Q
stringlengths
4
3.96k
A
stringlengths
1
3k
Result
stringclasses
4 values
Theorem 1. Let \( X \) be an arbitrary set, and \( \mathcal{C} \) a collection of subsets of \( X \), countably many of which cover \( X \) . Let \( \beta \) be a function from \( \mathcal{C} \) to \( {\mathbb{R}}^{ * } \) such that\n\n\[ \inf \{ \beta \left( C\right) : C \in \mathcal{C}\} = 0 \]\n\nThen the equation\n\n\[ \mu \left( A\right) = \inf \left\{ {\mathop{\sum }\limits_{{i = 1}}^{\infty }\beta \left( {C}_{i}\right) : A \subset \mathop{\bigcup }\limits_{{i = 1}}^{\infty }{C}_{i},{C}_{i} \in \mathcal{C}}\right\} \]\n\ndefines an outer measure on \( X \) .
Proof. Assume all the hypotheses. There are now three postulates for an outer measure to be verified. Our assumption about \( \beta \) implies that \( \beta \left( C\right) \geq 0 \) for all \( C \in \mathcal{C} \) . Therefore, \( \mu \left( A\right) \geq 0 \) for all \( A \) . Since \( \varnothing \subset C \) for all \( C \in \mathcal{C},\mu \left( \varnothing \right) \leq \beta \left( C\right) \) for all \( C \) . Taking an infimum yields \( \mu \left( \varnothing \right) \leq 0 \).\n\nIf \( A \subset B \) and \( B \subset \mathop{\bigcup }\limits_{{i = 1}}^{\infty }{C}_{i} \), then \( A \subset \mathop{\bigcup }\limits_{{i = 1}}^{\infty }{C}_{i} \) and \( \mu \left( A\right) \leq \mathop{\sum }\limits_{{i = 1}}^{\infty }\beta \left( {C}_{i}\right) \) . Taking an infimum over all countable covers of \( B \), we have \( \mu \left( A\right) \leq \mu \left( B\right) \).\n\nLet \( {A}_{i} \subset X\left( {i \in \mathbb{N}}\right) \) and let \( \varepsilon > 0 \) . By the definition of \( \mu \left( {A}_{i}\right) \) there exist \( {C}_{ij} \in \mathcal{C} \) such that \( {A}_{i} \subset \mathop{\bigcup }\limits_{{j = 1}}^{\infty }{C}_{ij} \) and \( \mathop{\sum }\limits_{{j = 1}}^{\infty }\beta \left( {C}_{ij}\right) \leq \mu \left( {A}_{i}\right) + \varepsilon /{2}^{i} \) . Since \( \mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i} \subset \mathop{\bigcup }\limits_{{i, j = 1}}^{\infty }{C}_{ij} \), we obtain\n\n\[ \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) \leq \mathop{\sum }\limits_{{i, j}}\beta \left( {C}_{ij}\right) \leq \mathop{\sum }\limits_{{i = 1}}^{\infty }\left\lbrack {\mu \left( {A}_{i}\right) + \varepsilon /{2}^{i}}\right\rbrack = \varepsilon + \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {A}_{i}\right) \]\n\nSince this is true for each positive \( \varepsilon \), we obtain \( \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) \leq \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {A}_{i}\right) \).
Yes
Lemma 1. If \( \left( {X,\mathcal{A}}\right) \) is a measurable space, then \( X \) and \( \varnothing \) belong to \( \mathcal{A} \) . Furthermore, \( \mathcal{A} \) is closed under countable intersections and set difference.
Proof. This is left to the reader as a problem.
No
Lemma 2. For any subset \( \mathcal{F} \) of \( {2}^{X} \) there is a smallest \( \sigma \) -algebra containing \( \mathcal{F} \) .
Proof. As Example 10 shows, there is certainly one \( \sigma \) -algebra containing \( \mathcal{F} \) . The smallest one will be the intersection of all the \( \sigma \) -algebras \( {\mathcal{A}}_{\nu } \) containing \( \mathcal{F} \) . It is only necessary to verify that \( \bigcap {\mathcal{A}}_{\nu } \) is a \( \sigma \) -algebra. If \( {A}_{i} \in \bigcap {\mathcal{A}}_{\nu } \), then \( {A}_{i} \in {\mathcal{A}}_{\nu } \) for all \( \nu \) . Since \( {\mathcal{A}}_{\nu } \) is a \( \sigma \) -algebra, \( \mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i} \in {\mathcal{A}}_{\nu } \) . Since this is true for all \( \nu ,\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i} \in \bigcap {\mathcal{A}}_{\nu } \) . A similar proof is needed for the other axiom.
Yes
Lemma 1. If \( \left( {X,\mathcal{A},\mu }\right) \) is a measure space, then:\n\n(1) \( \mu \left( A\right) \leq \mu \left( B\right) \) if \( A \in \mathcal{A}, B \in \mathcal{A} \), and \( A \subset B \) .\n\n(2) \( \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) \leq \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {A}_{i}\right) \) if \( {A}_{i} \in \mathcal{A} \) .
Proof. For (1), write\n\n\[ \mu \left( B\right) = \mu \left\lbrack {A \cup \left( {B \smallsetminus A}\right) }\right\rbrack = \mu \left( A\right) + \mu \left( {B \smallsetminus A}\right) \geq \mu \left( A\right) \]\n\nFor (2), we create a disjoint sequence of sets \( {B}_{i} \) by writing \( {B}_{1} = {A}_{1},{B}_{2} = \) \( {A}_{2} \smallsetminus {A}_{1},{B}_{3} = {A}_{3} \smallsetminus \left( {{A}_{1} \cup {A}_{2}}\right) \), and so on. (Try to remember this little trick.) These sets are in \( \mathcal{A} \) by Lemma 1 in the preceding section. Also, \( {B}_{i} \subset {A}_{i} \) and\n\n\[ \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) = \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{B}_{i}}\right) = \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {B}_{i}\right) \leq \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {A}_{i}\right) \]
Yes
Lemma 1. If \( \left( {X,\mathcal{A},\mu }\right) \) is a measure space, then:\n\n(1) \( \mu \left( A\right) \leq \mu \left( B\right) \) if \( A \in \mathcal{A}, B \in \mathcal{A} \), and \( A \subset B \) .\n\n(2) \( \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) \leq \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {A}_{i}\right) \) if \( {A}_{i} \in \mathcal{A} \) .
Proof. For (1), write\n\n\[ \mu \left( B\right) = \mu \left\lbrack {A \cup \left( {B \smallsetminus A}\right) }\right\rbrack = \mu \left( A\right) + \mu \left( {B \smallsetminus A}\right) \geq \mu \left( A\right) \]\n\nFor (2), we create a disjoint sequence of sets \( {B}_{i} \) by writing \( {B}_{1} = {A}_{1},{B}_{2} = \) \( {A}_{2} \smallsetminus {A}_{1},{B}_{3} = {A}_{3} \smallsetminus \left( {{A}_{1} \cup {A}_{2}}\right) \), and so on. (Try to remember this little trick.) These sets are in \( \mathcal{A} \) by Lemma 1 in the preceding section. Also, \( {B}_{i} \subset {A}_{i} \) and\n\n\[ \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) = \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{B}_{i}}\right) = \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {B}_{i}\right) \leq \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {A}_{i}\right) \]
Yes
Lemma 1. If \( \left( {X,\mathcal{A},\mu }\right) \) is a measure space, then:\n\n(1) \( \mu \left( A\right) \leq \mu \left( B\right) \) if \( A \in \mathcal{A}, B \in \mathcal{A} \), and \( A \subset B \) .\n\n(2) \( \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) \leq \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {A}_{i}\right) \) if \( {A}_{i} \in \mathcal{A} \) .
Proof. For (1), write\n\n\[ \mu \left( B\right) = \mu \left\lbrack {A \cup \left( {B \smallsetminus A}\right) }\right\rbrack = \mu \left( A\right) + \mu \left( {B \smallsetminus A}\right) \geq \mu \left( A\right) \]\n\nFor (2), we create a disjoint sequence of sets \( {B}_{i} \) by writing \( {B}_{1} = {A}_{1},{B}_{2} = \) \( {A}_{2} \smallsetminus {A}_{1},{B}_{3} = {A}_{3} \smallsetminus \left( {{A}_{1} \cup {A}_{2}}\right) \), and so on. (Try to remember this little trick.) These sets are in \( \mathcal{A} \) by Lemma 1 in the preceding section. Also, \( {B}_{i} \subset {A}_{i} \) and\n\n\[ \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) = \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{B}_{i}}\right) = \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {B}_{i}\right) \leq \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {A}_{i}\right) \]
Yes
Lemma 1. If \( \left( {X,\mathcal{A},\mu }\right) \) is a measure space, then:\n\n(1) \( \mu \left( A\right) \leq \mu \left( B\right) \) if \( A \in \mathcal{A}, B \in \mathcal{A} \), and \( A \subset B \) .\n\n(2) \( \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) \leq \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {A}_{i}\right) \) if \( {A}_{i} \in \mathcal{A} \) .
Proof. For (1), write\n\n\[ \mu \left( B\right) = \mu \left\lbrack {A \cup \left( {B \smallsetminus A}\right) }\right\rbrack = \mu \left( A\right) + \mu \left( {B \smallsetminus A}\right) \geq \mu \left( A\right) \]\n\nFor (2), we create a disjoint sequence of sets \( {B}_{i} \) by writing \( {B}_{1} = {A}_{1},{B}_{2} = \) \( {A}_{2} \smallsetminus {A}_{1},{B}_{3} = {A}_{3} \smallsetminus \left( {{A}_{1} \cup {A}_{2}}\right) \), and so on. (Try to remember this little trick.) These sets are in \( \mathcal{A} \) by Lemma 1 in the preceding section. Also, \( {B}_{i} \subset {A}_{i} \) and\n\n\[ \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) = \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{B}_{i}}\right) = \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {B}_{i}\right) \leq \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {A}_{i}\right) \]
Yes
Lemma 1. If \( \left( {X,\mathcal{A},\mu }\right) \) is a measure space, then:\n\n(1) \( \mu \left( A\right) \leq \mu \left( B\right) \) if \( A \in \mathcal{A}, B \in \mathcal{A} \), and \( A \subset B \) .\n\n(2) \( \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) \leq \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {A}_{i}\right) \) if \( {A}_{i} \in \mathcal{A} \) .
Proof. For (1), write\n\n\[ \mu \left( B\right) = \mu \left\lbrack {A \cup \left( {B \smallsetminus A}\right) }\right\rbrack = \mu \left( A\right) + \mu \left( {B \smallsetminus A}\right) \geq \mu \left( A\right) \]\n\nFor (2), we create a disjoint sequence of sets \( {B}_{i} \) by writing \( {B}_{1} = {A}_{1},{B}_{2} = \) \( {A}_{2} \smallsetminus {A}_{1},{B}_{3} = {A}_{3} \smallsetminus \left( {{A}_{1} \cup {A}_{2}}\right) \), and so on. (Try to remember this little trick.) These sets are in \( \mathcal{A} \) by Lemma 1 in the preceding section. Also, \( {B}_{i} \subset {A}_{i} \) and\n\n\[ \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) = \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{B}_{i}}\right) = \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {B}_{i}\right) \leq \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {A}_{i}\right) \]
Yes
Lemma 1. If \( \\left( {X,\\mathcal{A},\\mu }\\right) \) is a measure space, then:\n\n(1) \( \\mu \\left( A\\right) \\leq \\mu \\left( B\\right) \) if \( A \\in \\mathcal{A}, B \\in \\mathcal{A} \), and \( A \\subset B \) .\n\n(2) \( \\mu \\left( {\\mathop{\\bigcup }\\limits_{{i = 1}}^{\\infty }{A}_{i}}\\right) \\leq \\mathop{\\sum }\\limits_{{i = 1}}^{\\infty }\\mu \\left( {A}_{i}\\right) \) if \( {A}_{i} \\in \\mathcal{A} \) .
Proof. For (1), write\n\n\[ \n\\mu \\left( B\\right) = \\mu \\left\\lbrack {A \\cup \\left( {B \\smallsetminus A}\\right) }\\right\\rbrack = \\mu \\left( A\\right) + \\mu \\left( {B \\smallsetminus A}\\right) \\geq \\mu \\left( A\\right) \n\]\n\nFor (2), we create a disjoint sequence of sets \( {B}_{i} \) by writing \( {B}_{1} = {A}_{1},{B}_{2} = \) \( {A}_{2} \\smallsetminus {A}_{1},{B}_{3} = {A}_{3} \\smallsetminus \\left( {{A}_{1} \\cup {A}_{2}}\\right) \), and so on. (Try to remember this little trick.) These sets are in \( \\mathcal{A} \) by Lemma 1 in the preceding section. Also, \( {B}_{i} \\subset {A}_{i} \) and\n\n\[ \n\\mu \\left( {\\mathop{\\bigcup }\\limits_{{i = 1}}^{\\infty }{A}_{i}}\\right) = \\mu \\left( {\\mathop{\\bigcup }\\limits_{{i = 1}}^{\\infty }{B}_{i}}\\right) = \\mathop{\\sum }\\limits_{{i = 1}}^{\\infty }\\mu \\left( {B}_{i}\\right) \\leq \\mathop{\\sum }\\limits_{{i = 1}}^{\\infty }\\mu \\left( {A}_{i}\\right) \n\]
Yes
Lemma 2. Let \( \left( {X,\mathcal{A},\mu }\right) \) be a measure space. If \( {A}_{i} \in \mathcal{A} \) for \( i \in \mathbb{N} \) and \( {A}_{1} \subset {A}_{2} \subset \cdots \), then \( \mu \left( {A}_{n}\right) \uparrow \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) \) .
Proof. Let \( {A}_{0} = \varnothing \) and observe that \( {A}_{n} = \mathop{\bigcup }\limits_{{i = 1}}^{n}\left( {{A}_{i} \smallsetminus {A}_{i - 1}}\right) \) . It follows from the disjoint nature of the sets \( {A}_{i} \smallsetminus {A}_{i - 1} \) that \( \mu \left( {A}_{n}\right) = \mathop{\sum }\limits_{{i = 1}}^{n}\mu \left( {{A}_{i} \smallsetminus {A}_{i - 1}}\right) \) . Hence\n\n\[ \mu \left( {A}_{n}\right) \uparrow \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {{A}_{i} \smallsetminus {A}_{i - 1}}\right) = \mu \left\lbrack {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }\left( {{A}_{i} \smallsetminus {A}_{i - 1}}\right) }\right\rbrack = \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) \]
Yes
Theorem 2. Let \( \left( {X,\mathcal{A},\mu }\right) \) be a measure space. For \( S \in {2}^{X} \) define\n\n\[ \n{\mu }^{ * }\left( S\right) = \inf \{ \mu \left( A\right) : S \subset A \in \mathcal{A}\}\n\]\n\nThen \( {\mu }^{ * } \) is an outer measure whose restriction to \( \mathcal{A} \) is \( \mu \) . Furthermore, each set in \( \mathcal{A} \) is \( \mu \) -measurable.
Proof. I. Since \( \mu \) is nonnegative, so is \( {\mu }^{ * } \) . Since \( \varnothing \in \mathcal{A} \), we have \( 0 \leq {\mu }^{ * }\left( \varnothing \right) \leq \) \( \mu \left( \varnothing \right) = 0 \).\n\nII. If \( S \subset T \), then \( \{ A : S \subset A \in \mathcal{A}\} \) contains \( \{ A : T \subset A \in \mathcal{A}\} \) . Hence\n\n\[ \n{\mu }^{ * }\left( S\right) = \inf \{ \mu \left( A\right) : S \subset A \in \mathcal{A}\} \leq \inf \{ \mu \left( A\right) : T \subset A \in \mathcal{A}\} = {\mu }^{ * }\left( T\right)\n\]\n\nIII. If \( {S}_{i} \in {2}^{X} \) and \( \varepsilon > 0 \), select \( {A}_{i} \in \mathcal{A} \) so that \( {S}_{i} \subset {A}_{i} \) and \( {\mu }^{ * }\left( {S}_{i}\right) \geq \) \( \mu \left( {A}_{i}\right) - \varepsilon /{2}^{i} \) . Then \( \mathop{\bigcup }\limits_{{i = 1}}^{\infty }{S}_{i} \subset \mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i} \in \mathcal{A} \) . Consequently,\n\n\[ \n{\mu }^{ * }\left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{S}_{i}}\right) \leq \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) \leq \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {A}_{i}\right) \leq \mathop{\sum }\limits_{{i = 1}}^{\infty }\left\lbrack {{\mu }^{ * }\left( {S}_{i}\right) + \frac{\varepsilon }{{2}^{i}}}\right\rbrack\n\]\n\nSince \( \varepsilon \) can be any positive number, \( {\mu }^{ * }\left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{S}_{i}}\right) \leq \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {S}_{i}\right) \).\n\nIV. If \( S \in \mathcal{A} \) and \( S \subset A \in \mathcal{A} \), then \( {\mu }^{ * }\left( S\right) \leq \mu \left( S\right) \leq \mu \left( A\right) \) . Taking an infimum for all choices of \( A \), we get \( {\mu }^{ * }\left( S\right) \leq \mu \left( S\right) \leq {\mu }^{ * }\left( S\right) \) . This proves that \( {\mu }^{ * } \) is an extension of \( \mu \).\n\nV. To prove that each \( A \) in \( \mathcal{A} \) is \( {\mu }^{ * } \) -measurable, let \( S \) be any subset of \( X \) . Given \( \varepsilon > 0 \), we find \( B \in \mathcal{A} \) such that \( {\mu }^{ * }\left( S\right) \geq \mu \left( B\right) - \varepsilon \) and \( S \subset B \) . Then\n\n\[ \n{\mu }^{ * }\left( S\right) + \varepsilon \geq \mu \left( B\right) = \mu \left( {B \cap A}\right) + \mu \left( {B \smallsetminus A}\right) = {\mu }^{ * }\left( {B \cap A}\right) + {\mu }^{ * }\left( {B \smallsetminus A}\right)\n\]\n\n\[ \n\geq {\mu }^{ * }\left( {S \cap A}\right) + {\mu }^{ * }\left( {S \smallsetminus A}\right) \geq {\mu }^{ * }\left( S\right)\n\]\n\nThis calculation used Parts III and IV of the present proof. Since \( \varepsilon \) was arbitrary, \( {\mu }^{ * }\left( S\right) = {\mu }^{ * }\left( {S \cap A}\right) + {\mu }^{ * }\left( {S \smallsetminus A}\right) \) for all \( S \) . Hence \( A \) is \( {\mu }^{ * } \) -measurable.
Yes
Theorem 3. Under the same hypotheses as in Theorem 2, the outer measure \( {\mu }^{ * } \) is regular.
Proof. Let \( S \) be any subset of \( X \) . For each \( n \in \mathbb{N} \) select \( {A}_{n} \in \mathcal{A} \) so that \( S \subset {A}_{n} \) and \( {\mu }^{ * }\left( S\right) \geq \mu \left( {A}_{n}\right) - 1/n \) . Put \( A = \mathop{\bigcap }\limits_{{n = 1}}^{\infty }{A}_{n} \) . Since \( \mathcal{A} \) is a \( \sigma \) -algebra, \( A \in \mathcal{A} \) . (See Lemma 1 in the preceding section, page 384.) From the inclusion \( S \subset A \subset {A}_{n} \) we get\n\n\[ \n{\mu }^{ * }\left( S\right) \leq {\mu }^{ * }\left( A\right) = \mu \left( A\right) \leq \mu \left( {A}_{n}\right) < {\mu }^{ * }\left( S\right) + 1/n \n\]\n\nSince this is true for all \( n,{\mu }^{ * }\left( S\right) = {\mu }^{ * }\left( A\right) \) . By the preceding theorem, \( A \) is \( {\mu }^{ * } \) -measurable.
Yes
Theorem 1. The Lebesgue outer measure of an interval is its length.
Proof. Consider first a compact interval \( \left\lbrack {a, b}\right\rbrack \) . Since \( \left\lbrack {a, b}\right\rbrack \subset \left( {a - \varepsilon, b + \varepsilon }\right) \), we conclude from the definition (1) that \( \mu \left( \left\lbrack {a, b}\right\rbrack \right) \leq b - a + {2\varepsilon } \) for every positive \( \varepsilon \) . Hence \( \mu \left( \left\lbrack {a, b}\right\rbrack \right) \leq b - a \) . Suppose now that \( \mu \left( \left\lbrack {a, b}\right\rbrack \right) < b - a \) . Find intervals \( \left( {{a}_{i},{b}_{i}}\right) \) such that \( \left\lbrack {a, b}\right\rbrack \subset \mathop{\bigcup }\limits_{{i = 1}}^{\infty }\left( {{a}_{i},{b}_{i}}\right) \) and \( \mathop{\sum }\limits_{{i = 1}}^{\infty }\left| {{b}_{i} - {a}_{i}}\right| < b - a \) . We can assume \( {a}_{i} < {b}_{i} \) for all \( i \) . By compactness and renumbering we can get \( \left\lbrack {a, b}\right\rbrack \subset \mathop{\bigcup }\limits_{{i = 1}}^{n}\left( {{a}_{i},{b}_{i}}\right) \) . It follows that \( \mathop{\sum }\limits_{{i = 1}}^{n}\left( {{b}_{i} - {a}_{i}}\right) < b - a \) . By renumbering again we can assume \( a \in \left( {{a}_{1},{b}_{1}}\right) ,{b}_{1} \in \left( {{a}_{2},{b}_{2}}\right) ,{b}_{2} \in \left( {{a}_{3},{b}_{3}}\right) \), and so on. There must exist an index \( k \leq n \) such that \( b < {b}_{k} \) . Then we reach a contradiction:\n\n\[ b - a > \mathop{\sum }\limits_{{i = 1}}^{n}\left( {{b}_{i} - {a}_{i}}\right) \geq \mathop{\sum }\limits_{{i = 1}}^{k}\left( {{b}_{i} - {a}_{i}}\right) = {b}_{k} - {a}_{1} + \mathop{\sum }\limits_{{i = 1}}^{{k - 1}}\left( {{b}_{i} - {a}_{i + 1}}\right) > {b}_{k} - {a}_{1} > b - a \]\n\nIf \( J \) is a bounded interval of the type \( \left( {a, b}\right) ,(a, b\rbrack \), or \( \lbrack a, b) \), then from the inclusions\n\n\[ \left\lbrack {a + \varepsilon, b - \varepsilon }\right\rbrack \subset J \subset \left\lbrack {a - \varepsilon, b + \varepsilon }\right\rbrack \]\n\nwe obtain \( b - a - {2\varepsilon } \leq \mu \left( J\right) \leq b - a + {2\varepsilon } \) and \( \mu \left( J\right) = b - a \) .\n\nFinally, if \( J \) is an unbounded interval, then it contains intervals \( \left\lbrack {a, b}\right\rbrack \) of arbitrarily great length. Hence \( \mu \left( J\right) = \infty \) .
