1: %TCIDATA{Version=4.10.0.2363}

2: %TCIDATA{LaTeXparent=0,0,Advances2.tex}

3: 

4: 

5: \chapter{ Reverses for the Continuous Triangle Inequality}\label{ch4}

6: 

7: \section{Introduction}

8: 

9: Let $f:\left[ a,b\right] \rightarrow \mathbb{K}$, $\mathbb{K}=\mathbb{C}$ or

10: $\mathbb{R}$ be a Lebesgue integrable function. The following inequality,

11: which is the continuous version of the \textit{triangle inequality}%

12: \begin{equation}

13: \left\vert \int_{a}^{b}f\left( x\right) dx\right\vert \leq

14: \int_{a}^{b}\left\vert f\left( x\right) \right\vert dx,  \label{RTxx0.1.1}

15: \end{equation}%

16: plays a fundamental role in Mathematical Analysis and its applications.

17: 

18: It appears, see \cite[p. 492]{RTCMPF}, that the first reverse inequality for

19: (\ref{RTxx0.1.1}) was obtained by J. Karamata in his book from 1949, \cite%

20: {RTCK}. It can be stated as%

21: \begin{equation}

22: \cos \theta \int_{a}^{b}\left\vert f\left( x\right) \right\vert dx\leq

23: \left\vert \int_{a}^{b}f\left( x\right) dx\right\vert  \label{RTxx0.1.2}

24: \end{equation}%

25: provided%

26: \begin{equation*}

27: -\theta \leq \arg f\left( x\right) \leq \theta ,\ \ x\in \left[ a,b\right]

28: \end{equation*}%

29: for given $\theta \in \left( 0,\frac{\pi }{2}\right) .$

30: 

31: This integral inequality is the continuous version of a reverse inequality

32: for the generalised triangle inequality%

33: \begin{equation}

34: \cos \theta \sum_{i=1}^{n}\left\vert z_{i}\right\vert \leq \left\vert

35: \sum_{i=1}^{n}z_{i}\right\vert ,  \label{RTxx0.1.3}

36: \end{equation}%

37: provided%

38: \begin{equation*}

39: a-\theta \leq \arg \left( z_{i}\right) \leq a+\theta ,\ \ \text{for \ }i\in

40: \left\{ 1,\dots ,n\right\} ,

41: \end{equation*}%

42: where $a\in \mathbb{R}$ and $\theta \in \left( 0,\frac{\pi }{2}\right) ,$

43: which, as pointed out in \cite[p. 492]{RTCMPF}, was first discovered by M.

44: Petrovich in 1917, \cite{RTCP}, and, subsequently rediscovered by other

45: authors, including J. Karamata \cite[p. 300 -- 301]{RTCK}, H.S. Wilf \cite%

46: {RTCW}, and in an equivalent form, by M. Marden \cite{RTCM}.

47: 

48: The first to consider the problem in the more general case of Hilbert and

49: Banach spaces, were J.B. Diaz and F.T. Metcalf \cite{RTCDM} who showed that,

50: in an inner product space $H$ over the real or complex number field, the

51: following reverse of the triangle inequality holds%

52: \begin{equation}

53: r\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \leq \left\Vert

54: \sum_{i=1}^{n}x_{i}\right\Vert ,  \label{RTxx0.1.4}

55: \end{equation}%

56: provided%

57: \begin{equation*}

58: 0\leq r\leq \frac{\func{Re}\left\langle x_{i},a\right\rangle }{\left\Vert

59: x_{i}\right\Vert },\ \ \ \ \ i\in \left\{ 1,\dots ,n\right\} ,

60: \end{equation*}%

61: and $a\in H$ is a unit vector, i.e., $\left\Vert a\right\Vert =1.$ The case

62: of equality holds in (\ref{RTxx0.1.4}) if and only if%

63: \begin{equation}

64: \sum_{i=1}^{n}x_{i}=r\left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert

65: \right) a.  \label{RTxx0.1.5}

66: \end{equation}

67: 

68: A generalisation of this result for orthonormal families is also well known

69: \cite{RTCDM}:

70: 

71: Let $a_{1},\dots ,a_{m}$ be $m$ orthonormal vectors in $H.$ Suppose the

72: vectors $x_{1},\dots ,x_{n}\in H\backslash \left\{ 0\right\} $ satisfy%

73: \begin{equation*}

74: 0\leq r_{k}\leq \frac{\func{Re}\left\langle x_{i},a_{k}\right\rangle }{%

75: \left\Vert x_{i}\right\Vert },\ \ \ \ \ i\in \left\{ 1,\dots ,n\right\} ,\

76: k\in \left\{ 1,\dots ,m\right\} .

77: \end{equation*}%

78: Then%

79: \begin{equation*}

80: \left( \sum_{k=1}^{m}r_{k}^{2}\right) ^{\frac{1}{2}}\sum_{i=1}^{n}\left\Vert

81: x_{i}\right\Vert \leq \left\Vert \sum_{i=1}^{n}x_{i}\right\Vert ,

82: \end{equation*}%

83: where equality holds if and only if%

84: \begin{equation*}

85: \sum_{i=1}^{n}x_{i}=\left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right)

86: \sum_{k=1}^{m}r_{k}a_{k}.

87: \end{equation*}%

88: The main aim of this chapter is to survey some recent reverses of the

89: triangle inequality for Bochner integrable functions $f$ with values in

90: Hilbert spaces and defined on a compact interval $\left[ a,b\right] \subset

91: \mathbb{R}$. Applications for Lebesgue integrable complex-valued functions

92: are provided as well.

93: 

94: \section{Multiplicative Reverses}

95: 

96: \subsection{Reverses for a Unit Vector}

97: 

98: We recall that $f\in L\left( \left[ a,b\right] ;H\right) ,$ the space of

99: Bochner integrable functions with values in a Hilbert space $H,$ if and only

100: if $f:\left[ a,b\right] \rightarrow H$ is Bochner measurable on $\left[ a,b%

101: \right] $ and the Lebesgue integral $\int_{a}^{b}\left\Vert f\left( t\right)

102: \right\Vert dt$ is finite.

103: 

104: The following result holds \cite{RTCSSD1}:

105: 

106: \begin{theorem}[Dragomir, 2004]

107: \label{RTxxt1.2.1}If $f\in L\left( \left[ a,b\right] ;H\right) $ is such

108: that there exists a constant $K\geq 1$ and a vector $e\in H,$ $\left\Vert

109: e\right\Vert =1$ with%

110: \begin{equation}

111: \left\Vert f\left( t\right) \right\Vert \leq K\func{Re}\left\langle f\left(

112: t\right) ,e\right\rangle \ \ \ \text{for a.e. }t\in \left[ a,b\right] ,

113: \label{RTxx1.2.1}

114: \end{equation}%

115: then we have the inequality:%

116: \begin{equation}

117: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\leq K\left\Vert

118: \int_{a}^{b}f\left( t\right) dt\right\Vert .  \label{RTxx1.2.2}

119: \end{equation}%

120: The case of equality holds in (\ref{RTxx1.2.2}) if and only if%

121: \begin{equation}

122: \int_{a}^{b}f\left( t\right) dt=\frac{1}{K}\left( \int_{a}^{b}\left\Vert

123: f\left( t\right) \right\Vert dt\right) e.  \label{RTxx1.2.3}

124: \end{equation}

125: \end{theorem}

126: 

127: \begin{proof}

128: By the Schwarz inequality in inner product spaces, we have%

129: \begin{align}

130: \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert & =\left\Vert

131: \int_{a}^{b}f\left( t\right) dt\right\Vert \left\Vert e\right\Vert

132: \label{RTxx1.2.4} \\

133: & \geq \left\vert \left\langle \int_{a}^{b}f\left( t\right)

134: dt,e\right\rangle \right\vert \geq \left\vert \func{Re}\left\langle

135: \int_{a}^{b}f\left( t\right) dt,e\right\rangle \right\vert  \notag \\

136: & \geq \func{Re}\left\langle \int_{a}^{b}f\left( t\right) dt,e\right\rangle

137: =\int_{a}^{b}\func{Re}\left\langle f\left( t\right) ,e\right\rangle dt.

138: \notag

139: \end{align}%

140: From the condition (\ref{RTxx1.2.1}), on integrating over $\left[ a,b\right]

141: ,$ we deduce%

142: \begin{equation}

143: \int_{a}^{b}\func{Re}\left\langle f\left( t\right) ,e\right\rangle dt\geq

144: \frac{1}{K}\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt,

145: \label{RTxx1.2.5}

146: \end{equation}%

147: and thus, on making use of (\ref{RTxx1.2.4}) and (\ref{RTxx1.2.5}), we

148: obtain the desired inequality (\ref{RTxx1.2.2}).

149: 

150: If (\ref{RTxx1.2.3}) holds true, then, obviously%

151: \begin{equation*}

152: K\left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert =\left\Vert

153: e\right\Vert \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert

154: dt=\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt,

155: \end{equation*}%

156: showing that (\ref{RTxx1.2.2}) holds with equality.

157: 

158: If we assume that the equality holds in (\ref{RTxx1.2.2}), then by the

159: argument provided at the beginning of our proof, we must have equality in

160: each of the inequalities from (\ref{RTxx1.2.4}) and (\ref{RTxx1.2.5}).

161: 

162: Observe that in Schwarz's inequality $\left\Vert x\right\Vert \left\Vert

163: y\right\Vert \geq \func{Re}\left\langle x,y\right\rangle ,$ $x,y\in H,$ the

164: case of equality holds if and only if there exists a positive scalar $\mu $

165: such that $x=\mu e.$ Therefore, equality holds in the first inequality in (%

166: \ref{RTxx1.2.4}) iff $\int_{a}^{b}f\left( t\right) dt=\lambda e$, with $%

167: \lambda \geq 0$ $.$

168: 

169: If we assume that a strict inequality holds in (\ref{RTxx1.2.1}) on a subset

170: of nonzero Lebesgue measure in $\left[ a,b\right] ,$ then $$%

171: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt<K\int_{a}^{b}\func{Re}%

172: \left\langle f\left( t\right) ,e\right\rangle dt,$$ and by

173: (\ref{RTxx1.2.4}) we deduce a strict inequality in

174: (\ref{RTxx1.2.2}), which contradicts the

175: assumption. Thus, we must have $\left\Vert f\left( t\right) \right\Vert =K%

176: \func{Re}\left\langle f\left( t\right) ,e\right\rangle $ for a.e. $t\in %

177: \left[ a,b\right] .$

178: 

179: If we integrate this equality, we deduce%

180: \begin{align*}

181: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt& =K\int_{a}^{b}\func{%

182: Re}\left\langle f\left( t\right) ,e\right\rangle dt=K\func{Re}\left\langle

183: \int_{a}^{b}f\left( t\right) dt,e\right\rangle \\

184: & =K\func{Re}\left\langle \lambda e,e\right\rangle =\lambda K

185: \end{align*}%

186: giving%

187: \begin{equation*}

188: \lambda =\frac{1}{K}\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt,

189: \end{equation*}%

190: and thus the equality (\ref{RTxx1.2.3}) is necessary.

191: 

192: This completes the proof.

193: \end{proof}

194: 

195: A more appropriate result from an applications point of view is perhaps the

196: following result \cite{RTCSSD1}.

197: 

198: \begin{corollary}

199: \label{RTxxc1.2.2}Let $e$ be a unit vector in the Hilbert space $\left(

200: H;\left\langle \cdot ,\cdot \right\rangle \right) ,$ $\rho \in \left(

201: 0,1\right) $ and $f\in L\left( \left[ a,b\right] ;H\right) $ so that%

202: \begin{equation}

203: \left\Vert f\left( t\right) -e\right\Vert \leq \rho \ \ \ \text{for a.e. }%

204: t\in \left[ a,b\right] .  \label{RTxx1.2.6}

205: \end{equation}%

206: Then we have the inequality%

207: \begin{equation}

208: \sqrt{1-\rho ^{2}}\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\leq

209: \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert ,  \label{RTxx1.2.7}

210: \end{equation}%

211: with equality if and only if%

212: \begin{equation}

213: \int_{a}^{b}f\left( t\right) dt=\sqrt{1-\rho ^{2}}\left(

214: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right) e.

215: \label{RTxx1.2.8}

216: \end{equation}

217: \end{corollary}

218: 

219: \begin{proof}

220: From (\ref{RTxx1.2.6}), we have%

221: \begin{equation*}

222: \left\Vert f\left( t\right) \right\Vert ^{2}-2\func{Re}\left\langle f\left(

223: t\right) ,e\right\rangle +1\leq \rho ^{2},

224: \end{equation*}%

225: giving%

226: \begin{equation*}

227: \left\Vert f\left( t\right) \right\Vert ^{2}+1-\rho ^{2}\leq 2\func{Re}%

228: \left\langle f\left( t\right) ,e\right\rangle

229: \end{equation*}%

230: for a.e. $t\in \left[ a,b\right] .$

231: 

232: Dividing by $\sqrt{1-\rho ^{2}}>0,$ we deduce%

233: \begin{equation}

234: \frac{\left\Vert f\left( t\right) \right\Vert ^{2}}{\sqrt{1-\rho ^{2}}}+%

235: \sqrt{1-\rho ^{2}}\leq \frac{2\func{Re}\left\langle f\left( t\right)

236: ,e\right\rangle }{\sqrt{1-\rho ^{2}}}  \label{RTxx1.2.9}

237: \end{equation}%

238: for a.e. $t\in \left[ a,b\right] .$

239: 

240: On the other hand, by the elementary inequality%

241: \begin{equation*}

242: \frac{p}{\alpha }+q\alpha \geq 2\sqrt{pq},\ \ \ p,q\geq 0,\ \alpha >0

243: \end{equation*}%

244: we have%

245: \begin{equation}

246: 2\left\Vert f\left( t\right) \right\Vert \leq \frac{\left\Vert f\left(

247: t\right) \right\Vert ^{2}}{\sqrt{1-\rho ^{2}}}+\sqrt{1-\rho ^{2}}

248: \label{RTxx1.2.10}

249: \end{equation}%

250: for each $t\in \left[ a,b\right] .$

251: 

252: Making use of (\ref{RTxx1.2.9}) and (\ref{RTxx1.2.10}), we deduce%

253: \begin{equation*}

254: \left\Vert f\left( t\right) \right\Vert \leq \frac{1}{\sqrt{1-\rho ^{2}}}%

255: \func{Re}\left\langle f\left( t\right) ,e\right\rangle

256: \end{equation*}%

257: for a.e. $t\in \left[ a,b\right] .$

258: 

259: Applying Theorem \ref{RTxxt1.2.1} for $K=\frac{1}{\sqrt{1-\rho ^{2}}},$ we

260: deduce the desired inequality (\ref{RTxx1.2.7}).

261: \end{proof}

262: 

263: In the same spirit, we also have the following corollary \cite{RTCSSD1}.

264: 

265: \begin{corollary}

266: \label{RTxxc1.2.3}Let $e$ be a unit vector in $H$ and $M\geq m>0.$ If $f\in

267: L\left( \left[ a,b\right] ;H\right) $ is such that%

268: \begin{equation}

269: \func{Re}\left\langle Me-f\left( t\right) ,f\left( t\right) -me\right\rangle

270: \geq 0\ \ \   \label{RTxx1.2.9a}

271: \end{equation}%

272: or, equivalently,%

273: \begin{equation}

274: \left\Vert f\left( t\right) -\frac{M+m}{2}e\right\Vert \leq \frac{1}{2}%

275: \left( M-m\right) \ \ \   \label{RTxx1.2.10a}

276: \end{equation}%

277: for a.e. $t\in \left[ a,b\right] ,$ then we have the inequality%

278: \begin{equation}

279: \frac{2\sqrt{mM}}{M+m}\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert

280: dt\leq \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert ,

281: \label{RTxx1.2.11}

282: \end{equation}%

283: or, equivalently,%

284: \begin{align}

285: (0& \leq )\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt-\left\Vert

286: \int_{a}^{b}f\left( t\right) dt\right\Vert  \label{RTxx1.2.12} \\

287: & \leq \frac{\left( \sqrt{M}-\sqrt{m}\right) ^{2}}{M+m}\left\Vert

288: \int_{a}^{b}f\left( t\right) dt\right\Vert .  \notag

289: \end{align}%

290: The equality holds in (\ref{RTxx1.2.11}) (or in the second part of (\ref%

291: {RTxx1.2.12})) if and only if%

292: \begin{equation}

293: \int_{a}^{b}f\left( t\right) dt=\frac{2\sqrt{mM}}{M+m}\left(

294: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right) e.

295: \label{RTxx1.2.13}

296: \end{equation}

297: \end{corollary}

298: 

299: \begin{proof}

300: Firstly, we remark that if $x,z,Z\in H,$ then the following statements are

301: equivalent

302: 

303: \begin{enumerate}

304: \item[(i)] $\func{Re}\left\langle Z-x,x-z\right\rangle \geq 0$

305: 

306: and

307: 

308: \item[(ii)] $\left\Vert x-\frac{Z+z}{2}\right\Vert \leq \frac{1}{2}%

309: \left\Vert Z-z\right\Vert .$

310: \end{enumerate}

311: 

312: Using this fact, we may simply realise that (\ref{RTxx1.2.9}) and (\ref%

313: {RTxx1.2.10}) are equivalent.

314: 

315: Now, from (\ref{RTxx1.2.9}), we obtain%

316: \begin{equation*}

317: \left\Vert f\left( t\right) \right\Vert ^{2}+mM\leq \left( M+m\right) \func{%

318: Re}\left\langle f\left( t\right) ,e\right\rangle

319: \end{equation*}%

320: for a.e. $t\in \left[ a,b\right] .$ Dividing this inequality with $\sqrt{mM}%

321: >0,$ we deduce the following inequality that will be used in the sequel%

322: \begin{equation}

323: \frac{\left\Vert f\left( t\right) \right\Vert ^{2}}{\sqrt{mM}}+\sqrt{mM}\leq

324: \frac{M+m}{\sqrt{mM}}\func{Re}\left\langle f\left( t\right) ,e\right\rangle

325: \label{RTxx1.2.14}

326: \end{equation}%

327: for a.e. $t\in \left[ a,b\right] .$

328: 

329: On the other hand%

330: \begin{equation}

331: 2\left\Vert f\left( t\right) \right\Vert \leq \frac{\left\Vert f\left(

332: t\right) \right\Vert ^{2}}{\sqrt{mM}}+\sqrt{mM},  \label{RTxx1.2.15}

333: \end{equation}%

334: for any $t\in \left[ a,b\right] .$

335: 

336: Utilising (\ref{RTxx1.2.14}) and (\ref{RTxx1.2.15}), we may conclude with

337: the following inequality

338: \begin{equation*}

339: \left\Vert f\left( t\right) \right\Vert \leq \frac{M+m}{2\sqrt{mM}}\func{Re}%

340: \left\langle f\left( t\right) ,e\right\rangle ,

341: \end{equation*}%

342: for a.e. $t\in \left[ a,b\right] .$

343: 

344: Applying Theorem \ref{RTxxt1.2.1} for the constant $K:=\frac{m+M}{2\sqrt{mM}}%

345: \geq 1,$ we deduce the desired result.

346: \end{proof}

347: 

348: \subsection{Reverses for Orthonormal Families of Vectors}

349: 

350: The following result for orthonormal vectors in $H$ holds \cite{RTCSSD1}.

