} + \mathbf{v} + \mathbf{u} + \mathbf{v}; by Axioms 7, 8 and 5 it is also \mathbf{u} + \mathbf{u} + \mathbf{v} + \mathbf{v}, brackets dropped by Axiom 2. Cancelling \mathbf{u} on the left and \mathbf{v} on the right leaves \mathbf{v} + \mathbf{u} = \mathbf{u} + \mathbf{v}.
math.wiki · CC-BY-SA-4.0