Yes
Theorem 2. Every Borel set in \( \mathbb{R} \) is Lebesgue measurable.
Proof. (S.J. Bernau) The family of Borel sets is the smallest \( \sigma \) -algebra containing all the open sets. The Lebesgue measurable sets form a \( \sigma \) -algebra. Hence it suffices to prove that every open set is Lebesgue measurable.\n\nRecall that every open set in \( \mathbb{R} \) can be expressed as a countable union of open intervals \( \left( {a, b}\right) \) . Thus it suffices to prove that each interval \( \left( {a, b}\right) \) is Lebesgue measurable. We begin with an interval of the form \( \left( {a,\infty }\right) \), where \( a \in \mathbb{R} \) .\n\nTo prove that the open interval \( \left( {a,\infty }\right) \) is measurable, we must prove, for any set \( S \) in \( \mathbb{R} \), that\n\n(2)\n\n\[ \mu \left( S\right) \geq \mu \left\lbrack {S \cap \left( {a,\infty }\right) }\right\rbrack + \mu \left\lbrack {S \smallsetminus \left( {a,\infty }\right) }\right\rbrack \]\n\nLet us use the notation \( \left| I\right| \) for the length of an interval \( I \) . Given \( \varepsilon > 0 \), select open intervals \( {I}_{n} \) such that \( S \subset \mathop{\bigcup }\limits_{{n = 1}}^{\infty }{I}_{n} \) and \( \sum \left| {I}_{n}\right| < \mu \left( S\right) + \varepsilon \) . Define \( {J}_{n} = \) \( {I}_{n} \cap \left( {a,\infty }\right) ,{K}_{n} = {I}_{n} \cap \left( {-\infty, a}\right) \), and \( {K}_{0} = \left( {a - \varepsilon, a + \varepsilon }\right) \) . Then we have\n\n\[ S \cap \left( {a,\infty }\right) \subset \mathop{\bigcup }\limits_{{n = 1}}^{\infty }{J}_{n} \]\n\n\[ S \smallsetminus \left( {a,\infty }\right) = S \cap ( - \infty, a\rbrack \subset \mathop{\bigcup }\limits_{{n = 0}}^{\infty }{K}_{n} \]\n\n\[ {J}_{n} \cup {K}_{n} \subset {I}_{n}\;\text{ and }\;{J}_{n} \cap {K}_{n} = \varnothing \]\n\nConsequently,\n\n\[ \mu \left\lbrack {S \cap \left( {a,\infty }\right) }\right\rbrack + \mu \left\lbrack {S \smallsetminus \left( {a,\infty }\right) }\right\rbrack \leq \mathop{\sum }\limits_{{n = 1}}^{\infty }\left\{ {\left| {J}_{n}\right| + \left| {K}_{n}\right| }\right\} + \left| {K}_{0}\right| \]\n\n\[ \leq \mathop{\sum }\limits_{{n = 1}}^{\infty }\left| {I}_{n}\right| + {2\varepsilon } < \mu \left( S\right) + {3\varepsilon } \]\n\nBecause \( \varepsilon \) was arbitrary, this establishes Equation (2). Since the measurable sets make up a \( \sigma \) -algebra, each set of the form \( ( - \infty, b\rbrack = \mathbb{R} \smallsetminus \left( {b,\infty }\right) \) is measurable. Hence the set \( \left( {-\infty, b}\right) = \mathop{\bigcup }\limits_{{n = 1}}^{\infty }( - \infty, b - \frac{1}{n}\rbrack \) is measurable and so is \( \left( {a, b}\right) = \) \( \left( {-\infty, b}\right) \cap \left( {a,\infty }\right) \) .
Yes
Theorem 3. Lebesgue outer measure is invariant on the group \( \left( {\mathbb{R}, + }\right) \) .
Proof. The statement means that \( \mu \left( S\right) = \mu \left( {v + S}\right) \) for all \( S \in {2}^{\mathbb{R}} \) and all \( v \in \mathbb{R} \) . The translate \( v + S \) is defined to be \( \{ v + x : x \in S\} \) . Notice that the condition \( S \subset \mathop{\bigcup }\limits_{{i = 1}}^{\infty }\left( {{a}_{i},{b}_{i}}\right) \) is equivalent to the condition \( x + S \subset \mathop{\bigcup }\limits_{{i = 1}}^{\infty }\left( {x + {a}_{i}, x + {b}_{i}}\right) \) . Since the length of \( \left( {x + {a}_{i}, x + {b}_{i}}\right) \) is the same as the length of \( \left( {{a}_{i},{b}_{i}}\right) \), the definition of \( \mu \) gives equal values for \( \mu \left( S\right) \) and \( \mu \left( {x + S}\right) \) .
Yes
Theorem 4. There exists no translation-invariant measure \( \nu \) defined on \( {2}^{\mathbb{R}} \) such that \( 0 < \nu \left( \left\lbrack {0,1}\right\rbrack \right) < \infty \) . Consequently, there exist subsets of \( \mathbb{R} \) that are not Lebesgue measurable.
Proof. The second assertion follows from the first because if every set of reals were Lebesgue measurable, then Lebesgue measure would contradict the first assertion.\n\nTo prove the first assertion, suppose that a measure \( \nu \) exists as described. By the preceding lemma, the set \( P \) given there has the property\n\n\[ \left\lbrack {0,1}\right\rbrack \subset \mathop{\bigcup }\limits_{{i = 1}}^{\infty }\left( {{r}_{i} + P}\right) \subset \left\lbrack {-1,2}\right\rbrack \]\n\nAlso by the lemma, the sequence of sets \( {r}_{i} + P \) is disjoint. Consequently,\n\n\[ 0 < \nu \left( \left\lbrack {0,1}\right\rbrack \right) \leq \nu \left\lbrack {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }\left( {{r}_{i} + P}\right) }\right\rbrack = \mathop{\sum }\limits_{{i = 1}}^{\infty }\nu \left( {{r}_{i} + P}\right) = \mathop{\sum }\limits_{{i = 1}}^{\infty }\nu \left( P\right) \]\n\nTherefore, \( \nu \left( P\right) > 0,\sum \nu \left( P\right) = \infty \), and we have the contradiction\n\n\[ \infty = \nu \left\lbrack {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }\left( {{r}_{i} + P}\right) }\right\rbrack \leq \nu \left( \left\lbrack {-1,2}\right\rbrack \right) \leq {3\nu }\left( \left\lbrack {0,1}\right\rbrack \right) \]
Yes
Theorem 1. Let \( \left( {X,\mathcal{A}}\right) \) be a measurable space. A function \( f \) from \( X \) to the extended reals \( {\mathbb{R}}^{ * } \) is measurable if it has any one of the following properties:\na. \( {f}^{-1}(\left( {a,\infty \rbrack }\right) \in \mathcal{A} \) for each \( a \in {\mathbb{R}}^{ * }\)\nb. \( {f}^{-1}\left( \left\lbrack {a,\infty }\right\rbrack \right) \in \mathcal{A} \) for each \( a \in {\mathbb{R}}^{ * }\)\nc. \( {f}^{-1}\left( {\lbrack - \infty, a}\right) ) \in \mathcal{A} \) for each \( a \in {\mathbb{R}}^{ * }\)\nd. \( {f}^{-1}\left( \left\lbrack {-\infty, a}\right\rbrack \right) \in \mathcal{A} \) for each \( a \in {\mathbb{R}}^{ * }\)\ne. \( {f}^{-1}\left( \left( {a, b}\right) \right) \in \mathcal{A} \) for all \( a \) and \( b \) in \( {\mathbb{R}}^{ * }\)\nf. \( {f}^{-1}\left( O\right) \in \mathcal{A} \) for each open set \( O \) in \( {\mathbb{R}}^{ * } \)
Proof. We shall prove that each condition implies the one following it, and that \( \mathbf{f} \) implies that \( f \) is measurable. That \( \mathbf{a} \) implies \( \mathbf{b} \) follows from the equation \( {f}^{-1}\left( \left\lbrack {a,\infty }\right\rbrack \right) = \mathop{\bigcap }\limits_{{n = 1}}^{\infty }{f}^{-1}\left( \left( {a - \frac{1}{n},\infty }\right\rbrack \right) \) and from the properties of a \( \sigma \) -algebra. That \( \mathbf{b} \) implies \( \mathbf{c} \) follows in the same manner from the equation \( {f}^{-1}\left( {\lbrack - \infty, a}\right) ) = \) \( X \smallsetminus {f}^{-1}\left( \left\lbrack {a,\infty }\right\rbrack \right) \) . That \( \mathbf{c} \) implies \( \mathbf{d} \) follows from the equation \( {f}^{-1}\left( \left\lbrack {-\infty, a}\right\rbrack \right) = \) \( \mathop{\bigcap }\limits_{{n = 1}}^{\infty }{f}^{-1}\left( {\lbrack - \infty, a + \frac{1}{n}}\right) ) \) . That \( \mathbf{d} \) implies \( \mathbf{e} \) follows from writing \( {f}^{-1}\left( \left( {a, b}\right) \right) = \) \( \mathop{\bigcup }\limits_{{n = 1}}^{\infty }{f}^{-1}\left( {\lbrack - \infty, b - \frac{1}{n}}\right) ) \smallsetminus {f}^{-1}\left( \left\lbrack {-\infty, a}\right\rbrack \right) \) . That e implies \( \mathbf{f} \) is a consequence of the theorem that each open set in \( {\mathbb{R}}^{ * } \) is a countable union of intervals of the form \( \left( {a, b}\right) \), where \( a \) and \( b \) are in \( {\mathbb{R}}^{ * } \) . To complete the proof, assume condition f. Let \( \mathcal{S} \) be the family of all sets \( S \) contained in \( {\mathbb{R}}^{ * } \) such that \( {f}^{-1}\left( S\right) \in \mathcal{A} \) . It is straightforward to verify that \( \mathcal{S} \) is a \( \sigma \) -algebra. By hypothesis, each open set in \( {\mathbb{R}}^{ * } \) belongs to \( \mathcal{S} \) . Hence \( \mathcal{S} \) contains the \( \sigma \) -algebra of Borel sets. Consequently, \( {f}^{-1}\left( B\right) \in \mathcal{A} \) for each Borel set \( B \), and \( f \) is measurable.
Yes
Theorem 3. Let \( \\left( {X,\\mathcal{A},\\mu }\\right) \) be a complete measure space, as defined in Section 8.2, page 387. If \( f \) is a measurable function and if \( f\\left( x\\right) = g\\left( x\\right) \) almost everywhere, then \( g \) is measurable.
Proof. Define \( A = \\{ x : f\\left( x\\right) \\neq g\\left( x\\right) \\} \) . Then \( A \) is measurable and \( \\mu \\left( A\\right) = 0 \) . Also, \( X \\smallsetminus A \) is measurable. For \( a \\in {\\mathbb{R}}^{ * } \) we write\n\n\[ \n{g}^{-1}(\\left( {a,\\infty \\rbrack }\\right) = \\left\\{ {{g}^{-1}(\\left( {a,\\infty \\rbrack }\\right) \\cap \\left( {X \\smallsetminus A}\\right) }\\right\\} \\cup \\left\\{ {{g}^{-1}(\\left( {a,\\infty \\rbrack }\\right) \\cap A}\\right\\} \n\]\n\nOn the right side of this equation we see the union of two sets. The first of these is measurable because it is \( {f}^{-1}\\left( \\left( {a,\\infty }\\right) \\right) \\smallsetminus A \) . The second set is measurable because it is a subset of a set of measure 0 , and the measure space is complete.
Yes
Theorem 4 (Egorov’s Theorem). Let \( \left( {X,\mathcal{A},\mu }\right) \) be a measure space such that \( \mu \left( X\right) < \infty \) . For a sequence of finite-valued measurable functions \( f,{f}_{1},{f}_{2},\ldots \) these properties are equivalent:\n\na. \( {f}_{n} \rightarrow f \) almost everywhere\n\nb. \( {f}_{n} \rightarrow f \) almost uniformly
Proof. Assume that \( \mathbf{b} \) is true. For each \( m \) in \( \mathbb{N} \) there is a measurable set \( {A}_{m} \) such that \( \mu \left( {A}_{m}\right) < 1/m \), and on \( X \smallsetminus {A}_{m},{f}_{n}\left( x\right) \rightarrow f\left( x\right) \) uniformly. Define \( A = \mathop{\bigcap }\limits_{{m = 1}}^{\infty }{A}_{m} \) . Then \( \mu \left( A\right) = 0 \) because \( A \subset {A}_{m} \) for all \( m \) . Also, \( {f}_{n}\left( x\right) \rightarrow f\left( x\right) \) on \( X \smallsetminus A \) because \( {f}_{n}\left( x\right) \rightarrow f\left( x\right) \) for \( x \in X \smallsetminus {A}_{m} \) and \( X \smallsetminus A = X \smallsetminus \bigcap {A}_{m} = \) \( \bigcup \left( {X \smallsetminus {A}_{m}}\right) \) . Thus \( \mathbf{a} \) is true.\n\nNow assume that \( \mathbf{a} \) is true. Let \( {g}_{n} = f - {f}_{n} \) . By altering \( {g}_{n} \) on a set of measure 0, we can assume that \( {g}_{n}\left( x\right) \rightarrow 0 \) everywhere. Next, we define \( {A}_{n}^{m} = \left\{ {x : \left| {{g}_{i}\left( x\right) }\right| \leq 1/m\text{for}i \geq n}\right\} \) . Thus \( {A}_{1}^{m} \subset {A}_{2}^{m} \subset \cdots \) For each \( x \) there is an index \( n \) such that \( x \in {A}_{n}^{m} \) ; in other words, \( x \in \mathop{\bigcup }\limits_{{n = 1}}^{\infty }{A}_{n}^{m} \) and \( X \subset \mathop{\bigcup }\limits_{{n = 1}}^{\infty }{A}_{n}^{m} \) . Since \( X \) has finite measure, \( \mu \left( {X \smallsetminus {A}_{n}^{m}}\right) \rightarrow 0 \) as \( n \rightarrow \infty \) . (See Lemma 2 in Section 8.2, page 387.) Let \( \varepsilon > 0 \) . For each \( m \), let \( {n}_{m} \) be an integer such that \( \mu \left( {X \smallsetminus {A}_{{n}_{m}}^{m}}\right) < \varepsilon /{2}^{m} \) . Define \( A = \mathop{\bigcap }\limits_{{m = 1}}^{\infty }{A}_{{n}_{m}}^{m} \) . Then\n\n\[ \mu \left( {X \smallsetminus A}\right) = \mu \left( {X \smallsetminus \mathop{\bigcap }\limits_{{m = 1}}^{\infty }{A}_{{n}_{m}}^{m}}\right) = \mu \left( {\mathop{\bigcup }\limits_{{m = 1}}^{\infty }\left( {X \smallsetminus {A}_{{n}_{m}}^{m}}\right) }\right) \leq \mathop{\sum }\limits_{{m = 1}}^{\infty }\mu \left( {X \smallsetminus {A}_{{m}_{n}}^{m}}\right) < \varepsilon \]\n\nOn \( A,{g}_{i}\left( x\right) \rightarrow 0 \) uniformly. Indeed, for \( x \in A \) we have (for all \( m \) )\n\n\[ i \geq {n}_{m} \Rightarrow \left| {{g}_{i}\left( x\right) }\right| \leq \frac{1}{m} \]
Yes
Theorem 5. Let \( \\left( {X,\\mathcal{A}}\\right) \) be a measurable space, and \( f \) any nonnegative measurable function. Then there exists a sequence of nonnegative simple functions \( {g}_{n} \) such that \( {g}_{n}\\left( x\\right) \\uparrow f\\left( x\\right) \) for each \( x \) . If \( f \) is bounded, this sequence can be constructed so that \( {g}_{n} \\uparrow f \) uniformly.
Proof. ([HewS], page 159.) Define\n\n\[ \n{A}_{i}^{n} = \\left\\{ {x \\in X : \\frac{i}{{2}^{n}} \\leq f\\left( x\\right) < \\frac{i + 1}{{2}^{n}}}\\right\\} \\;\\left( {0 \\leq i < n{2}^{n}}\\right) \n\]\n\n\[ \n{B}^{n} = \\{ x \\in X : f\\left( x\\right) \\geq n\\} \n\]\n\n\[ \n{g}_{n} = \\mathop{\\sum }\\limits_{i}\\frac{i}{{2}^{n}}{x}_{{A}_{i}^{n}} + n{x}_{{B}^{n}} \n\]\n\nThe sets \( {A}_{i}^{n} \) and \( {B}^{n} \) are measurable, by Theorem 1. Hence \( {g}_{n} \) is a simple function. The definition of \( {g}_{n} \) shows directly that \( {g}_{n} \\leq f \) . In order to verify that \( {g}_{n}\\left( x\\right) \) converges to \( f\\left( x\\right) \) for each \( x \), consider first the case when \( f\\left( x\\right) \\neq \\infty \) . For large \( n \) and a suitable \( i, x \\in {A}_{i}^{n} \) . Then \( f\\left( x\\right) - {g}_{n}\\left( x\\right) < \\frac{i + 1}{{2}^{n}} - \\frac{i}{{2}^{n}} = \\frac{1}{{2}^{n}} \) . On the other hand, if \( f\\left( x\\right) = + \\infty \), then \( {g}_{n}\\left( x\\right) = n \\rightarrow f\\left( x\\right) \) .\n\nFor the monotonicity of \( {g}_{n}\\left( x\\right) \) as a function of \( n \) for \( x \) fixed, first verify (Problem 3) that (for \( i < n{2}^{n} \) ) \( {A}_{i}^{n} = {A}_{2i}^{n + 1} \\cup {A}_{{2i} + 1}^{n + 1} \) . If \( x \\in {A}_{2i}^{n + 1} \) , then \( {g}_{n + 1}\\left( x\\right) = {2i}/{2}^{n + 1} = i/{2}^{n} = {g}_{n}\\left( x\\right) \) . If \( x \\in {A}_{{2i} + 1}^{n + 1} \), then \( {g}_{n + 1}\\left( x\\right) = \) \( \\left( {{2i} + 1}\\right) /{2}^{n + 1} \\geq {2i}/{2}^{n + 1} = {g}_{n}\\left( x\\right) \) . If \( x \\in {B}^{n} \), then \( f\\left( x\\right) \\geq n \), and therefore \( x \\in \\mathop{\\bigcup }\\limits_{{i \\geq n{2}^{n + 1}}}{A}_{i}^{n + 1} \\cup {B}^{n + 1} \) . It follows that \( {g}_{n + 1}\\left( x\\right) \\geq n = {g}_{n}\\left( x\\right) \) .\n\nFinally, if \( f \) is bounded by \( m \), then for \( n \\geq m \) we have \( 0 \\leq f\\left( x\\right) - {g}_{n}\\left( x\\right) \\leq \) \( {2}^{-n} \) . In this case the convergence is uniform.
Yes
Lemma 1. Let \( f = \mathop{\sum }\limits_{{i = 1}}^{n}{\alpha }_{i}{x}_{{A}_{i}} \), where we assume only that the sets \( {A}_{i} \) are mutually disjoint measurable sets. Then \( \int f = \) \( \mathop{\sum }\limits_{{i = 1}}^{n}{\alpha }_{i}\mu \left( {A}_{i}\right) . \)
Proof. The function \( f \) is simple, and its range contains at most \( n \) elements. Let \( \left\{ {{\beta }_{1},\ldots ,{\beta }_{k}}\right\} \) be the range of \( f \), and let \( {B}_{i} = {f}^{-1}\left( \left\{ {\beta }_{i}\right\} \right) \) . Then \( f = \mathop{\sum }\limits_{{i = 1}}^{k}{\beta }_{i}{x}_{{B}_{i}} \) , and this representation is canonical; i.e., it conforms to the requirements of Equation (2). Putting \( {J}_{i} = \left\{ {j : {\alpha }_{j} = {\beta }_{i}}\right\} \), we have\n\n\[ \int f = \mathop{\sum }\limits_{{i = 1}}^{k}{\beta }_{i}\mu \left( {B}_{i}\right) = \mathop{\sum }\limits_{{i = 1}}^{k}{\beta }_{i}\mu \left( {\mathop{\bigcup }\limits_{{j \in {J}_{i}}}{A}_{j}}\right) = \mathop{\sum }\limits_{{i = 1}}^{k}\mathop{\sum }\limits_{{j \in {J}_{i}}}{\beta }_{i}\mu \left( {A}_{j}\right) \]\n\n\[ = \mathop{\sum }\limits_{{i = 1}}^{k}\mathop{\sum }\limits_{{j \in {J}_{i}}}{\alpha }_{j}\mu \left( {A}_{j}\right) = \mathop{\sum }\limits_{{j = 1}}^{n}{\alpha }_{j}\mu \left( {A}_{j}\right) \]
Yes
Lemma 2. If \( g \) and \( f \) are simple functions such that \( g \leq f \), then \( \int g \leq \int f \) .
Proof. Start with canonical representations, as described following Equation (2):\n\n\[ g = \mathop{\sum }\limits_{{i = 1}}^{n}{\alpha }_{i}{x}_{{A}_{i}}\;f = \mathop{\sum }\limits_{{j = 1}}^{k}{\beta }_{j}{x}_{{B}_{j}} \]\n\nThen we have (non-canonical) representations conforming to Lemma 1:\n\n\[ g = \mathop{\sum }\limits_{{i = 1}}^{n}\mathop{\sum }\limits_{{j = 1}}^{k}{\alpha }_{i}{X}_{{A}_{i} \cap {B}_{j}}\;f = \mathop{\sum }\limits_{{j = 1}}^{k}\mathop{\sum }\limits_{{i = 1}}^{n}{\beta }_{j}{X}_{{A}_{i} \cap {B}_{j}} \]\n\nSince \( g \leq f \), we have \( {\alpha }_{i} \leq {\beta }_{j} \) whenever \( {A}_{i} \cap {B}_{j} \neq \varnothing \) . By Lemma 1\n\n\[ \int g = \mathop{\sum }\limits_{{i = 1}}^{n}\mathop{\sum }\limits_{{j = 1}}^{k}{\alpha }_{i}\mu \left( {{A}_{i} \cap {B}_{j}}\right) \;\int f = \mathop{\sum }\limits_{{j = 1}}^{k}\mathop{\sum }\limits_{{i = 1}}^{n}{\beta }_{j}\mu \left( {{A}_{j} \cap {B}_{j}}\right) \]\n\nHence\n\n\[ \int f - \int g = \mathop{\sum }\limits_{{i = 1}}^{n}\mathop{\sum }\limits_{{j = 1}}^{k}\left( {{\beta }_{j} - {\alpha }_{i}}\right) \mu \left( {{A}_{i} \cap {B}_{j}}\right) \geq 0 \]
Yes
Lemma 3. If \( f \) is a nonnegative simple function, then its integral as given in Equation (2) equals its integral as given in Equation (3).