351: 

352: \begin{theorem}[Dragomir, 2004]

353: \label{RTxxt1.3.1}Let $\left\{ e_{1},\dots ,e_{n}\right\} $ be a family of

354: orthonormal vectors in $H,$ $k_{i}\geq 0,$ $i\in \left\{ 1,\dots ,n\right\} $

355: and $f\in L\left( \left[ a,b\right] ;H\right) $ such that%

356: \begin{equation}

357: k_{i}\left\Vert f\left( t\right) \right\Vert \leq \func{Re}\left\langle

358: f\left( t\right) ,e_{i}\right\rangle   \label{RTxx1.3.1}

359: \end{equation}%

360: for each $i\in \left\{ 1,\dots ,n\right\} $ and for a.e. $t\in \left[ a,b%

361: \right] .$

362: 

363: Then%

364: \begin{equation}

365: \left( \sum_{i=1}^{n}k_{i}^{2}\right) ^{\frac{1}{2}}\int_{a}^{b}\left\Vert

366: f\left( t\right) \right\Vert dt\leq \left\Vert \int_{a}^{b}f\left( t\right)

367: dt\right\Vert ,  \label{RTxx1.3.2}

368: \end{equation}%

369: where equality holds if and only if%

370: \begin{equation}

371: \int_{a}^{b}f\left( t\right) dt=\left( \int_{a}^{b}\left\Vert f\left(

372: t\right) \right\Vert dt\right) \sum_{i=1}^{n}k_{i}e_{i}.  \label{RTxx1.3.3}

373: \end{equation}

374: \end{theorem}

375: 

376: \begin{proof}

377: By Bessel's inequality applied for $\int_{a}^{b}f\left( t\right) dt$ and the

378: orthonormal vectors $\left\{ e_{1},\dots ,e_{n}\right\} ,$ we have%

379: \begin{align}

380: \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert ^{2}& \geq

381: \sum_{i=1}^{n}\left\vert \left\langle \int_{a}^{b}f\left( t\right)

382: dt,e_{i}\right\rangle \right\vert ^{2}  \label{RTxx1.3.4} \\

383: & \geq \sum_{i=1}^{n}\left[ \func{Re}\left\langle \int_{a}^{b}f\left(

384: t\right) dt,e_{i}\right\rangle \right] ^{2}  \notag \\

385: & =\sum_{i=1}^{n}\left[ \int_{a}^{b}\func{Re}\left\langle f\left( t\right)

386: ,e_{i}\right\rangle dt\right] ^{2}.  \notag

387: \end{align}%

388: Integrating (\ref{RTxx1.3.1}), we get for each $i\in \left\{ 1,\dots

389: ,n\right\} $%

390: \begin{equation*}

391: 0\leq k_{i}\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\leq

392: \int_{a}^{b}\func{Re}\left\langle f\left( t\right) ,e_{i}\right\rangle dt,

393: \end{equation*}%

394: implying%

395: \begin{equation}

396: \sum_{i=1}^{n}\left[ \int_{a}^{b}\func{Re}\left\langle f\left( t\right)

397: ,e_{i}\right\rangle dt\right] ^{2}\geq \sum_{i=1}^{n}k_{i}^{2}\left(

398: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right) ^{2}.

399: \label{RTxx1.3.5}

400: \end{equation}%

401: On making use of (\ref{RTxx1.3.4}) and (\ref{RTxx1.3.5}), we deduce%

402: \begin{equation*}

403: \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert ^{2}\geq

404: \sum_{i=1}^{n}k_{i}^{2}\left( \int_{a}^{b}\left\Vert f\left( t\right)

405: \right\Vert dt\right) ^{2},

406: \end{equation*}%

407: which is clearly equivalent to (\ref{RTxx1.3.2}).

408: 

409: If (\ref{RTxx1.3.3}) holds true, then%

410: \begin{align*}

411: \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert & =\left(

412: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right) \left\Vert

413: \sum_{i=1}^{n}k_{i}e_{i}\right\Vert \\

414: & =\left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right)

415: \left[ \sum_{i=1}^{n}k_{i}^{2}\left\Vert e_{i}\right\Vert ^{2}\right] ^{%

416: \frac{1}{2}} \\

417: & =\left( \sum_{i=1}^{n}k_{i}^{2}\right) ^{\frac{1}{2}}\int_{a}^{b}\left%

418: \Vert f\left( t\right) \right\Vert dt,

419: \end{align*}%

420: showing that (\ref{RTxx1.3.2}) holds with equality.

421: 

422: Now, suppose that there is an $i_{0}\in \left\{ 1,\dots ,n\right\} $ for

423: which%

424: \begin{equation*}

425: k_{i_{0}}\left\Vert f\left( t\right) \right\Vert <\func{Re}\left\langle

426: f\left( t\right) ,e_{i_{0}}\right\rangle

427: \end{equation*}%

428: on a subset of nonzero Lebesgue measure in $\left[ a,b\right] .$ Then

429: obviously%

430: \begin{equation*}

431: k_{i_{0}}\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt<\int_{a}^{b}%

432: \func{Re}\left\langle f\left( t\right) ,e_{i_{0}}\right\rangle dt,

433: \end{equation*}%

434: and using the argument given above, we deduce%

435: \begin{equation*}

436: \left( \sum_{i=1}^{n}k_{i}^{2}\right) ^{\frac{1}{2}}\int_{a}^{b}\left\Vert

437: f\left( t\right) \right\Vert dt<\left\Vert \int_{a}^{b}f\left( t\right)

438: dt\right\Vert .

439: \end{equation*}%

440: Therefore, if the equality holds in (\ref{RTxx1.3.2}), we must have%

441: \begin{equation}

442: k_{i}\left\Vert f\left( t\right) \right\Vert =\func{Re}\left\langle f\left(

443: t\right) ,e_{i}\right\rangle   \label{RTxx1.3.6}

444: \end{equation}%

445: for each $i\in \left\{ 1,\dots ,n\right\} $ and a.e. $t\in \left[ a,b\right]

446: .$

447: 

448: Also, if the equality holds in (\ref{RTxx1.3.2}), then we must have equality

449: in all inequalities (\ref{RTxx1.3.4}), this means that%

450: \begin{equation}

451: \int_{a}^{b}f\left( t\right) dt=\sum_{i=1}^{n}\left\langle

452: \int_{a}^{b}f\left( t\right) dt,e_{i}\right\rangle e_{i}  \label{RTxx1.3.7}

453: \end{equation}%

454: and

455: \begin{equation}

456: \func{Im}\left\langle \int_{a}^{b}f\left( t\right) dt,e_{i}\right\rangle =0%

457: \text{ \ for each \ }i\in \left\{ 1,\dots ,n\right\} .  \label{RTxx1.3.8}

458: \end{equation}%

459: Using (\ref{RTxx1.3.6}) and (\ref{RTxx1.3.8}) in (\ref{RTxx1.3.7}), we deduce%

460: \begin{align*}

461: \int_{a}^{b}f\left( t\right) dt& =\sum_{i=1}^{n}\func{Re}\left\langle

462: \int_{a}^{b}f\left( t\right) dt,e_{i}\right\rangle e_{i} \\

463: & =\sum_{i=1}^{n}\int_{a}^{b}\func{Re}\left\langle f\left( t\right)

464: ,e_{i}\right\rangle e_{i}dt \\

465: & =\sum_{i=1}^{n}\left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert

466: dt\right) k_{i}e_{i} \\

467: & =\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert

468: dt\sum_{i=1}^{n}k_{i}e_{i},

469: \end{align*}%

470: and the condition (\ref{RTxx1.3.3}) is necessary.

471: 

472: This completes the proof.

473: \end{proof}

474: 

475: The following two corollaries are of interest \cite{RTCSSD1}.

476: 

477: \begin{corollary}

478: \label{RTxxc1.3.2}Let $\left\{ e_{1},\dots ,e_{n}\right\} $ be a family of

479: orthonormal vectors in $H,$ $\rho _{i}\in \left( 0,1\right) ,$ $i\in \left\{

480: 1,\dots ,n\right\} $ and $f\in L\left( \left[ a,b\right] ;H\right) $ such

481: that:%

482: \begin{equation}

483: \left\Vert f\left( t\right) -e_{i}\right\Vert \leq \rho _{i}\text{ \ for \ }%

484: i\in \left\{ 1,\dots ,n\right\} \text{ \ and a.e. }t\in \left[ a,b\right] .

485: \label{RTxx1.3.9}

486: \end{equation}%

487: Then we have the inequality%

488: \begin{equation*}

489: \left( n-\sum_{i=1}^{n}\rho _{i}^{2}\right) ^{\frac{1}{2}}\int_{a}^{b}\left%

490: \Vert f\left( t\right) \right\Vert dt\leq \left\Vert \int_{a}^{b}f\left(

491: t\right) dt\right\Vert ,

492: \end{equation*}%

493: with equality if and only if%

494: \begin{equation*}

495: \int_{a}^{b}f\left( t\right) dt=\int_{a}^{b}\left\Vert f\left( t\right)

496: \right\Vert dt\sum_{i=1}^{n}\left( 1-\rho _{i}^{2}\right) ^{1/2}e_{i}.

497: \end{equation*}

498: \end{corollary}

499: 

500: \begin{proof}

501: From the proof of Theorem \ref{RTxxt1.2.1}, we know that (\ref{RTxx1.3.3})

502: implies the inequality%

503: \begin{equation*}

504: \sqrt{1-\rho _{i}^{2}}\left\Vert f\left( t\right) \right\Vert \leq \func{Re}%

505: \left\langle f\left( t\right) ,e_{i}\right\rangle ,\ \ \ i\in \left\{

506: 1,\dots ,n\right\} ,\text{ \ for a.e. }t\in \left[ a,b\right] .

507: \end{equation*}%

508: Now, applying Theorem \ref{RTxxt1.3.1} for $k_{i}:=\sqrt{1-\rho _{i}^{2}},$ $%

509: i\in \left\{ 1,\dots ,n\right\} $, we deduce the desired result.

510: \end{proof}

511: 

512: A different results is incorporated in (see \cite{RTCSSD1}):

513: 

514: \begin{corollary}

515: \label{RTxxc1.3.3}Let $\left\{ e_{1},\dots ,e_{n}\right\} $ be a family of

516: orthonormal vectors in $H,$ $M_{i}\geq m_{i}>0,$ $i\in \left\{ 1,\dots

517: ,n\right\} $ and $f\in L\left( \left[ a,b\right] ;H\right) $ such that%

518: \begin{equation}

519: \func{Re}\left\langle M_{i}e_{i}-f\left( t\right) ,f\left( t\right)

520: -m_{i}e_{i}\right\rangle \geq 0  \label{RTxx1.3.10}

521: \end{equation}%

522: or, equivalently,%

523: \begin{equation*}

524: \left\Vert f\left( t\right) -\frac{M_{i}+m_{i}}{2}e_{i}\right\Vert \leq

525: \frac{1}{2}\left( M_{i}-m_{i}\right)

526: \end{equation*}%

527: for \ $i\in \left\{ 1,\dots ,n\right\} $ \ and a.e. $t\in \left[ a,b\right]

528: . $ Then we have the reverse of the continuous triangle inequality%

529: \begin{equation*}

530: \left[ \sum_{i=1}^{n}\frac{4m_{i}M_{i}}{\left( m_{i}+M_{i}\right) ^{2}}%

531: \right] ^{\frac{1}{2}}\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert

532: dt\leq \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert ,

533: \end{equation*}%

534: with equality if and only if%

535: \begin{equation*}

536: \int_{a}^{b}f\left( t\right) dt=\int_{a}^{b}\left\Vert f\left( t\right)

537: \right\Vert dt\left( \sum_{i=1}^{n}\frac{2\sqrt{m_{i}M_{i}}}{m_{i}+M_{i}}%

538: e_{i}\right) .

539: \end{equation*}

540: \end{corollary}

541: 

542: \begin{proof}

543: From the proof of Corollary \ref{RTxxc1.3.2}, we know (\ref{RTxx1.3.10})

544: implies that%

545: \begin{equation*}

546: \frac{2\sqrt{m_{i}M_{i}}}{m_{i}+M_{i}}\left\Vert f\left( t\right)

547: \right\Vert \leq \func{Re}\left\langle f\left( t\right) ,e_{i}\right\rangle

548: ,\ \ \ i\in \left\{ 1,\dots ,n\right\} \text{ \ and a.e. }t\in \left[ a,b%

549: \right] .

550: \end{equation*}%

551: Now, applying Theorem \ref{RTxxt1.3.1} for $k_{i}:=\frac{2\sqrt{m_{i}M_{i}}}{%

552: m_{i}+M_{i}},$ $i\in \left\{ 1,\dots ,n\right\} ,$ we deduce the desired

553: result.

554: \end{proof}

555: 

556: \section{Some Additive Reverses}

557: 

558: \subsection{The Case of a Unit Vector}

559: 

560: The following result holds \cite{RTCSSD2}.

561: 

562: \begin{theorem}[Dragomir, 2004]

563: \label{RTxxt2.2.1}If $f\in L\left( \left[ a,b\right] ;H\right) $ is such

564: that there exists a vector $e\in H,$ $\left\Vert e\right\Vert =1$ and $k:%

565: \left[ a,b\right] \rightarrow \lbrack 0,\infty ),$ a Lebesgue integrable

566: function with%

567: \begin{equation}

568: \left\Vert f\left( t\right) \right\Vert -\func{Re}\left\langle f\left(

569: t\right) ,e\right\rangle \leq k\left( t\right) \ \ \ \text{for a.e. }t\in %

570: \left[ a,b\right] ,  \label{RTxx2.2.1}

571: \end{equation}%

572: then we have the inequality:%

573: \begin{equation}

574: \left( 0\leq \right) \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert

575: dt-\left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert \leq

576: \int_{a}^{b}k\left( t\right) dt.  \label{RTxx2.2.2}

577: \end{equation}%

578: The equality holds in (\ref{RTxx2.2.2}) if and only if%

579: \begin{equation}

580: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\geq

581: \int_{a}^{b}k\left( t\right) dt  \label{RTxx2.2.3}

582: \end{equation}%

583: and%

584: \begin{equation}

585: \int_{a}^{b}f\left( t\right) dt=\left( \int_{a}^{b}\left\Vert f\left(

586: t\right) \right\Vert dt-\int_{a}^{b}k\left( t\right) dt\right) e.

587: \label{RTxx2.2.4}

588: \end{equation}

589: \end{theorem}

590: 

591: \begin{proof}

592: If we integrate the inequality (\ref{RTxx2.2.1}), we get%

593: \begin{equation}

594: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\leq \func{Re}%

595: \left\langle \int_{a}^{b}f\left( t\right) dt,e\right\rangle

596: +\int_{a}^{b}k\left( t\right) dt.  \label{RTxx2.2.5}

597: \end{equation}%

598: By Schwarz's inequality for $e$ and $\int_{a}^{b}f\left( t\right) dt,$ we

599: have%

600: \begin{align}

601: & \func{Re}\left\langle \int_{a}^{b}f\left( t\right) dt,e\right\rangle

602: \label{RTxx2.2.6} \\

603: & \leq \left\vert \func{Re}\left\langle \int_{a}^{b}f\left( t\right)

604: dt,e\right\rangle \right\vert \leq \left\vert \left\langle

605: \int_{a}^{b}f\left( t\right) dt,e\right\rangle \right\vert  \notag \\

606: & \leq \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert \left\Vert

607: e\right\Vert =\left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert .  \notag

608: \end{align}%

609: Making use of (\ref{RTxx2.2.5}) and (\ref{RTxx2.2.6}), we deduce the desired

610: inequality (\ref{RTxx2.2.2}).

611: 

612: If (\ref{RTxx2.2.3}) and (\ref{RTxx2.2.4}) hold true, then%

613: \begin{align*}

614: \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert & =\left\vert

615: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt-\int_{a}^{b}k\left(

616: t\right) dt\right\vert \left\Vert e\right\Vert \\

617: & =\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert

618: dt-\int_{a}^{b}k\left( t\right) dt

619: \end{align*}%

620: and the equality holds true in (\ref{RTxx2.2.2}).

621: 

622: Conversely, if the equality holds in (\ref{RTxx2.2.2}), then, obviously (\ref%

623: {RTxx2.2.3}) is valid and we need only to prove (\ref{RTxx2.2.4}).

624: 

625: If $\left\Vert f\left( t\right) \right\Vert -\func{Re}\left\langle f\left(

626: t\right) ,e\right\rangle <k\left( t\right) $ on a subset of nonzero Lebesgue

627: measure in $\left[ a,b\right] ,$ then (\ref{RTxx2.2.5}) holds as a strict

628: inequality, implying that (\ref{RTxx2.2.2}) also holds as a strict

629: inequality. Therefore, if we assume that equality holds in (\ref{RTxx2.2.2}%

630: ), then we must have%

631: \begin{equation}

632: \left\Vert f\left( t\right) \right\Vert =\func{Re}\left\langle f\left(

633: t\right) ,e\right\rangle +k\left( t\right) \ \ \text{for a.e.\ \ }t\in \left[

634: a,b\right] .  \label{RTxx2.2.7}

635: \end{equation}

636: 

637: It is well known that in Schwarz's inequality $\left\Vert x\right\Vert

638: \left\Vert y\right\Vert \geq \func{Re}\left\langle x,y\right\rangle $ the

639: equality holds iff there exists a $\lambda \geq 0$ such that $x=\lambda y.$

640: Therefore, if we assume that the equality holds in all of (\ref{RTxx2.2.6}),

641: then there exists a $\lambda \geq 0$ such that%

642: \begin{equation}

643: \int_{a}^{b}f\left( t\right) dt=\lambda e.  \label{RTxx2.2.8}

644: \end{equation}%

645: Integrating (\ref{RTxx2.2.7}) on $\left[ a,b\right] ,$ we deduce%

646: \begin{equation*}

647: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt=\func{Re}\left\langle

648: \int_{a}^{b}f\left( t\right) dt,e\right\rangle +\int_{a}^{b}k\left( t\right)

649: dt,

650: \end{equation*}%

651: and thus, by (\ref{RTxx2.2.8}), we get%

652: \begin{equation*}

653: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt=\lambda \left\Vert

654: e\right\Vert ^{2}+\int_{a}^{b}k\left( t\right) dt,

655: \end{equation*}%

656: giving $\lambda =\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert

657: dt-\int_{a}^{b}k\left( t\right) dt.$

658: 

659: Using (\ref{RTxx2.2.8}), we deduce (\ref{RTxx2.2.4}) and the theorem is

660: completely proved.

661: \end{proof}

662: 

663: The following corollary may be useful for applications \cite{RTCSSD2}.

