Published online by Cambridge University Press: 24 October 2008
This paper is a continuation of three others under the same title. The paragraphs are numbered following on to those of the third paper.
† Steen, , Proc. London Math. Soc. (2), 41 (1936), 361–92CrossRefGoogle Scholar; 43 (1937), 529–43; 44 (1938), 398–411.
‡ This is required in § XXIII, Cor. (iv), etc.
† Annals of Math. 37 (1936), 116–229Google Scholar.
‡ Banach, , Opérations linéaires (1932), p. 53.Google Scholar
† Math. Ann. 102 (1929), 370–427.Google Scholar
‡ Hausdorff, , Mengenlehre, 2nd ed. (1927), p. 228.Google Scholar
§ Kantorovitch, , C.R. Acad. Sci. U.R.S.S. 14 (1937), 535.Google Scholar
† Take B*A, A*B for A, B respectively in the first part of the proof.
† If A is real and λ+(A) ≥ O for all λ+, then A ≥ O. For A = A + − A −, A + ≥ O, A − ≥ O, A +A − = O. λ+(A −A) is a positive linear functional equal to therefore A − = O.
‡ This is possible, since as n → ∞.
§ We have, by Cor. (x), If ∥ A ∥ = a, then A* A ≤ a 2I, whence Similarly [Cor. (x) with Thus | λ+(A) | ≤ 2aλ+(I), and so ‖λ+ ‖ ≤ 2λ+(I). By hypothesis, given k < 0, we can choose n and λ+ so that | λ+(A n) | > kλ+(I); hence we can choose n and λ+ so that | λ+(A 2)| > ½K‖ λ+‖. Clearly we can take ‖ λ+ ‖ to be any positive number (by multiplying λ+ by a suitable constant). Thus if we fix then given k′ > 0 we can find n and λ+ so that | λ+(A n) | > k′. We need only take
† A′ belongs to the same association as the A n, since O = A n: A m → A′: A m(s) as n → ∞ (Cor. xvii), and O → O(s).
† For definitions of these terms see Ore, , Annals of Math. 36 (1935), 406–37.CrossRefGoogle Scholar
‡ Rank has only been defined for real operators. Left and right ranks are defined for any operators.
† This takes the place of lemma 7.3.1 in the paper “On rings of operators”, and Cor. (xiii) takes the place of lemma 6.2.2.
† We omit proofs that can be obtained directly from the paper “On rings of operators”. This is lemma 6. 1. 1. Cor. (iv) is lemma 6.2.3.
‡ V* × E = V* × E″ × E = V* × E″ = V*.
† 6·3 to Th. VIII.
‡ If E and F are projections, then (E vF) − F ≼ E. For (E vF) − F is the left rank of E − F × E. We have {(E vF) − F} × {E − F × E} = E − F × E. Hence (E vF) − F ≥ left rank of E − F × E. Let the left rank be (E vF) − F − G, where G is a projection and GF = O, G < E vF. Then G × (E − F × E) = O, i.e. G × E = O; hence G × (E vF) = O, thus G = O. Hence