Proof. Since \( f \) itself is simple, the expression on the right of Equation (3) is at least \( \int f \) . On the other hand, if \( g \) is simple and if \( g \leq f \), then by Lemma 2, \( \int g \leq \int f \) . By taking a supremum, we see that the right side of Equation (3) is at most \( \int f \) .
Yes
Lemma 4. If \( f \) and \( g \) are nonnegative simple functions, then \( f\left( {f + g}\right) = \int f + \int g \) .
Proof. Proceed exactly as in the proof of Lemma 2. Then\n\n\[ g + f = \mathop{\sum }\limits_{{i = 1}}^{n}\mathop{\sum }\limits_{{j = 1}}^{k}\left( {{\alpha }_{i} + {\beta }_{j}}\right) {X}_{{A}_{i} \cap {B}_{j}} \]\n\nBy the disjoint nature of the family \( \left\{ {{A}_{i} \cap {B}_{j}}\right\} \) we have\n\n\[ \int \left( {g + f}\right) = \mathop{\sum }\limits_{{i = 1}}^{n}\mathop{\sum }\limits_{{j = 1}}^{k}\left( {{\alpha }_{i} + {\beta }_{j}}\right) \mu \left( {{A}_{i} \cap {B}_{j}}\right) \]\n\nThis is the same as \( \int g + \int f \), as we see from an equation in the proof of Lemma 2.
Yes
Lemma 5. For two measurable functions \( f \) and \( g \), the condition \( 0 \leq f \leq g \) implies \( 0 \leq \int f \leq \int g \) .
Proof. Since 0 is a simple function, the definition of \( \int f \) in Equation (3) gives \( \int f \geq \int 0 = 0 \) . If \( h \) is a simple function such that \( h \leq f \), then \( h \leq g \) and \( \int h \leq \int g \) by the definition of \( \int g \) . In this last inequality, take the supremum in \( h \) to get \( \int f \leq \int g \) .
Yes
Theorem 1. Monotone Convergence Theorem. is a sequence of measurable functions such that \( 0 \leq {f}_{n} \uparrow f \), then \( 0 \leq \int {f}_{n} \uparrow \int f \)
Proof. (Rudin) Since \( 0 \leq {f}_{n} \leq {f}_{n + 1} \leq f \), we have \( 0 \leq \int {f}_{n} \leq \int {f}_{n + 1} \leq \int f \) by Lemma 5. Hence \( \lim \int {f}_{n} \) exists and is no greater than \( \int f \) . For the reverse inequality, let \( 0 < \theta < 1 \) and let \( g \) be a simple function satisfying \( 0 \leq g \leq f \) . Put \( {A}_{n} = \left\{ {x : {f}_{n}\left( x\right) \geq {\theta g}\left( x\right) }\right\} \) . If \( f\left( x\right) = 0 \), then \( g\left( x\right) = {f}_{n}\left( x\right) = 0 \), and \( x \in {A}_{n} \) for all \( n \) . If \( f\left( x\right) > 0 \), then eventually \( {f}_{n}\left( x\right) \geq {\theta g}\left( x\right) \) . Hence \( x \in \mathop{\bigcup }\limits_{{n = 1}}^{\infty }{A}_{n} \) and \( X = \mathop{\bigcup }\limits_{{n = 1}}^{\infty }{A}_{n} \) . Also, we have \( {A}_{n} \subset {A}_{n + 1} \) for all \( n \) . By Lemma 2 in Section 8.2, page 387, we have, for any measurable set \( E \) ,\n\n(4)\n\n\[ \mu \left( {{A}_{n} \cap E}\right) \uparrow \mu \left( E\right) \]\n\nFrom this it is easy to prove that \( \int g{x}_{{A}_{n}} \uparrow \int g \) . Indeed, we write \( g = \) \( \mathop{\sum }\limits_{{i = 1}}^{m}{\lambda }_{i}{x}_{{E}_{i}} \) ( \( {E}_{i} \) being mutually disjoint) and observe that\n\n\[ \int g{X}_{{A}_{n}} = \int \mathop{\sum }\limits_{{i = 1}}^{m}{\lambda }_{i}{X}_{{A}_{n}}{X}_{{E}_{i}} = \int \mathop{\sum }\limits_{{i = 1}}^{m}{\lambda }_{i}{X}_{{A}_{n} \cap {E}_{i}} = \mathop{\sum }\limits_{{i = 1}}^{m}{\lambda }_{i}\mu \left( {{A}_{n} \cap {E}_{i}}\right) \]\n\nAs \( n \uparrow \infty \), we have \( \mu \left( {{A}_{n} \cap {E}_{i}}\right) \uparrow \mu \left( {E}_{i}\right) \) by Equation (4). Since the coefficients \( {\lambda }_{i} \) are nonnegative, \( \int g{\mathbf{X}}_{n} \uparrow \mathop{\sum }\limits_{{i = 1}}^{m}{\lambda }_{i}\mu \left( {E}_{i}\right) = \int g \) . We have proved that\n\n\[ \theta \int g = \mathop{\lim }\limits_{n}\int {\theta g}{x}_{{A}_{n}} \leq \mathop{\lim }\limits_{n}\int {f}_{n} \]\n\nSince this is true for any \( \theta \) in \( \left( {0,1}\right) \), one concludes that \( \int g \leq \mathop{\lim }\limits_{n}\int {f}_{n} \) . In this inequality take a supremum over all simple \( g \) for which \( 0 \leq g \leq f \), arriving at \( \int f \leq \mathop{\lim }\limits_{n}\int {f}_{n} \)
Yes
Theorem 2. For nonnegative measurable functions \( f \) and \( g \) we have \( \int \left( {f + g}\right) = \int f + \int g \) .
Proof. By Theorem 5 in Section 8.4, page 397, there exist nonnegative simple functions \( {f}_{n} \uparrow f \) and \( {g}_{n} \uparrow g \) . Then \( {f}_{n} + {g}_{n} \uparrow f + g \) . By Theorem 1 (the Monotone Convergence Theorem) and Lemma 4 above, we have \[ \int \left( {f + g}\right) = \mathop{\lim }\limits_{n}\int \left( {{f}_{n} + {g}_{n}}\right) = \lim \left\lbrack {\int {f}_{n}+\int {g}_{n}}\right\rbrack = \int f + \int g \]
Yes
Theorem 3. Let \( f \) be nonnegative and measurable. The conditions\n\n\( \int f = 0 \) and \( f\left( x\right) = 0 \) almost everywhere are equivalent.
Proof. Let \( A = \{ x : f\left( x\right) > 0\} \) and \( B = X \smallsetminus A \) . If \( f\left( x\right) = 0 \) almost everywhere, then \( \mu \left( A\right) = 0 \) . Hence\n\n\[ \n\int f = \int \left( {f{x}_{A} + f{x}_{B}}\right) = \int f{x}_{A} + \int f{x}_{B} \n\]\n\n\[ \n\leq \int \infty {x}_{A} + \int 0{x}_{B} = \infty \mu \left( A\right) + {0\mu }\left( B\right) = 0 \n\]\n\nFor the other implication, assume \( \int f = 0.\; \) Define \( \;{A}_{n} = \{ x : f\left( x\right) > \frac{1}{n}\} . \) Then \( A = \mathop{\bigcup }\limits_{{n = 1}}^{\infty }{A}_{n} \) . Since \( \frac{1}{n}{x}_{{A}_{n}} \) is a simple function bounded above by \( f \) we have\n\n\[ \n0 = \int f \geq \int \frac{1}{n}{x}_{{A}_{n}} = \frac{1}{n}\mu \left( {A}_{n}\right) \n\]\n\nThus \( \mu \left( {A}_{n}\right) = 0 \) for all \( n \) and \( \mu \left( A\right) = 0 \) by Lemma 1 in Section 8.2, page 386. ∎
Yes
Theorem 4. Fatou's Lemma. For a sequence of nonnegative measurable functions, \( \int \left( {\liminf {f}_{n}}\right) \leq \liminf \int {f}_{n} \) .
Proof. Recall that the limit infimum of a sequence of real numbers \( \left\lbrack {x}_{n}\right\rbrack \) is defined to be \( \mathop{\lim }\limits_{{n \rightarrow \infty }}\mathop{\inf }\limits_{{i \geq n}}{x}_{i} \) . The limit infimum of a sequence of real-valued functions is defined pointwise: \( \left( {\lim \inf {f}_{n}}\right) \left( x\right) = \lim \inf {f}_{n}\left( x\right) = \lim {g}_{n}\left( x\right) \), where \( {g}_{n}\left( x\right) = \mathop{\inf }\limits_{{i \geq n}}{f}_{i}\left( x\right) \) . Observe that \( {g}_{n - 1}\left( x\right) \leq {g}_{n}\left( x\right) \leq {f}_{n}\left( x\right) \) and that \( {g}_{n} \uparrow \) lim inf \( {f}_{n} \) . Hence by Theorem 1 (The Monotone Convergence Theorem)\n\n\[ \int \left( {\lim \inf {f}_{n}}\right) = \int \lim {g}_{n} = \lim \int {g}_{n} = \lim \inf \int {g}_{n} \leq \lim \inf \int {f}_{n} \]
Yes
Theorem 5. If \( f \) and \( g \) are nonnegative measurable functions that are equal almost everywhere, then \( \int f = \int g \) .
Proof. Let \( A = \{ x : f\left( x\right) = g\left( x\right) \} \) and \( B = X \smallsetminus A \) . Then\n\n\[ 0 \leq \int f{x}_{B} \leq \int \infty {x}_{B} = \infty \mu \left( B\right) = \infty 0 = 0 \]\n\nSimilarly, \( \int g{\mathbf{X}}_{B} = 0 \) . Hence\n\n\[ \int f = \int (f{X}_{X} + f{X}_{B}) = \int f{X}_{A} = \int g{X}_{A} = \int (g{X}_{A} + g{X}_{B}) = \int g \]
Yes
Lemma 1. A function \( f \) is integrable if and only if its positive and negative parts, \( {f}^{ + } \) and \( {f}^{ - } \), are integrable.
Proof. Assume that \( f \) is integrable. Then it is measurable, and the measurability of \( {f}^{ + } \) follows from the fact that \( \left\{ {x : {f}^{ + }\left( x\right) \geq a}\right\} \) is \( X \) when \( a \leq 0 \) and is \( \{ x : f\left( x\right) \geq a\} \) when \( a > 0 \) . The finiteness of the integral of \( \left| {f}^{ + }\right| \) is immediate from the inequality \( \left| {{f}^{ + }\left( x\right) }\right| \leq \left| {f\left( x\right) }\right| \) . The remainder of the proof involves similar elementary ideas.
No
Theorem 1. The set \( {L}^{1}\left( {X,\mathcal{A},\mu }\right) \) is a linear space, and the integral is a linear functional on it.
Proof. Let \( f \) and \( g \) be members of \( {L}^{1} \). To show that \( f + g \in {L}^{1} \), write \( h = f + g \), and \[ {h}^{ + } - {h}^{ - } = h = {f}^{ + } - {f}^{ - } + {g}^{ + } - {g}^{ - } \] From this it follows that \[ {h}^{ + } + {f}^{ - } + {g}^{ - } = {h}^{ - } + {f}^{ + } + {g}^{ + } \] Since these are all nonnegative functions, Theorem 2 of Section 8.5, page 402, is applicable, and \[ \int {h}^{ + } + \int {f}^{ - } + \int {g}^{ - } = \int {h}^{ - } + \int {f}^{ + } + \int {g}^{ + } \] Therefore, by Lemma 1, \[ \int (f + g) = \int h = \int {h}^{ + } - \int {h}^{ - } = \int {f}^{ + } - \int {f}^{ - } + \int {g}^{ + } - \int {g}^{ - } = \int f + \int g \] With this equation now established, we use Lemma 5 in Section 8.5 (page 401) to write \[ \int \left| {f + g}\right| \leq \int \left( {\left| f\right| + \left| g\right| }\right) = \int \left| f\right| + \int \left| g\right| < \infty \] For scalar multiplication, observe first that if \( \lambda \geq 0 \) and \( f \geq 0 \), then the definition of the integral in Equation (3) of Section 8.5 (page 400) gives \( \int {\lambda f} = \lambda \int f \). If \( f \geq 0 \) and \( \lambda < 0 \), then \[ \int {\lambda f} = \int {\left( \lambda f\right) }^{ + } - \int {\left( \lambda f\right) }^{ - } = - \int {\left( \lambda f\right) }^{ - } = - \int \left( {-\lambda {f}^{ + }}\right) = \lambda \int {f}^{ + } = \lambda \int f \] In the general case, we use what has already been proved: \[ \int {\lambda f} = \int \left\lbrack {\lambda {f}^{ + } + \left( {-\lambda }\right) {f}^{ - }}\right\rbrack = \int \lambda {f}^{ + } + \int - \lambda {f}^{ - } = \lambda \int {f}^{ + } - \lambda \int {f}^{ - } \] \[ = \lambda \left\lbrack {\int {f}^{ + }-\int {f}^{ - }}\right\rbrack = \lambda \int f \] The finiteness of the integral is now trivial: \[ \int \left| {\lambda f}\right| = \int \left| \lambda \right| \left| f\right| = \left| \lambda \right| \int \left| f\right| < \infty \]
Yes
Theorem 2. Dominated Convergence Theorem. \( g,{f}_{1},{f}_{2},\ldots \) be functions in \( {L}^{1}\left( {X,\mathcal{A},\mu }\right) \) such that \( \left| {f}_{n}\right| \leq g \) . If the sequence \( \left\lbrack {f}_{n}\right\rbrack \) converges pointwise to a function \( f \), then \( f \in {L}^{1} \) and \( \int {f}_{n} \rightarrow \int f \) .
Proof. The functions \( {f}_{n} + g \) are nonnegative. By Fatou’s Lemma (Theorem 4 in Section 8.5, page 403) and by the preceding theorem,\n\n\[ \int g + \int f = \int \left( {g + f}\right) = \int \liminf \left( {g + {f}_{n}}\right) \leq \liminf \int \left( {g + {f}_{n}}\right) \]\n\n\[ = \lim \inf \left\lbrack {\int g+\int {f}_{n}}\right\rbrack = \int g + \lim \inf \int {f}_{n} \]\n\nSince \( \int g < \infty \), we conclude that \( \int f \leq \liminf \int {f}_{n} \) . Since \( - f \) and \( - {f}_{n} \) satisfy the hypotheses of our theorem, the same conclusion can be drawn for them: \( \int - f \leq \liminf \int - {f}_{n} \) . This is equivalent to \( - \int f \leq - \limsup \int {f}_{n} \) and to \( \int f \geq \) lim sup \( \int {f}_{n} \) . Putting this all together produces\n\n\[ \lim \inf \int {f}_{n} \leq \lim \sup \int {f}_{n} \leq \int f \leq \lim \inf \int {f}_{n} \]
Yes
Theorem 3. Let \( f \) be Lebesgue integrable on the real line. For any positive \( \varepsilon \) there exist a simple function \( g \), a step function \( h \) and a continuous function \( k \) having compact support such that\n\n\[ \int \left| {f - g}\right| < \varepsilon \;\int \left| {f - h}\right| < \varepsilon \;\int \left| {f - k}\right| < \varepsilon \]
Proof. By Lemma \( 1,{f}^{ + } \) and \( {f}^{ - } \) are integrable. By the definition of the integral, Equation (3) in Section 8.5, page 400, there exist simple functions \( {g}_{1} \) and \( {g}_{2} \) such that \( {g}_{1} \leq {f}^{ + },{g}_{2} \leq {f}^{ - },\int {f}^{ + } < \int {g}_{1} + \varepsilon \), and \( \int {f}^{ - } < \int {g}_{2} + \varepsilon \) . Then \( {g}_{1} - {g}_{2} \) is a simple function such that\n\n\[ \int \left| {f - {g}_{1} + {g}_{2}}\right| \leq \int \left| {{f}^{ + } - {g}_{1}}\right| + \int \left| {{f}^{ - } - {g}_{2}}\right| = \int \left( {{f}^{ + } - {g}_{1}}\right) + \int \left( {{f}^{ - } - {g}_{2}}\right) < {2\varepsilon } \]\n\nIn order to establish the second part of the theorem, it now suffices to prove it in the special case that \( f \) is an integrable simple function. It is therefore a linear combination of characteristic functions of measurable sets of finite measure. It then suffices to prove this part of the theorem when \( f = {x}_{A} \) for some measurable set \( A \) having finite measure. By the definition of Lebesgue measure, there is a countable family of open intervals \( \left\{ {I}_{n}\right\} \) that cover \( A \) and satisfy \( \mu \left( A\right) \leq \mathop{\sum }\limits_{{n = 1}}^{\infty }\mu \left( {I}_{n}\right) < \mu \left( A\right) + \varepsilon \) . There is no loss of generality in assuming that the family \( \left\{ {I}_{n}\right\} \) is disjoint, because if two of these intervals have a point in common, their union is a single open interval. Since the series \( \sum \mu \left( {I}_{n}\right) \) converges, there is an index \( m \) such that \( \mathop{\sum }\limits_{{n = m + 1}}^{\infty }\mu \left( {I}_{n}\right) < \varepsilon \) . Put \( B = \mathop{\bigcup }\limits_{{n = 1}}^{m}{I}_{n} \) , \( E = \mathop{\bigcup }\limits_{{n = m + 1}}^{\infty }{I}_{n}, h = {\mathbf{X}}_{B} \), and \( \varphi = {\mathbf{X}}_{E} \) . Then \( h \) is a step function. Since \( A \subset B \cup E \), we have \( f \leq h + \varphi \) . Then\n\n\[ \left| {h - f}\right| \leq \left| {h + \varphi - f}\right| + \left| \varphi \right| = \left( {h + \varphi - f}\right) - \varphi \]\n\nConsequently,\n\n\[ \int \left| {h - f}\right| \leq \int \left( {h + \varphi - f}\right) + \int \varphi \]\n\n\[ = \mu \left( B\right) + \mu \left( E\right) - \mu \left( A\right) + \mu \left( E\right) \leq {2\varepsilon } \]\n\nFor the third part of the proof it suffices to consider an \( f \) that is an integrable step function. For this, in turn, it is enough to prove that the characteristic function of a single compact interval can be approximated in \( {L}^{1} \) by a continuous function that vanishes outside that interval. This can certainly be done with a piecewise linear function.
Yes
Theorem 1. Hölder’s Inequality. Let \( 1 \leq p \leq \infty ,\frac{1}{p} + \frac{1}{q} = 1 \) , \( f \in {L}^{p} \), and \( g \in {L}^{q} \) . Then \( {fg} \in {L}^{1} \) and \[ \int \left| {fg}\right| = \parallel {fg}{\parallel }_{1} \leq \parallel f{\parallel }_{p}\parallel g{\parallel }_{q} \]
Proof. The seminorms involved here are homogeneous: \( \parallel {\lambda f}\parallel = \left| \lambda \right| \parallel f\parallel \) . Consequently, it will suffice to establish Equation (3) in the special case when \( \parallel f{\parallel }_{p} = \) \( \parallel g{\parallel }_{q} = 1 \) . At first, let \( p = 1 \) and \( q = \infty \) . Since \( g \in {L}^{\infty } \), we have \( \left| {g\left( x\right) }\right| \leq M \) a.e. for some \( M \) . From this it follows that \( \int \left| {fg}\right| \leq M\int \left| f\right| = M\parallel f{\parallel }_{1} \) . By taking the infimum for all \( M \), we obtain \( \parallel {fg}{\parallel }_{1} \leq \parallel f{\parallel }_{1}\parallel g{\parallel }_{\infty } \) . Suppose now that \( p > 1 \) . We prove first that if \( a > 0, b > 0 \), and \( 0 \leq t \leq 1 \) , then \( {a}^{t}{b}^{1 - t} \leq {ta} + \left( {1 - t}\right) b \) . The accompanying Figure 8.1 shows the functions of \( t \) on the two sides of this inequality (when \( a = 2 \) and \( b = {12} \) ). It is clear that we should prove convexity of the function \( \varphi \left( t\right) = {a}^{t}{b}^{1 - t} \) . This requires that we prove \( {\varphi }^{\prime \prime }\left( t\right) \geq 0 \) . Since \( \log \varphi \left( t\right) = t\log a + \left( {1 - t}\right) \log b \), we have \[ \frac{{\varphi }^{\prime }\left( t\right) }{\varphi \left( t\right) } = \log a - \log b = c \] whence \( {\varphi }^{\prime \prime }\left( t\right) = c{\varphi }^{\prime }\left( t\right) = {c}^{2}\varphi \left( t\right) \geq 0 \) . Now let \( a = {\left| f\left( x\right) \right| }^{p}, b = {\left| g\left( x\right) \right| }^{q}, t = 1/p,1 - t = 1/q \) . Our inequality yields then \( \left| {f\left( x\right) g\left( x\right) }\right| \leq \frac{1}{p}{\left| f\left( x\right) \right| }^{p} + \frac{1}{q}{\left| g\left( x\right) \right| }^{q} \) . By hypothesis, the functions on the right in this inequality belong to \( {L}^{1} \) . Hence by integrating we obtain \[ \parallel {fg}{\parallel }_{1} = \int \left| {fg}\right| \leq \frac{1}{p}\int {\left| f\right| }^{p} + \frac{1}{q}\int {\left| g\right| }^{q} = \frac{1}{p} + \frac{1}{q} = 1 = \parallel f{\parallel }_{p}\parallel g{\parallel }_{q} \]
Yes
Theorem 2. Minkowski’s Inequality Let \( 1 \leq p \leq \infty \) . If \( f \) and \( g \) belong to \( {L}^{p} \), then so does \( f + g \), and \[ \parallel f + g{\parallel }_{p} \leq \parallel f{\parallel }_{p} + \parallel g{\parallel }_{p} \]
Proof. The cases \( p = 1 \) and \( p = \infty \) are special. For the first of these cases, just write \[ \int \left| {f + g}\right| \leq \int \left( {\left| f\right| + \left| g\right| }\right) = \int \left| f\right| + \int \left| g\right| \] For \( p = \infty \), select constants \( M \) and \( N \) for which \( \left| {f\left( x\right) }\right| \leq M \) a.e. and \( \left| {g\left( x\right) }\right| \leq N \) a.e. Then \( \left| {f\left( x\right) + g\left( x\right) }\right| \leq M + N \) a.e. This proves that \( f + g \in {L}^{\infty } \) and that \( \parallel f + g{\parallel }_{\infty } \leq M + N \) . By taking infima we get \( \parallel f + g{\parallel }_{\infty } \leq \parallel f{\parallel }_{\infty } + \parallel g{\parallel }_{\infty } \) . Now let \( 1 < p < \infty \) . From the observation that \( \left| {f + g}\right| \leq 2\max \{ \left| f\right| ,\left| g\right| \} \), we have \[ {\left| f + g\right| }^{p} \leq {2}^{p}\max \left\{ {{\left| f\right| }^{p},{\left| g\right| }^{p}}\right\} \leq {2}^{p}\left( {{\left| f\right| }^{p} + {\left| g\right| }^{p}}\right) \] This establishes that \( f + g \in {L}^{p} \) . Next, write \[ {\left| f + g\right| }^{p} = \left| {f + g}\right| {\left| f + g\right| }^{p - 1} \leq \left| f\right| {\left| f + g\right| }^{p - 1} + \left| g\right| {\left| f + g\right| }^{p - 1} \] Since \( \left| {f + g}\right| \in {L}^{p} \), we can infer that \( {\left| f + g\right| }^{p - 1} \in {L}^{q} \) (where \( \frac{1}{p} + \frac{1}{q} = 1 \) ) because \[ \int {\left| f + g\right| }^{\left( {p - 1}\right) q} = \int {\left| f + g\right| }^{p} < \infty \] By the homogeneity of Minkowski’s inequality, we may assume that \( \parallel f + g{\parallel }_{p} = 1 \) . Observe now that Hölder’s Inequality is applicable to the product \( \left| f\right| {\left| f + g\right| }^{p - 1} \) and to the product \( \left| g\right| {\left| f + g\right| }^{p - 1} \) . Consequently, \[ 1 = \int {\left| f + g\right| }^{p} \leq \int \left| f\right| {\left| f + g\right| }^{p - 1} + \int \left| g\right| {\left| f + g\right| }^{p - 1} \] \[ \leq {\begin{Vmatrix}f\end{Vmatrix}}_{p}{\begin{Vmatrix}{\left| f + g\right| }^{p - 1}\end{Vmatrix}}_{q} + {\begin{Vmatrix}g\end{Vmatrix}}_{p}{\begin{Vmatrix}{\left| f + g\right| }^{p - 1}\end{Vmatrix}}_{q} \] This is equivalent to \[ 1 \leq \left\{ {\parallel f{\parallel }_{p} + \parallel g{\parallel }_{p}}\right\} \parallel f + g{\parallel }_{p}^{p/q} = \parallel f{\parallel }_{p} + \parallel g{\parallel }_{p} \]
Yes
Theorem 3. The Riesz-Fischer Theorem. \( E \)\n\n\( {L}^{p}\left( {X,\mathcal{A},\mu }\right) \), where \( 1 \leq p \leq \infty \), is complete.