664: 

665: \begin{corollary}

666: \label{RTxxc2.2.2}If $f\in L\left( \left[ a,b\right] ;H\right) $ is such

667: that there exists a vector $e\in H,$ $\left\Vert e\right\Vert =1$ and $\rho

668: \in \left( 0,1\right) $ such that

669: \begin{equation}

670: \left\Vert f\left( t\right) -e\right\Vert \leq \rho \ \ \ \text{for a.e. }%

671: t\in \left[ a,b\right] ,  \label{RTxx2.2.9}

672: \end{equation}%

673: then we have the inequality%

674: \begin{align}

675: (0& \leq )\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt-\left\Vert

676: \int_{a}^{b}f\left( t\right) dt\right\Vert  \label{RTxx2.2.10} \\

677: & \leq \frac{\rho ^{2}}{\sqrt{1-\rho ^{2}}\left( 1+\sqrt{1-\rho ^{2}}\right)

678: }\func{Re}\left\langle \int_{a}^{b}f\left( t\right) dt,e\right\rangle  \notag

679: \\

680: & \left( \leq \frac{\rho ^{2}}{\sqrt{1-\rho ^{2}}\left( 1+\sqrt{1-\rho ^{2}}%

681: \right) }\left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert \right) .

682: \notag

683: \end{align}%

684: The equality holds in (\ref{RTxx2.2.10}) if and only if%

685: \begin{equation}

686: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\geq \frac{\rho ^{2}}{%

687: \sqrt{1-\rho ^{2}}\left( 1+\sqrt{1-\rho ^{2}}\right) }\func{Re}\left\langle

688: \int_{a}^{b}f\left( t\right) dt,e\right\rangle  \label{RTxx2.2.11}

689: \end{equation}%

690: and%

691: \begin{multline}

692: \int_{a}^{b}f\left( t\right) dt  \label{RTxx2.2.12} \\

693: =\left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt-\frac{\rho

694: ^{2}}{\sqrt{1-\rho ^{2}}\left( 1+\sqrt{1-\rho ^{2}}\right) }\func{Re}%

695: \left\langle \int_{a}^{b}f\left( t\right) dt,e\right\rangle \right) e.

696: \end{multline}

697: \end{corollary}

698: 

699: \begin{proof}

700: Firstly, note that (\ref{RTxx2.2.3}) is equivalent to%

701: \begin{equation*}

702: \left\Vert f\left( t\right) \right\Vert ^{2}+1-\rho ^{2}\leq 2\func{Re}%

703: \left\langle f\left( t\right) ,e\right\rangle ,

704: \end{equation*}%

705: giving%

706: \begin{equation*}

707: \frac{\left\Vert f\left( t\right) \right\Vert ^{2}}{\sqrt{1-\rho ^{2}}}+%

708: \sqrt{1-\rho ^{2}}\leq \frac{2\func{Re}\left\langle f\left( t\right)

709: ,e\right\rangle }{\sqrt{1-\rho ^{2}}}

710: \end{equation*}%

711: for a.e. $t\in \left[ a,b\right] .$

712: 

713: Since, obviously%

714: \begin{equation*}

715: 2\left\Vert f\left( t\right) \right\Vert \leq \frac{\left\Vert f\left(

716: t\right) \right\Vert ^{2}}{\sqrt{1-\rho ^{2}}}+\sqrt{1-\rho ^{2}}

717: \end{equation*}%

718: for any $t\in \left[ a,b\right] ,$ then we deduce the inequality%

719: \begin{equation*}

720: \left\Vert f\left( t\right) \right\Vert \leq \frac{\func{Re}\left\langle

721: f\left( t\right) ,e\right\rangle }{\sqrt{1-\rho ^{2}}}\ \ \ \text{for a.e. }%

722: t\in \left[ a,b\right] ,

723: \end{equation*}%

724: which is clearly equivalent to%

725: \begin{equation*}

726: \left\Vert f\left( t\right) \right\Vert -\func{Re}\left\langle f\left(

727: t\right) ,e\right\rangle \leq \frac{\rho ^{2}}{\sqrt{1-\rho ^{2}}\left( 1+%

728: \sqrt{1-\rho ^{2}}\right) }\func{Re}\left\langle f\left( t\right)

729: ,e\right\rangle

730: \end{equation*}%

731: for a.e. $t\in \left[ a,b\right] .$

732: 

733: Applying Theorem \ref{RTxxt2.2.1} for $k\left( t\right) :=\frac{\rho ^{2}}{%

734: \sqrt{1-\rho ^{2}}\left( 1+\sqrt{1-\rho ^{2}}\right) }\func{Re}\left\langle

735: f\left( t\right) ,e\right\rangle ,$ we deduce the desired result.

736: \end{proof}

737: 

738: In the same spirit, we also have the following corollary \cite{RTCSSD2}.

739: 

740: \begin{corollary}

741: \label{RTxxc2.2.3}If $f\in L\left( \left[ a,b\right] ;H\right) $ is such

742: that there exists a vector $e\in H,$ $\left\Vert e\right\Vert =1$ and $M\geq

743: m>0$ such that either%

744: \begin{equation}

745: \func{Re}\left\langle Me-f\left( t\right) ,f\left( t\right) -me\right\rangle

746: \geq 0\ \ \   \label{RTxx2.2.13}

747: \end{equation}%

748: or, equivalently,%

749: \begin{equation}

750: \left\Vert f\left( t\right) -\frac{M+m}{2}e\right\Vert \leq \frac{1}{2}%

751: \left( M-m\right) \ \   \label{RTxx2.2.14}

752: \end{equation}%

753: $\ $for a.e. $t\in \left[ a,b\right] ,$ then we have the inequality%

754: \begin{align}

755: (0& \leq )\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt-\left\Vert

756: \int_{a}^{b}f\left( t\right) dt\right\Vert  \label{RTxx2.2.15} \\

757: & \leq \frac{\left( \sqrt{M}-\sqrt{m}\right) ^{2}}{2\sqrt{mM}}\func{Re}%

758: \left\langle \int_{a}^{b}f\left( t\right) dt,e\right\rangle  \notag \\

759: & \left( \leq \frac{\left( \sqrt{M}-\sqrt{m}\right) ^{2}}{2\sqrt{mM}}%

760: \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert \right) .  \notag

761: \end{align}%

762: The equality holds in (\ref{RTxx2.2.15}) if and only if%

763: \begin{equation*}

764: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\geq \frac{\left(

765: \sqrt{M}-\sqrt{m}\right) ^{2}}{2\sqrt{mM}}\func{Re}\left\langle

766: \int_{a}^{b}f\left( t\right) dt,e\right\rangle

767: \end{equation*}%

768: and%

769: \begin{equation*}

770: \int_{a}^{b}f\left( t\right) dt=\left( \int_{a}^{b}\left\Vert f\left(

771: t\right) \right\Vert dt-\frac{\left( \sqrt{M}-\sqrt{m}\right) ^{2}}{2\sqrt{mM%

772: }}\func{Re}\left\langle \int_{a}^{b}f\left( t\right) dt,e\right\rangle

773: \right) e.

774: \end{equation*}

775: \end{corollary}

776: 

777: \begin{proof}

778: Observe that (\ref{RTxx2.2.13}) is clearly equivalent to%

779: \begin{equation*}

780: \left\Vert f\left( t\right) \right\Vert ^{2}+mM\leq \left( M+m\right) \func{%

781: Re}\left\langle f\left( t\right) ,e\right\rangle

782: \end{equation*}%

783: for a.e. $t\in \left[ a,b\right] ,$ giving the inequality%

784: \begin{equation*}

785: \frac{\left\Vert f\left( t\right) \right\Vert ^{2}}{\sqrt{mM}}+\sqrt{mM}\leq

786: \frac{M+m}{\sqrt{mM}}\func{Re}\left\langle f\left( t\right) ,e\right\rangle

787: \end{equation*}%

788: for a.e. $t\in \left[ a,b\right] .$

789: 

790: Since, obviously,%

791: \begin{equation*}

792: 2\left\Vert f\left( t\right) \right\Vert \leq \frac{\left\Vert f\left(

793: t\right) \right\Vert ^{2}}{\sqrt{mM}}+\sqrt{mM}

794: \end{equation*}%

795: for any $t\in \left[ a,b\right] ,$ hence we deduce the inequality%

796: \begin{equation*}

797: \left\Vert f\left( t\right) \right\Vert \leq \frac{M+m}{\sqrt{mM}}\func{Re}%

798: \left\langle f\left( t\right) ,e\right\rangle \ \ \text{for a.e. }t\in \left[

799: a,b\right] ,

800: \end{equation*}%

801: which is clearly equivalent to%

802: \begin{equation*}

803: \left\Vert f\left( t\right) \right\Vert -\func{Re}\left\langle f\left(

804: t\right) ,e\right\rangle \leq \frac{\left( \sqrt{M}-\sqrt{m}\right) ^{2}}{2%

805: \sqrt{mM}}\func{Re}\left\langle f\left( t\right) ,e\right\rangle

806: \end{equation*}%

807: for a.e. $t\in \left[ a,b\right] .$

808: 

809: Finally, applying Theorem \ref{RTxxt2.2.1}, we obtain the desired result.

810: \end{proof}

811: 

812: We can state now (see also \cite{RTCSSD2}):

813: 

814: \begin{corollary}

815: \label{RTxxc2.2.4}If $f\in L\left( \left[ a,b\right] ;H\right) $ and $r\in

816: L_{2}\left( \left[ a,b\right] ;H\right) ,$ $e\in H,$ $\left\Vert

817: e\right\Vert =1$ are such that

818: \begin{equation}

819: \left\Vert f\left( t\right) -e\right\Vert \leq r\left( t\right) \ \ \ \text{%

820: for a.e. }t\in \left[ a,b\right] ,  \label{RTxx2.2.16}

821: \end{equation}%

822: then we have the inequality%

823: \begin{equation}

824: \left( 0\leq \right) \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert

825: dt-\left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert \leq \frac{1}{2}%

826: \int_{a}^{b}r^{2}\left( t\right) dt.  \label{RTxx2.2.17}

827: \end{equation}%

828: The equality holds in (\ref{RTxx2.2.17}) if and only if%

829: \begin{equation*}

830: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\geq \frac{1}{2}%

831: \int_{a}^{b}r^{2}\left( t\right) dt

832: \end{equation*}%

833: and%

834: \begin{equation*}

835: \int_{a}^{b}f\left( t\right) dt=\left( \int_{a}^{b}\left\Vert f\left(

836: t\right) \right\Vert dt-\frac{1}{2}\int_{a}^{b}r^{2}\left( t\right)

837: dt\right) e.

838: \end{equation*}

839: \end{corollary}

840: 

841: \begin{proof}

842: The condition (\ref{RTxx2.2.16}) is obviously equivalent to%

843: \begin{equation*}

844: \left\Vert f\left( t\right) \right\Vert ^{2}+1\leq 2\func{Re}\left\langle

845: f\left( t\right) ,e\right\rangle +r^{2}\left( t\right)

846: \end{equation*}%

847: for a.e. $t\in \left[ a,b\right] .$

848: 

849: Using the elementary inequality%

850: \begin{equation*}

851: 2\left\Vert f\left( t\right) \right\Vert \leq \left\Vert f\left( t\right)

852: \right\Vert ^{2}+1,\ \ t\in \left[ a,b\right] ,

853: \end{equation*}%

854: we deduce%

855: \begin{equation*}

856: \left\Vert f\left( t\right) \right\Vert -\func{Re}\left\langle f\left(

857: t\right) ,e\right\rangle \leq \frac{1}{2}r^{2}\left( t\right)

858: \end{equation*}%

859: for a.e. $t\in \left[ a,b\right] .$

860: 

861: Applying Theorem \ref{RTxxt2.2.1} for $k\left( t\right) :=\frac{1}{2}%

862: r^{2}\left( t\right) ,$ $t\in \left[ a,b\right] $, we deduce the desired

863: result.

864: \end{proof}

865: 

866: Finally, we may state and prove the following result as well \cite{RTCSSD2}.

867: 

868: \begin{corollary}

869: \label{RTxxc2.2.5}If $f\in L\left( \left[ a,b\right] ;H\right) $, $e\in H,$ $%

870: \left\Vert e\right\Vert =1$ and $M,m:\left[ a,b\right] \rightarrow \lbrack

871: 0,\infty )$ with $M\geq m$ a.e. on $\left[ a,b\right] ,$ are such that $%

872: \frac{\left( M-m\right) ^{2}}{M+m}\in L\left[ a,b\right] $ and either%

873: \begin{equation}

874: \left\Vert f\left( t\right) -\frac{M\left( t\right) +m\left( t\right) }{2}%

875: e\right\Vert \leq \frac{1}{2}\left[ M\left( t\right) -m\left( t\right) %

876: \right]  \label{RTxx2.2.18}

877: \end{equation}%

878: or, equivalently,%

879: \begin{equation}

880: \func{Re}\left\langle M\left( t\right) e-f\left( t\right) ,f\left( t\right)

881: -m\left( t\right) e\right\rangle \geq 0\ \ \   \label{RTxx2.2.19}

882: \end{equation}%

883: for a.e. $t\in \left[ a,b\right] ,$ then we have the inequality%

884: \begin{equation}

885: \left( 0\leq \right) \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert

886: dt-\left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert \leq \frac{1}{4}%

887: \int_{a}^{b}\frac{\left[ M\left( t\right) -m\left( t\right) \right] ^{2}}{%

888: M\left( t\right) +m\left( t\right) }dt.  \label{RTxx2.2.20}

889: \end{equation}%

890: The equality holds in (\ref{RTxx2.2.20}) if and only if%

891: \begin{equation*}

892: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\geq \frac{1}{4}%

893: \int_{a}^{b}\frac{\left[ M\left( t\right) -m\left( t\right) \right] ^{2}}{%

894: M\left( t\right) +m\left( t\right) }dt

895: \end{equation*}%

896: and%

897: \begin{equation*}

898: \int_{a}^{b}f\left( t\right) dt=\left( \int_{a}^{b}\left\Vert f\left(

899: t\right) \right\Vert dt-\frac{1}{4}\int_{a}^{b}\frac{\left[ M\left( t\right)

900: -m\left( t\right) \right] ^{2}}{M\left( t\right) +m\left( t\right) }%

901: dt\right) e.

902: \end{equation*}

903: \end{corollary}

904: 

905: \begin{proof}

906: The condition (\ref{RTxx2.2.18}) is equivalent to%

907: \begin{multline*}

908: \left\Vert f\left( t\right) \right\Vert ^{2}+\left( \frac{M\left( t\right)

909: +m\left( t\right) }{2}\right) ^{2} \\

910: \leq 2\left( \frac{M\left( t\right) +m\left( t\right) }{2}\right) \func{Re}%

911: \left\langle f\left( t\right) ,e\right\rangle +\frac{1}{4}\left[ M\left(

912: t\right) -m\left( t\right) \right] ^{2}

913: \end{multline*}%

914: for a.e. $t\in \left[ a,b\right] ,$ and since%

915: \begin{equation*}

916: 2\left( \frac{M\left( t\right) +m\left( t\right) }{2}\right) \left\Vert

917: f\left( t\right) \right\Vert \leq \left\Vert f\left( t\right) \right\Vert

918: ^{2}+\left( \frac{M\left( t\right) +m\left( t\right) }{2}\right) ^{2},\ \ \

919: t\in \left[ a,b\right]

920: \end{equation*}%

921: hence%

922: \begin{equation*}

923: \left\Vert f\left( t\right) \right\Vert -\func{Re}\left\langle f\left(

924: t\right) ,e\right\rangle \leq \frac{1}{4}\frac{\left[ M\left( t\right)

925: -m\left( t\right) \right] ^{2}}{M\left( t\right) +m\left( t\right) }

926: \end{equation*}%

927: for a.e. $t\in \left[ a,b\right] .$

928: 

929: Now, applying Theorem \ref{RTxxt2.2.1} for $k\left( t\right) :=\frac{1}{4}%

930: \frac{\left[ M\left( t\right) -m\left( t\right) \right] ^{2}}{M\left(

931: t\right) +m\left( t\right) },$ $t\in \left[ a,b\right] $, we deduce the

932: desired inequality.

933: \end{proof}

934: 

935: \subsection{Additive Reverses for Orthonormal Families}

936: 

937: The following reverse of the continuous triangle inequality for vector

938: valued integrals holds \cite{RTCSSD2}.

939: 

940: \begin{theorem}[Dragomir, 2004]

941: \label{RTxxt2.3.1}Let $f\in L\left( \left[ a,b\right] ;H\right) ,$ where $H$

942: is a Hilbert space over the real or complex number field $\mathbb{K}$, $%

943: \left\{ e_{i}\right\} _{i\in \left\{ 1,\dots ,n\right\} }$ an orthonormal

944: family in $H$ and $M_{i}\in L\left[ a,b\right] ,$ $i\in \left\{ 1,\dots

945: ,n\right\} .$ If we assume that%

946: \begin{equation}

947: \left\Vert f\left( t\right) \right\Vert -\func{Re}\left\langle f\left(

948: t\right) ,e_{i}\right\rangle \leq M_{i}\left( t\right) \ \ \text{for a.e. }%

949: t\in \left[ a,b\right] ,  \label{RTxx2.3.1}

950: \end{equation}%

951: then we have the inequality%

952: \begin{equation}

953: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\leq \frac{1}{\sqrt{n}}%

954: \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert +\frac{1}{n}%

955: \sum_{i=1}^{n}\int_{a}^{b}M_{i}\left( t\right) dt.  \label{RTxx2.3.2}

956: \end{equation}%

957: The equality holds in (\ref{RTxx2.3.2}) if and only if%

958: \begin{equation}

959: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\geq \frac{1}{n}%

960: \sum_{i=1}^{n}\int_{a}^{b}M_{i}\left( t\right) dt  \label{RTxx2.3.3}

961: \end{equation}%

962: and

963: \begin{equation}

964: \int_{a}^{b}f\left( t\right) dt=\left( \int_{a}^{b}\left\Vert f\left(

965: t\right) \right\Vert dt-\frac{1}{n}\sum_{i=1}^{n}\int_{a}^{b}M_{i}\left(

966: t\right) dt\right) \sum_{i=1}^{n}e_{i}.  \label{RTxx2.3.4}

967: \end{equation}

968: \end{theorem}

969: 

970: \begin{proof}

971: If we integrate the inequality (\ref{RTxx2.3.1}) on $\left[ a,b\right] ,$ we

972: get%

973: \begin{equation*}

974: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\leq \func{Re}%

975: \left\langle \int_{a}^{b}f\left( t\right) dt,e_{i}\right\rangle

976: +\int_{a}^{b}M_{i}\left( t\right) dt

977: \end{equation*}%

978: for each $i\in \left\{ 1,\dots ,n\right\} .$ Summing these inequalities over

979: $i$ from $1$ to $n,$ we deduce%

980: \begin{equation}

981: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\leq \frac{1}{n}\func{%

982: Re}\left\langle \int_{a}^{b}f\left( t\right)

983: dt,\sum_{i=1}^{n}e_{i}\right\rangle +\frac{1}{n}\sum_{i=1}^{n}%

984: \int_{a}^{b}M_{i}\left( t\right) dt.  \label{RTxx2.3.5}

985: \end{equation}%

986: By Schwarz's inequality for $\int_{a}^{b}f\left( t\right) dt$ and $%

987: \sum_{i=1}^{n}e_{i},$ we have%

988: \begin{align}

989: & \func{Re}\left\langle \int_{a}^{b}f\left( t\right)

990: dt,\sum_{i=1}^{n}e_{i}\right\rangle  \label{RTxx2.3.6} \\

991: & \leq \left\vert \func{Re}\left\langle \int_{a}^{b}f\left( t\right)

992: dt,\sum_{i=1}^{n}e_{i}\right\rangle \right\vert \leq \left\vert \left\langle

993: \int_{a}^{b}f\left( t\right) dt,\sum_{i=1}^{n}e_{i}\right\rangle \right\vert

994: \notag \\

995: & \leq \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert \left\Vert

996: \sum_{i=1}^{n}e_{i}\right\Vert =\sqrt{n}\left\Vert \int_{a}^{b}f\left(

997: t\right) dt\right\Vert ,  \notag

998: \end{align}%

999: since%

1000: \begin{equation*}

1001: \left\Vert \sum_{i=1}^{n}e_{i}\right\Vert =\sqrt{\left\Vert

1002: \sum_{i=1}^{n}e_{i}\right\Vert ^{2}}=\sqrt{\sum_{i=1}^{n}\left\Vert

1003: e_{i}\right\Vert ^{2}}=\sqrt{n}.

