A Course in Mathematical Analysis: Volume 2, Metric and Topological Spaces, Functions of a Vector Variable

D ) (that is, j = ±1 for 1 ≤ j ≤ d) such that f (ej ) = j eσ(j) , f (−ej ) = − j eσ(j) for 1 ≤ j ≤ d. (c) express that f is linear, in order that f (x) = dj=1 j xj eσ(j) for x ∈ l1d . eleven. five. three via contemplating vectors of the shape ( 1 , . . . , d ), the place j = ±1 for d into itself is an isometry 1 ≤ j ≤ d, exhibit that if a mapping f of l∞ and if f (0) = zero then there exists a permutation σ of {1, . . . , d} and a call of symptoms ( 1 , . . . , d ) (that is, j = ±1 for 1 ≤ j ≤ d) such that f (ej ) = j eσ(j) , f (−ej ) = − j eσ(j) for 1 ≤ j ≤ d. exhibit that f is linear. eleven. 6 *The Mazur–Ulam theorem* (This part may be passed over on a first analyzing. ) consider that (E, . E ) and (F, . F ) are actual normed areas and that J : E → F is an isometry. permit L = T−J(0) ◦J, the place T−J(0) is the interpretation of F mapping J(0) to zero. hence L(x) = J(x)−J(0), in order that L is an isometry of E into F , with L(0) = zero. Our significant objective is to teach that if L is surjective, then it needs to be linear. This extends the result of routines eleven. five. 2 and eleven. five. three of the former part. Theorem eleven. 6. 1 (The Mazur–Ulam theorem) If L : E → F is an isometry of a true normed house (E, . E ) onto a true normed area (F, . F ) with L(0) = zero, then L is a linear mapping. on the way to turn out this, we introduce a few rules in regards to the geometry of metric areas, of curiosity of their personal correct. First, feel that x, y, z are components of a metric house. we are saying that y is among x and z if d(x, y) + d(y, z) = d(x, z), and we are saying that y is midway among x and z if d(x, y) = 324 Metric areas and normed areas d(y, z) = 12 d(x, z). We denote the set of issues midway among x and z by means of H(x, z). The set H(x, z) should be • empty, for instance if (X, d) has the discrete metric, • a singleton set, for instance if (X, d) is the true line R, with the standard metric, whilst H(x, z) = { 12 (x + z)}, 2 (R), x = (−1, 0), and • or might comprise a couple of element: permit (X, d) = l∞ z = (1, 0); then H(x, z) = {(0, y) : −1 ≤ y ≤ 1}. If (E, . E ) is a normed area, then 12 (x + z) ∈ H(x, z), yet H(x, z) may well comprise different issues, because the final instance exhibits. The set H(x, z) is often bounded, given that if y, y ∈ H(x, z) then d(y, y ) ≤ d(y, x) + d(x, y ) = d(x, z). consider is a bounded subset of a metric area (X, d). will we find a different aspect in A that is the centre of A, in a few metric feel? commonly, the reply needs to be ‘no’, in view that, for instance, in a metric area with the discrete metric, there is not any visible exact aspect. in some cases, in spite of the fact that, the answer's ‘yes’. First, permit κ(A) = {x ∈ A : d(x, y) ≤ 12 diam (A) for all y ∈ A}; κ(A) is the principal center of A. back, κ(A) could be empty, might encompass one element (which may then be the centre of A) or could include a couple of aspect; for instance, if 2 (R) : −1 ≤ x ≤ 1, − 12 ≤ y ≤ 12 }, A = {(x, y) ∈ l∞ then κ(A) = {(0, y) : − 12 ≤ y ≤ 12 }. observe even though that diam (κ(A)) ≤ 1 2 diam A. this means that we iterate the process: we set κ1 (A) = κ(A), and if κn (A) = ∅ we set κn+1 (A) = κ(κn (A)). There are then 3 attainable results: κn (A) = ∅ for a few n ∈ N; • κn (A) = ∅ for all n ∈ N, yet ∩∞ n=1 (κn (A)) = ∅; • κn (A) = ∅ for all n ∈ N, and ∩∞ n=1 (κn (A)) = ∅.