Proof. The case \( p = \infty \) is special and is addressed first. Let \( \left\lbrack {f}_{n}\right\rbrack \) be a Cauchy sequence in \( {L}^{\infty } \) . Define\n\n\[ \n{E}_{nm} = \left\{ {x : \left| {{f}_{n}\left( x\right) - {f}_{m}\left( x\right) }\right| > {\begin{Vmatrix}{f}_{n} - {f}_{m}\end{Vmatrix}}_{\infty }}\right\} \n\]\n\nBy Problem 1, these sets all have measure 0 . Hence the same is true of their union, \( E \) . If \( x \in X \smallsetminus E \), then \( \left| {{f}_{n}\left( x\right) - {f}_{m}\left( x\right) }\right| \leq {\begin{Vmatrix}{f}_{n} - {f}_{m}\end{Vmatrix}}_{\infty } \), and thus \( \left\lbrack {{f}_{n}\left( x\right) }\right\rbrack \) is a Cauchy sequence in \( \mathbb{R} \) for each \( x \in X \smallsetminus E \) . This sequence converges to a number that we may denote by \( f\left( x\right) \) . Define \( f\left( x\right) = 0 \) for \( x \in E \) . On \( X \smallsetminus E \) , \( \left| {f\left( x\right) }\right| = \lim \left| {{f}_{n}\left( x\right) }\right| \leq \lim {\begin{Vmatrix}{f}_{n}\end{Vmatrix}}_{\infty } < \infty \) . (Use the fact that a Cauchy sequence in a metric space is bounded.) Thus, \( f \in {L}^{\infty } \) . To prove that \( {\begin{Vmatrix}{f}_{n} - f\end{Vmatrix}}_{\infty } \rightarrow 0 \) , let \( \varepsilon > 0 \) and select \( N \) so that \( {\begin{Vmatrix}{f}_{n} - {f}_{m}\end{Vmatrix}}_{\infty } < \varepsilon \) when \( n > m > N \) . Then \( \left| {{f}_{n}\left( x\right) - {f}_{m}\left( x\right) }\right| < \varepsilon \) on \( X \smallsetminus E \), and \( \left| {f\left( x\right) - {f}_{m}\left( x\right) }\right| \leq \varepsilon \) for \( m > N.\)
No
Theorem 1. If \( \\left( {X,\\mathcal{A},\\mu }\\right) \) is a measure space, and if \( f \) is a nonnegative measurable function, then the equation\n\n(1)\n\n\[ \n\\nu \\left( A\\right) = {\\int }_{A}{fd\\mu }\\;\\left( {A \\in \\mathcal{A}}\\right) \n\]\n\ndefines a measure \( \\nu \) that is absolutely continuous with respect to \( \\mu \) .
Proof. The postulates for a measure are quickly verified.\n\n(a) \( \\nu \\left( \\varnothing \\right) = {\\int }_{\\varnothing }f = \\int f{x}_{\\varnothing } = \\int 0 = 0 \)\n\n(b) \( \\nu \\left( A\\right) \\geq 0 \) because \( f \\geq 0 \)\n\n(c) If \( \\left\\lbrack {A}_{i}\\right\\rbrack \) is a disjoint sequence of measurable sets, then\n\n\[ \n\\nu \\left( {\\mathop{\\bigcup }\\limits_{{i = 1}}^{\\infty }{A}_{i}}\\right) = {\\int }_{\\cup {A}_{i}}f = \\int f{x}_{\\cup {A}_{i}} = \\int f\\sum {x}_{{A}_{i}} = \\int \\sum f{x}_{{A}_{i}} \n\]\n\n\[ \n= \\int \\mathop{\\lim }\\limits_{n}\\mathop{\\sum }\\limits_{{i = 1}}^{n}f{x}_{{A}_{i}} = \\mathop{\\lim }\\limits_{n}\\int \\mathop{\\sum }\\limits_{{i = 1}}^{n}f{x}_{{A}_{i}} = \\mathop{\\lim }\\limits_{n}\\mathop{\\sum }\\limits_{{i = 1}}^{n}\\int f{x}_{{A}_{i}} \n\]\n\n\[ \n= \\mathop{\\sum }\\limits_{{i = 1}}^{\\infty }\\nu \\left( {A}_{i}\\right) \n\]\nThis calculation used the Monotone Convergence Theorem (Section 8.5, page 401). The absolute continuity of \( \\nu \) is clear: if \( \\mu \\left( A\\right) = 0 \), then \( \\nu \\left( A\\right) = {\\int }_{A}f = 0 \) .
Yes
Theorem 2. Radon-Nikodym Theorem. Let \( \mu \) and \( \nu \) be \( \sigma \) -finite measures on a measurable space \( \left( {X,\mathcal{A}}\right) \) . If \( \nu \) is absolutely continuous with respect to \( \mu \), then there exists a nonnegative measurable function \( h \), determined uniquely up to a set of \( \mu \) -measure 0, such that \( \nu \left( A\right) = \) \( {\int }_{A}{hd\mu } \) for all \( A \in \mathcal{A} \) .
Proof. We prove the theorem first under the assumption that \( \mu \left( X\right) < \infty \) and \( \nu \left( X\right) < \infty \) . Consider the Hilbert space \( {L}^{2} = {L}^{2}\left( {X,\mathcal{A},\mu + \nu }\right) \) . For any \( f \) in \( {L}^{2} \), define \( \Phi \left( f\right) = \int {fd\mu } \) . It is easily verified that \( \Phi \) is a linear functional on \( {L}^{2} \) . Furthermore it is bounded (continuous) because by the Hölder Inequality (Theorem 1 in Section 8.7, page 409)\n\n\[ \left| {\Phi \left( f\right) }\right| = \left| {\int f \cdot {1d\mu }}\right| \leq \int \left| f\right| \cdot {1d}\left( {\mu + \nu }\right) \leq \parallel f{\parallel }_{2}\parallel 1{\parallel }_{2} \]\n\nBy the Riesz Representation Theorem for Hilbert space (Section 2.3, page 81) there exists an element \( {h}_{0} \) in \( {L}^{2} \) such that\n\n\[ \Phi \left( f\right) = \int f{h}_{0}d\left( {\mu + \nu }\right) \;f \in {L}^{2} \]\n\nThis means that \( \int {fd\mu } = \int f{h}_{0}d\left( {\mu + \nu }\right) \), whence\n\n\[ \int f\left( {1 - {h}_{0}}\right) {d\mu } = \int f{h}_{0}{d\nu } \]\n\nLet \( B = \left\{ {x : {h}_{0}\left( x\right) \leq 0}\right\} \) . Then \( 1 - {h}_{0} \geq 1 \) on \( B \), and consequently\n\n\[ 0 \leq \mu \left( B\right) \leq \int {x}_{B}\left( {1 - {h}_{0}}\right) {d\mu } = \int {x}_{B}{h}_{0}{d\nu } \leq 0 \]\n\nThus \( \mu \left( B\right) = 0 \) and \( {h}_{0}\left( x\right) > 0 \) a.e. (with respect to \( \mu \) ). Since \( \nu \ll \mu \), we have \( \nu \left( B\right) = 0 \) also. Hence for any \( A \in \mathcal{A} \),\n\n\[ \nu \left( A\right) = \int {\mathbf{X}}_{A}{d\nu } = \int {h}_{0}^{-1}{\mathbf{X}}_{A}{h}_{0}{d\nu } = \int {h}_{0}^{-1}{\mathbf{X}}_{A}\left( {1 - {h}_{0}}\right) {d\mu } \]\n\n\[ = {\int }_{A}{h}_{0}^{-1}\left( {1 - {h}_{0}}\right) {d\mu } = {\int }_{A}{hd\mu }\;\left( {h = {h}_{0}^{-1}\left( {1 - {h}_{0}}\right) }\right) \]\nTo see that \( h \geq 0 \) a.e., with respect to \( \mu \), write \( A = \{ x : h\left( x\right) < 0\} \), so that \( 0 \leq \nu \left( A\right) = {\int }_{A}{hd\mu } \leq 0 \), whence \( \mu \left( A\right) = 0 \) .
Yes
Jordan Decomposition. The difference of two measures (defined on the same \( \sigma \) -algebra), one of which is finite, is a signed measure. Conversely, every signed measure \( \mu \) is the difference of two measures \( {\mu }^{ + } \) and \( {\mu }^{ - } \), one of which is finite. Furthermore, we may require these two measures to be mutually singular, and in that case they are uniquely determined by \( \mu \) .
For the first assertion, let \( {\mu }_{1} \) and \( {\mu }_{2} \) be measures, and suppose that \( {\mu }_{1} \) is finite. Put \( \mu = {\mu }_{1} - {\mu }_{2} \) . To see that \( \mu \) is a signed measure, note first that \( \mu \) does not assume the value \( + \infty \) . Next, we have \( \mu \left( \varnothing \right) = 0 \) since \( {\mu }_{1} \) and \( {\mu }_{2} \) have this property. Finally, let \( \left\{ {A}_{i}\right\} \) be a disjoint sequence of measurable sets. Then\n\n\[ \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) = {\mu }_{1}\left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) - {\mu }_{2}\left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) \]\n\n\[ = \mathop{\sum }\limits_{{i = 1}}^{\infty }{\mu }_{1}\left( {A}_{i}\right) - \mathop{\sum }\limits_{{i = 1}}^{\infty }{\mu }_{2}\left( {A}_{i}\right) \]\n\n\[ = \mathop{\lim }\limits_{n}\mathop{\sum }\limits_{{i = 1}}^{n}{\mu }_{1}\left( {A}_{i}\right) - \mathop{\lim }\limits_{n}\mathop{\sum }\limits_{{i = 1}}^{n}{\mu }_{2}\left( {A}_{i}\right) \]\n\n\[ = \mathop{\lim }\limits_{n}\left\lbrack {\mathop{\sum }\limits_{{i = 1}}^{n}{\mu }_{1}\left( {A}_{i}\right) - \mathop{\sum }\limits_{{i = 1}}^{n}{\mu }_{2}\left( {A}_{i}\right) }\right\rbrack \]\n\n\[ = \mathop{\lim }\limits_{n}\mathop{\sum }\limits_{{i = 1}}^{n}\left\lbrack {{\mu }_{1}\left( {A}_{i}\right) - {\mu }_{2}\left( {A}_{i}\right) }\right\rbrack = \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {A}_{i}\right) \]\n\nNotice that on the second line of this calculation the first sum is finite, although the second may be infinite.
Yes
Theorem 2. Radon-Nikodym Theorem for Signed Measures. Let \( \left( {X,\mathcal{A},\mu }\right) \) be a \( \sigma \) -finite measure space. If \( \nu \) is a finite-valued signed measure that is absolutely continuous with respect to \( \mu \), then there is a measurable function \( h \) such that for all \( A \in \mathcal{A},\nu \left( A\right) = {\int }_{A}{hd\mu } \) .
Proof. By the preceding theorem, there exist measures \( {\nu }^{ + } \) and \( {\nu }^{ - } \) such that \( \nu = {\nu }^{ + } - {\nu }^{ - } \) and \( {\nu }^{ + } \bot {\nu }^{ - } \) . Consequently, there exists a measurable set \( P \) for which \( {\nu }^{ + }\left( {X \smallsetminus P}\right) = 0 = {\nu }^{ - }\left( P\right) \) . If \( A \) is a measurable set satisfying \( \mu \left( A\right) = 0 \) , then \( \mu \left( {A \cap P}\right) = 0 \) and \( \nu \left( {A \cap P}\right) = 0 \), by the absolute continuity. Hence\n\n\[ \n{\nu }^{ + }\left( A\right) = {\nu }^{ + }\left( {A \cap P}\right) + {\nu }^{ + }\left( {A \smallsetminus P}\right) = {\nu }^{ + }\left( {A \cap P}\right) \n\]\n\n\[ \n= \left( {\nu + {\nu }^{ - }}\right) \left( {A \cap P}\right) = \nu \left( {A \cap P}\right) = 0 \n\]\n\nThis establishes that \( {\nu }^{ + } \) is absolutely continuous with respect to \( \mu \) . It follows that \( {\nu }^{ - } \) is also absolutely continuous with respect to \( \mu \) . By the earlier Radon-Nikodym Theorem (Theorem 2 in Section 8.8, page 414), there exist nonnegative measurable functions \( {h}_{1} \) and \( {h}_{2} \) such that for \( A \) in \( \mathcal{A} \) ,\n\n\[ \n{\nu }^{ + }\left( A\right) = {\int }_{A}{h}_{1}{d\mu }\;{\nu }^{ - }\left( A\right) = {\int }_{A}{h}_{2}{d\mu } \n\]\n\nIt follows that \( {h}_{1} \) and \( {h}_{2} \) are finite almost everywhere. Thus, there is nothing suspicious in the equation\n\n\[ \n\nu \left( A\right) = {\nu }^{ + }\left( A\right) - {\nu }^{ - }\left( A\right) = {\int }_{A}{h}_{1}{d\mu } - {\int }_{A}{h}_{2}{d\mu } = {\int }_{A}\left( {{h}_{1} - {h}_{2}}\right) {d\mu } \n\]
Yes
Theorem 3. The Hahn Decomposition. If \( \mu \) is a signed measure on the measurable space \( \left( {X,\mathcal{A}}\right) \), then there is a decomposition of \( X \) into a disjoint pair of measurable sets \( N \) and \( P \) such that \( \mu \left( A\right) \geq 0 \) when \( A \subset P \) and \( \mu \left( A\right) \leq 0 \) when \( A \subset N \) .
Proof. Left as a problem.
No
Lemma 1. Let \( \\left( {X,\\mathcal{A}}\\right) \) and \( \\left( {Y,\\mathcal{B}}\\right) \) be two measurable spaces. If\n\n\( E \\in \\mathcal{A} \\otimes \\mathcal{B} \), then \( {E}_{x} \\in \\mathcal{B} \) for all \( x \\in X \) and \( {E}^{y} \\in \\mathcal{A} \) for all \( y \\in Y \) .
Proof. Define\n\n\\[ \n\\mathcal{M} = \\left\\{ {E : E \\subset X \\times Y\\text{ and }{E}^{y} \\in \\mathcal{A}\\text{ for all }y \\in Y}\\right\\} \n\\]\n\nWe shall prove that \( \\mathcal{M} \) is a \( \\sigma \) -algebra containing all rectangles. From this it will follow that \( \\mathcal{M} \\supset \\mathcal{A} \\otimes \\mathcal{B} \), since the latter is the smallest \( \\sigma \) -algebra containing all rectangles. Then, if \( E \\in \\mathcal{A} \\otimes \\mathcal{B} \), we can conclude that \( E \\in \\mathcal{M} \) and that \( {E}^{y} \\in \\mathcal{A} \) for each \( y \) . Now consider any rectangle \( E = A \\times B \) . If \( y \\in B \), then \( {E}^{y} = A \\in \\mathcal{A} \) . If \( y \\notin B \), then \( {E}^{y} = \\varnothing \\in \\mathcal{A} \) . Thus in all cases \( {E}^{y} \\in \\mathcal{A} \) and \( E \\in \\mathcal{M} \) . Next, let \( E \) be any member of \( \\mathcal{M} \) . The equation\n\n(1)\n\n\\[ \n{\\left\\[ \\left( X \\times Y\\right) \\smallsetminus E\\right\\rbrack }^{y} = X \\smallsetminus {E}^{y} \n\\]\n\nshows that \( \\left( {X \\times Y}\\right) \\smallsetminus E \) belongs to \( \\mathcal{M} \) . If \( {E}_{i} \\in \\mathcal{M} \), then by the equation\n\n(2)\n\n\\[ \n{\\left\\[ \\mathop{\\bigcup }\\limits_{{i = 1}}^{\\infty }{E}_{i}\\right\\rbrack }^{y} = \\mathop{\\bigcup }\\limits_{{i = 1}}^{\\infty }{E}_{i}^{y} \n\\]\n\nwe see that \( \\bigcup {E}_{i} \\in \\mathcal{M} \) .
Yes
Lemma 2. The collection of all unions of finite disjoint families of rectangles constructed from a pair of \( \sigma \) -algebras is an algebra.
Proof. Let \( \mathcal{C} \) be the collection referred to, and let \( E \) and \( F \) be members of \( \mathcal{C} \). Then \( E \) and \( F \) have expressions \( E = \mathop{\bigcup }\limits_{{i = 1}}^{n}\left( {{A}_{i} \times {B}_{i}}\right) \) and \( F = \mathop{\bigcup }\limits_{{j = 1}}^{m}\left( {{C}_{j} \times {D}_{j}}\right) \), both being unions of disjoint families. Since\n\n\[ E \cap F = \mathop{\bigcup }\limits_{{i = 1}}^{n}\mathop{\bigcup }\limits_{{j = 1}}^{m}\left\lbrack {\left( {{A}_{i} \cap {C}_{j}}\right) \times \left( {{B}_{i} \cap {D}_{j}}\right) }\right\rbrack \]\n\nwe see that \( E \cap F \in \mathcal{C} \), and that \( \mathcal{C} \) is closed under the taking of intersections. From the equation\n\n\[ \left( {X \times Y}\right) \smallsetminus \left( {A \times B}\right) = \left\lbrack {\left( {X \smallsetminus A}\right) \times B}\right\rbrack \cup \left\lbrack {X \times \left( {Y \smallsetminus B}\right) }\right\rbrack \]\n\nwe get\n\n\[ \left( {X \times Y}\right) \smallsetminus E = \left( {X \times Y}\right) \smallsetminus \mathop{\bigcup }\limits_{{i = 1}}^{n}\left( {{A}_{i} \times {B}_{i}}\right) = \mathop{\bigcap }\limits_{{i = 1}}^{n}\left\lbrack {\left( {X \times Y}\right) \smallsetminus \left( {{A}_{i} \times {B}_{i}}\right) }\right\rbrack \]\n\n\[ = \mathop{\bigcap }\limits_{{i = 1}}^{n}\left\{ {\left\lbrack {\left( {X \smallsetminus {A}_{i}}\right) \times {B}_{i}}\right\rbrack \cup \left\lbrack {X \times \left( {Y \smallsetminus {B}_{i}}\right) }\right\rbrack }\right\} \]\n\nThis shows that the complement of \( E \) belongs to \( \mathcal{C} \), because \( \mathcal{C} \) is closed under finite intersections. By the de Morgan identities, \( \mathcal{C} \) is closed under unions.
Yes
Lemma 3. In any measure space \( \left( {X,\mathcal{A},\mu }\right) \) the following are true for measurable sets \( {A}_{i} \) :\n\n(1) If \( {A}_{1} \subset {A}_{2} \subset \cdots \), then \( \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) = \mathop{\lim }\limits_{n}\mu \left( {A}_{n}\right) \)\n\n(2) If \( {A}_{1} \supset {A}_{2} \supset \cdots \) and \( \mu \left( {A}_{1}\right) < \infty \), then \( \mu \left( {\mathop{\bigcap }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) = \mathop{\lim }\limits_{n}\mu \left( {A}_{n}\right) \)
Proof. Assume the hypothesis in (1), and define \( {B}_{n} = {A}_{n} \smallsetminus {A}_{n - 1} \) . The sequence \( \left\{ {B}_{n}\right\} \) is disjoint, and consequently,\n\n\[ \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) = \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{B}_{i}}\right) = \mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {B}_{i}\right) = \mathop{\lim }\limits_{{n \rightarrow \infty }}\mathop{\sum }\limits_{{i = 1}}^{n}\mu \left( {B}_{i}\right) \]\n\n\[ = \mathop{\lim }\limits_{{n \rightarrow \infty }}\mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{n}{B}_{i}}\right) = \mathop{\lim }\limits_{{n \rightarrow \infty }}\mu \left( {A}_{n}\right) \]\n\nTo establish (2), assume its hypothesis. Then \( \left\{ {{A}_{1} \smallsetminus {A}_{n}}\right\} \) is an increasing sequence, and by part (1) we have\n\n\[ \mu \left( {A}_{1}\right) - \mu \left( {\mathop{\bigcap }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) = \mu \left( {{A}_{1} \smallsetminus \mathop{\bigcap }\limits_{{i = 1}}^{\infty }{A}_{i}}\right) = \mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }\left( {{A}_{1} \smallsetminus {A}_{i}}\right) }\right) \]\n\n\[ = \mathop{\lim }\limits_{{n \rightarrow \infty }}\mu \left( {{A}_{1} \smallsetminus {A}_{n}}\right) = \mathop{\lim }\limits_{{n \rightarrow \infty }}\left( {\mu \left( {A}_{1}\right) - \mu \left( {A}_{n}\right) }\right) \]\n\n\[ = \mu \left( {A}_{1}\right) - \mathop{\lim }\limits_{{n \rightarrow \infty }}\mu \left( {A}_{n}\right) \]
Yes
Lemma 5. If \( \left( {X,\mathcal{A},\mu }\right) \) and \( \left( {Y,\mathcal{B},\nu }\right) \) are \( \sigma \) -finite measure spaces, then so is \( \left( {X \times Y,\mathcal{A} \otimes \mathcal{B},\mu \otimes \nu }\right) \) .