1004: \end{equation*}%

1005: Making use of (\ref{RTxx2.3.5}) and (\ref{RTxx2.3.6}), we deduce the desired

1006: inequality (\ref{RTxx2.3.2}).

1007: 

1008: If (\ref{RTxx2.3.3}) and (\ref{RTxx2.3.4}) hold, then

1009: \begin{align*}

1010: \frac{1}{\sqrt{n}}\left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert & =%

1011: \frac{1}{\sqrt{n}}\left\vert \int_{a}^{b}\left\Vert f\left( t\right)

1012: \right\Vert dt-\frac{1}{n}\sum_{i=1}^{n}\int_{a}^{b}M_{i}\left( t\right)

1013: dt\right\vert \left\Vert \sum_{i=1}^{n}e_{i}\right\Vert \\

1014: & =\left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt-\frac{1}{n}%

1015: \sum_{i=1}^{n}\int_{a}^{b}M_{i}\left( t\right) dt\right)

1016: \end{align*}%

1017: and the equality in (\ref{RTxx2.3.2}) holds true.

1018: 

1019: Conversely, if the equality holds in (\ref{RTxx2.3.2}), then, obviously, (%

1020: \ref{RTxx2.3.3}) is valid.

1021: 

1022: Taking into account the argument presented above for the previous result (%

1023: \ref{RTxx2.3.2}), it is obvious that, if the equality holds in (\ref%

1024: {RTxx2.3.2}), then it must hold in (\ref{RTxx2.3.1}) for a.e. $t\in \left[

1025: a,b\right] $ and for each $i\in \left\{ 1,\dots ,n\right\} $ and also the

1026: equality must hold in any of the inequalities in (\ref{RTxx2.3.6}).

1027: 

1028: It is well known that in Schwarz's inequality $\func{Re}\left\langle

1029: u,v\right\rangle \leq \left\Vert u\right\Vert \left\Vert v\right\Vert ,$ the

1030: equality occurs if and only if $u=\lambda v$ with $\lambda \geq 0,$

1031: consequently, the equality holds in all inequalities from (\ref{RTxx2.3.6})

1032: simultaneously iff there exists a $\mu \geq 0$ with%

1033: \begin{equation}

1034: \mu \sum_{i=1}^{n}e_{i}=\int_{a}^{b}f\left( t\right) dt.  \label{RTxx2.3.7}

1035: \end{equation}%

1036: If we integrate the equality in (\ref{RTxx2.3.1}) and sum over $i,$ we deduce%

1037: \begin{equation}

1038: n\int_{a}^{b}f\left( t\right) dt=\func{Re}\left\langle \int_{a}^{b}f\left(

1039: t\right) dt,\sum_{i=1}^{n}e_{i}\right\rangle

1040: +\sum_{i=1}^{n}\int_{a}^{b}M_{i}\left( t\right) dt.  \label{RTxx2.3.8}

1041: \end{equation}%

1042: Replacing $\int_{a}^{b}f\left( t\right) dt$ from (\ref{RTxx2.3.7}) into (\ref%

1043: {RTxx2.3.8}), we deduce%

1044: \begin{align}

1045: n\int_{a}^{b}f\left( t\right) dt& =\mu \left\Vert

1046: \sum_{i=1}^{n}e_{i}\right\Vert ^{2}+\sum_{i=1}^{n}\int_{a}^{b}M_{i}\left(

1047: t\right) dt  \label{RTxx2.3.9} \\

1048: & =\mu n+\sum_{i=1}^{n}\int_{a}^{b}M_{i}\left( t\right) dt.  \notag

1049: \end{align}

1050: 

1051: Finally, we note that (\ref{RTxx2.3.7}) and (\ref{RTxx2.3.9}) will produce

1052: the required identity (\ref{RTxx2.3.4}), and the proof is complete.

1053: \end{proof}

1054: 

1055: The following corollaries may be of interest for applications \cite{RTCSSD2}.

1056: 

1057: \begin{corollary}

1058: \label{RTxxc2.3.2}Let $f\in L\left( \left[ a,b\right] ;H\right) ,$ $\left\{

1059: e_{i}\right\} _{i\in \left\{ 1,\dots ,n\right\} }$ an orthonormal family in $%

1060: H$ and $\rho _{i}\in \left( 0,1\right) ,$ $i\in \left\{ 1,\dots ,n\right\} $

1061: such that%

1062: \begin{equation}

1063: \left\Vert f\left( t\right) -e_{i}\right\Vert \leq \rho _{i}\ \ \text{for

1064: a.e. }t\in \left[ a,b\right] .  \label{RTxx2.3.10}

1065: \end{equation}%

1066: Then we have the inequalities:%

1067: \begin{align}

1068: & \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\leq \frac{1}{\sqrt{n%

1069: }}\left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert  \label{RTxx2.3.11a}

1070: \\

1071: & +\func{Re}\left\langle \int_{a}^{b}f\left( t\right) dt,\frac{1}{n}%

1072: \sum_{i=1}^{n}\frac{\rho _{i}^{2}}{\sqrt{1-\rho _{i}^{2}}\left( 1+\sqrt{%

1073: 1-\rho _{i}^{2}}\right) }e_{i}\right\rangle  \notag \\

1074: & \leq \frac{1}{\sqrt{n}}\left\Vert \int_{a}^{b}f\left( t\right)

1075: dt\right\Vert  \notag \\

1076: & \times \left[ 1+\left( \frac{1}{n}\sum_{i=1}^{n}\frac{\rho _{i}^{2}}{\sqrt{%

1077: 1-\rho _{i}^{2}}\left( 1+\sqrt{1-\rho _{i}^{2}}\right) }\right) ^{\frac{1}{2}%

1078: }\right] .  \notag

1079: \end{align}%

1080: The equality holds in the first inequality in (\ref{RTxx2.3.11a}) if and

1081: only if%

1082: \begin{equation*}

1083: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\geq \func{Re}%

1084: \left\langle \int_{a}^{b}f\left( t\right) dt,\frac{1}{n}\sum_{i=1}^{n}\frac{%

1085: \rho _{i}^{2}}{\sqrt{1-\rho _{i}^{2}}\left( 1+\sqrt{1-\rho _{i}^{2}}\right) }%

1086: e_{i}\right\rangle

1087: \end{equation*}%

1088: and%

1089: \begin{multline*}

1090: \int_{a}^{b}f\left( t\right) dt \\

1091: =\left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt-\func{Re}%

1092: \left\langle \int_{a}^{b}f\left( t\right) dt,\frac{1}{n}\sum_{i=1}^{n}\frac{%

1093: \rho _{i}^{2}}{\sqrt{1-\rho _{i}^{2}}\left( 1+\sqrt{1-\rho _{i}^{2}}\right) }%

1094: e_{i}\right\rangle \right) \\

1095: \times \sum_{i=1}^{n}e_{i}.

1096: \end{multline*}

1097: \end{corollary}

1098: 

1099: \begin{proof}

1100: As in the proof of Corollary \ref{RTxxc2.2.2}, the assumption (\ref%

1101: {RTxx2.3.10}) implies%

1102: \begin{equation*}

1103: \left\Vert f\left( t\right) \right\Vert -\func{Re}\left\langle f\left(

1104: t\right) ,e_{i}\right\rangle \leq \frac{\rho _{i}^{2}}{\sqrt{1-\rho _{i}^{2}}%

1105: \left( \sqrt{1-\rho _{i}^{2}}+1\right) }\func{Re}\left\langle f\left(

1106: t\right) ,e_{i}\right\rangle

1107: \end{equation*}%

1108: for a.e. $t\in \left[ a,b\right] $ and for each $i\in \left\{ 1,\dots

1109: ,n\right\} .$

1110: 

1111: Now, if we apply Theorem \ref{RTxxt2.3.1} for

1112: \begin{equation*}

1113: M_{i}\left( t\right) :=\frac{\rho _{i}^{2}\func{Re}\left\langle f\left(

1114: t\right) ,e_{i}\right\rangle }{\sqrt{1-\rho _{i}^{2}}\left( \sqrt{1-\rho

1115: _{i}^{2}}+1\right) },\ \ i\in \left\{ 1,\dots ,n\right\} ,\ \ t\in \left[ a,b%

1116: \right] ,

1117: \end{equation*}%

1118: we deduce the first inequality in (\ref{RTxx2.3.11a}).

1119: 

1120: By Schwarz's inequality in $H,$ we have%

1121: \begin{align*}

1122: & \func{Re}\left\langle \int_{a}^{b}f\left( t\right) dt,\frac{1}{n}%

1123: \sum_{i=1}^{n}\frac{\rho _{i}^{2}}{\sqrt{1-\rho _{i}^{2}}\left( 1+\sqrt{%

1124: 1-\rho _{i}^{2}}\right) }e_{i}\right\rangle \\

1125: & \leq \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert \left\Vert

1126: \frac{1}{n}\sum_{i=1}^{n}\frac{\rho _{i}^{2}}{\sqrt{1-\rho _{i}^{2}}\left( 1+%

1127: \sqrt{1-\rho _{i}^{2}}\right) }e_{i}\right\Vert \\

1128: & =\frac{1}{n}\left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert \left(

1129: \sum_{i=1}^{n}\left[ \frac{\rho _{i}^{2}}{\sqrt{1-\rho _{i}^{2}}\left( 1+%

1130: \sqrt{1-\rho _{i}^{2}}\right) }\right] ^{2}\right) ^{\frac{1}{2}},

1131: \end{align*}%

1132: which implies the second inequality in (\ref{RTxx2.3.11a}).

1133: \end{proof}

1134: 

1135: The second result is incorporated in \cite{RTCSSD2}:

1136: 

1137: \begin{corollary}

1138: \label{RTxxc2.3.3}Let $f\in L\left( \left[ a,b\right] ;H\right) ,$ $\left\{

1139: e_{i}\right\} _{i\in \left\{ 1,\dots ,n\right\} }$ an orthonormal family in $%

1140: H$ and $M_{i}\geq m_{i}>0$ such that either%

1141: \begin{equation}

1142: \func{Re}\left\langle M_{i}e_{i}-f\left( t\right) ,f\left( t\right)

1143: -m_{i}e_{i}\right\rangle \geq 0\   \label{RTxx2.3.12a}

1144: \end{equation}%

1145: or, equivalently,%

1146: \begin{equation*}

1147: \left\Vert f\left( t\right) -\frac{M_{i}+m_{i}}{2}\cdot e_{i}\right\Vert

1148: \leq \frac{1}{2}\left( M_{i}-m_{i}\right) \

1149: \end{equation*}%

1150: for a.e. $t\in \left[ a,b\right] $ \ and each \ $i\in \left\{ 1,\dots

1151: ,n\right\} .$Then we have

1152: \begin{align}

1153: & \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\leq \frac{1}{\sqrt{n%

1154: }}\left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert  \label{RTxx2.3.13}

1155: \\

1156: & +\func{Re}\left\langle \int_{a}^{b}f\left( t\right) dt,\frac{1}{n}%

1157: \sum_{i=1}^{n}\frac{\left( \sqrt{M_{i}}-\sqrt{m_{i}}\right) ^{2}}{2\sqrt{%

1158: m_{i}M_{i}}}e_{i}\right\rangle  \notag \\

1159: & \leq \frac{1}{\sqrt{n}}\left\Vert \int_{a}^{b}f\left( t\right)

1160: dt\right\Vert \left[ 1+\left( \frac{1}{n}\sum_{i=1}^{n}\frac{\left( \sqrt{%

1161: M_{i}}-\sqrt{m_{i}}\right) ^{4}}{4m_{i}M_{i}}\right) ^{\frac{1}{2}}\right] .

1162: \notag

1163: \end{align}%

1164: The equality holds in the first inequality in (\ref{RTxx2.3.13}) if and only

1165: if%

1166: \begin{equation*}

1167: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\geq \func{Re}%

1168: \left\langle \int_{a}^{b}f\left( t\right) dt,\frac{1}{n}\sum_{i=1}^{n}\frac{%

1169: \left( \sqrt{M_{i}}-\sqrt{m_{i}}\right) ^{2}}{2\sqrt{m_{i}M_{i}}}%

1170: e_{i}\right\rangle

1171: \end{equation*}%

1172: and%

1173: \begin{multline*}

1174: \int_{a}^{b}f\left( t\right) dt \\

1175: =\left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt-\func{Re}%

1176: \left\langle \int_{a}^{b}f\left( t\right) dt,\frac{1}{n}\sum_{i=1}^{n}\frac{%

1177: \left( \sqrt{M_{i}}-\sqrt{m_{i}}\right) ^{2}}{2\sqrt{m_{i}M_{i}}}%

1178: e_{i}\right\rangle \right) \\

1179: \times \sum_{i=1}^{n}e_{i}.

1180: \end{multline*}

1181: \end{corollary}

1182: 

1183: \begin{proof}

1184: As in the proof of Corollary \ref{RTxxc2.2.3}, from (\ref{RTxx2.3.12a}), we

1185: have%

1186: \begin{equation*}

1187: \left\Vert f\left( t\right) \right\Vert -\func{Re}\left\langle f\left(

1188: t\right) ,e_{i}\right\rangle \leq \frac{\left( \sqrt{M_{i}}-\sqrt{m_{i}}%

1189: \right) ^{2}}{2\sqrt{m_{i}M_{i}}}\func{Re}\left\langle f\left( t\right)

1190: ,e_{i}\right\rangle

1191: \end{equation*}%

1192: for a.e. $t\in \left[ a,b\right] $ and $i\in \left\{ 1,\dots ,n\right\} .$

1193: 

1194: Applying Theorem \ref{RTxxt2.3.1} for%

1195: \begin{equation*}

1196: M_{i}\left( t\right) :=\frac{\left( \sqrt{M_{i}}-\sqrt{m_{i}}\right) ^{2}}{2%

1197: \sqrt{m_{i}M_{i}}}\func{Re}\left\langle f\left( t\right) ,e_{i}\right\rangle

1198: ,\ \ \text{ }t\in \left[ a,b\right] ,\ i\in \left\{ 1,\dots ,n\right\} ,

1199: \end{equation*}%

1200: we deduce the desired result.

1201: \end{proof}

1202: 

1203: In a different direction, we may state the following result as well \cite%

1204: {RTCSSD2}.

1205: 

1206: \begin{corollary}

1207: \label{RTxxc2.3.4}Let $f\in L\left( \left[ a,b\right] ;H\right) ,$ $\left\{

1208: e_{i}\right\} _{i\in \left\{ 1,\dots ,n\right\} }$ an orthonormal family in $%

1209: H$ and $r_{i}\in L^{2}\left( \left[ a,b\right] \right) ,$ $i\in \left\{

1210: 1,\dots ,n\right\} $ such that%

1211: \begin{equation*}

1212: \left\Vert f\left( t\right) -e_{i}\right\Vert \leq r_{i}\left( t\right) \ \

1213: \ \text{for a.e. }t\in \left[ a,b\right] \text{ \ and \ }i\in \left\{

1214: 1,\dots ,n\right\} .

1215: \end{equation*}%

1216: Then we have the inequality%

1217: \begin{equation}

1218: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\leq \frac{1}{\sqrt{n}}%

1219: \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert +\frac{1}{2n}%

1220: \sum_{i=1}^{n}\left( \int_{a}^{b}r_{i}^{2}\left( t\right) dt\right) .

1221: \label{RTxx2.3.15}

1222: \end{equation}%

1223: The equality holds in (\ref{RTxx2.3.15}) if and only if%

1224: \begin{equation*}

1225: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\geq \frac{1}{2n}%

1226: \sum_{i=1}^{n}\left( \int_{a}^{b}r_{i}^{2}\left( t\right) dt\right)

1227: \end{equation*}%

1228: and%

1229: \begin{equation*}

1230: \int_{a}^{b}f\left( t\right) dt=\left[ \int_{a}^{b}\left\Vert f\left(

1231: t\right) \right\Vert dt-\frac{1}{n}\sum_{i=1}^{n}\left(

1232: \int_{a}^{b}r_{i}^{2}\left( t\right) dt\right) \right] \sum_{i=1}^{n}e_{i}.

1233: \end{equation*}

1234: \end{corollary}

1235: 

1236: \begin{proof}

1237: As in the proof of Corollary \ref{RTxxc2.2.4}, from (\ref{RTxx2.2.16}), we

1238: deduce that%

1239: \begin{equation}

1240: \left\Vert f\left( t\right) \right\Vert -\func{Re}\left\langle f\left(

1241: t\right) ,e_{i}\right\rangle \leq \frac{1}{2}r_{i}^{2}\left( t\right)

1242: \label{RTxx2.3.16}

1243: \end{equation}%

1244: for a.e. $t\in \left[ a,b\right] $ and $i\in \left\{ 1,\dots ,n\right\} .$

1245: 

1246: Applying Theorem \ref{RTxxt2.3.1} for%

1247: \begin{equation*}

1248: M_{i}\left( t\right) :=\frac{1}{2}r_{i}^{2}\left( t\right) ,\ \ t\in \left[

1249: a,b\right] ,\ i\in \left\{ 1,\dots ,n\right\} ,

1250: \end{equation*}%

1251: we get the desired result.

1252: \end{proof}

1253: 

1254: Finally, the following result holds \cite{RTCSSD2}.

1255: 

1256: \begin{corollary}

1257: \label{RTxxc2.3.5}Let $f\in L\left( \left[ a,b\right] ;H\right) ,$ $\left\{

1258: e_{i}\right\} _{i\in \left\{ 1,\dots ,n\right\} }$ an orthonormal family in $%

1259: H$, $M_{i},m_{i}:\left[ a,b\right] \rightarrow \lbrack 0,\infty )$ with $%

1260: M_{i}\geq m_{i}$ a.e. on $\left[ a,b\right] $ and $\frac{\left(

1261: M_{i}-m_{i}\right) ^{2}}{M_{i}+m_{i}}\in L\left[ a,b\right] ,$ and either%

1262: \begin{multline}

1263: \left\Vert f\left( t\right) -\frac{M_{i}\left( t\right) +m_{i}\left(

1264: t\right) }{2}e_{i}\right\Vert \leq \frac{1}{2}\left[ M_{i}\left( t\right)

1265: -m_{i}\left( t\right) \right] ^{2}\ \   \label{RTxx2.3.17} \\

1266: \end{multline}%

1267: or, equivalently,%

1268: \begin{equation*}

1269: \func{Re}\left\langle M_{i}\left( t\right) e_{i}-f\left( t\right) ,f\left(

1270: t\right) -m_{i}\left( t\right) e_{i}\right\rangle \geq 0\ \ \

1271: \end{equation*}%

1272: for a.e. $t\in \left[ a,b\right] $ and any\ $i\in \left\{ 1,\dots ,n\right\}

1273: $, then we have the inequality%

1274: \begin{multline}

1275: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\leq \frac{1}{\sqrt{n}}%

1276: \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert  \label{RTxx2.3.18} \\

1277: +\frac{1}{4n}\sum_{i=1}^{n}\left( \int_{a}^{b}\frac{\left[ M_{i}\left(

1278: t\right) -m_{i}\left( t\right) \right] ^{2}}{M_{i}\left( t\right)

1279: +m_{i}\left( t\right) }dt\right) .