Proof. It is clear that the set function \( \phi \) has the property \( \phi \left( \varnothing \right) = 0 \) and the property \( \phi \left( E\right) \geq 0 \) . If \( \left\{ {E}_{i}\right\} \) is a disjoint sequence of sets in \( \mathcal{A} \otimes \mathcal{B} \), then \( \left\{ {E}_{i}^{y}\right\} \) is a disjoint sequence in \( \mathcal{A} \) . Hence, by the Dominated Convergence Theorem, \[ \phi \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{E}_{i}}\right) = {\int }_{Y}\mu \left( {\left( \mathop{\bigcup }\limits_{{i = 1}}^{\infty }{E}_{i}\right) }^{y}\right) {d\nu }\left( y\right) = {\int }_{Y}\mu \left( {\mathop{\bigcup }\limits_{{i = 1}}^{\infty }{E}_{i}^{y}}\right) {d\nu }\left( y\right) \] \[ = {\int }_{Y}\mathop{\sum }\limits_{{i = 1}}^{\infty }\mu \left( {E}_{i}^{y}\right) {d\nu }\left( y\right) = \mathop{\sum }\limits_{{i = 1}}^{\infty }{\int }_{Y}\mu \left( {E}_{i}^{y}\right) {d\nu }\left( y\right) = \mathop{\sum }\limits_{{i = 1}}^{\infty }\phi \left( {E}_{i}\right) \] Thus \( \phi \) is a measure. For the \( \sigma \) -finiteness, observe that if \( X = \mathop{\bigcup }\limits_{{n = 1}}^{\infty }{A}_{n} \) and \( Y = \mathop{\bigcup }\limits_{{n = 1}}^{\infty }{B}_{n} \), where \( {A}_{1} \subset {A}_{2} \subset \cdots \) and \( {B}_{1} \subset {B}_{2} \subset \cdots \), then \( X \times Y = \) \( \mathop{\bigcup }\limits_{{n = 1}}^{\infty }\left( {{A}_{n} \times {B}_{n}}\right) \) . If, further, \( \mu \left( {A}_{n}\right) < \infty \) and \( \nu \left( {B}_{n}\right) < \infty \) for all \( n \), then we have \( \phi \left( {{A}_{n} \times {B}_{n}}\right) = \mu \left( {A}_{n}\right) \nu \left( {B}_{n}\right) < \infty \) .
Yes
Theorem 2. Second Fubini Theorem. Let \( \\left( {X,\\mathcal{A},\\mu }\\right) \) and \( \\left( {Y,\\mathcal{B},\\nu }\\right) \) be two \( \\sigma \) -finite measure spaces. Let \( f \) be a nonnegative function on \( X \\times Y \) that is measurable with respect to \( \\left( {X \\times Y,\\mathcal{A} \\otimes \\mathcal{B}}\\right) \) . Then\n\n(1) For each \( x, y \\mapsto f\\left( {x, y}\\right) \) is measurable\n\n(2) For each \( y, x \\mapsto f\\left( {x, y}\\right) \) is measurable\n\n(3) \( y \\mapsto {\\int }_{X}f\\left( {x, y}\\right) {d\\mu }\\left( x\\right) \) is measurable\n\n(4) \( x \\mapsto {\\int }_{Y}f\\left( {x, y}\\right) {d\\nu }\\left( y\\right) \) is measurable\n\n\\left( 5\\right) \\;{\\int }_{X \\times Y}f\\left( {x, y}\\right) \\;{d\\phi } = {\\int }_{X}{\\int }_{Y}f\\left( {x, y}\\right) \\;{d\\nu }\\;{d\\mu } = {\\int }_{Y}{\\int }_{X}f\\left( {x, y}\\right) \\;{d\\mu }\\;{d\\nu }
Proof. If \( f \) is the characteristic function of a measurable set \( E \), then (1) is true because \( f\\left( {x, y}\\right) = {x}_{{E}_{x}}\\left( y\\right) \) . Part (2) is true by the symmetry in the situation. Since\n\n\\[ \n{\\int }_{X}f\\left( {x, y}\\right) {d\\mu }\\left( x\\right) = {\\int }_{X}{x}_{E}\\left( {x, y}\\right) {d\\mu }\\left( x\\right) = {\\int }_{X}{x}_{E}y\\left( x\\right) {d\\mu }\\left( x\\right) = \\mu \\left( {E}^{y}\\right) \n\\]\n\nthe preceding lemma asserts that (3) is also true in this case. Part (4) is true by symmetry. For part (5), write\n\n\\[ \n{\\int }_{X \\times Y}f\\left( {x, y}\\right) {d\\phi } = \\phi \\left( E\\right) = {\\int }_{Y}\\mu \\left( {E}^{y}\\right) {d\\nu }\\left( y\\right) \n\\]\n\n\\[ \n= {\\int }_{Y}{\\int }_{X}f\\left( {x, y}\\right) {d\\mu }\\left( x\\right) {d\\nu }\\left( y\\right) \n\\]\n\nThe other equality is similar. Thus, Theorem 2 is true when \( f \) is the characteristic function of a measurable set.\n\nIf \( f \) is a simple function, then \( f \) has properties (1) to (5) by the linearity of the integrals.\n\nIf \( f \) is an arbitrary nonnegative measurable function, then there exist simple functions \( {f}_{n} \) such that \( {f}_{n} \\uparrow f \) . Since the limit of a sequence of measurable functions is measurable, \( f \) has properties (1) to (4). By the Monotone Convergence Theorem, property (5) follows for \( f \) .
Yes
Theorem 3. Fubini's Theorem for Complex Measures. Let \( \left( {X,\mathcal{A}}\right) \) and \( \left( {Y,\mathcal{B}}\right) \) be two measurable spaces, and let \( \mu \) and \( \nu \) be complex measures on \( X \) and \( Y \), respectively. Let \( f \) be a complex-valued measurable function on \( X \times Y \) . If \( {\int }_{Y}{\int }_{X}\left| {f\left( {x, y}\right) }\right| d\left| \mu \right| d\left| \nu \right| < \infty \), then\n\n\[ \n{\int }_{Y}{\int }_{X}f\left( {x, y}\right) {d\mu }\left( x\right) {d\nu }\left( y\right) = {\int }_{X}{\int }_{Y}f\left( {x, y}\right) {d\nu }\left( y\right) {d\mu }\left( x\right) \n\]
This theorem is to be found in [DS].
No
Corollary 1. \( {A}^{-1} \) is an open set in \( A \) and \( x \mapsto {x}^{-1} \) is a continuous map of \( {A}^{-1} \) to itself.
Proof. To see that \( {A}^{-1} \) is open, choose an invertible element \( {x}_{0} \) and an arbitrary element \( h \in A \) . We have \( {x}_{0} + h = {x}_{0}\left( {\mathbf{1} + {x}_{0}^{-1}h}\right) \) . So if \( \begin{Vmatrix}{{x}_{0}^{-1}h}\end{Vmatrix} < 1 \) then by the preceding theorem \( {x}_{0} + h \) is invertible. In particular, if \( \parallel h\parallel < \) \( {\begin{Vmatrix}{x}_{0}^{-1}\end{Vmatrix}}^{-1} \), then this condition is satisfied, proving that \( {x}_{0} + h \) is invertible when \( \parallel h\parallel \) is sufficiently small.\n\nSupposing that \( h \) has been so chosen, we can write\n\n\[{\left( {x}_{0} + h\right) }^{-1} - {x}_{0}^{-1} = {\left( {x}_{0}\left( \mathbf{1} + {x}_{0}^{-1}h\right) \right) }^{-1} - {x}_{0}^{-1} = \left\lbrack {{\left( \mathbf{1} + {x}_{0}^{-1}h\right) }^{-1} - \mathbf{1}}\right\rbrack \cdot {x}_{0}^{-1}.\n\]\n\nThus for \( \parallel h\parallel < {\begin{Vmatrix}{x}_{0}^{-1}\end{Vmatrix}}^{-1} \) we have\n\n\[ \begin{Vmatrix}{{\left( {x}_{0} + h\right) }^{-1} - {x}_{0}^{-1}}\end{Vmatrix} \leq \begin{Vmatrix}{{\left( \mathbf{1} + {x}_{0}^{-1}h\right) }^{-1} - \mathbf{1}}\end{Vmatrix} \cdot \begin{Vmatrix}{x}_{0}^{-1}\end{Vmatrix} \leq \frac{\begin{Vmatrix}{{x}_{0}^{-1}h}\end{Vmatrix} \cdot \begin{Vmatrix}{x}_{0}^{-1}\end{Vmatrix}}{1 - \begin{Vmatrix}{{x}_{0}^{-1}h}\end{Vmatrix}},\]\n\nand the last term obviously tends to zero as \( \parallel h\parallel \rightarrow 0 \) .
Yes
Proposition 1.6.2. For every \( x \in A,\sigma \left( x\right) \) is a closed subset of the disk \( \{ z \in \mathbb{C} : \left| z\right| \leq \parallel x\parallel \} \)
Proof. The complement of the spectrum is given by\n\n\[ \mathbb{C} \smallsetminus \sigma \left( x\right) = \left\{ {\lambda \in \mathbb{C} : x - \lambda \in {A}^{-1}}\right\} .\n\]\n\nSince \( {A}^{-1} \) is open and the map \( \lambda \in \mathbb{C} \mapsto x - \lambda \in A \) is continuous, the complement of \( \sigma \left( x\right) \) must be open.\n\nTo prove the second assertion, we will show that no complex number \( \lambda \) with \( \left| \lambda \right| > \parallel x\parallel \) can belong to \( \sigma \left( x\right) \) . Indeed, for such a \( \lambda \) the formula\n\n\[ x - \lambda = \left( {-\lambda }\right) \left( {1 - {\lambda }^{-1}x}\right) \]\n\n\ntogether with the fact that \( \begin{Vmatrix}{{\lambda }^{-1}x}\end{Vmatrix} < 1 \), implies that \( x - \lambda \) is invertible.
Yes
Corollary 1. An element \( x \) of a unital Banach algebra \( A \) is quasinilpotent iff \( \sigma \left( x\right) = \{ 0\} \) .
Proof. \( x \) is quasinilpotent \( \Leftrightarrow r\left( x\right) = 0 \Leftrightarrow \sigma \left( x\right) = \{ 0\} \) .
Yes
Proposition 1.8.2. Let \( A \) be a Banach algebra with normalized unit \( \mathbf{1} \) and let \( I \) be a proper ideal in \( A \) . Then for every \( z \in I \) we have \( \parallel \mathbf{1} + z\parallel \geq 1 \) . In particular, the closure of a proper ideal is a proper ideal.
Proof. If there is an element \( z \in I \) with \( \parallel \mathbf{1} + z\parallel < 1 \), then by Theorem 1.5.2 \( z \) must be invertible in \( A \) ; hence \( \mathbf{1} = {z}^{-1}z \in I \), which implies that \( I \) cannot be a proper ideal. The second assertion follows from the continuity of the norm; if \( \parallel \mathbf{1} + z\parallel \geq 1 \) for all \( z \in I \), then \( \parallel \mathbf{1} + z\parallel \geq 1 \) persists for all \( z \) in the closure of \( I \) .
Yes
Proposition 1.8.4. Every bounded homomorphism of Banach algebras \( \omega : A \rightarrow B \) has a unique factorization \( \omega = \dot{\omega } \circ \pi \), where \( \dot{\omega } \) is an injective homomorphism of \( A/\ker \omega \) to \( B \) and \( \pi : A \rightarrow A/\ker \omega \) is the natural projection. One has \( \parallel \dot{\omega }\parallel = \parallel \omega \parallel \) .
Proof. The assertions in the first sentence are straightforward, and we prove \( \parallel \dot{\omega }\parallel = \parallel \omega \parallel \) . From the factorization \( \omega = \dot{\omega } \circ \pi \) and the fact that \( \parallel \pi \parallel \leq 1 \) we have \( \parallel \omega \parallel \leq \parallel \dot{\omega }\parallel \) ; the opposite inequality follows from\n\n\[ \parallel \dot{\omega }\left( \dot{x}\right) \parallel = \parallel \omega \left( x\right) \parallel = \parallel \omega \left( {x + z}\right) \parallel \leq \parallel \omega \parallel \parallel x + z\parallel ,\;z \in \ker \omega ,\]\n\nafter the infimum is taken over all \( z \in \ker \omega \) .
Yes
Proposition 1.9.3. In its relative weak*-topology, \( \operatorname{sp}\left( A\right) \) is a compact Hausdorff space.
Proof. It suffices to show that \( \operatorname{sp}\left( A\right) \) is a weak*-closed subset of the unit ball of the dual of \( A \) . Notice that a linear functional \( f : A \rightarrow \mathbb{C} \) belongs to \( \operatorname{sp}\left( A\right) \) iff \( \parallel f\parallel \leq 1, f\left( \mathbf{1}\right) = 1 \), and \( f\left( {yz}\right) = f\left( y\right) f\left( z\right) \) for all \( y, z \in A \) . These conditions obviously define a weak*-closed subset of the unit ball of \( {A}^{\prime } \) .
Yes
Proposition 1.11.1. Let \( B \) be a Banach subalgebra of \( A \) that contains the unit of \( A \) . For every element \( x \in B \) we have \( {\sigma }_{A}\left( x\right) \subseteq {\sigma }_{B}\left( x\right) \) .
Proof. This is an immediate consequence of the fact that invertible elements of \( B \) are invertible elements of \( A \) .
No
Corollary 1. Let \( {\mathbf{1}}_{A} \in B \subseteq A \) be as above, let \( x \in A \), and let \( \Omega \) be a bounded component of \( \mathbf{C} \smallsetminus {\sigma }_{A}\left( x\right) \) . Then either \( \Omega \cap {\sigma }_{B}\left( x\right) = \varnothing \) or \( \Omega \subseteq {\sigma }_{B}\left( x\right) \) .
Proof. Let \( \Omega \) be a hole of \( {\sigma }_{A}\left( x\right) \) . Consider \( X = \Omega \cap {\sigma }_{B}\left( x\right) \) as a closed subspace of the topological space \( \Omega \) . Since \( \Omega \) is an open set in \( \mathbb{C} \), the boundary \( {\partial }_{\Omega }X \) of \( X \) relative to \( \Omega \) is contained in\n\n\[ \n\partial {\sigma }_{B}\left( x\right) \subseteq {\sigma }_{A}\left( x\right) \subseteq \mathbb{C} \smallsetminus \Omega .\n\]\n\nHence \( {\partial }_{\Omega }X = \varnothing \) . Lemma 1.11.4 implies that either \( X = \varnothing \) or \( X = \Omega \), as asserted.
Yes
For every bounded complex-valued sesquilinear form \( \left\lbrack {\cdot , \cdot }\right\rbrack \) on \( H \) there is a unique bounded operator \( A \) on \( H \) such that\n\n\[ \left\lbrack {\xi ,\eta }\right\rbrack = \langle {A\xi },\eta \rangle ,\;\xi ,\eta \in H. \]
Proof. Fix a vector \( \xi \in H \) and consider the linear functional \( f \) defined on \( H \) by \( f\left( \eta \right) = \overline{\left\lbrack \xi ,\eta \right\rbrack } \), the bar denoting complex conjugation. Since \( f \) is a bounded linear functional, the Riesz lemma in its above form implies that there is a unique vector \( {A\xi } \in H \) satisfying \( f\left( \eta \right) = \langle \eta ,{A\xi }\rangle \) ; and after taking the complex conjugate we find that the function \( A : H \rightarrow H \) that we have defined must satisfy\n\n\[ \left\lbrack {\xi ,\eta }\right\rbrack = \langle {A\xi },\eta \rangle ,\;\xi ,\eta \in H. \]\n\nIt is straightforward to verify that this formula implies that \( A \) is a linear transformation. It is bounded because\n\n\[ \mathop{\sup }\limits_{{\parallel \xi \parallel \leq 1}}\parallel {A\xi }\parallel = \mathop{\sup }\limits_{{\parallel \xi \parallel \leq 1,\parallel \eta \parallel \leq 1}}\left| \left\lbrack {\xi ,\eta }\right\rbrack \right| < \infty . \]\n\nThe uniqueness of the operator \( A \) is evident from the uniqueness assertion of the Riesz lemma for linear functionals.
Yes
Corollary 1. Let \( H, K \) be Hilbert spaces and let \( A \in \mathcal{B}\left( {H, K}\right) \) be a bounded operator from \( H \) to \( K \) . There is a unique operator \( {A}^{ * } \in \mathcal{B}\left( {K, H}\right) \) satisfying\n\n\[ \langle {A\xi },\eta {\rangle }_{K} = {\left\langle \xi ,{A}^{ * }\eta \right\rangle }_{H},\;\xi \in H,\;\eta \in K. \]
Proof. One simply applies the above results to the bounded sesquilin-ear form \( \left\lbrack {\cdot , \cdot }\right\rbrack \) defined on \( K \times H \) by \( \left\lbrack {\eta ,\xi }\right\rbrack = \langle \eta ,{A\xi }\rangle \) to deduce the existence of a unique operator \( {A}^{ * } \in \mathcal{B}\left( {K, H}\right) \) satisfying \( {\left\langle {A}^{ * }\eta ,\xi \right\rangle }_{H} = \langle \eta ,{A\xi }{\rangle }_{K} \), and then takes the complex conjugate of both sides.
Yes
Proposition 2.2.3. Let \( x, y \) be elements of a unital Banach algebra \( A \) satisfying \( {xy} = {yx} \). Then \( {e}^{x + y} = {e}^{x}{e}^{y} \).
Proof. Using formula (2.3), we have\n\n\[ \n{e}^{x}{e}^{y} = \mathop{\sum }\limits_{{p, q = 0}}^{\infty }\frac{1}{p!}{x}^{p}\frac{1}{q!}{y}^{q} = \mathop{\sum }\limits_{{n = 0}}^{\infty }\left( {\mathop{\sum }\limits_{{p + q = n}}\frac{1}{p!q!}{x}^{p}{y}^{q}}\right) .\n\]\n\nSince \( {xy} = {yx} \), the proof of the binomial theorem applies here to give\n\n\[ \n{\left( x + y\right) }^{n} = \mathop{\sum }\limits_{{k = 0}}^{n}\left( \begin{array}{l} n \\ k \end{array}\right) {x}^{k}{y}^{n - k} = n!\mathop{\sum }\limits_{{p + q = n}}\frac{1}{p!q!}{x}^{p}{y}^{q}\n\]\n\nhence the right side of the preceding formula becomes\n\n\[ \n\mathop{\sum }\limits_{{n = 0}}^{\infty }\frac{1}{n!}{\left( x + y\right) }^{n} = {e}^{x + y}\n\]
Yes
Corollary 1. Let \( A \) be a (perhaps noncommutative) unital \( {C}^{ * } \) -algebra. Then the spectrum of any self-adjoint element \( x \) of \( A \) is real.
Proof. Choose an element \( x = {x}^{ * } \) of \( A \), and let \( B \) be the norm-closure of the set of all polynomials in \( x \) . Then \( B \) is a commutative \( {C}^{ * } \) -subalgebra of \( A \) that contains the unit of \( A \), hence \( {\sigma }_{A}\left( x\right) \subseteq {\sigma }_{B}\left( x\right) \) . On the other hand, Theorem 2.2.4 implies that \( \omega \left( x\right) \) is real for every \( \omega \in \operatorname{sp}\left( B\right) \), and hence \( {\operatorname{sp}}_{A}\left( x\right) \subseteq {\sigma }_{B}\left( x\right) = \{ \omega \left( x\right) : \omega \in \operatorname{sp}\left( B\right) \} \subseteq \mathbb{R}. \
Yes
Corollary 2. Let \( A \) be a unital \( {C}^{ * } \) -algebra and let \( B \subseteq A \) be a \( {C}^{ * } \) - subalgebra of \( A \) that contains the unit of \( A \) . Then for every \( x \in B \) we have \( {\sigma }_{B}\left( x\right) = {\sigma }_{A}\left( x\right) \) . In particular, for every self-adjoint \( x \in A \) , \[ \parallel x\parallel = r\left( x\right) . \]
Proof. We know that \( {\sigma }_{A}\left( x\right) \subseteq {\sigma }_{B}\left( x\right) \) in general, and to prove the opposite inclusion it suffices to show that for any element \( x \in B \) which is invertible in \( A \) one has \( {x}^{-1} \in B \) . Fix such an \( x \) . Then \( {x}^{ * }x \) is a self-adjoint element of \( B \) that is also invertible in \( A \) . By the preceding corollary, \( {\sigma }_{B}\left( {{x}^{ * }x}\right) \) is real. In particular, every point of \( {\sigma }_{B}\left( {{x}^{ * }x}\right) \) is a boundary point. By Theorem 1.11.3, \( {\sigma }_{B}\left( {{x}^{ * }x}\right) = \) \( \partial {\sigma }_{B}\left( {{x}^{ * }x}\right) \subseteq {\sigma }_{A}\left( {{x}^{ * }x}\right) \) . Since \( 0 \notin {\sigma }_{A}\left( {{x}^{ * }x}\right) ,0 \notin {\sigma }_{B}\left( {{x}^{ * }x}\right) \), and hence \( {x}^{ * }x \) is invertible in \( B \), equivalently, \( {\left( {x}^{ * }x\right) }^{-1} \in B \) . Obviously, \( {\left( {x}^{ * }x\right) }^{-1}{x}^{ * } \) is a left inverse of \( x \) ; hence \( {x}^{-1} = {\left( {x}^{ * }x\right) }^{-1}{x}^{ * } \) must belong to \( B \) . The assertion that \( \parallel x\parallel = r\left( x\right) \) follows after an application of Theorem 2.2.4 to the \( {C}^{ * } \) -subalgebra of \( A \) generated by \( x \) and 1 .
Yes
Proposition 2.5.4. Every nonunital Banach \( * \) -algebra can be embedded as a maximal ideal of codimension 1 in a unital Banach \( * \) -algebra for which \( \parallel \mathbf{1}\parallel = 1 \) .