1280: \end{multline}%

1281: The equality holds in (\ref{RTxx2.3.18}) if and only if%

1282: \begin{equation*}

1283: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\geq \frac{1}{4n}%

1284: \sum_{i=1}^{n}\left( \int_{a}^{b}\frac{\left[ M_{i}\left( t\right)

1285: -m_{i}\left( t\right) \right] ^{2}}{M_{i}\left( t\right) +m_{i}\left(

1286: t\right) }dt\right)

1287: \end{equation*}%

1288: and%

1289: \begin{multline*}

1290: \int_{a}^{b}f\left( t\right) dt \\

1291: =\left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt-\frac{1}{4n}%

1292: \sum_{i=1}^{n}\left( \int_{a}^{b}\frac{\left[ M_{i}\left( t\right)

1293: -m_{i}\left( t\right) \right] ^{2}}{M_{i}\left( t\right) +m_{i}\left(

1294: t\right) }dt\right) \right) \sum_{i=1}^{n}e_{i}.

1295: \end{multline*}

1296: \end{corollary}

1297: 

1298: \begin{proof}

1299: As in the proof of Corollary \ref{RTxxc2.2.5}, (\ref{RTxx2.3.17}), implies

1300: that%

1301: \begin{equation*}

1302: \left\Vert f\left( t\right) \right\Vert -\func{Re}\left\langle f\left(

1303: t\right) ,e_{i}\right\rangle \leq \frac{1}{4}\cdot \frac{\left[ M_{i}\left(

1304: t\right) -m_{i}\left( t\right) \right] ^{2}}{M_{i}\left( t\right)

1305: +m_{i}\left( t\right) }\

1306: \end{equation*}%

1307: for a.e. $t\in \left[ a,b\right] $ and $i\in \left\{ 1,\dots ,n\right\} .$

1308: 

1309: Applying Theorem \ref{RTxxt2.3.1} for%

1310: \begin{equation*}

1311: M_{i}\left( t\right) :=\frac{1}{4}\cdot \frac{\left[ M_{i}\left( t\right)

1312: -m_{i}\left( t\right) \right] ^{2}}{M_{i}\left( t\right) +m_{i}\left(

1313: t\right) },\text{ \ \ }t\in \left[ a,b\right] \text{, \ }i\in \left\{

1314: 1,\dots ,n\right\} ,

1315: \end{equation*}%

1316: we deduce the desired result.

1317: \end{proof}

1318: 

1319: \section{Quadratic Reverses of the Triangle Inequality}

1320: 

1321: \subsection{Additive Reverses}

1322: 

1323: The following lemma holds \cite{RTCSSD3}.

1324: 

1325: \begin{lemma}[Dragomir, 2004]

1326: \label{RTxxl3.2.1}Let $f\in L\left( \left[ a,b\right] ;H\right) $ be such

1327: that there exists a function $k:\Delta \subset \mathbb{R}^{2}\rightarrow

1328: \mathbb{R}$, $\Delta :=\left\{ \left( t,s\right) |a\leq t\leq s\leq

1329: b\right\} $ with the property that $k\in L\left( \Delta \right) $ and%

1330: \begin{equation}

1331: \left( 0\leq \right) \left\Vert f\left( t\right) \right\Vert \left\Vert

1332: f\left( s\right) \right\Vert -\func{Re}\left\langle f\left( t\right)

1333: ,f\left( s\right) \right\rangle \leq k\left( t,s\right) ,  \label{RTxx3.2.1}

1334: \end{equation}%

1335: for a.e. $\left( t,s\right) \in \Delta .$ Then we have the following

1336: quadratic reverse of the continuous triangle inequality:%

1337: \begin{equation}

1338: \left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right)

1339: ^{2}\leq \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert

1340: ^{2}+2\iint_{\Delta }k\left( t,s\right) dtds.  \label{RTxx3.2.2}

1341: \end{equation}%

1342: The case of equality holds in (\ref{RTxx3.2.2}) if and only if it holds in (%

1343: \ref{RTxx3.2.1}) for a.e. $\left( t,s\right) \in \Delta .$

1344: \end{lemma}

1345: 

1346: \begin{proof}

1347: We observe that the following identity holds%

1348: \begin{align}

1349: & \left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right)

1350: ^{2}-\left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert ^{2}

1351: \label{RTxx3.2.3} \\

1352: & =\int_{a}^{b}\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert

1353: \left\Vert f\left( s\right) \right\Vert dtds-\left\langle

1354: \int_{a}^{b}f\left( t\right) dt,\int_{a}^{b}f\left( s\right) ds\right\rangle

1355: \notag \\

1356: & =\int_{a}^{b}\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert

1357: \left\Vert f\left( s\right) \right\Vert dtds-\int_{a}^{b}\int_{a}^{b}\func{Re%

1358: }\left\langle f\left( t\right) ,f\left( s\right) \right\rangle dtds  \notag

1359: \\

1360: & =\int_{a}^{b}\int_{a}^{b}\left[ \left\Vert f\left( t\right) \right\Vert

1361: \left\Vert f\left( s\right) \right\Vert -\func{Re}\left\langle f\left(

1362: t\right) ,f\left( s\right) \right\rangle \right] dtds:=I.  \notag

1363: \end{align}%

1364: Now, observe that for any $\left( t,s\right) \in \left[ a,b\right] \times %

1365: \left[ a,b\right] ,$ we have%

1366: \begin{multline*}

1367: \qquad \left\Vert f\left( t\right) \right\Vert \left\Vert f\left(

1368: s\right) \right\Vert -\func{Re}\left\langle f\left( t\right)

1369: ,f\left( s\right)

1370: \right\rangle \\

1371: =\left\Vert f\left( s\right) \right\Vert \left\Vert f\left(

1372: t\right) \right\Vert -\func{Re}\left\langle f\left( s\right)

1373: ,f\left( t\right) \right\rangle\qquad

1374: \end{multline*}%

1375: and thus

1376: \begin{equation}

1377: I=2\iint_{\Delta }\left[ \left\Vert f\left( t\right) \right\Vert \left\Vert

1378: f\left( s\right) \right\Vert -\func{Re}\left\langle f\left( t\right)

1379: ,f\left( s\right) \right\rangle \right] dtds.  \label{RTxx3.2.4}

1380: \end{equation}%

1381: Using the assumption (\ref{RTxx3.2.1}), we deduce%

1382: \begin{equation*}

1383: \iint_{\Delta }\left[ \left\Vert f\left( t\right) \right\Vert \left\Vert

1384: f\left( s\right) \right\Vert -\func{Re}\left\langle f\left( t\right)

1385: ,f\left( s\right) \right\rangle \right] dtds\leq \iint_{\Delta }k\left(

1386: t,s\right) dtds,

1387: \end{equation*}%

1388: and, by the identities (\ref{RTxx3.2.3}) and (\ref{RTxx3.2.4}), we deduce

1389: the desired inequality (\ref{RTxx3.2.2}).

1390: 

1391: The case of equality is obvious and we omit the details.

1392: \end{proof}

1393: 

1394: \begin{remark}

1395: \label{RTxxr3.2.2}From (\ref{RTxx3.2.2}) one may deduce a coarser inequality

1396: that can be useful in some applications. It is as follows:%

1397: \begin{equation*}

1398: \left( 0\leq \right) \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert

1399: dt-\left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert \leq \sqrt{2}\left(

1400: \iint_{\Delta }k\left( t,s\right) dtds\right) ^{\frac{1}{2}}.

1401: \end{equation*}

1402: \end{remark}

1403: 

1404: \begin{remark}

1405: \label{RTxxr3.2.3}If the condition (\ref{RTxx3.2.1}) is replaced with the

1406: following refinement of the Schwarz inequality%

1407: \begin{equation}

1408: \left( 0\leq \right) k\left( t,s\right) \leq \left\Vert f\left( t\right)

1409: \right\Vert \left\Vert f\left( s\right) \right\Vert -\func{Re}\left\langle

1410: f\left( t\right) ,f\left( s\right) \right\rangle  \label{RTxx3.2.5}

1411: \end{equation}%

1412: for a.e. $\left( t,s\right) \in \Delta ,$ then the following refinement of

1413: the quadratic triangle inequality is valid%

1414: \begin{align}

1415: \left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right) ^{2}&

1416: \geq \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert

1417: ^{2}+2\iint_{\Delta }k\left( t,s\right) dtds  \label{RTxx3.2.6} \\

1418: & \left( \geq \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert

1419: ^{2}\right) .  \notag

1420: \end{align}%

1421: The equality holds in (\ref{RTxx3.2.6}) iff the case of equality holds in (%

1422: \ref{RTxx3.2.5}) for a.e. $\left( t,s\right) \in \Delta .$

1423: \end{remark}

1424: 

1425: The following result holds \cite{RTCSSD3}.

1426: 

1427: \begin{theorem}[Dragomir, 2004]

1428: \label{RTxxt3.2.4}Let $f\in L\left( \left[ a,b\right] ;H\right) $ be such

1429: that there exists $M\geq 1\geq m\geq 0$ such that either%

1430: \begin{equation}

1431: \func{Re}\left\langle Mf\left( s\right) -f\left( t\right) ,f\left( t\right)

1432: -mf\left( s\right) \right\rangle \geq 0\text{ \ }  \label{RTxx3.2.7}

1433: \end{equation}%

1434: or, equivalently,%

1435: \begin{equation}

1436: \left\Vert f\left( t\right) -\frac{M+m}{2}f\left( s\right) \right\Vert \leq

1437: \frac{1}{2}\left( M-m\right) \left\Vert f\left( s\right) \right\Vert \text{

1438: \ }  \label{RTxx3.2.8}

1439: \end{equation}%

1440: for a.e. $\left( t,s\right) \in \Delta .$ Then we have the inequality:%

1441: \begin{multline}

1442: \left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right)

1443: ^{2}\leq \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert ^{2}

1444: \label{RTxx3.2.9} \\

1445: +\frac{1}{2}\cdot \frac{\left( M-m\right) ^{2}}{M+m}\int_{a}^{b}\left(

1446: s-a\right) \left\Vert f\left( s\right) \right\Vert ^{2}ds.

1447: \end{multline}%

1448: The case of equality holds in (\ref{RTxx3.2.9}) if and only if%

1449: \begin{equation}

1450: \left\Vert f\left( t\right) \right\Vert \left\Vert f\left( s\right)

1451: \right\Vert -\func{Re}\left\langle f\left( t\right) ,f\left( s\right)

1452: \right\rangle =\frac{1}{4}\cdot \frac{\left( M-m\right) ^{2}}{M+m}\left\Vert

1453: f\left( s\right) \right\Vert ^{2}  \label{RTxx3.2.10}

1454: \end{equation}%

1455: for a.e. $\left( t,s\right) \in \Delta .$

1456: \end{theorem}

1457: 

1458: \begin{proof}

1459: Taking the square in (\ref{RTxx3.2.8}), we get%

1460: \begin{multline*}

1461: \left\Vert f\left( t\right) \right\Vert ^{2}+\left( \frac{M+m}{2}\right)

1462: ^{2}\left\Vert f\left( s\right) \right\Vert ^{2} \\

1463: \leq 2\func{Re}\left\langle f\left( t\right) ,\frac{M+m}{2}f\left( s\right)

1464: \right\rangle +\frac{1}{4}\left( M-m\right) ^{2}\left\Vert f\left( s\right)

1465: \right\Vert ^{2},

1466: \end{multline*}%

1467: for a.e. $\left( t,s\right) \in \Delta ,$ and obviously, since%

1468: \begin{equation*}

1469: 2\left( \frac{M+m}{2}\right) \left\Vert f\left( t\right) \right\Vert

1470: \left\Vert f\left( s\right) \right\Vert \leq \left\Vert f\left( t\right)

1471: \right\Vert ^{2}+\left( \frac{M+m}{2}\right) ^{2}\left\Vert f\left( s\right)

1472: \right\Vert ^{2},

1473: \end{equation*}%

1474: we deduce that%

1475: \begin{multline*}

1476: 2\left( \frac{M+m}{2}\right) \left\Vert f\left( t\right) \right\Vert

1477: \left\Vert f\left( s\right) \right\Vert \\

1478: \leq 2\func{Re}\left\langle f\left( t\right) ,\frac{M+m}{2}f\left( s\right)

1479: \right\rangle +\frac{1}{4}\left( M-m\right) ^{2}\left\Vert f\left( s\right)

1480: \right\Vert ^{2},

1481: \end{multline*}%

1482: giving the much simpler inequality:%

1483: \begin{equation}

1484: \left\Vert f\left( t\right) \right\Vert \left\Vert f\left( s\right)

1485: \right\Vert -\func{Re}\left\langle f\left( t\right) ,f\left( s\right)

1486: \right\rangle \leq \frac{1}{4}\cdot \frac{\left( M-m\right) ^{2}}{M+m}%

1487: \left\Vert f\left( s\right) \right\Vert ^{2}  \label{RTxx3.2.11}

1488: \end{equation}%

1489: for a.e. $\left( t,s\right) \in \Delta .$

1490: 

1491: Applying Lemma \ref{RTxxl3.2.1} for $k\left( t,s\right) :=\frac{1}{4}\cdot

1492: \frac{\left( M-m\right) ^{2}}{M+m}\left\Vert f\left( s\right) \right\Vert

1493: ^{2},$ we deduce%

1494: \begin{multline}

1495: \left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right)

1496: ^{2}\leq \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert ^{2}

1497: \label{RTxx3.2.12} \\

1498: +\frac{1}{2}\cdot \frac{\left( M-m\right) ^{2}}{M+m}\iint_{\Delta

1499: }\left\Vert f\left( s\right) \right\Vert ^{2}ds

1500: \end{multline}%

1501: with equality if and only if (\ref{RTxx3.2.11}) holds for a.e. $\left(

1502: t,s\right) \in \Delta .$

1503: 

1504: Since%

1505: \begin{equation*}

1506: \iint_{\Delta }\left\Vert f\left( s\right) \right\Vert

1507: ^{2}ds=\int_{a}^{b}\left( \int_{a}^{s}\left\Vert f\left( s\right)

1508: \right\Vert ^{2}dt\right) ds=\int_{a}^{b}\left( s-a\right) \left\Vert

1509: f\left( s\right) \right\Vert ^{2}ds,

1510: \end{equation*}%

1511: then by (\ref{RTxx3.2.12}) we deduce the desired result (\ref{RTxx3.2.9}).

1512: \end{proof}

1513: 

1514: Another result which is similar to the one above is incorporated in the

1515: following theorem \cite{RTCSSD3}.

1516: 

1517: \begin{theorem}[Dragomir, 2004]

1518: \label{RTxxt3.2.5}With the assumptions of Theorem \ref{RTxxt3.2.4}, we have%

1519: \begin{multline}

1520: \left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right)

1521: ^{2}-\left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert ^{2}

1522: \label{RTxx3.2.13} \\

1523: \leq \frac{\left( \sqrt{M}-\sqrt{m}\right) ^{2}}{2\sqrt{Mm}}\left\Vert

1524: \int_{a}^{b}f\left( t\right) dt\right\Vert ^{2}

1525: \end{multline}%

1526: or, equivalently,%

1527: \begin{equation}

1528: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\leq \left( \frac{M+m}{%

1529: 2\sqrt{Mm}}\right) ^{\frac{1}{2}}\left\Vert \int_{a}^{b}f\left( t\right)

1530: dt\right\Vert .  \label{RTxx3.2.14}

1531: \end{equation}%

1532: The case of equality holds in (\ref{RTxx3.2.13}) or (\ref{RTxx3.2.14}) if

1533: and only if

1534: \begin{equation}

1535: \left\Vert f\left( t\right) \right\Vert \left\Vert f\left( s\right)

1536: \right\Vert =\frac{M+m}{2\sqrt{Mm}}\func{Re}\left\langle f\left( t\right)

1537: ,f\left( s\right) \right\rangle ,  \label{RTxx3.2.14'}

1538: \end{equation}%

1539: for a.e. $\left( t,s\right) \in \Delta .$

1540: \end{theorem}

1541: 

1542: \begin{proof}

1543: From (\ref{RTxx3.2.7}), we deduce%

1544: \begin{equation*}

1545: \left\Vert f\left( t\right) \right\Vert ^{2}+Mm\left\Vert f\left( s\right)

1546: \right\Vert ^{2}\leq \left( M+m\right) \func{Re}\left\langle f\left(

1547: t\right) ,f\left( s\right) \right\rangle

1548: \end{equation*}%

1549: for a.e. $\left( t,s\right) \in \Delta .$ Dividing by $\sqrt{Mm}>0,$ we

1550: deduce%

1551: \begin{equation*}

1552: \frac{\left\Vert f\left( t\right) \right\Vert ^{2}}{\sqrt{Mm}}+\sqrt{Mm}%

1553: \left\Vert f\left( s\right) \right\Vert ^{2}\leq \frac{M+m}{\sqrt{Mm}}\func{%

1554: Re}\left\langle f\left( t\right) ,f\left( s\right) \right\rangle

1555: \end{equation*}%

1556: and, obviously, since%

1557: \begin{equation*}

1558: 2\left\Vert f\left( t\right) \right\Vert \left\Vert f\left( s\right)

1559: \right\Vert \leq \frac{\left\Vert f\left( t\right) \right\Vert ^{2}}{\sqrt{Mm%

1560: }}+\sqrt{Mm}\left\Vert f\left( s\right) \right\Vert ^{2},

1561: \end{equation*}%

1562: hence%

1563: \begin{equation*}

1564: \left\Vert f\left( t\right) \right\Vert \left\Vert f\left( s\right)

1565: \right\Vert \leq \frac{M+m}{\sqrt{Mm}}\func{Re}\left\langle f\left( t\right)

1566: ,f\left( s\right) \right\rangle

1567: \end{equation*}%

1568: for a.e. $\left( t,s\right) \in \Delta ,$ giving%

1569: \begin{equation*}

1570: \left\Vert f\left( t\right) \right\Vert \left\Vert f\left( s\right)

1571: \right\Vert -\func{Re}\left\langle f\left( t\right) ,f\left( s\right)

1572: \right\rangle \leq \frac{\left( \sqrt{M}-\sqrt{m}\right) ^{2}}{2\sqrt{Mm}}%

1573: \func{Re}\left\langle f\left( t\right) ,f\left( s\right) \right\rangle .