Proof. Let \( A \) be a nonunital Banach \( * \) -algebra. The vector space \( A \oplus \mathbb{C} \) can be made into a \( * \) -algebra \( \widetilde{A} \) by introducing the operations\n\n\[ \left( {a,\lambda }\right) \cdot \left( {b,\mu }\right) = \left( {{ab} + {\lambda b} + {\mu a},{\lambda \mu }}\right) ,\;{\left( a,\lambda \right) }^{ * } = \left( {{a}^{ * },\bar{\lambda }}\right) .\n\]\n\nThe element \( \mathbf{1} = \left( {0,1}\right) \) is a unit for \( \widetilde{A} \), and we have \( \left( {a,\lambda }\right) = a + \lambda \mathbf{1} \) . Obviously, \( A \) is a maximal ideal of codimension 1 in \( \widetilde{A} \) . \( \widetilde{A} \) becomes a Banach \( * \) -algebra by way of the norm \( \parallel \left( {a,\lambda }\right) \parallel = \parallel a\parallel + \left| \lambda \right| \), with respect to which the inclusion map of \( A \) in \( \widetilde{A} \) is an isometric \( * \) -homomorphism.
Yes
Proposition 2.7.1. A spectral measure \( P \) has the following properties:\n\n(1) \( {E}_{1} \subseteq {E}_{2} \Rightarrow P\left( {E}_{1}\right) \leq P\left( {E}_{2}\right) \) .\n\n(2) \( E \cap F = \varnothing \Rightarrow P\left( E\right) \bot P\left( F\right) \) .\n\n(3) For every \( E, F \in \mathcal{B}, P\left( {E \cap F}\right) = P\left( E\right) P\left( F\right) \) .
Proof. The first assertion follows from finite additivity of \( P \), together with the decomposition \( F = E \cup \left( {F \smallsetminus E}\right) \) and the fact that \( P\left( {F \smallsetminus E}\right) \geq 0 \) .\n\nFor (2), we can write\n\n\[ \mathbf{1} = P\left( {E \cup \left( {\mathbb{C} \smallsetminus E}\right) }\right) = P\left( E\right) + P\left( {\mathbb{C} \smallsetminus E}\right) . \]\n\nHence by \( \left( 1\right), P\left( F\right) \leq P\left( {\mathbb{C} \smallsetminus E}\right) = \mathbf{1} - P\left( E\right) \), the latter being the projection onto \( P\left( E\right) {H}^{ \bot } \) .\n\nTo deduce (3) from (2), one can write \( P\left( E\right) = P\left( {E \cap F}\right) + P\left( {E \smallsetminus F}\right) \), \( P\left( F\right) = P\left( {E \cap F}\right) + P\left( {F \smallsetminus E}\right) \), and observe that because of (2), \( P\left( {E \cap F}\right) \), \( P\left( {E \smallsetminus F}\right) \), and \( P\left( {F \smallsetminus E}\right) \) are mutually orthogonal projections.\n\nThese observations imply that the projections \( P\left( {E}_{1}\right), P\left( {E}_{2}\right) ,\ldots \) appearing on the right of (2.17) are mutually orthogonal, so that the infinite sum has a clear meaning.
Yes
Proposition 2.8.1. Let \( N \) be a normal operator acting on an infinite-dimensional Hilbert space \( H \) . For every accumulation point \( \lambda \in \sigma \left( N\right) \) there is an orthonormal sequence \( {\xi }_{1},{\xi }_{2},\ldots \) in \( H \) such that\n\n\[ \mathop{\lim }\limits_{{n \rightarrow \infty }}\begin{Vmatrix}{N{\xi }_{n} - \lambda {\xi }_{n}}\end{Vmatrix} = 0 \]
Proof. By the Spectral Theorem we may assume that \( H = {L}^{2}\left( {X,\mu }\right) \) has been coordinatized by a \( \sigma \) -finite measure space and that \( N = {M}_{f} \) is multiplication by an \( {L}^{\infty } \) function. By Theorem 2.1.4 the spectrum of \( N \) is the essential range \( \Lambda \) of \( f \) .\n\nSince \( \lambda \) is an accumulation point of \( \Lambda \), we can find a sequence of distinct points \( {\lambda }_{n} \in \Lambda \) that converges to \( \lambda \) . For each \( n \) choose \( {\epsilon }_{n} > 0 \) small enough that \( {\epsilon }_{n} \rightarrow 0 \) and the disks \( {D}_{n} = \left\{ {z \in \mathbb{C} : \left| {z - {\lambda }_{n}}\right| < {\epsilon }_{n}}\right\}, n = 1,2,\ldots \) , are mutually disjoint. For each \( n \) the set \( \left\{ {p \in X : f\left( p\right) \in {D}_{n}}\right\} \) has positive measure because \( {\lambda }_{n} \) belongs to the essential range of \( f \) ; and by \( \sigma \) -finiteness there is a subset \( {E}_{n} \subseteq \left\{ {p \in X : f\left( p\right) \in {D}_{n}}\right\} \) of finite positive measure, \( n = 1,2,\ldots \) Considered as elements of \( {L}^{2}\left( {X,\mu }\right) \), the characteristic functions \( {\chi }_{{E}_{1}},{\chi }_{{E}_{2}},\ldots \) are mutually orthogonal because the sets \( {E}_{1},{E}_{2},\ldots \) are mutually disjoint. Moreover,\n\n\[ \left| {f - \lambda }\right| \cdot {\chi }_{{E}_{n}} \leq \left( {\left| {f - {\lambda }_{n}}\right| + \left| {{\lambda }_{n} - \lambda }\right| }\right) {\chi }_{{E}_{n}} \leq \left( {{\epsilon }_{n} + \left| {{\lambda }_{n} - \lambda }\right| }\right) {\chi }_{{E}_{n}}. \]\n\nIt follows that\n\n\[ \begin{Vmatrix}{\left( {N - \lambda }\right) {\chi }_{{E}_{n}}}\end{Vmatrix} \leq \left( {{\epsilon }_{n} + \left| {{\lambda }_{n} - \lambda }\right| }\right) {\begin{Vmatrix}{\chi }_{{E}_{n}}\end{Vmatrix}}_{2} = \left( {{\epsilon }_{n} + \left| {{\lambda }_{n} - \lambda }\right| }\right) \mu {\left( {E}_{n}\right) }^{1/2}, \]\n\nand the orthonormal sequence can be taken as \( {\xi }_{n} = \mu {\left( {E}_{n}\right) }^{-1/2}{\chi }_{{E}_{n}}, n = \) \( 1,2,\ldots \) .
Yes
Proposition 2.8.3. The trace has the following properties:\n\n(1) \( \operatorname{trace}{A}^{ * }A = \operatorname{trace}A{A}^{ * } \), for any \( A \in \mathcal{B}\left( H\right) \) .
Proof. For (1), consider the double sequence of nonnegative terms \( {\left| \left\langle A{e}_{p},{e}_{q}\right\rangle \right| }^{2} = {\left| \left\langle {e}_{p},{A}^{ * }{e}_{q}\right\rangle \right| }^{2}, p, q = 1,2,\ldots \) Summing first on \( q \) and then on \( p \), we obtain\n\n\[ \mathop{\sum }\limits_{{p = 1}}^{\infty }\mathop{\sum }\limits_{{q = 1}}^{\infty }{\left| \left\langle A{e}_{p},{e}_{q}\right\rangle \right| }^{2} = \mathop{\sum }\limits_{{p = 1}}^{\infty }{\begin{Vmatrix}A{e}_{p}\end{Vmatrix}}^{2} \]\n\nwhile summing in the opposite order gives\n\n\[ \mathop{\sum }\limits_{{q = 1}}^{\infty }\mathop{\sum }\limits_{{p = 1}}^{\infty }{\left| \left\langle {e}_{p},{A}^{ * }{e}_{q}\right\rangle \right| }^{2} = \mathop{\sum }\limits_{{q = 1}}^{\infty }{\begin{Vmatrix}{A}^{ * }{e}_{q}\end{Vmatrix}}^{2}. \]\n\nSince the sum of a nonnegative double sequence is independent of the order of summation, this proves (1).
Yes
Proposition 2.8.4. Every Hilbert-Schmidt operator \( A \) is compact, and satisfies \( \parallel A{\parallel }^{2} \leq \operatorname{trace}{A}^{ * }A \) .
Proof. We first prove the inequality \( \parallel A{\parallel }^{2} \leq \operatorname{trace}{A}^{ * }A \) . Indeed, for every unit vector \( e \) we can find an orthonormal basis \( {e}_{1},{e}_{2},\ldots \) starting with \( {e}_{1} = e \) . Hence \( \parallel {Ae}{\parallel }^{2} \leq \mathop{\sum }\limits_{n}{\begin{Vmatrix}A{e}_{n}\end{Vmatrix}}^{2} = \operatorname{trace}{A}^{ * }A \), and since \( e \) is arbitrary we obtain\n\n\[ \parallel A{\parallel }^{2} = \mathop{\sup }\limits_{{\parallel e\parallel = 1}}\parallel {Ae}{\parallel }^{2} \leq \operatorname{trace}{A}^{ * }A. \]\n\nTo see that every Hilbert-Schmidt operator \( A \) is compact, fix an orthonormal basis \( {e}_{1},{e}_{2},\ldots \) and for every \( n \geq 1 \) let \( {Q}_{n} \) be the projection onto the subspace spanned by \( {e}_{n + 1},{e}_{n + 2},\ldots \) Obviously, \( {F}_{n} = A\left( {1 - {Q}_{n}}\right) \) is a finite-rank operator, and by the preceding paragraph we have\n\n\[ {\begin{Vmatrix}A - {F}_{n}\end{Vmatrix}}^{2} = {\begin{Vmatrix}A{Q}_{n}\end{Vmatrix}}^{2} \leq \operatorname{trace}\left( {{Q}_{n}{A}^{ * }A{Q}_{n}}\right) = \mathop{\sum }\limits_{{k = n + 1}}^{\infty }{\begin{Vmatrix}A{e}_{k}\end{Vmatrix}}^{2}. \]\n\nThe right side tends to 0 as \( n \rightarrow \infty \) because \( \mathop{\sum }\limits_{k}{\begin{Vmatrix}A{e}_{k}\end{Vmatrix}}^{2} < \infty \) . Hence \( A = \mathop{\lim }\limits_{n}{F}_{n} \) is the norm limit of a sequence of finite-rank operators, and is therefore compact.
Yes
Proposition 2.8.6. Let \( \\left( {X,\\mu }\\right) \) be a separable \( \\sigma \) -finite measure space. For every function \( k \\in {L}^{2}\\left( {X \\times X,\\mu \\times \\mu }\\right) \) there is a unique bounded operator \( {A}_{k} \) on \( {L}^{2}\\left( {X,\\mu }\\right) \) satisfying
\[ {A}_{k}\\xi \\left( x\\right) = {\\int }_{X}k\\left( {x, y}\\right) \\xi \\left( y\\right) {d\\mu }\\left( y\\right) ,\\;\\xi \\in {L}^{2}\\left( {X,\\mu }\\right) . \] The map \( k \\mapsto {A}_{k} \) is an isometric isomorphism of the Hilbert space \( {L}^{2}(X \\times \) \( X,\\mu \\times \\mu \) ) onto the Hilbert space \( {\\mathcal{L}}^{2} \) of all Hilbert-Schmidt operators on \( {L}^{2}\\left( {X,\\mu }\\right) \) .
Yes
The involution in \( {A}^{e} \) satisfies \( \begin{Vmatrix}{{X}^{ * }X}\end{Vmatrix} = \parallel X{\parallel }^{2} \) for every \( X \in {A}^{e} \), and \( {A}^{e} \) is closed in the operator norm of \( \mathcal{B}\left( A\right) \) ; hence it is a unital \( {C}^{ * } \) -algebra. Moreover, the regular representation is an isometric *-isomorphism of \( A \) onto a maximal ideal of codimension one in \( {A}^{e} \).
Notice first that \( \begin{Vmatrix}{L}_{a}\end{Vmatrix} = \parallel a\parallel \) for every \( a \in A \) . Indeed, \( \leq \) is true for any Banach algebra, and the opposite inequality follows for an element \( a \) of norm 1 because\n\n\[ \begin{Vmatrix}{L}_{a}\end{Vmatrix} \geq \begin{Vmatrix}{{L}_{a}\left( {a}^{ * }\right) }\end{Vmatrix} = \begin{Vmatrix}{a{a}^{ * }}\end{Vmatrix} = {\begin{Vmatrix}{a}^{ * }\end{Vmatrix}}^{2} = \parallel a{\parallel }^{2} = 1. \]\n\nThe set \( \left\{ {{L}_{a} : a \in A}\right\} \) is obviously an ideal in \( {A}^{e} \) of codimension at most one. If the codimension were zero, then the identity operator would have the form \( {L}_{f} \) for some element \( f \in A \) ; that would imply \( f \) was a unit for \( A \) , contrary to hypothesis. Hence \( \left\{ {{L}_{a} : a \in A}\right\} \) has codimension one. Since \( L \) is an isometry, this ideal must be closed in the operator norm of \( \mathcal{B}\left( A\right) \) ; and since \( {A}^{e} \) is obtained from this ideal by adjoining the one-dimensional space spanned by 1, it follows that \( {A}^{e} \) must also be norm closed.\n\nIt remains to show that the involution in \( {A}^{e} \) satisfies \( \begin{Vmatrix}{{X}^{ * }X}\end{Vmatrix} = \parallel X{\parallel }^{2} \) . By Remark 2.9.1, it is enough to verify the inequality \( \parallel X{\parallel }^{2} \leq \begin{Vmatrix}{{X}^{ * }X}\end{Vmatrix} \) for \( X = {L}_{a} + ╏ \) in \( {A}^{e} \) . For such an \( X \), we have\n\n\[ \parallel X{\parallel }^{2} = \mathop{\sup }\limits_{{\parallel b\parallel \leq 1}}{\begin{Vmatrix}\left( {L}_{a} + \lambda \mathbf{1}\right) \left( b\right) \end{Vmatrix}}^{2} = \mathop{\sup }\limits_{{\parallel b\parallel \leq 1}}\parallel {ab} + {\lambda b}{\parallel }^{2} \]\n\n\[ = \mathop{\sup }\limits_{{\parallel b\parallel \leq 1}}\begin{Vmatrix}{{\left( ab + \lambda b\right) }^{ * }\left( {{ab} + {\lambda b}}\right) }\end{Vmatrix} = \mathop{\sup }\limits_{{\parallel b\parallel \leq 1}}\begin{Vmatrix}{{b}^{ * }\left( {{X}^{ * }X\left( b\right) }\right) }\end{Vmatrix} \]\n\n\[ \leq \mathop{\sup }\limits_{{\parallel b\parallel \leq 1}}\begin{Vmatrix}{{X}^{ * }X\left( b\right) }\end{Vmatrix} \leq \begin{Vmatrix}{{X}^{ * }X}\end{Vmatrix}. \]
Yes
Proposition 2.9.3. Every \( * \) -homomorphism \( \pi : A \rightarrow B \) of \( {C}^{ * } \) -algebras has norm at most \( 1 \) . If \( \pi \) has trivial kernel, then it is an isometry.
Proof. Suppose first that \( A \) has a unit \( {\mathbf{1}}_{A} \) . By passing from \( B \) to the closure of the \( * \) -subalgebra \( \pi \left( A\right) \) if necessary, we may assume that \( \pi \left( A\right) \) is dense in \( B \) . In this case, \( \pi \left( {\mathbf{1}}_{A}\right) \) is the unit \( {\mathbf{1}}_{B} \) of \( B \) . Thus we may argue as we did for nondegenerate representations. For example, since \( \pi \) must map invertible elements of \( A \) to invertible elements of \( B \), it follows that \( \sigma \left( {\pi \left( x\right) }\right) \subseteq \sigma \left( x\right) \) for every element \( x \in A \) . Corollary 2 of Theorem 2.2.4 implies that for self-adjoint elements \( x \in A \) we have\n\n\[ \parallel \pi \left( x\right) \parallel = r\left( {\pi \left( x\right) }\right) \leq r\left( x\right) = \parallel x\parallel \]\n\nso that for general elements \( z \in A \) we have\n\n\[ \parallel \pi \left( z\right) {\parallel }^{2} = \begin{Vmatrix}{\pi {\left( z\right) }^{ * }\pi \left( x\right) }\end{Vmatrix} = \begin{Vmatrix}{\pi \left( {{z}^{ * }z}\right) }\end{Vmatrix} \leq \begin{Vmatrix}{{z}^{ * }z}\end{Vmatrix} = \parallel z{\parallel }^{2}. \]\n\nIf, in addition, \( \pi \) has trivial kernel, then we claim that \( \parallel \pi \left( x\right) \parallel = \parallel x\parallel \) for every \( x \in A \) . As above, this reduces to the case where \( x = {x}^{ * } \) is selfadjoint; and by Corollary 2 of Theorem 2.2.4 it is enough to show that \( x \) and \( \pi \left( x\right) \) have the same spectrum when \( x = {x}^{ * } \) . We have already seen that \( \sigma \left( {\pi \left( x\right) }\right) \subseteq \sigma \left( x\right) \) . For the opposite inclusion, suppose that \( \lambda \) is a point of \( \sigma \left( x\right) \) that does not belong to \( \sigma \left( {\pi \left( x\right) }\right) \) . There is a continuous function \( f : \sigma \left( x\right) \rightarrow \mathbb{R} \) such that \( f \) vanishes on \( \sigma \left( {\pi \left( x\right) }\right) \) and \( f\left( \lambda \right) \neq 0 \) . Since \( f = 0 \) on \( \sigma \left( {\pi \left( x\right) }\right) \), we must have \( f\left( {\pi \left( x\right) }\right) = 0 \) . Notice that \( f\left( {\pi \left( x\right) }\right) = \pi \left( {f\left( x\right) }\right) \) (this is obvious if \( f \) is a polynomial, and it follows for general continuous \( f \) by an application of the Weierstrass approximation theorem and the previously established fact that \( \pi \) is a bounded linear map of \( A \) to \( B \) ). But \( \pi \left( {f\left( x\right) }\right) = 0 \) implies that \( f\left( x\right) = 0 \) because \( \pi \) has trivial kernel; in turn, \( f\left( x\right) = 0 \) implies that \( f = 0 \) on \( \sigma \left( x\right) \), contradicting the fact that \( f\left( \lambda \right) \neq 0 \) .
Yes
Corollary 1. Let \( A \) be a complex algebra with involution. If there is a norm on \( A \) that makes it into a \( {C}^{ * } \) -algebra, then that norm is unique.
Proof. Let \( \parallel \cdot {\parallel }_{1} \) and \( \parallel \cdot {\parallel }_{2} \) be two (complete) Banach algebra norms on \( A \) satisfying \( {\begin{Vmatrix}{x}^{ * }x\end{Vmatrix}}_{k} = \parallel x{\parallel }_{k}^{2} \) for \( x \in A \), and let \( {A}_{k} \) be the algebra \( A \) considered as a \( {C}^{ * } \) -algebra in each norm respectively, \( k = 1,2 \) . The identity map of \( A \) can be regarded as a \( * \) -isomorphism of \( {A}_{1} \) onto \( {A}_{2} \) . By Proposition 2.9.3 this map must be an isometry; hence \( \parallel x{\parallel }_{1} = \parallel x{\parallel }_{2} \) for all \( x \in A \) .
Yes
Corollary 1. A bounded operator \( T \) belongs to \( \mathcal{F}\left( E\right) \) iff there is an operator \( S \in \mathcal{B}\left( E\right) \) such that \( \mathbf{1} - {ST} \) and \( \mathbf{1} - {TS} \) are both compact.
Proof of Corollary 1. If \( \dot{T} \) is invertible in \( \mathcal{B}\left( E\right) /\mathcal{K}\left( E\right) \), then its inverse is an element of the form \( \dot{S} \) for some \( S \in \mathcal{B}\left( E\right) \), and the operators \( 1 - {ST} \) and \( 1 - {TS} \) must be compact because they map to 0 in the quotient algebra. The converse follows immediately from Atkinson's theorem.
No
Corollary 2. The set \( \mathcal{F}\left( E\right) \) of Fredholm operators is open in the norm topology of \( \mathcal{B}\left( E\right) \), it is stable under compact perturbations, it contains all invertible operators of \( \mathcal{B}\left( E\right) \), and it is closed under operator multiplication.
Proof of Corollary 2. Atkinson’s theorem implies that \( \mathcal{F}\left( E\right) \) is the inverse image of the general linear group of \( \mathcal{B}\left( E\right) /\mathcal{K}\left( E\right) \) under the continuous homomorphism \( T \mapsto T \) ; hence these assertions all follow from the fact that the set of invertible elements of a unital Banach algebra \( A \) forms a group that is open in the norm topology of \( A \) .
Yes
Corollary 2 (Stability of index). For every Fredholm operator \( A \in \) \( \mathcal{B}\left( E\right) \) and compact operator \( K \) ,\n\n\[ \text{ind}\left( {A + K}\right) = \text{ind}A\text{.} \]
Proof. By Atkinson’s theorem there is a Fredholm operator \( B \in \mathcal{B}\left( E\right) \) such that \( {AB} = 1 + L \) with \( L \in \mathcal{K}\left( E\right) \) . We have \( \left( {A + K}\right) B = 1 + {L}^{\prime } \) where \( {L}^{\prime } = L - {KB} \in \mathcal{K}\left( E\right) \) . As we have already pointed out, the Fredholm alternative implies that \( \operatorname{ind}\left( {\mathbf{1} + L}\right) = \operatorname{ind}\left( {\mathbf{1} + {L}^{\prime }}\right) = 0 \) ; hence ind \( {AB} = \) ind \( \left( {A + K}\right) B = 0 \) . Using Corollary 1 one has\n\n\[ \operatorname{ind}\left( {A + K}\right) + \operatorname{ind}B = \operatorname{ind}\left( {A + K}\right) B = \operatorname{ind}{AB} = \operatorname{ind}A + \operatorname{ind}B \]\n\nand the formula follows after one cancels the integer ind \( B \) .
Yes
Corollary 3 (Continuity of index). Given a Fredholm operator \( A \) , let \( {A}_{1},{A}_{2},\ldots \) be a sequence of bounded operators that converges to \( A \) , \( \mathop{\lim }\limits_{{n \rightarrow \infty }}\begin{Vmatrix}{{A}_{n} - A}\end{Vmatrix} = 0 \) . There is an \( {n}_{0} \) such that for \( n \geq {n}_{0},{A}_{n} \) is a Fredholm operator with ind \( {A}_{n} = \operatorname{ind}A \) .