1574: \end{equation*}%

1575: Applying Lemma \ref{RTxxl3.2.1} for $k\left( t,s\right) :=\frac{\left( \sqrt{%

1576: M}-\sqrt{m}\right) ^{2}}{\sqrt{Mm}}\func{Re}\left\langle f\left( t\right)

1577: ,f\left( s\right) \right\rangle ,$ we deduce%

1578: \begin{multline}

1579: \left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right)

1580: ^{2}\leq \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert ^{2}

1581: \label{RTxx3.2.15} \\

1582: +\frac{\left( \sqrt{M}-\sqrt{m}\right) ^{2}}{2\sqrt{Mm}}\func{Re}%

1583: \left\langle f\left( t\right) ,f\left( s\right) \right\rangle .

1584: \end{multline}%

1585: On the other hand, since%

1586: \begin{equation*}

1587: \func{Re}\left\langle f\left( t\right) ,f\left( s\right) \right\rangle =%

1588: \func{Re}\left\langle f\left( s\right) ,f\left( t\right) \right\rangle \text{

1589: \ for any \ }\left( t,s\right) \in \left[ a,b\right] ^{2},

1590: \end{equation*}%

1591: hence%

1592: \begin{align*}

1593: \iint_{\Delta }\func{Re}\left\langle f\left( t\right) ,f\left( s\right)

1594: \right\rangle dtds& =\frac{1}{2}\int_{a}^{b}\int_{a}^{b}\func{Re}%

1595: \left\langle f\left( t\right) ,f\left( s\right) \right\rangle dtds \\

1596: & =\frac{1}{2}\func{Re}\left\langle \int_{a}^{b}f\left( t\right)

1597: dt,\int_{a}^{b}f\left( s\right) ds\right\rangle \\

1598: & =\frac{1}{2}\left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert ^{2}

1599: \end{align*}%

1600: and thus, from (\ref{RTxx3.2.15}), we get (\ref{RTxx3.2.13}).

1601: 

1602: The equivalence between (\ref{RTxx3.2.13}) and (\ref{RTxx3.2.14}) is obvious

1603: and we omit the details.

1604: \end{proof}

1605: 

1606: \subsection{Related Results}

1607: 

1608: The following result also holds \cite{RTCSSD3}.

1609: 

1610: \begin{theorem}[Dragomir, 2004]

1611: \label{RTxxt3.3.1}Let $f\in L\left( \left[ a,b\right] ;H\right) $ and $%

1612: \gamma ,\Gamma \in \mathbb{R}$ be such that either%

1613: \begin{equation}

1614: \func{Re}\left\langle \Gamma f\left( s\right) -f\left( t\right) ,f\left(

1615: t\right) -\gamma f\left( s\right) \right\rangle \geq 0\text{ \ }

1616: \label{RTxx3.3.1}

1617: \end{equation}%

1618: or, equivalently,%

1619: \begin{equation}

1620: \left\Vert f\left( t\right) -\frac{\Gamma +\gamma }{2}f\left( s\right)

1621: \right\Vert \leq \frac{1}{2}\left\vert \Gamma -\gamma \right\vert \left\Vert

1622: f\left( s\right) \right\Vert \text{ }  \label{RTxx3.3.2}

1623: \end{equation}%

1624: \ for a.e. \ $\left( t,s\right) \in \Delta .$ Then we have the inequality:%

1625: \begin{equation}

1626: \int_{a}^{b}\left[ \left( b-s\right) +\gamma \Gamma \left( s-a\right) \right]

1627: \left\Vert f\left( s\right) \right\Vert ^{2}ds\leq \frac{\Gamma +\gamma }{2}%

1628: \left\Vert \int_{a}^{b}f\left( s\right) ds\right\Vert ^{2}.

1629: \label{RTxx3.3.3}

1630: \end{equation}%

1631: The case of equality holds in (\ref{RTxx3.3.3}) if and only if the case of

1632: equality holds in either (\ref{RTxx3.3.1}) or (\ref{RTxx3.3.2}) for a.e. $%

1633: \left( t,s\right) \in \Delta $.

1634: \end{theorem}

1635: 

1636: \begin{proof}

1637: The inequality (\ref{RTxx3.3.1}) is obviously equivalent to%

1638: \begin{equation}

1639: \left\Vert f\left( t\right) \right\Vert ^{2}+\gamma \Gamma \left\Vert

1640: f\left( s\right) \right\Vert ^{2}\leq \left( \Gamma +\gamma \right) \func{Re}%

1641: \left\langle f\left( t\right) ,f\left( s\right) \right\rangle

1642: \label{RTxx3.3.4}

1643: \end{equation}%

1644: for a.e. $\left( t,s\right) \in \Delta .$

1645: 

1646: Integrating (\ref{RTxx3.3.4}) on $\Delta ,$ we deduce%

1647: \begin{multline}

1648: \int_{a}^{b}\left( \int_{a}^{s}\left\Vert f\left( t\right) \right\Vert

1649: ^{2}dt\right) ds+\gamma \Gamma \int_{a}^{b}\left( \left\Vert f\left(

1650: s\right) \right\Vert ^{2}\int_{a}^{s}dt\right) ds  \label{RTxx3.3.5} \\

1651: =\left( \Gamma +\gamma \right) \int_{a}^{b}\left( \int_{a}^{s}\func{Re}%

1652: \left\langle f\left( t\right) ,f\left( s\right) \right\rangle dt\right) ds.

1653: \end{multline}%

1654: It is easy to see, on integrating by parts, that%

1655: \begin{align*}

1656: \int_{a}^{b}\left( \int_{a}^{s}\left\Vert f\left( t\right) \right\Vert

1657: ^{2}dt\right) ds& =s\left. \int_{a}^{s}\left\Vert f\left( t\right)

1658: \right\Vert ^{2}dt\right\vert _{a}^{b}-\int_{a}^{b}s\left\Vert f\left(

1659: s\right) \right\Vert ^{2}ds \\

1660: & =b\int_{a}^{s}\left\Vert f\left( s\right) \right\Vert

1661: ^{2}ds-\int_{a}^{b}s\left\Vert f\left( s\right) \right\Vert ^{2}ds \\

1662: & =\int_{a}^{b}\left( b-s\right) \left\Vert f\left( s\right) \right\Vert

1663: ^{2}ds

1664: \end{align*}%

1665: and

1666: \begin{equation*}

1667: \int_{a}^{b}\left( \left\Vert f\left( s\right) \right\Vert

1668: ^{2}\int_{a}^{s}dt\right) ds=\int_{a}^{b}\left( s-a\right) \left\Vert

1669: f\left( s\right) \right\Vert ^{2}ds.

1670: \end{equation*}%

1671: Since%

1672: \begin{align*}

1673: \frac{d}{ds}\left( \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert

1674: ^{2}\right) & =\frac{d}{ds}\left\langle \int_{a}^{s}f\left( t\right)

1675: dt,\int_{a}^{s}f\left( t\right) dt\right\rangle \\

1676: & =\left\langle f\left( s\right) ,\int_{a}^{s}f\left( t\right)

1677: dt\right\rangle +\left\langle \int_{a}^{s}f\left( t\right) dt,f\left(

1678: s\right) \right\rangle \\

1679: & =2\func{Re}\left\langle \int_{a}^{s}f\left( t\right) dt,f\left( s\right)

1680: \right\rangle ,

1681: \end{align*}%

1682: hence%

1683: \begin{align*}

1684: \int_{a}^{b}\left( \int_{a}^{s}\func{Re}\left\langle f\left( t\right)

1685: ,f\left( s\right) \right\rangle dt\right) ds& =\int_{a}^{b}\func{Re}%

1686: \left\langle \int_{a}^{s}f\left( t\right) dt,f\left( s\right) \right\rangle

1687: ds \\

1688: & =\frac{1}{2}\int_{a}^{b}\frac{d}{ds}\left( \left\Vert \int_{a}^{s}f\left(

1689: t\right) dt\right\Vert ^{2}\right) ds \\

1690: & =\frac{1}{2}\left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert ^{2}.

1691: \end{align*}%

1692: Utilising (\ref{RTxx3.3.5}), we deduce the desired inequality (\ref%

1693: {RTxx3.3.3}).

1694: 

1695: The case of equality is obvious and we omit the details.

1696: \end{proof}

1697: 

1698: \begin{remark}

1699: Consider the function $\varphi \left( s\right) :=\left( b-s\right) +\gamma

1700: \Gamma \left( s-a\right) ,$ $s\in \left[ a,b\right] .$ Obviously,%

1701: \begin{equation*}

1702: \varphi \left( s\right) =\left( \Gamma \gamma -1\right) s+b-\gamma \Gamma a.

1703: \end{equation*}%

1704: Observe that, if $\Gamma \gamma \geq 1,$ then%

1705: \begin{equation*}

1706: b-a=\varphi \left( a\right) \leq \varphi \left( s\right) \leq \varphi \left(

1707: b\right) =\gamma \Gamma \left( b-a\right) ,\ \ \ \ s\in \left[ a,b\right]

1708: \end{equation*}%

1709: and, if $\Gamma \gamma <1,$ then%

1710: \begin{equation*}

1711: \gamma \Gamma \left( b-a\right) \leq \varphi \left( s\right) \leq b-a,\ \ \

1712: \ s\in \left[ a,b\right] .

1713: \end{equation*}

1714: \end{remark}

1715: 

1716: Taking into account the above remark, we may state the following corollary

1717: \cite{RTCSSD3}.

1718: 

1719: \begin{corollary}

1720: \label{RTxxc3.3.2}Assume that $f,\gamma ,\Gamma $ are as in Theorem \ref%

1721: {RTxxt3.3.1}.

1722: 

1723: \begin{enumerate}

1724: \item[a)] If $\Gamma \gamma \geq 1,$ then we have the inequality%

1725: \begin{equation*}

1726: \left( b-a\right) \int_{a}^{b}\left\Vert f\left( s\right) \right\Vert

1727: ^{2}ds\leq \frac{\Gamma +\gamma }{2}\left\Vert \int_{a}^{b}f\left( s\right)

1728: ds\right\Vert ^{2}.

1729: \end{equation*}

1730: 

1731: \item[b)] If $0<\Gamma \gamma <1,$ then we have the inequality%

1732: \begin{equation*}

1733: \gamma \Gamma \left( b-a\right) \int_{a}^{b}\left\Vert f\left( s\right)

1734: \right\Vert ^{2}ds\leq \frac{\Gamma +\gamma }{2}\left\Vert

1735: \int_{a}^{b}f\left( s\right) ds\right\Vert ^{2}.

1736: \end{equation*}

1737: \end{enumerate}

1738: \end{corollary}

1739: 

1740: \section{Refinements for Complex Spaces}

1741: 

1742: \subsection{The Case of a Unit Vector}

1743: 

1744: The following result holds \cite{RTCSSD4}.

1745: 

1746: \begin{theorem}[Dragomir, 2004]

1747: \label{RTxxt4.2.1}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

1748: \right) $ be a complex Hilbert space. If $f\in L\left( \left[ a,b\right]

1749: ;H\right) $ is such that there exists $k_{1},k_{2}\geq 0$ with%

1750: \begin{equation}

1751: k_{1}\left\Vert f\left( t\right) \right\Vert \leq \func{Re}\left\langle

1752: f\left( t\right) ,e\right\rangle ,\ \ k_{2}\left\Vert f\left( t\right)

1753: \right\Vert \leq \func{Im}\left\langle f\left( t\right) ,e\right\rangle

1754: \label{RTxx4.2.1}

1755: \end{equation}%

1756: for a.e. $t\in \left[ a,b\right] ,$ where $e\in H,$ $\left\Vert e\right\Vert

1757: =1,$ is given, then%

1758: \begin{equation}

1759: \sqrt{k_{1}^{2}+k_{2}^{2}}\int_{a}^{b}\left\Vert f\left( t\right)

1760: \right\Vert dt\leq \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert .

1761: \label{RTxx4.2.2}

1762: \end{equation}%

1763: The case of equality holds in (\ref{RTxx4.2.2}) if and only if%

1764: \begin{equation}

1765: \int_{a}^{b}f\left( t\right) dt=\left( k_{1}+ik_{2}\right) \left(

1766: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right) e.

1767: \label{RTxx4.2.3}

1768: \end{equation}

1769: \end{theorem}

1770: 

1771: \begin{proof}

1772: Using the Schwarz inequality $\left\Vert u\right\Vert \left\Vert

1773: v\right\Vert \geq \left\vert \left\langle u,v\right\rangle \right\vert ,$ $%

1774: u,v\in H;$ in the complex Hilbert space $\left( H;\left\langle \cdot ,\cdot

1775: \right\rangle \right) ,$ we have%

1776: \begin{align}

1777: & \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert ^{2}

1778: \label{RTxx4.2.4}

1779:  =\left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert ^{2}\left\Vert

1780: e\right\Vert ^{2}  \\

1781: & \geq \left\vert \left\langle \int_{a}^{b}f\left( t\right)

1782: dt,e\right\rangle \right\vert ^{2}=\left\vert \int_{a}^{b}\left\langle

1783: f\left( t\right) ,e\right\rangle dt\right\vert ^{2}  \notag \\

1784: & =\left\vert \int_{a}^{b}\func{Re}\left\langle f\left( t\right)

1785: ,e\right\rangle dt+i\left( \int_{a}^{b}\func{Im}\left\langle f\left(

1786: t\right) ,e\right\rangle dt\right) \right\vert ^{2}  \notag \\

1787: & =\left( \int_{a}^{b}\func{Re}\left\langle f\left( t\right) ,e\right\rangle

1788: dt\right) ^{2}+\left( \int_{a}^{b}\func{Im}\left\langle f\left( t\right)

1789: ,e\right\rangle dt\right) ^{2}.  \notag

1790: \end{align}%

1791: Now, on integrating (\ref{RTxx4.2.1}), we deduce%

1792: \begin{align}

1793: k_{1}\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt& \leq

1794: \int_{a}^{b}\func{Re}\left\langle f\left( t\right) ,e\right\rangle dt,

1795: \label{RTxx4.2.5} \\

1796: k_{2}\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt& \leq

1797: \int_{a}^{b}\func{Im}\left\langle f\left( t\right) ,e\right\rangle dt  \notag

1798: \end{align}%

1799: implying%

1800: \begin{equation}

1801: \left( \int_{a}^{b}\func{Re}\left\langle f\left( t\right) ,e\right\rangle

1802: dt\right) ^{2}\geq k_{1}^{2}\left( \int_{a}^{b}\left\Vert f\left( t\right)

1803: \right\Vert dt\right) ^{2}  \label{RTxx4.2.6}

1804: \end{equation}%

1805: and%

1806: \begin{equation}

1807: \left( \int_{a}^{b}\func{Im}\left\langle f\left( t\right) ,e\right\rangle

1808: dt\right) ^{2}\geq k_{2}^{2}\left( \int_{a}^{b}\left\Vert f\left( t\right)

1809: \right\Vert dt\right) ^{2}.  \label{RTxx4.2.7}

1810: \end{equation}%

1811: If we add (\ref{RTxx4.2.6}) and (\ref{RTxx4.2.7}) and use (\ref{RTxx4.2.4}),

1812: we deduce the desired inequality (\ref{RTxx4.2.2}).

1813: 

1814: Further, if (\ref{RTxx4.2.3}) holds, then obviously%

1815: \begin{align*}

1816: \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert & =\left\vert

1817: k_{1}+ik_{2}\right\vert \left( \int_{a}^{b}\left\Vert f\left( t\right)

1818: \right\Vert dt\right) \left\Vert e\right\Vert \\

1819: & =\sqrt{k_{1}^{2}+k_{2}^{2}}\int_{a}^{b}\left\Vert f\left( t\right)

1820: \right\Vert dt,

1821: \end{align*}%

1822: and the equality case holds in (\ref{RTxx4.2.2}).

1823: 

1824: Before we prove the reverse implication, let us observe that, for $x\in H$

1825: and $e\in H,$ $\left\Vert e\right\Vert =1,$ the following identity is valid%

1826: \begin{equation*}

1827: \left\Vert x-\left\langle x,e\right\rangle e\right\Vert ^{2}=\left\Vert

1828: x\right\Vert ^{2}-\left\vert \left\langle x,e\right\rangle \right\vert ^{2},

1829: \end{equation*}%

1830: therefore $\left\Vert x\right\Vert =\left\vert \left\langle x,e\right\rangle

1831: \right\vert $ if and only if $x=\left\langle x,e\right\rangle e.$

1832: 

1833: If we assume that equality holds in (\ref{RTxx4.2.2}), then the case of

1834: equality must hold in all the inequalities required in the argument used to

1835: prove the inequality (\ref{RTxx4.2.2}). Therefore, we must have%

1836: \begin{equation}

1837: \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert =\left\vert

1838: \left\langle \int_{a}^{b}f\left( t\right) dt,e\right\rangle \right\vert

1839: \label{RTxx4.2.7a}

1840: \end{equation}%

1841: and

1842: \begin{equation}

1843: k_{1}\left\Vert f\left( t\right) \right\Vert =\func{Re}\left\langle f\left(

1844: t\right) ,e\right\rangle ,\ \ \ k_{2}\left\Vert f\left( t\right) \right\Vert

1845: =\func{Im}\left\langle f\left( t\right) ,e\right\rangle  \label{RTxx4.2.8}

1846: \end{equation}%

1847: for a.e. $t\in \left[ a,b\right] .$

1848: 

1849: From (\ref{RTxx4.2.7a}) we deduce%

1850: \begin{equation}

1851: \int_{a}^{b}f\left( t\right) dt=\left\langle \int_{a}^{b}f\left( t\right)

1852: dt,e\right\rangle e,  \label{RTxx4.2.9}

1853: \end{equation}%

1854: and from (\ref{RTxx4.2.8}), by multiplying the second equality with $i,$ the

1855: imaginary unit, and integrating both equations on $\left[ a,b\right] ,$ we

1856: deduce%

1857: \begin{equation}

1858: \left( k_{1}+ik_{2}\right) \int_{a}^{b}\left\Vert f\left( t\right)

1859: \right\Vert dt=\left\langle \int_{a}^{b}f\left( t\right) dt,e\right\rangle .

1860: \label{RTxx4.2.10}

1861: \end{equation}%

1862: Finally, by (\ref{RTxx4.2.9}) and (\ref{RTxx4.2.10}), we deduce the desired

1863: equality (\ref{RTxx4.2.3}).

1864: \end{proof}

1865: 

1866: The following corollary is of interest \cite{RTCSSD4}.

1867: 

1868: \begin{corollary}

1869: \label{RTxxc4.2.2}Let $e$ be a unit vector in the complex Hilbert space $%

1870: \left( H;\left\langle \cdot ,\cdot \right\rangle \right) $ and $\eta

1871: _{1},\eta _{2}\in \left( 0,1\right) .$ If $f\in L\left( \left[ a,b\right]

1872: ;H\right) $ is such that%

1873: \begin{equation}

1874: \left\Vert f\left( t\right) -e\right\Vert \leq \eta _{1},\left\Vert f\left(

1875: t\right) -ie\right\Vert \leq \eta _{2}\text{ }  \label{RTxx4.2.11a}

1876: \end{equation}%

1877: \ for a.e. \ $t\in \left[ a,b\right] ,$ then we have the inequality%

1878: \begin{equation}

1879: \sqrt{2-\eta _{1}^{2}-\eta _{2}^{2}}\int_{a}^{b}\left\Vert f\left( t\right)

1880: \right\Vert dt\leq \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert .

1881: \label{RTxx4.2.12}

1882: \end{equation}%

1883: The case of equality holds in (\ref{RTxx4.2.12}) if and only if%

1884: \begin{equation}

1885: \int_{a}^{b}f\left( t\right) dt=\left( \sqrt{1-\eta _{1}^{2}}+i\sqrt{1-\eta

1886: _{2}^{2}}\right) \left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert

1887: dt\right) e.  \label{RTxx4.2.13}

1888: \end{equation}

1889: \end{corollary}

1890: 

1891: \begin{proof}

1892: From the first inequality in (\ref{RTxx4.2.11a}) we deduce, by taking the

1893: square, that%

1894: \begin{equation*}

1895: \left\Vert f\left( t\right) \right\Vert ^{2}+1-\eta _{1}^{2}\leq 2\func{Re}%

1896: \left\langle f\left( t\right) ,e\right\rangle ,

1897: \end{equation*}%

1898: implying%

1899: \begin{equation}

1900: \frac{\left\Vert f\left( t\right) \right\Vert ^{2}}{\sqrt{1-\eta _{1}^{2}}}+%

1901: \sqrt{1-\eta _{1}^{2}}\leq \frac{2\func{Re}\left\langle f\left( t\right)

1902: ,e\right\rangle }{\sqrt{1-\eta _{1}^{2}}}  \label{RTxx4.2.14}

1903: \end{equation}%

1904: for a.e. $t\in \left[ a,b\right] .$

1905: 

1906: Since, obviously%

1907: \begin{equation}

1908: 2\left\Vert f\left( t\right) \right\Vert \leq \frac{\left\Vert f\left(

1909: t\right) \right\Vert ^{2}}{\sqrt{1-\eta _{1}^{2}}}+\sqrt{1-\eta _{1}^{2}},

1910: \label{RTxx4.2.15}

1911: \end{equation}%

1912: hence, by (\ref{RTxx4.2.14}) and (\ref{RTxx4.2.15}) we get%

1913: \begin{equation}

1914: 0\leq \sqrt{1-\eta _{1}^{2}}\left\Vert f\left( t\right) \right\Vert \leq

1915: \func{Re}\left\langle f\left( t\right) ,e\right\rangle  \label{RTxx4.2.16}

1916: \end{equation}%

1917: for a.e. $t\in \left[ a,b\right] .$

1918: 

1919: From the second inequality in (\ref{RTxx4.2.11a}) we deduce%

1920: \begin{equation*}

1921: 0\leq \sqrt{1-\eta _{2}^{2}}\left\Vert f\left( t\right) \right\Vert \leq

1922: \func{Re}\left\langle f\left( t\right) ,ie\right\rangle

1923: \end{equation*}%

1924: for a.e. $t\in \left[ a,b\right] .$ Since%

1925: \begin{equation*}

1926: \func{Re}\left\langle f\left( t\right) ,ie\right\rangle =\func{Im}%

1927: \left\langle f\left( t\right) ,e\right\rangle

1928: \end{equation*}%

1929: hence%

1930: \begin{equation}

1931: 0\leq \sqrt{1-\eta _{2}^{2}}\left\Vert f\left( t\right) \right\Vert \leq

1932: \func{Im}\left\langle f\left( t\right) ,e\right\rangle  \label{RTxx4.2.17}

1933: \end{equation}%

1934: for a.e. $t\in \left[ a,b\right] .$

1935: 

1936: Now, observe from (\ref{RTxx4.2.16}) and (\ref{RTxx4.2.17}), that the

1937: condition (\ref{RTxx4.2.1}) of Theorem \ref{RTxxt4.2.1} is satisfied for $%

1938: k_{1}=\sqrt{1-\eta _{1}^{2}},$ $k_{2}=\sqrt{1-\eta _{2}^{2}}\in \left(

1939: 0,1\right) ,$ and thus the corollary is proved.

1940: \end{proof}

1941: 

1942: The following corollary may be stated as well \cite{RTCSSD4}.

1943: 

1944: \begin{corollary}

1945: \label{RTxxc4.2.3}Let $e$ be a unit vector in the complex Hilbert space $%

1946: \left( H;\left\langle \cdot ,\cdot \right\rangle \right) $ and $M_{1}\geq

1947: m_{1}>0,$ $M_{2}\geq m_{2}>0.$ If $f\in L\left( \left[ a,b\right] ;H\right) $

1948: is such that either%

1949: \begin{align}

1950: \func{Re}\left\langle M_{1}e-f\left( t\right) ,f\left( t\right)

1951: -m_{1}e\right\rangle & \geq 0,\text{ }  \label{RTxx4.2.18} \\

1952: \func{Re}\left\langle M_{2}ie-f\left( t\right) ,f\left( t\right)

1953: -m_{2}ie\right\rangle & \geq 0  \notag

1954: \end{align}%

1955: or, equivalently,%

1956: \begin{align}

1957: \left\Vert f\left( t\right) -\frac{M_{1}+m_{1}}{2}e\right\Vert & \leq \frac{1%

1958: }{2}\left( M_{1}-m_{1}\right) ,  \label{RTxx4.2.19} \\

1959: \left\Vert f\left( t\right) -\frac{M_{2}+m_{2}}{2}ie\right\Vert & \leq \frac{%

1960: 1}{2}\left( M_{2}-m_{2}\right) ,  \notag

1961: \end{align}%

1962: for a.e. $t\in \left[ a,b\right] ,$ then we have the inequality%

1963: \begin{multline}

1964: 2\left[ \frac{m_{1}M_{1}}{\left( M_{1}+m_{1}\right) ^{2}}+\frac{m_{2}M_{2}}{%

1965: \left( M_{2}+m_{2}\right) ^{2}}\right] ^{\frac{1}{2}}\int_{a}^{b}\left\Vert

1966: f\left( t\right) \right\Vert dt  \label{RTxx4.2.20} \\

1967: \leq \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert .\qquad

1968: \end{multline}%

1969: The equality holds in (\ref{RTxx4.2.20}) if and only if%

1970: \begin{equation}

1971: \int_{a}^{b}f\left( t\right) dt=2\left( \frac{\sqrt{m_{1}M_{1}}}{M_{1}+m_{1}}%

1972: +i\frac{\sqrt{m_{2}M_{2}}}{M_{2}+m_{2}}\right) \left( \int_{a}^{b}\left\Vert

1973: f\left( t\right) \right\Vert dt\right) e.  \label{RTxx4.2.21}

1974: \end{equation}

1975: \end{corollary}

1976: 

1977: \begin{proof}

1978: From the first inequality in (\ref{RTxx4.2.18}), we get%

1979: \begin{equation*}

1980: \left\Vert f\left( t\right) \right\Vert ^{2}+m_{1}M_{1}\leq \left(

1981: M_{1}+m_{1}\right) \func{Re}\left\langle f\left( t\right) ,e\right\rangle

1982: \end{equation*}%

1983: implying%

1984: \begin{equation}

1985: \frac{\left\Vert f\left( t\right) \right\Vert ^{2}}{\sqrt{m_{1}M_{1}}}+\sqrt{%

1986: m_{1}M_{1}}\leq \frac{M_{1}+m_{1}}{\sqrt{m_{1}M_{1}}}\func{Re}\left\langle

1987: f\left( t\right) ,e\right\rangle  \label{RTxx4.2.22}

1988: \end{equation}%

1989: for a.e. $t\in \left[ a,b\right] .$

1990: 

1991: Since, obviously,%

1992: \begin{equation}

1993: 2\left\Vert f\left( t\right) \right\Vert \leq \frac{\left\Vert f\left(

1994: t\right) \right\Vert ^{2}}{\sqrt{m_{1}M_{1}}}+\sqrt{m_{1}M_{1}},

1995: \label{RTxx4.2.23}

1996: \end{equation}%

1997: hence, by (\ref{RTxx4.2.22}) and (\ref{RTxx4.2.23})%

1998: \begin{equation}

1999: 0\leq \frac{2\sqrt{m_{1}M_{1}}}{M_{1}+m_{1}}\left\Vert f\left( t\right)

2000: \right\Vert \leq \func{Re}\left\langle f\left( t\right) ,e\right\rangle

2001: \label{RTxx4.2.24}

2002: \end{equation}%

2003: for a.e. $t\in \left[ a,b\right] .$

2004: 

2005: Using the same argument as in the proof of Corollary \ref{RTxxc4.2.2}, we

2006: deduce the desired inequality. We omit the details.

2007: \end{proof}

2008: 

2009: \subsection{The Case of Orthonormal Vectors}

2010: 

2011: The following result holds \cite{RTCSSD4}.

2012: 

2013: \begin{theorem}[Dragomir, 2004]

2014: \label{RTxxt4.3.2}Let $\left\{ e_{1},\dots ,e_{n}\right\} $ be a family of

2015: orthonormal vectors in the complex Hilbert space $\left( H;\left\langle

2016: \cdot ,\cdot \right\rangle \right) $. If $k_{j},h_{j}\geq 0,$ $j\in \left\{

2017: 1,\dots ,n\right\} $ and $f\in L\left( \left[ a,b\right] ;H\right) $ are

2018: such that%

2019: \begin{equation}

2020: k_{j}\left\Vert f\left( t\right) \right\Vert \leq \func{Re}\left\langle

2021: f\left( t\right) ,e_{j}\right\rangle ,\quad h_{j}\left\Vert f\left( t\right)

2022: \right\Vert \leq \func{Im}\left\langle f\left( t\right) ,e_{j}\right\rangle

2023: \label{RTxx4.3.4}

2024: \end{equation}%

2025: for each $j\in \left\{ 1,\dots ,n\right\} $ and a.e. $t\in \left[ a,b\right]

2026: ,$ then%

2027: \begin{equation}

2028: \left[ \sum_{j=1}^{n}\left( k_{j}^{2}+h_{j}^{2}\right) \right] ^{\frac{1}{2}%

2029: }\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\leq \left\Vert

2030: \int_{a}^{b}f\left( t\right) dt\right\Vert .  \label{RTxx4.3.5}

2031: \end{equation}%

2032: The case of equality holds in (\ref{RTxx4.3.5}) if and only if%

2033: \begin{equation}

2034: \int_{a}^{b}f\left( t\right) dt=\left( \int_{a}^{b}\left\Vert f\left(

2035: t\right) \right\Vert dt\right) \sum_{j=1}^{n}\left( k_{j}+ih_{j}\right)

2036: e_{j}.  \label{RTxx4.3.6}

2037: \end{equation}

2038: \end{theorem}

2039: 

2040: \begin{proof}

2041: Before we prove the theorem, let us recall that, if $x\in H$ and $%

2042: e_{1},\dots ,e_{n}$ are orthonormal vectors, then the following identity

2043: holds true:%

2044: \begin{equation}

2045: \left\Vert x-\sum_{j=1}^{n}\left\langle x,e_{j}\right\rangle

2046: e_{j}\right\Vert ^{2}=\left\Vert x\right\Vert ^{2}-\sum_{j=1}^{n}\left\vert

2047: \left\langle x,e_{j}\right\rangle \right\vert ^{2}.  \label{RTxx4.3.7}

2048: \end{equation}%

2049: As a consequence of this identity, we have the \textit{Bessel inequality}%

2050: \begin{equation}

2051: \sum_{j=1}^{n}\left\vert \left\langle x,e_{j}\right\rangle \right\vert

2052: ^{2}\leq \left\Vert x\right\Vert ^{2},x\in H,  \label{RTxx4.3.8}

2053: \end{equation}%

2054: in which, the case of equality holds if and only if

2055: \begin{equation}

2056: x=\sum_{j=1}^{n}\left\langle x,e_{j}\right\rangle e_{j}.  \label{RTxx4.3.9}

2057: \end{equation}%

2058: Now, applying Bessel's inequality for $x=\int_{a}^{b}f\left( t\right) dt,$

2059: we have successively%

2060: \begin{align}

2061: & \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert ^{2}

2062: \label{RTxx4.3.10} \\

2063: & \geq \sum_{j=1}^{n}\left\vert \left\langle \int_{a}^{b}f\left( t\right)

2064: dt,e_{j}\right\rangle \right\vert ^{2}=\sum_{j=1}^{n}\left\vert

2065: \int_{a}^{b}\left\langle f\left( t\right) ,e_{j}\right\rangle dt\right\vert

2066: ^{2}  \notag \displaybreak\\

2067: & =\sum_{j=1}^{n}\left\vert \int_{a}^{b}\func{Re}\left\langle f\left(

2068: t\right) ,e_{j}\right\rangle dt+i\left( \int_{a}^{b}\func{Im}\left\langle

2069: f\left( t\right) ,e_{j}\right\rangle dt\right) \right\vert ^{2}  \notag \\

2070: & =\sum_{j=1}^{n}\left[ \left( \int_{a}^{b}\func{Re}\left\langle f\left(

2071: t\right) ,e_{j}\right\rangle dt\right) ^{2}+\left( \int_{a}^{b}\func{Im}%

2072: \left\langle f\left( t\right) ,e_{j}\right\rangle dt\right) ^{2}\right] .

2073: \notag

2074: \end{align}%

2075: Integrating (\ref{RTxx4.3.4}) on $\left[ a,b\right] ,$ we get%

2076: \begin{equation}

2077: \int_{a}^{b}\func{Re}\left\langle f\left( t\right) ,e_{j}\right\rangle

2078: dt\geq k_{j}\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt

2079: \label{RTxx4.3.11}

2080: \end{equation}%

2081: and%

2082: \begin{equation}

2083: \int_{a}^{b}\func{Im}\left\langle f\left( t\right) ,e_{j}\right\rangle

2084: dt\geq h_{j}\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt

2085: \label{RTxx4.3.12}

2086: \end{equation}%

2087: for each $j\in \left\{ 1,\dots ,n\right\} .$

2088: 

2089: Squaring and adding the above two inequalities (\ref{RTxx4.3.11}) and (\ref%

2090: {RTxx4.3.12}), we deduce%

2091: \begin{multline*}

2092: \sum_{j=1}^{n}\left[ \left( \int_{a}^{b}\func{Re}\left\langle f\left(

2093: t\right) ,e_{j}\right\rangle dt\right) ^{2}+\left( \int_{a}^{b}\func{Im}%

2094: \left\langle f\left( t\right) ,e_{j}\right\rangle dt\right) ^{2}\right] \\

2095: \geq \sum_{j=1}^{n}\left( k_{j}^{2}+h_{j}^{2}\right) \left(

2096: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right) ^{2},

2097: \end{multline*}%

2098: which combined with (\ref{RTxx4.3.10}) will produce the desired inequality (%

2099: \ref{RTxx4.3.5}).

2100: 

2101: Now, if (\ref{RTxx4.3.6}) holds true, then%

2102: \begin{align*}

2103: \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert & =\left(

2104: \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right) \left\Vert

2105: \sum_{j=1}^{n}\left( k_{j}+ih_{j}\right) e_{j}\right\Vert \\

2106: & =\left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right)

2107: \left( \left\Vert \sum_{j=1}^{n}\left( k_{j}+ih_{j}\right) e_{j}\right\Vert

2108: ^{2}\right) ^{\frac{1}{2}} \\

2109: & =\left( \int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\right)

2110: \left[ \sum_{j=1}^{n}\left( k_{j}^{2}+h_{j}^{2}\right) \right] ^{\frac{1}{2}%

2111: },

2112: \end{align*}%

2113: and the case of equality holds in (\ref{RTxx4.3.5}).

2114: 

2115: Conversely, if the equality holds in (\ref{RTxx4.3.5}), then it must hold in

2116: all the inequalities used to prove (\ref{RTxx4.3.5}) and therefore we must

2117: have%

2118: \begin{equation}

2119: \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert

2120: ^{2}=\sum_{j=1}^{n}\left\vert \left\langle \int_{a}^{b}f\left( t\right)

2121: dt,e_{j}\right\rangle \right\vert ^{2}  \label{RTxx4.3.13}

2122: \end{equation}%

2123: and%

2124: \begin{equation}

2125: k_{j}\left\Vert f\left( t\right) \right\Vert =\func{Re}\left\langle f\left(

2126: t\right) ,e_{j}\right\rangle \text{ \ and \ }h_{j}\left\Vert f\left(

2127: t\right) \right\Vert =\func{Re}\left\langle f\left( t\right)

2128: ,e_{j}\right\rangle  \label{RTxx4.3.14}

2129: \end{equation}%

2130: for each $j\in \left\{ 1,\dots ,n\right\} $ and a.e. $t\in \left[ a,b\right]

2131: .$

2132: 

2133: From (\ref{RTxx4.3.13}), on using the identity (\ref{RTxx4.3.9}), we deduce

2134: that%

2135: \begin{equation}

2136: \int_{a}^{b}f\left( t\right) dt=\sum_{j=1}^{n}\left\langle

2137: \int_{a}^{b}f\left( t\right) dt,e_{j}\right\rangle e_{j}.  \label{RTxx4.3.15}

2138: \end{equation}%

2139: Now, multiplying the second equality in (\ref{RTxx4.3.14}) with the

2140: imaginary unit $i,$ integrating both inequalities on $\left[ a,b\right] $

2141: and summing them up, we get%

2142: \begin{equation}

2143: \left( k_{j}+ih_{j}\right) \int_{a}^{b}\left\Vert f\left( t\right)

2144: \right\Vert dt=\left\langle \int_{a}^{b}f\left( t\right)

2145: dt,e_{j}\right\rangle  \label{RTxx4.3.16}

2146: \end{equation}%

2147: for each $j\in \left\{ 1,\dots ,n\right\} .$

2148: 

2149: Finally, utilising (\ref{RTxx4.3.15}) and (\ref{RTxx4.3.16}), we deduce (\ref%

2150: {RTxx4.3.6}) and the theorem is proved.

2151: \end{proof}

2152: 

2153: The following corollaries are of interest \cite{RTCSSD4}.

2154: 

2155: \begin{corollary}

2156: \label{RTxxc4.3.2}Let $e_{1},\dots ,e_{m}$ be orthonormal vectors in the

2157: complex Hilbert space $\left( H;\left\langle \cdot ,\cdot \right\rangle

2158: \right) $ and $\rho _{k},\eta _{k}\in \left( 0,1\right) ,$ $k\in \left\{

2159: 1,\dots ,n\right\} .$ If $f\in L\left( \left[ a,b\right] ;H\right) $ is such

2160: that%

2161: \begin{equation*}

2162: \left\Vert f\left( t\right) -e_{k}\right\Vert \leq \rho _{k},\qquad

2163: \left\Vert f\left( t\right) -ie_{k}\right\Vert \leq \eta _{k}

2164: \end{equation*}%

2165: for each $k\in \left\{ 1,\dots ,n\right\} $ and for a.e. $t\in \left[ a,b%

2166: \right] ,$ then we have the inequality%

2167: \begin{equation}

2168: \left[ \sum_{k=1}^{n}\left( 2-\rho _{k}^{2}-\eta _{k}^{2}\right) \right] ^{%

2169: \frac{1}{2}}\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt\leq

2170: \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert .  \label{RTxx4.3.17}

2171: \end{equation}%

2172: The case of equality holds in (\ref{RTxx4.3.17}) if and only if%

2173: \begin{multline}

2174: \quad \int_{a}^{b}f\left( t\right) dt\\=\left(

2175: \int_{a}^{b}\left\Vert f\left(

2176: t\right) \right\Vert dt\right) \sum_{k=1}^{n}\left( \sqrt{1-\rho _{k}^{2}}+i%

2177: \sqrt{1-\eta _{k}^{2}}\right) e_{k}.  \label{RTxx4.3.18}\quad

2178: \end{multline}

2179: \end{corollary}

2180: 

2181: The proof follows by Theorem \ref{RTxxt4.3.2} and is similar to the one from

2182: Corollary \ref{RTxxc4.2.2}. We omit the details.