Proof. By Atkinson’s theorem, \( \mathcal{F}\left( E\right) \) is open, so that \( {A}_{n} \in \mathcal{F}\left( E\right) \) for sufficiently large \( n \) . We can also find a Fredholm operator \( B \) such that \( {AB} = \mathbf{1} + K \) with \( K \) compact. Writing \( {A}_{n} = A + {T}_{n} \) with \( \begin{Vmatrix}{T}_{n}\end{Vmatrix} \rightarrow 0 \) as \( n \rightarrow \infty \), we can find \( {n}_{0} \) so that, for \( n \geq {n}_{0},\begin{Vmatrix}{{T}_{n}B}\end{Vmatrix} < 1 \) and hence \( \mathbf{1} + {T}_{n}B \) is invertible. For such \( n \), we have\n\n\[ \operatorname{ind}{A}_{n} + \operatorname{ind}B = \operatorname{ind}\left( {A + {T}_{n}}\right) B = \operatorname{ind}\left( {\mathbf{1} + {T}_{n}B + K}\right) .\n\]\n\nThe right side vanishes because \( 1 + {T}_{n}B + K \) is a compact perturbation of an invertible operator (see Exercise (1) below). On the other hand,\n\n\[ \operatorname{ind}A + \operatorname{ind}B = \operatorname{ind}{AB} = \operatorname{ind}\left( {\mathbf{1} + K}\right) = 0 \]\n\nby the Fredholm alternative; hence ind \( {A}_{n} = - \operatorname{ind}B = \operatorname{ind}A \) for sufficiently large \( n \) .
Yes
Corollary 2. Let \( \left\{ {{e}_{n} : n \in \mathbb{Z}}\right\} \) be a bilateral orthonormal basis for a Hilbert space \( H \), and let \( U \) be the bilateral shift defined on \( H \) by \( U{e}_{n} = {e}_{n + 1} \) , \( n \in \mathbb{Z} \) . Then the von Neumann algebra \( {W}^{ * }\left( U\right) \) generated by \( U \) is maximal abelian, and consists of all operators in \( \mathcal{B}\left( H\right) \) that commute with \( U \) .
Proof. We have seen that \( U \) is unitarily equivalent to the multiplication operator \( {M}_{\zeta } \) acting on \( {L}^{2}\left( \mathbb{T}\right) \) by \( {M}_{\zeta }\xi \left( z\right) = \zeta \left( z\right) \xi \left( z\right) ,\zeta \) being the current variable \( \zeta \left( z\right) = z, z \in \mathbb{T} \) . Since \( {M}_{\zeta } \) is unitary, any operator commuting with it must also commute with its adjoint \( {M}_{\zeta }^{ * } = {M}_{\zeta }^{-1} \) . On the other hand, since \( \zeta \) separates points of \( \mathbb{T} \), it follows from Theorem 4.1.3 that any operator commuting with \( \left\{ {{M}_{\zeta },{M}_{\zeta }^{ * }}\right\} \) must belong to the multiplication algebra of \( {L}^{2}\left( \mathbb{T}\right) \), and that the multiplication algebra coincides with the von Neumann algebra generated by \( U \) .
Yes
Proposition 4.2.1. \( {H}^{\infty } = \left\{ {\phi \in {L}^{\infty } : {M}_{\phi }{H}^{2} \subseteq {H}^{2}}\right\} \) .
Proof. Let \( \phi \in {L}^{\infty } \) . If \( \phi \in {H}^{\infty } \), then for \( n \geq 0 \) ,\n\n\[ \n{M}_{\phi }{\zeta }^{n} = {\zeta }^{n} \cdot \phi \in {\zeta }^{n} \cdot {H}^{2} \subseteq {H}^{2} \n\] \n\nhence \( {M}_{\phi } \) leaves \( {H}^{2} = \left\lbrack {1,\zeta ,{\zeta }^{2},\ldots }\right\rbrack \) invariant.\n\nConversely, if \( {M}_{\phi }{H}^{2} \subseteq {H}^{2} \), then \( \phi = {M}_{\phi }1 \in {H}^{2} \) .
Yes
Proposition 4.2.3. Let \( A \) be a bounded operator on \( {H}^{2} \) . The matrix of A relative to the natural basis \( \left\{ {{\zeta }^{n} : n = 0,1,2,\ldots }\right\} \) is a Toeplitz matrix iff \( {S}^{ * }{AS} = A \) .
Proof. The hypothesis on the matrix entries \( {a}_{ij} = \left\langle {A{\zeta }^{j},{\zeta }^{i}}\right\rangle \) of \( A \) is equivalent to requiring\n\n\[ \n{a}_{i + 1, j + 1} = {a}_{ij},\;i, j = 0,1,2,\ldots \n\] \n\nNoting that \( S{\zeta }^{n} = {\zeta }^{n + 1} \) for \( n \geq 0 \) we find that this is equivalent to the requirement that \n\n\[ \n\langle {S}^{ * }{AS}{\zeta }^{j},{\zeta }^{i}\rangle = \langle {AS}{\zeta }^{j}, S{\zeta }^{i}\rangle = \langle A{\zeta }^{j + 1},{\zeta }^{i + 1}\rangle = \langle A{\zeta }^{j},{\zeta }^{i}\rangle \n\] \n\nfor all \( i, j \geq 0 \) ; hence it is equivalent to the formula \( {S}^{ * }{AS} = A \) .
Yes
Every Toeplitz operator \( {T}_{\phi },\phi \in {L}^{\infty } \), satisfies\n\n\[ \inf \left\{ {\begin{Vmatrix}{{T}_{\phi } + K}\end{Vmatrix} : K \in \mathcal{K}}\right\} = \begin{Vmatrix}{T}_{\phi }\end{Vmatrix} = \parallel \phi {\parallel }_{\infty }.\n\]\n\nIn particular, the only compact Toeplitz operator is 0.
Proof. Let \( S \) be the unilateral shift acting on \( {H}^{2} \) by \( S{\zeta }^{n} = {\zeta }^{n + 1}, n \geq 0 \) . It suffices to show that for any operator \( A \in \mathcal{B}\left( {H}^{2}\right) \) satisfying \( {S}^{ * }{AS} = A \) and for any compact operator \( K \) we have\n\n\[ \parallel A + K\parallel \geq \parallel A\parallel \text{.}\]\n\nThe hypothesis \( {S}^{ * }{AS} = A \) implies that \( {S}^{*n}A{S}^{n} = A \) for every \( n = 1,2,\ldots \) ; noting that \( {P}_{n} = {S}^{n}{S}^{*n} \) is the projection onto \( \left\lbrack {{\zeta }^{n},{\zeta }^{n + 1},\ldots }\right\rbrack \) we have\n\n\[ \parallel A + K\parallel \geq \begin{Vmatrix}{{P}_{n}\left( {A + K}\right) {P}_{n}}\end{Vmatrix} = \begin{Vmatrix}{{S}^{*n}\left( {A + K}\right) {S}^{n}}\end{Vmatrix} = \begin{Vmatrix}{A + {S}^{*n}K{S}^{n}}\end{Vmatrix}.\n\nThe norm of the compression of \( K \) to the subspace \( \left\lbrack {{\zeta }^{n},{\zeta }^{n + 1},\ldots }\right\rbrack \) is given by \( \begin{Vmatrix}{{P}_{n}K{P}_{n}}\end{Vmatrix} = \begin{Vmatrix}{{S}^{*n}K{S}^{n}}\end{Vmatrix} \), which tends to 0 as \( n \rightarrow \infty \) because \( K \) is compact and \( {P}_{n} \downarrow 0 \) . Thus\n\n\[ \parallel A + K\parallel \geq \mathop{\lim }\limits_{{n \rightarrow \infty }}\begin{Vmatrix}{A + {S}^{*n}K{S}^{n}}\end{Vmatrix} = \parallel A\parallel \]\n\nas asserted.
Yes
Proposition 4.3.1. Let \( f, g \in {L}^{\infty } \) . If one of the functions \( f, g \) is continuous, then \( {T}_{fg} - {T}_{f}{T}_{g} \in \mathcal{K} \) .
Proof. Since \( {T}_{fg}^{ * } = {T}_{\bar{f}\bar{g}} \) and \( {\left( {T}_{f}{T}_{g}\right) }^{ * } = {T}_{\bar{g}}{T}_{\bar{f}} \), it suffices to prove the following assertion: If \( f \in C\left( \mathbb{T}\right) \) and \( g \in {L}^{\infty } \), then \( {T}_{fg} - {T}_{f}{T}_{g} \in \mathcal{K} \) . Moreover, since \( C\left( \mathbb{T}\right) \) is the norm-closed linear span of the monomials \( {\zeta }^{n}, n \in \mathbb{Z} \) , and \( \mathcal{K} \) is a norm-closed linear space, we may reduce to the case \( f = {\zeta }^{n} \) and \( g \in {L}^{\infty }, n \in \mathbb{Z} \) .\n\nIf \( n \geq 0 \), then \( {\zeta }^{n} \in {H}^{\infty } \), so that by (4.3.1) we have \( {T}_{f{\zeta }^{n}} = {T}_{f}{T}_{{\zeta }^{n}} \) . Thus \( {T}_{fg} - {T}_{f}{T}_{g} = 0 \) in this case.\n\nIf \( n < 0 \), say \( n = - m \) with \( m \geq 1 \), then \( {\zeta }^{n} \) is the complex conjugate of the \( {H}^{\infty } \) function \( {\zeta }^{m} \), and another application of (4.3.1) gives \( {T}_{f{\zeta }^{n}} = \) \( {T}_{{\zeta }^{n}}{T}_{f} = {S}^{*m}{T}_{f} \) . Noting that \( {S}^{*m}{T}_{f}{S}^{m} = {T}_{f} \) (by iterating the basic formula \( {S}^{ * }{T}_{f}S = {T}_{f} \) valid for any Toeplitz operator) we can write\n\n\[ \n{T}_{f}{T}_{{\zeta }^{n}} = {T}_{f}{S}^{*m} = {S}^{*m}{T}_{f}{S}^{m}{S}^{*m} = {S}^{*m}{T}_{f} - {S}^{*m}{T}_{f}\left( {\mathbf{1} - {S}^{m}{S}^{*m}}\right) .\n\]\n\nHence\n\n\[ \n{T}_{f{\zeta }^{n}} - {T}_{f}{T}_{{\zeta }^{n}} = {S}^{*m}{T}_{f} - {T}_{f}{S}^{*m} = - {S}^{*m}{T}_{f}\left( {\mathbf{1} - {S}^{m}{S}^{*m}}\right) ,\n\]\n\nwhich is a finite-rank operator, since \( 1 - {S}^{m}{S}^{*m} \) is the projection onto \( \left\lbrack {1,\zeta ,{\zeta }^{2},\ldots ,{\zeta }^{m - 1}}\right\rbrack . \)
Yes
Corollary 1. The Fredholm operators in \( \mathcal{T} \) are precisely the operators of the form \( {T}_{f} + K \) where \( f \) is an invertible symbol in \( C{\left( \mathbb{T}\right) }^{-1} \) and \( K \in \mathcal{K} \) .
Consider a Fredholm operator in \( \mathcal{T} \), say \( {T}_{f} + K \) where \( f \in C\left( \mathbb{T}\right) \) has no zeros on the circle and \( K \) is a compact operator. By the stability results of Chapter 3 we see that \( {T}_{f} \) is also a Fredholm operator and\n\n\[ \text{ind}\left( {{T}_{f} + K}\right) = \operatorname{ind}\left( {T}_{f}\right) \text{.} \]\n\nWe know that for \( f = \zeta ,{T}_{f} \) is the shift; hence ind \( \left( {T}_{f}\right) = - 1 \) . However, we still lack tools for computing the index of more general Toeplitz operators with symbols in \( C{\left( \mathbb{T}\right) }^{-1} \) . This issue will be taken up in the following section.
No
Proposition 4.4.1. For every function \( F \in C\left\lbrack {0,1}\right\rbrack \) such that \( F\left( t\right) \neq 0 \) for every \( t \in \left\lbrack {0,1}\right\rbrack \), there is a function \( G \in C\left\lbrack {0,1}\right\rbrack \) such that\n\n\[ F\left( t\right) = {e}^{G\left( t\right) },\;0 \leq t \leq 1. \]
Proof. On the domain \( \{ z \in \mathbb{C} : \left| {z - 1}\right| < 1\} \), let \( \log z \) be the principal branch of the logarithm,\n\n\[ \log z = - \mathop{\sum }\limits_{{n = 1}}^{\infty }\frac{{\left( 1 - z\right) }^{n}}{n} \]\n\nThe log function is holomorphic, satisfies \( \log 1 = 0 \), and of course \( {e}^{\log z} = z \) for \( \left| {z - 1}\right| < 1 \) . Let\n\n\[ M = \mathop{\sup }\limits_{{0 \leq t \leq 1}}{\left| F\left( t\right) \right| }^{-1} \]\n\nBy uniform continuity of \( F \), we can find a finite partition of the interval \( \left\lbrack {0,1}\right\rbrack ,0 = {t}_{0} < {t}_{1} < \cdots < {t}_{n} = 1 \), such that\n\n\[ \mathop{\sup }\limits_{{{t}_{k - 1} \leq t \leq {t}_{k}}}\left| {F\left( t\right) - F\left( {t}_{k - 1}\right) }\right| < \frac{1}{2M}. \]\n\nIt follows that for \( k = 1,\ldots, n \) and \( t \in \left\lbrack {{t}_{k - 1},{t}_{k}}\right\rbrack \) ,\n\n(4.10)\n\n\[ \left| {1 - \frac{F\left( t\right) }{F\left( {t}_{k - 1}\right) }}\right| = \frac{\left| F\left( t\right) - F\left( {t}_{k - 1}\right) \right| }{\left| F\left( {t}_{k - 1}\right) \right| } \leq \frac{1}{{2M}\left| {F\left( {t}_{k - 1}\right) }\right| } \leq \frac{1}{2} < 1. \]\n\nSetting\n\n\[ {G}_{k}\left( t\right) = \log \left( {F\left( t\right) /F\left( {t}_{k - 1}\right) }\right) ,\;{t}_{k - 1} \leq t \leq {t}_{k}, \]\n\nwe find that \( {G}_{k} \) is continuous, \( {G}_{k}\left( {t}_{k - 1}\right) = 0,{G}_{k}\left( {t}_{k}\right) = \log \left( {F\left( {t}_{k}\right) /F\left( {t}_{k - 1}\right) }\right) \) , and it satisfies\n\n\[ F\left( t\right) = F\left( {t}_{k - 1}\right) {e}^{{G}_{k}\left( t\right) } \]\n\nthroughout the interval \( \left\lbrack {{t}_{k - 1},{t}_{k}}\right\rbrack \) . There is an obvious way to piece the \( {G}_{k} \) together so as to obtain a continuous function \( G : \left\lbrack {0,1}\right\rbrack \rightarrow \mathbb{C} \), namely \( G\left( t\right) = {G}_{1}\left( t\right) \) for \( t \in \left\lbrack {0,{t}_{1}}\right\rbrack \) and, for \( k = 2,\ldots, n \),\n\n\[ G\left( t\right) = {G}_{1}\left( {t}_{1}\right) + \cdots + {G}_{k - 1}\left( {t}_{k - 1}\right) + {G}_{k}\left( t\right) ,\;t \in \left\lbrack {{t}_{k - 1},{t}_{k}}\right\rbrack \]\n\nIt follows that\n\n\[ F\left( t\right) = F\left( 0\right) {e}^{G\left( t\right) },\;0 \leq t \leq 1. \]\n\nWriting \( F\left( 0\right) \in {\mathbb{C}}^{ \times } \) as an exponential \( F\left( 0\right) = {e}^{{z}_{0}} \), we obtain a continuous function \( \widetilde{G} \) satisfying \( F = {e}^{\widetilde{G}} \) by way of \( \widetilde{G}\left( t\right) = G\left( t\right) + {z}_{0} \) .
Yes
Proposition 4.4.2. For \( f, g \in G = C{\left( \mathbb{T}\right) }^{-1} \) ,\n\n(1) \( \# \left( {fg}\right) = \# \left( f\right) + \# \left( g\right) \) .\n\n(2) \( \# \left( f\right) = n \in \mathbb{Z} \) iff there is a function \( h \in C\left( \mathbb{T}\right) \) such that \( f = {\zeta }^{n}{e}^{h} \) .
Proof. For (1), pick continuous functions \( F, G : \left\lbrack {0,1}\right\rbrack \rightarrow \mathbb{C} \) such that\n\n\[ f\left( {e}^{2\pi it}\right) = {e}^{{2\pi iF}\left( t\right) },\;g\left( {e}^{2\pi it}\right) = {e}^{{2\pi iG}\left( t\right) },\;t \in \left\lbrack {0,1}\right\rbrack .\n\]\n\nThen\n\n\[ f\left( {e}^{2\pi it}\right) g\left( {e}^{2\pi it}\right) = {e}^{{2\pi i}\left( {F\left( t\right) + G\left( t\right) }\right) },\;t \in \left\lbrack {0,1}\right\rbrack ,\n\]\n\nand the winding number of \( {fg} \) is given by\n\n\[ F\left( 1\right) + G\left( 1\right) - \left( {F\left( 0\right) + G\left( 0\right) }\right) = \# \left( f\right) + \# \left( g\right) .\n\]\n\nFor (2), consider first the case \( n = 0 \) . If \( f = {e}^{h} \) is the exponential of a function \( h \in C\left( \mathbb{T}\right) \), then we have\n\n(4.13)\n\n\[ f\left( {e}^{2\pi it}\right) = {e}^{{2\pi iF}\left( t\right) },\;0 \leq t \leq 1\n\]\n\nwhere \( F\left( t\right) = {\left( 2\pi \right) }^{-1}h\left( {e}^{2\pi it}\right) \) . Clearly, \( F\left( 1\right) = F\left( 0\right) \), so that \( \# \left( f\right) = F\left( 1\right) - \) \( F\left( 0\right) = 0 \) . Conversely, if \( \# \left( f\right) = 0 \), then there is a function \( F \in C\left\lbrack {0,1}\right\rbrack \) such that (4.13) is satisfied and \( F\left( 1\right) - F\left( 0\right) = \# \left( f\right) = 0 \) . Since \( F \) is periodic, we have \( f = {e}^{h} \), where \( h \in C\left( \mathbb{T}\right) \) is the function \( h\left( {e}^{2\pi it}\right) = {2\pi iF}\left( t\right) ,0 \leq t \leq 1 \) .\n\nTo deal with the case of arbitrary \( n \in \mathbb{Z} \) note first that \( \# \left( \zeta \right) = 1 \) . Indeed, this is immediate from the fact that\n\n\[ \zeta \left( {e}^{2\pi it}\right) = {e}^{2\pi it},\;0 \leq t \leq 1.\n\]\n\nFrom the property (1) it follows that \( \# \left( {\zeta }^{n}\right) = n \) for every \( n \in \mathbb{Z} \) ; hence \( \# \left( {{\zeta }^{n}{e}^{h}}\right) = \# \left( {\zeta }^{n}\right) + \# \left( {e}^{h}\right) = n \), as asserted. Conversely, if \( f \in C\left( \mathbb{T}\right) \) satisfies \( \# \left( f\right) = n \), consider \( g = {\zeta }^{-n}f \in C{\left( \mathbb{T}\right) }^{-1} \) . Using (1) again we have \( \# \left( g\right) = 0 \) , and by the preceding paragraph there is an \( h \in C\left( \mathbb{T}\right) \) such that \( g = {e}^{h} \) . Thus \( f = {\zeta }^{n}g = {\zeta }^{n}{e}^{h} \) has the asserted form.
Yes
Proposition 4.7.1. Every positive linear functional \( \rho \) on \( A \) satisfies the Schwarz inequality\n\n(4.19)\n\n\[{\left| \rho \left( {y}^{ * }x\right) \right| }^{2} \leq \rho \left( {{x}^{ * }x}\right) \rho \left( {{y}^{ * }y}\right)\]\n\nand moreover, \( \parallel \rho \parallel = \rho \left( \mathbf{1}\right) \) . In particular, every state of \( A \) has norm 1 .
Proof. Considering \( A \) as a complex vector space,\n\n\[x, y \in A \mapsto \left\lbrack {x, y}\right\rbrack = \rho \left( {{y}^{ * }x}\right)\]\n\ndefines a sesquilinear form which is positive semidefinite in the sense that \( \left\lbrack {x, x}\right\rbrack \geq 0 \) for every \( x \) . The argument that establishes the Schwarz inequality for complex inner product spaces applies verbatim in this context, and we deduce (4.19) from \( {\left| \left\lbrack x, y\right\rbrack \right| }^{2} \leq \left\lbrack {x, x}\right\rbrack \left\lbrack {y, y}\right\rbrack \) .\n\nClearly, \( \rho \left( \mathbf{1}\right) = \rho \left( {{\mathbf{1}}^{ * }\mathbf{1}}\right) \geq 0 \), and we claim that \( \parallel \rho \parallel \leq \rho \left( \mathbf{1}\right) \) . Indeed, for every \( x \in A \) the Schwarz inequality (4.19) implies\n\n\[{\left| \rho \left( x\right) \right| }^{2} = \left| {\rho \left( {{\mathbf{1}}^{ * }x}\right) }\right| \leq \rho \left( {{x}^{ * }x}\right) \rho \left( \mathbf{1}\right) .\n\nIf, in addition, \( \parallel x\parallel \leq 1 \), then \( {x}^{ * }x \) is a self-adjoint element in \( A \) of norm at most 1; consequently, \( 1 - {x}^{ * }x \) must have a self-adjoint square root \( y \in A \) (see Exercise (2b) below). It follows that \( \rho \left( {\mathbf{1} - {x}^{ * }x}\right) = \rho \left( {y}^{2}\right) \geq 0 \), i.e., \( 0 \leq \rho \left( {{x}^{ * }x}\right) \leq \rho \left( \mathbf{1}\right) \) . Substitution into the previous inequality gives \( {\left| \rho \left( x\right) \right| }^{2} \leq \) \( \rho \left( {{x}^{ * }x}\right) \rho \left( \mathbf{1}\right) \leq \rho {\left( \mathbf{1}\right) }^{2} \), and \( \parallel \rho \parallel \leq \rho \left( \mathbf{1}\right) \) follows. Since the inequality \( \parallel \rho \parallel \geq \rho \left( \mathbf{1}\right) \) is obvious, we conclude that \( \parallel \rho \parallel = \rho \left( \mathbf{1}\right) \) .
No
Corollary 1. Let \( \rho \) be a linear functional on a unital \( {C}^{ * } \) -algebra \( A \) satisfying \( \parallel \rho \parallel = \rho \left( \mathbf{1}\right) = 1 \) . Then \( \rho \) is a state.
Proof. We have to show that \( \rho \left( {{a}^{ * }a}\right) \geq 0 \) for every \( a \in A \) . By Theorem 4.8.3 it is enough to show that for every self-adjoint element \( x \in A \) having nonnegative spectrum, we have \( \rho \left( x\right) \geq 0 \) . More generally, we claim that for every normal element \( z \in A \) ,\n\n\[ \rho \left( z\right) \in \overline{\operatorname{conv}}\sigma \left( z\right) \]\n\nTo see this, let \( B \) be the commutative \( {C}^{ * } \) -subalgebra generated by \( z \) and 1 . The restriction \( {\rho }_{0} \) of \( \rho \) to \( B \) satisfies the same hypotheses \( \begin{Vmatrix}{\rho }_{0}\end{Vmatrix} = {\rho }_{0}\left( \mathbf{1}\right) = 1 \) . By Theorem 2.2.4, \( B \) is isometrically \( * \) -isomorphic to \( C\left( X\right) \), and for \( C\left( X\right) \) this is the result of Lemma 1.10.3.
Yes
For every element \( x \) in a unital \( {C}^{ * } \) -algebra \( A \) there is a state \( \rho \) such that \( \rho \left( {{x}^{ * }x}\right) = \parallel x{\parallel }^{2} \) .
Proof. Consider the self-adjoint element \( y = {x}^{ * }x \), and let \( B \) be the sub \( {C}^{ * } \) -algebra generated by \( y \) and the identity. Again, since \( B \cong C\left( X\right) \) there is a complex homomorphism \( \omega \in \operatorname{sp}\left( B\right) \) such that \( \omega \left( y\right) = \parallel y\parallel \) . Let \( \rho \) be any extension of \( \omega \) to a linear functional on \( A \) with \( \parallel \rho \parallel = \parallel \omega \parallel = 1 \) . We also have \( \rho \left( \mathbf{1}\right) = \omega \left( \mathbf{1}\right) = 1 \) . Thus \( \parallel \rho \parallel = \rho \left( \mathbf{1}\right) = 1 \), and the preceding corollary implies that \( \rho \) is a state.
Yes
Theorem. A connected algebraic group \( G \) is semisimple if and only if \( \mathfrak{g} \) is semisimple.
Proof. Let \( \mathfrak{g} \) be semisimple. If \( N \) is a closed connected commutative normal subgroup of \( G \), then \( n \) is a commutative ideal of \( \mathfrak{g} \) (use the easy half of Theorem 13.3 and the remarks above). So \( \mathfrak{n} = 0 \), forcing \( N = e \). Conversely, let \( G \) be semisimple and let \( \mathfrak{n} \) be a commutative ideal of \( \mathfrak{g} \). Define \( H = {C}_{G}{\left( \mathfrak{n}\right) }^{ \circ } \). Then \( \mathfrak{h} = {\mathfrak{c}}_{\mathfrak{g}}\left( \mathfrak{n}\right) \) (this follows from Theorem 13.2, cf. Exercise 2). Since \( \mathfrak{n} \) is an ideal, the Jacobi identity shows at once that \( \mathfrak{h} \) is also an ideal (including \( n \), since \( n \) is commutative, as part of its center). Thanks to Theorem 13.3, \( H \) must be normal in \( G \); as a result \( Z = Z{\left( H\right) }^{ \circ } \) is also normal in \( G \). By Theorem 13.4(b), \( \mathfrak{z} \) is the center of \( \mathfrak{h} \), and therefore includes \( \mathfrak{n} \). But \( G \) is semisimple, so \( Z = e,\mathfrak{z} = 0 \). This forces \( \mathfrak{n} = 0 \). \( ▱ \)
Yes
The extension of \( {\varphi }_{T} : T \rightarrow H \) to an isomorphism \( \varphi : G \rightarrow H \) was carried out under several assumptions involving the canonical rank 1 or 2 subsystems \( {\Phi }_{\alpha \beta } \) of \( \Phi \left( {\alpha ,\beta \in \Delta }\right) \) . These assumptions were formulated as follows in (32.4):\n\n(A) \( {\varphi }_{T}\left( {t}_{\alpha \beta }\right) = {\left( {h}_{\alpha }{h}_{\beta }\right) }^{m\left( {\alpha ,\beta }\right) } \) for all \( \alpha ,\beta \in \Delta \), where \( {t}_{\alpha \beta } = {\left( {n}_{\alpha }{n}_{\beta }\right) }^{m\left( {\alpha ,\beta }\right) }, m\left( {\alpha ,\beta }\right) = \) order of \( {\sigma }_{\alpha }{\sigma }_{\beta } \) in \( W \) .\n\n(B) For each pair \( \alpha ,\beta \in \Delta \), there is an isomorphism \( {\varphi }_{\alpha \beta } \) (of algebraic groups) from \( {U}_{\alpha \beta } \) onto the corresponding subgroup of \( H \), such that \( {\varphi }_{\alpha \beta } \) extends both \( {\varphi }_{\alpha } \) and \( {\varphi }_{\beta } \) .\n\n(C) Let \( \alpha ,\beta \in \Delta \) . If \( \gamma \in {\Phi }_{\alpha \beta }^{ + } \) with \( \gamma \neq \alpha \), then Int \( {h}_{\alpha } \circ {\varphi }_{\alpha \beta } \) agrees on \( {U}_{\gamma } \) with \( {\varphi }_{\alpha \beta } \circ \operatorname{Int}{n}_{\alpha } \) .\n\nNotice that when \( \alpha = \beta \) ,(B) and (C) are vacuous, while (A) is satisfied by virtue of the classification of rank 1 groups in (32.3).
Ostensibly these three statements involve the pair of groups \( G, H \) . But to verify them it is actually sufficient to work inside \( G \) alone. All we have to show is that the choices made \( \left( {{\varepsilon }_{\alpha },{\varepsilon }_{-\alpha }}\right. \), hence \( \left. {{n}_{\alpha }\text{, for}\alpha \in \Delta }\right) \) completely determine the elements \( {t}_{\alpha \beta } \), the commutation in \( {U}_{\alpha \beta } \), and the action of Int \( \alpha \) on certain root groups. Then the corresponding choices in \( H \) must lead to corresponding results. (This approach amounts to providing generators and relations for \( G \) .) With this in mind, let us reformulate (A),(B),(C).\n\nTo prove (A) it will be enough to express \( {t}_{\alpha \beta } \) in terms of \( {t}_{\alpha } \) and \( {t}_{\beta } \) in a way depending only on which root system of rank 2 is involved, since we already know that \( {t}_{\alpha } \) and \( {t}_{\beta } \) are well determined by the given data.\n\nFor (B) we shall prescribe \( {\varepsilon }_{\gamma } : {\mathbf{G}}_{a} \rightarrow {U}_{\gamma } \) for each \( \gamma \in {\Phi }_{\alpha \beta }^{ + } \) in a way depending only on the root system, then show that the constants occurring in the commutator formulas (Lemma 32.5) for all pairs of roots in \( {\Phi }_{\alpha \beta }^{ + } \) are completely determined. So the group-theoretic structure of \( {U}_{\alpha \beta } \) will just depend on the root system. On the other hand, it is easy to construct an isomorphism of varieties from \( {U}_{\alpha \beta } \) onto the corresponding subgroup of \( H \) (cf. (32.5)); the use of the \( {\varepsilon }_{\gamma } \) and corresponding \( {\varepsilon }_{\gamma }^{\prime } \), will insure that this is a group homomorphism. As an example of how \( {\varepsilon }_{\gamma } \) is to be prescribed, consider type \( {A}_{2} \), where the positive roots are \( \alpha ,\beta ,\alpha + \beta \) . We are already given \( {\varepsilon }_{\alpha } \) and \( {n}_{\beta } \), so we just set \( {\varepsilon }_{\alpha + \beta } = {\varepsilon }_{\alpha } \circ {n}_{\beta } \) .
Yes
Proposition 1.1. Let \( f, g \in A \) with \( g \neq 0 \) . Then there exist elements \( q, r \in A \) such that \( f = {qg} + r \) and \( r \) is either 0 or \( \deg \left( r\right) < \deg \left( g\right) \) . Moreover, \( q \) and \( r \) are uniquely determined by these conditions.
Proof. Let \( n = \deg \left( f\right), m = \deg \left( g\right) ,\alpha = \operatorname{sgn}\left( f\right) ,\beta = \operatorname{sgn}\left( g\right) \) . We give the proof by induction on \( n = \deg \left( f\right) \) . If \( n < m \), set \( q = 0 \) and \( r = f \) . If \( n \geq m \), we note that \( {f}_{1} = f - \alpha {\beta }^{-1}{T}^{n - m}g \) has smaller degree than \( f \) . By induction, there exist \( {q}_{1},{r}_{1} \in A \) such that \( {f}_{1} = {q}_{1}g + {r}_{1} \) with \( {r}_{1} \) being either 0 or with degree less than \( \deg \left( g\right) \) . In this case, set \( q = \alpha {\beta }^{-1}{T}^{n - m} + {q}_{1} \) and \( r = {r}_{1} \) and we are done.\n\nIf \( f = {qg} + r = {q}^{\prime }g + {r}^{\prime } \), then \( g \) divides \( r - {r}^{\prime } \) and by degree considerations we see \( r = {r}^{\prime } \) . In this case, \( {qg} = {q}^{\prime }g \) so \( q = {q}^{\prime } \) and the uniqueness is established.\n\nThis proposition shows that \( A \) is a Euclidean domain and thus a principal ideal domain and a unique factorization domain. It also allows a quick proof of the finiteness of the residue class rings.
Yes
Proposition 1.2. Suppose \( g \in A \) and \( g \neq 0 \) . Then \( A/{gA} \) is a finite ring with \( {q}^{\deg \left( g\right) } \) elements.
Proof. Let \( m = \deg \left( g\right) \) . By Proposition 1.1 one easily verifies that \( \{ r \in \) \( A \mid \deg \left( r\right) < m\} \) is a complete set of representatives for \( A/{gA} \) . Such elements look like\n\n\[ r = {\alpha }_{0}{T}^{m - 1} + {\alpha }_{1}{T}^{m - 2} + \cdots + {\alpha }_{m - 1}\;\text{ with }{\alpha }_{i} \in \mathbb{F}. \]\n\nSince the \( {\alpha }_{i} \) vary independently through \( \mathbb{F} \) there are \( {q}^{m} \) such polynomials and the result follows.
Yes
Proposition 1.3. The group of units in \( A \) is \( {\mathbb{F}}^{ * } \) . In particular, it is a finite cyclic group with \( q - 1 \) elements.
Proof. The only thing left to prove is the cyclicity of \( {\mathbb{F}}^{ * } \) . This follows from the very general fact that a finite subgroup of the multiplicative group of a field is cyclic.
No
Proposition 1.4. Let \( {m}_{1},{m}_{2},\ldots ,{m}_{t} \) be elements of \( A \) which are pairwise relatively prime. Let \( m = {m}_{1}{m}_{2}\ldots {m}_{t} \) and \( {\phi }_{i} \) be the natural homomorphism from \( A/{mA} \) to \( A/{m}_{i}A \) . Then, the map \( \phi : A/{mA} \rightarrow A/{m}_{1}A \oplus A/{m}_{2}A \oplus \cdots \oplus A/{m}_{t}A \) given by \[ \phi \left( a\right) = \left( {{\phi }_{1}\left( a\right) ,{\phi }_{2}\left( a\right) ,\ldots ,{\phi }_{t}\left( a\right) }\right) \] is a ring isomorphism.
Proof. This is a standard result which holds in any principal ideal domain (properly formulated it holds in much greater generality).
No
Proposition 1.5. Let \( P \in A \) be an irreducible polynomial. Then, \( {\left( A/PA\right) }^{ * } \) is a cyclic group with \( \left| P\right| - 1 \) elements.
Proof. Since \( A \) is a principal ideal domain, \( {PA} \) is a maximal ideal and so \( A/{PA} \) is a field. A finite subgroup of the multiplicative group of a field is cyclic. Thus \( {\left( A/PA\right) }^{ * } \) is cyclic. That the order of this group is \( \left| P\right| - 1 \) is immediate.
Yes
Let \( P \in A \) be an irreducible polynomial and \( e \) a positive integer. The order of \( {\left( A/{P}^{e}A\right) }^{ * } \) is \( {\left| P\right| }^{e - 1}\left( {\left| P\right| - 1}\right) \) . Let \( {\left( A/{P}^{e}A\right) }^{\left( 1\right) } \) be the kernel of the natural map from \( {\left( A/{P}^{e}A\right) }^{ * } \) to \( {\left( A/PA\right) }^{ * } \) . It is a p-group of order \( {\left| P\right| }^{e - 1} \) . As e tends to infinity, the minimal number of generators of \( {\left( A/{P}^{e}A\right) }^{\left( 1\right) } \) tends to infinity.
The ring \( A/{P}^{e}A \) has only one maximal ideal \( {PA}/{P}^{e}A \) which has \( {\left| P\right| }^{e - 1} \) elements. Thus, \( {\left( A/{P}^{e}A\right) }^{ * } = A/{P}^{e}A - {PA}/{P}^{e}A \) has \( {\left| P\right| }^{e} - {\left| P\right| }^{e - 1} = \) \( {\left| P\right| }^{e - 1}\left( {\left| P\right| - 1}\right) \) number of elements. Since \( {\left( A/{P}^{e}A\right) }^{ * } \rightarrow {\left( A/PA\right) }^{ * } \) is onto, and the latter group has \( \left| P\right| - 1 \) elements the assertion about the size of \( {\left( A/{P}^{e}A\right) }^{\left( 1\right) } \) follows. It remains to prove the assertion about the minimal number of generators.\n\nIt is instructive to first consider the case \( e = 2 \) . Every element in \( {\left( A/{P}^{2}A\right) }^{\left( 1\right) } \) can be represented by a polynomial of the form \( a = 1 + {bP} \) . Since we are working in characteristic \( p \) we have \( {a}^{p} = 1 + {b}^{p}{P}^{p} \equiv 1 \) \( \left( {\;\operatorname{mod}\;{P}^{2}}\right) \) . So, we have a group of order \( \left| P\right| \) with exponent \( p \) . If \( q = {p}^{f} \) it follows that \( {\left( A/{P}^{2}A\right) }^{\left( 1\right) } \) is a direct sum of \( f\deg \left( P\right) \) number of copies of \( \mathbb{Z}/p\mathbb{Z} \) . This is a cyclic group only under the very restrictive conditions that \( q = p \) and \( \deg \left( P\right) = 1 \) .\n\nTo deal with the general case, suppose that \( s \) is the smallest integer such that \( {p}^{s} \geq e \) . Since \( {\left( 1 + bP\right) }^{{p}^{s}} = 1 + {\left( bP\right) }^{{p}^{s}} \equiv 1\left( {\;\operatorname{mod}\;{P}^{e}}\right) \) we have that raising to the \( {p}^{s} \) -power annihilates \( G = {\left( A/{P}^{e}A\right) }^{\left( 1\right) } \) . Let \( d \) be the minimal number of generators of this group. It follows that there is an onto map from \( {\left( \mathbb{Z}/{p}^{s}\mathbb{Z}\right) }^{d} \) onto \( G \) . Thus, \( {p}^{ds} \geq {p}^{f\deg \left( P\right) \left( {e - 1}\right) } \) and so\n\n\[ d \geq \frac{f\deg \left( P\right) \left( {e - 1}\right) }{s}. \]\n\nSince \( s \) is the smallest integer bigger than or equal to \( {\log }_{p}\left( e\right) \) it is clear that \( d \rightarrow \infty \) as \( e \rightarrow \infty \).
Yes
Proposition 1.7.\n\n\[ \Phi \left( f\right) = \left| f\right| \mathop{\prod }\limits_{{P \mid f}}\left( {1 - \frac{1}{\left| P\right| }}\right) \]
Proof. Let \( f = \alpha {P}_{1}^{{e}_{1}}{P}_{2}^{{e}_{2}}\ldots {P}_{t}^{{e}_{t}} \) be the prime decomposition of \( f \) . By the corollary to Propositions 1.4 and by Proposition 1.6, we see that\n\n\[ \Phi \left( f\right) = \mathop{\prod }\limits_{{i = 1}}^{t}\Phi \left( {P}_{i}^{{e}_{i}}\right) = \mathop{\prod }\limits_{{i = 1}}^{t}\left( {{\left| {P}_{i}\right| }^{{e}_{i}} - {\left| {P}_{i}\right| }^{{e}_{i} - 1}}\right) ,\]\n\nfrom which the result follows immediately.
Yes
Proposition 1.8. If \( f \in A, f \neq 0 \), and \( a \in A \) is relatively prime to \( f \), i.e., \( \left( {a, f}\right) = 1 \), then\n\n\[ \n{a}^{\Phi \left( f\right) } \equiv 1\;\left( {\;\operatorname{mod}\;f}\right) \n\]
Proof. The group \( {\left( A/fA\right) }^{ * } \) has \( \Phi \left( f\right) \) elements. The coset of \( a \) modulo \( f,\bar{a} \) , lies in this group. Thus, \( {\bar{a}}^{\Phi \left( f\right) } = \overline{1} \) and this is equivalent to the congruence in the proposition.
Yes
Proposition 1.9. Let \( P \in A \) be irreducible of degree \( d \) . Suppose \( X \) is an indeterminate. Then,\n\n\[ \n{X}^{\left| P\right| - 1} - 1 \equiv \mathop{\prod }\limits_{{0 \leq \deg \left( f\right) < d}}\left( {X - f}\right) \;\left( {\;\operatorname{mod}\;P}\right) .\n\]
Proof. Recall that \( \{ f \in A \mid \deg \left( f\right) < d\} \) is a set of representatives for the cosets of \( A/{PA} \) . If we throw out \( f = 0 \) we get a set of representatives for \( {\left( A/PA\right) }^{ * } \) . We find\n\n\[ \n{X}^{\left| P\right| - 1} - \overline{1} = \mathop{\prod }\limits_{{0 \leq \deg \left( f\right) < d}}\left( {X - \bar{f}}\right)\n\]\n\nwhere the bars denote cosets modulo \( P \) . This follows from the corollary to Proposition 1.8 since both sides of the equation are monic polynomials in \( X \) with the same set of roots in the field \( A/{PA} \) . Since there are \( \left| P\right| - 1 \) roots and the difference of the two sides has degree less than \( \left| P\right| - 1 \), the difference of the two sides must be 0 . The congruence in the Proposition is equivalent to this assertion.
Yes
Corollary 1. Let \( d \) divide \( \left| P\right| - 1 \) . The congruence \( {X}^{d} \equiv 1\left( {\;\operatorname{mod}\;P}\right) \) has exactly \( d \) solutions. Equivalently, the equation \( {X}^{d} = \overline{1} \) has exactly \( d \) solutions in \( {\left( A/PA\right) }^{ * } \) .
Proof. We prove the second assertion. Since \( d\left| \right| P \mid - 1 \) it follows that \( {X}^{d} - 1 \) divides \( {X}^{\left| P\right| - 1} - 1 \) . By the proposition, the latter polynomial splits as a product of distinct linear factors. Thus so does the former polynomial. This establishes the result.
No
Corollary 2. With the same notation,\n\n\[ \mathop{\prod }\limits_{{0 \leq \deg \left( f\right) < \deg P}}f \equiv - 1\;\left( {\;\operatorname{mod}\;P}\right) \]
Proof. Just set \( X = 0 \) in the proposition. If the characteristic of \( \mathbb{F} \) is odd \( \left| P\right| - 1 \) is even and the result follows. If the characteristic is 2 then the result also follows since in characteristic 2 we have \( - 1 = 1 \) .
Yes
Proposition 1.10. Let \( P \) be irreducible and \( a \in A \) not divisible by \( P \) . Assume \( d \) divides \( \left| P\right| - 1 \) . The congruence \( {X}^{d} \equiv a\left( {\;\operatorname{mod}\;{P}^{e}}\right) \) is solvable if and only if\n\n\[ \n{a}^{\frac{\left| P\right| - 1}{d}} \equiv 1\;\left( {\;\operatorname{mod}\;P}\right) \n\]\n\nThere are \( \frac{\Phi \left( {P}^{e}\right) }{d}d \) -th power residues modulo \( {P}^{e} \) .
Proof. Assume to begin with that \( e \doteq 1 \) .\n\nIf \( {b}^{d} \equiv a\left( {\;\operatorname{mod}\;P}\right) \), then \( {a}^{\frac{\left| P\right| - 1}{d}} \equiv {b}^{\left| P\right| - 1} \equiv 1\left( {\;\operatorname{mod}\;P}\right) \) by the corollary to Proposition 1.8. This shows the condition is necessary. To show it is sufficient recall that by Corollary 1 to Proposition 1.9 all the \( d \) -th roots of unity are in the field \( A/{PA} \) . Consider the homomorphism from \( {\left( A/PA\right) }^{ * } \) to itself given by raising to the \( d \) -th power. It’s kernel has order \( d \) and its image is the \( d \) -th powers. Thus, there are precisely \( \frac{\left| P\right| - 1}{d}d \) -th powers in \( {\left( A/PA\right) }^{ * } \) . We have seen that they all satisfy \( {X}^{\frac{\left| P\right| - 1}{d}} - 1 = 0 \) . Thus, they are precisely the roots of this equation. This proves all assertions in the case \( e = 1 \) .\n\nTo deal with the remaining cases, we employ a little group theory. The natural map (i.e., reduction modulo \( P \) ) is a homomorphism from \( {\left( A/{P}^{e}A\right) }^{ * } \) onto \( {\left( A/PA\right) }^{ * } \) and the kernel is a \( p \) -group as follows from Proposition 1.6. Since the order of \( {\left( A/PA\right) }^{ * } \) is \( \left| P\right| - 1 \) which is prime to \( p \) it follows that \( {\left( A/{P}^{e}A\right) }^{ * } \) is the direct product of a \( p \) -group and a copy of \( {\left( A/PA\right) }^{ * } \) . Since \( \left( {d, p}\right) = 1 \), raising to the \( d \) -th power in an abelian \( p \) -group is an automorphism. Thus, \( a \in A \) is a \( d \) -th power modulo \( {P}^{e} \) if and only if it is a \( d \) -th power modulo \( P \) . The latter has been shown to hold if and only if \( {a}^{\frac{\left| P\right| - 1}{d}} \equiv 1\left( {\;\operatorname{mod}\;P}\right) \) . Now consider the homomorphism from \( {\left( A/{P}^{e}A\right) }^{ * } \) to itself given by raising to the \( d \) -th power. It easily follows from what has been said that the kernel has \( d \) elements and the image is the subgroup of \( d \) -th powers. It follows that the latter group has order \( \frac{\widetilde{\Phi }\left( {P}^{e}\right) }{d} \) . This concludes the proof.
Yes
Proposition 2.1.\n\n\[ \mathop{\sum }\limits_{{d \mid n}}d{a}_{d} = {q}^{n} \]
This formula is often attributed to Richard Dedekind. It is interesting to note that it appears, with essentially the above proof, in a manuscript of C.F. Gauss (unpublished in his lifetime), \
No