2183: 

2184: Next, the following result may be stated \cite{RTCSSD4}:

2185: 

2186: \begin{corollary}

2187: \label{RTxxc4.3.3}Let $e_{1},\dots ,e_{m}$ be as in Corollary \ref%

2188: {RTxxc4.3.2} and $M_{k}\geq m_{k}>0,$ $N_{k}\geq n_{k}>0,$ $k\in \left\{

2189: 1,\dots ,n\right\} .$ If $f\in L\left( \left[ a,b\right] ;H\right) $ is such

2190: that either%

2191: \begin{align*}

2192: \func{Re}\left\langle M_{k}e_{k}-f\left( t\right) ,f\left( t\right)

2193: -m_{k}e_{k}\right\rangle & \geq 0,\  \\

2194: \func{Re}\left\langle N_{k}ie_{k}-f\left( t\right) ,f\left( t\right)

2195: -n_{k}ie_{k}\right\rangle & \geq 0

2196: \end{align*}%

2197: or, equivalently,%

2198: \begin{align*}

2199: \left\Vert f\left( t\right) -\frac{M_{k}+m_{k}}{2}e_{k}\right\Vert & \leq

2200: \frac{1}{2}\left( M_{k}-m_{k}\right) ,\  \\

2201: \left\Vert f\left( t\right) -\frac{N_{k}+n_{k}}{2}ie_{k}\right\Vert & \leq

2202: \frac{1}{2}\left( N_{k}-n_{k}\right)

2203: \end{align*}%

2204: for each $k\in \left\{ 1,\dots ,n\right\} $ and a.e. $t\in \left[ a,b\right]

2205: ,$ then we have the inequality%

2206: \begin{multline}

2207: 2\left\{ \sum_{k=1}^{m}\left[ \frac{m_{k}M_{k}}{\left( M_{k}+m_{k}\right)

2208: ^{2}}+\frac{n_{k}N_{k}}{\left( N_{k}+n_{k}\right) ^{2}}\right] \right\} ^{%

2209: \frac{1}{2}}\int_{a}^{b}\left\Vert f\left( t\right) \right\Vert dt

2210: \label{RTxx4.3.19} \\

2211: \leq \left\Vert \int_{a}^{b}f\left( t\right) dt\right\Vert .

2212: \end{multline}%

2213: The case of equality holds in (\ref{RTxx4.3.19}) if and only if%

2214: \begin{multline}

2215: \quad \int_{a}^{b}f\left( t\right) dt=2\left(

2216: \int_{a}^{b}\left\Vert f\left(

2217: t\right) \right\Vert dt\right)   \label{RTxx4.3.20} \\

2218: \times \sum_{k=1}^{n}\left( \frac{\sqrt{m_{k}M_{k}}}{M_{k}+m_{k}}+i\frac{%

2219: \sqrt{n_{k}N_{k}}}{N_{k}+n_{k}}\right) e_{k}.\quad

2220: \end{multline}

2221: \end{corollary}

2222: 

2223: The proof employs Theorem \ref{RTxxt4.3.2} and is similar to the one in

2224: Corollary \ref{RTxxc4.2.3}. We omit the details.

2225: 

2226: \section{Applications for Complex-Valued Functions}

2227: 

2228: The following proposition holds \cite{RTCSSD1}.

2229: 

2230: \begin{proposition}

2231: \label{RTxxp1.4.1}If $f:\left[ a,b\right] \rightarrow \mathbb{C}$ is a

2232: Lebesgue integrable function with the property that there exists a constant $%

2233: K\geq 1$ such that%

2234: \begin{equation}

2235: \left\vert f\left( t\right) \right\vert \leq K\left[ \alpha \func{Re}f\left(

2236: t\right) +\beta \func{Im}f\left( t\right) \right]  \label{RTxx1.4.1}

2237: \end{equation}%

2238: for a.e. $t\in \left[ a,b\right] ,$ where $\alpha ,\beta \in \mathbb{R}$, $%

2239: \alpha ^{2}+\beta ^{2}=1$ are given, then we have the following reverse of

2240: the continuous triangle inequality:%

2241: \begin{equation}

2242: \int_{a}^{b}\left\vert f\left( t\right) \right\vert dt\leq K\left\vert

2243: \int_{a}^{b}f\left( t\right) dt\right\vert .  \label{RTxx1.4.2}

2244: \end{equation}%

2245: The case of equality holds in (\ref{RTxx1.4.2}) if and only if%

2246: \begin{equation*}

2247: \int_{a}^{b}f\left( t\right) dt=\frac{1}{K}\left( \alpha +i\beta \right)

2248: \int_{a}^{b}\left\vert f\left( t\right) \right\vert dt.

2249: \end{equation*}

2250: \end{proposition}

2251: 

2252: The proof is obvious by Theorem \ref{RTxxt1.2.1}, and we omit the details.

2253: 

2254: \begin{remark}

2255: \label{RTxxr1.4.2}If in the above Proposition \ref{RTxxp1.4.1} we choose $%

2256: \alpha =1,$ $\beta =0,$ then the condition (\ref{RTxx1.4.1}) for $\func{Re}%

2257: f\left( t\right) >0$ is equivalent to%

2258: \begin{equation*}

2259: \left[ \func{Re}f\left( t\right) \right] ^{2}+\left[ \func{Im}f\left(

2260: t\right) \right] ^{2}\leq K^{2}\left[ \func{Re}f\left( t\right) \right] ^{2}

2261: \end{equation*}%

2262: or with the inequality:%

2263: \begin{equation*}

2264: \frac{\left\vert \func{Im}f\left( t\right) \right\vert }{\func{Re}f\left(

2265: t\right) }\leq \sqrt{K^{2}-1}.

2266: \end{equation*}%

2267: Now, if we assume that%

2268: \begin{equation}

2269: \left\vert \arg f\left( t\right) \right\vert \leq \theta ,\ \ \ \theta \in

2270: \left( 0,\frac{\pi }{2}\right) ,  \label{RTxx1.4.3}

2271: \end{equation}%

2272: then, for $\func{Re}f\left( t\right) >0,$%

2273: \begin{equation*}

2274: \left\vert \tan \left[ \arg f\left( t\right) \right] \right\vert =\frac{%

2275: \left\vert \func{Im}f\left( t\right) \right\vert }{\func{Re}f\left( t\right)

2276: }\leq \tan \theta ,

2277: \end{equation*}%

2278: and if we choose $K=\frac{1}{\cos \theta }>1,$ then%

2279: \begin{equation*}

2280: \sqrt{K^{2}-1}=\tan \theta ,

2281: \end{equation*}%

2282: and by Proposition \ref{RTxxp1.4.1}, we deduce%

2283: \begin{equation}

2284: \cos \theta \int_{a}^{b}\left\vert f\left( t\right) \right\vert dt\leq

2285: \left\vert \int_{a}^{b}f\left( t\right) dt\right\vert ,  \label{RTxx1.4.4}

2286: \end{equation}%

2287: which is exactly the Karamata inequality (\ref{RTxx0.1.2}) from the

2288: Introduction.

2289: \end{remark}

2290: 

2291: Obviously, the result from Proposition \ref{RTxxp1.4.1} is more

2292: comprehensive since for other values of $\left( \alpha ,\beta \right) \in

2293: \mathbb{R}^{2}$ with $\alpha ^{2}+\beta ^{2}=1$ we can get different

2294: sufficient conditions for the function $f$ such that the inequality (\ref%

2295: {RTxx1.4.2}) holds true.

2296: 

2297: A different sufficient condition in terms of complex disks is incorporated

2298: in the following proposition \cite{RTCSSD1}.

2299: 

2300: \begin{proposition}

2301: \label{RTxxp1.4.3}Let $e=\alpha +i\beta $ with $\alpha ^{2}+\beta ^{2}=1,$ $%

2302: r\in \left( 0,1\right) $ and $f:\left[ a,b\right] \rightarrow \mathbb{C}$ a

2303: Lebesgue integrable function such that%

2304: \begin{equation}

2305: f\left( t\right) \in \bar{D}\left( e,r\right) :=\left\{ z\in \mathbb{C}|%

2306: \text{ }\left\vert z-e\right\vert \leq r\right\} \text{ \ \ for a.e. }t\in %

2307: \left[ a,b\right] .  \label{RTxx1.4.5}

2308: \end{equation}%

2309: Then we have the inequality%

2310: \begin{equation}

2311: \sqrt{1-r^{2}}\int_{a}^{b}\left\vert f\left( t\right) \right\vert dt\leq

2312: \left\vert \int_{a}^{b}f\left( t\right) dt\right\vert .  \label{RTxx1.4.6}

2313: \end{equation}%

2314: The case of equality holds in (\ref{RTxx1.4.6}) if and only if%

2315: \begin{equation*}

2316: \int_{a}^{b}f\left( t\right) dt=\sqrt{1-r^{2}}\left( \alpha +i\beta \right)

2317: \int_{a}^{b}\left\vert f\left( t\right) \right\vert dt.

2318: \end{equation*}

2319: \end{proposition}

2320: 

2321: The proof follows by Corollary \ref{RTxxc1.2.2} and we omit the details.

2322: 

2323: Further, we may state the following proposition as well \cite{RTCSSD1}.

2324: 

2325: \begin{proposition}

2326: \label{RTxxp1.4.4}Let $e=\alpha +i\beta $ with $\alpha ^{2}+\beta ^{2}=1$

2327: and $M\geq m>0.$ If $f:\left[ a,b\right] \rightarrow \mathbb{C}$ is such that%

2328: \begin{equation}

2329: \func{Re}\left[ \left( Me-f\left( t\right) \right) \left( \overline{f\left(

2330: t\right) }-m\overline{e}\right) \right] \geq 0\text{ \ \ for a.e. }t\in %

2331: \left[ a,b\right] ,  \label{RTxx1.4.7}

2332: \end{equation}%

2333: or, equivalently,%

2334: \begin{equation}

2335: \left\vert f\left( t\right) -\frac{M+m}{2}e\right\vert \leq \frac{1}{2}%

2336: \left( M-m\right) \text{ \ \ for a.e. }t\in \left[ a,b\right] ,

2337: \label{RTxx1.4.8}

2338: \end{equation}%

2339: then we have the inequality%

2340: \begin{equation}

2341: \frac{2\sqrt{mM}}{M+m}\int_{a}^{b}\left\vert f\left( t\right) \right\vert

2342: dt\leq \left\vert \int_{a}^{b}f\left( t\right) dt\right\vert ,

2343: \label{RTxx1.4.9}

2344: \end{equation}%

2345: or, equivalently,%

2346: \begin{align}

2347: (0& \leq )\int_{a}^{b}\left\vert f\left( t\right) \right\vert dt-\left\vert

2348: \int_{a}^{b}f\left( t\right) dt\right\vert  \label{RTxx1.4.10} \\

2349: & \leq \frac{\left( \sqrt{M}-\sqrt{m}\right) ^{2}}{M+m}\left\vert

2350: \int_{a}^{b}f\left( t\right) dt\right\vert .  \notag

2351: \end{align}%

2352: The equality holds in (\ref{RTxx1.4.9}) (or in the second part of (\ref%

2353: {RTxx1.4.10})) if and only if%

2354: \begin{equation*}

2355: \int_{a}^{b}f\left( t\right) dt=\frac{2\sqrt{mM}}{M+m}\left( \alpha +i\beta

2356: \right) \int_{a}^{b}\left\vert f\left( t\right) \right\vert dt.

2357: \end{equation*}

2358: \end{proposition}

2359: 

2360: The proof follows by Corollary \ref{RTxxc1.2.3} and we omit the details.

2361: 

2362: \begin{remark}

2363: \label{RTxxr1.4.5}Since%

2364: \begin{align*}

2365: Me-f\left( t\right) & =M\alpha -\func{Re}f\left( t\right) +i\left[ M\beta -%

2366: \func{Im}f\left( t\right) \right] , \\

2367: \overline{f\left( t\right) }-m\overline{e}& =\func{Re}f\left( t\right)

2368: -m\alpha -i\left[ \func{Im}f\left( t\right) -m\beta \right]

2369: \end{align*}%

2370: hence%

2371: \begin{multline}

2372: \func{Re}\left[ \left( Me-f\left( t\right) \right) \left( \overline{f\left(

2373: t\right) }-m\overline{e}\right) \right]  \label{RTxx1.4.11} \\

2374: =\left[ M\alpha -\func{Re}f\left( t\right) \right] \left[ \func{Re}f\left(

2375: t\right) -m\alpha \right] \\

2376: +\left[ M\beta -\func{Im}f\left( t\right) \right] \left[ \func{Im}f\left(

2377: t\right) -m\beta \right] .

2378: \end{multline}%

2379: It is obvious that, if%

2380: \begin{equation}

2381: m\alpha \leq \func{Re}f\left( t\right) \leq M\alpha \text{ \ \ for a.e. }%

2382: t\in \left[ a,b\right] ,  \label{RTxx1.4.12}

2383: \end{equation}%

2384: and

2385: \begin{equation}

2386: m\beta \leq \func{Im}f\left( t\right) \leq M\beta \text{ \ \ for a.e. }t\in %

2387: \left[ a,b\right] ,  \label{RTxx1.4.13}

2388: \end{equation}%

2389: then, by (\ref{RTxx1.4.11}),%

2390: \begin{equation*}

2391: \func{Re}\left[ \left( Me-f\left( t\right) \right) \left( \overline{f\left(

2392: t\right) }-m\overline{e}\right) \right] \geq 0\text{ \ \ for a.e. }t\in %

2393: \left[ a,b\right] ,

2394: \end{equation*}%

2395: and then either (\ref{RTxx1.4.9}) or (\ref{RTxx1.4.12}) hold true.

2396: \end{remark}

2397: 

2398: We observe that the conditions (\ref{RTxx1.4.12}) and (\ref{RTxx1.4.13}) are

2399: very easy to verify in practice and may be useful in various applications

2400: where reverses of the continuous triangle inequality are required.

2401: 

2402: \begin{remark}

2403: Similar results may be stated for functions $f:\left[ a,b\right] \rightarrow

2404: \mathbb{R}^{n}$ or $f:\left[ a,b\right] \rightarrow H,$ with $H$ particular

2405: instances of Hilbert spaces of significance in applications, but we leave

2406: them to the interested reader.

2407: \end{remark}

2408: 

2409: Let $e=\alpha +i\beta $ $\left( \alpha ,\beta \in \mathbb{R}\right) $ be a

2410: complex number with the property that $\left\vert e\right\vert =1,$ i.e., $%

2411: \alpha ^{2}+\beta ^{2}=1.$ The following proposition concerning a reverse of

2412: the continuous triangle inequality for complex-valued functions may be

2413: stated \cite{RTCSSD2}:

2414: 

2415: \begin{proposition}

2416: \label{RTxxp.2.4.1} Let $f:\left[ a,b\right] \rightarrow \mathbb{C}$ be a

2417: Lebesgue integrable function with the property that there exists a constant $%

2418: \rho \in \left( 0,1\right) $ such that

2419: \begin{equation}

2420: \left\vert f\left( t\right) -e\right\vert \leq \rho \text{ for a.e. }t\in %

2421: \left[ a,b\right] ,  \label{RTxxe.2.4.1}

2422: \end{equation}%

2423: where $e$ has been defined above. Then we have the following reverse of the

2424: continuous triangle inequality

2425: \begin{align}

2426: (0& \leq )\int_{a}^{b}\left\vert f\left( t\right) \right\vert dt-\left\vert

2427: \int_{a}^{b}f\left( t\right) dt\right\vert  \label{RTxxe.2.4.2} \\

2428: & \leq \frac{\rho ^{2}}{\sqrt{1-\rho ^{2}}\left( 1+\sqrt{1-\rho ^{2}}\right)

2429: }  \notag \\

2430: & \qquad \times \left[ \alpha \int_{a}^{b}\func{Re}f\left( t\right) dt+\beta

2431: \int_{a}^{b}\func{Im}f\left( t\right) dt\right] .  \notag

2432: \end{align}

2433: \end{proposition}

2434: 

2435: The proof follows by Corollary \ref{RTxxc2.2.2}, and the details are omitted.

2436: 

2437: On the other hand, the following result is perhaps more useful for

2438: applications \cite{RTCSSD2}:

2439: 

2440: \begin{proposition}

2441: \label{RTxxp.2.4.2} Assume that $f$ and $e$ are as in Proposition \ref%

2442: {RTxxp.2.4.1}. If there exists the constants $M\geq m>0$ such that either

2443: \begin{equation}

2444: \func{Re}\left[ \left( Me-f\left( t\right) \right) \left( \overline{f\left(

2445: t\right) }-m\overline{e}\right) \right] \geq 0  \label{RTxxe.2.4.3}

2446: \end{equation}%

2447: or, equivalently,

2448: \begin{equation}

2449: \left\vert f\left( t\right) -\frac{M+m}{2}e\right\vert \leq \frac{1}{2}%

2450: \left( M-m\right)  \label{RTxxe.2.4.4}

2451: \end{equation}%

2452: for a.e. $t\in \left[ a,b\right] ,$ holds, then

2453: \begin{align}

2454: (0& \leq )\int_{a}^{b}\left\vert f\left( t\right) \right\vert dt-\left\vert

2455: \int_{a}^{b}f\left( t\right) dt\right\vert  \label{RTxxe.2.4.5} \\

2456: & \leq \frac{\left( \sqrt{M}-\sqrt{m}\right) ^{2}}{2\sqrt{Mm}}\left[ \alpha

2457: \int_{a}^{b}\func{Re}f\left( t\right) dt+\beta \int_{a}^{b}\func{Im}f\left(

2458: t\right) dt\right] .  \notag

2459: \end{align}

2460: \end{proposition}

2461: 

2462: The proof may be done on utilising Corollary \ref{RTxxc2.2.3}, but we omit

2463: the details

2464: 

2465: Subsequently, on making use of Corollary \ref{RTxxc2.2.5}, one may state the

2466: following result as well \cite{RTCSSD2}:

2467: 

2468: \begin{proposition}

2469: \label{RTxxp.2.4.3} Let $f$ be as in Proposition \ref{RTxxp.2.4.1} and the

2470: measurable functions $K,k:\left[ a,b\right] \rightarrow \lbrack 0,\infty )$

2471: with the property that

2472: \begin{equation*}

2473: \frac{\left( K-k\right) ^{2}}{K+k}\in L\left[ a,b\right]

2474: \end{equation*}%

2475: and

2476: \begin{equation*}

2477: \alpha k\left( t\right) \leq \func{Re}f\left( t\right) \leq \alpha K\left(

2478: t\right) \text{ and }\beta k\left( t\right) \leq \func{Im}f\left( t\right)

2479: \leq \beta K\left( t\right)

2480: \end{equation*}%

2481: for a.e. $t\in \left[ a,b\right] ,$ where $\alpha ,\beta $ are assumed to be

2482: positive and satisfying the condition $\alpha ^{2}+\beta ^{2}=1$. Then the

2483: following reverse of the continuous triangle inequality is valid:

2484: \begin{align*}

2485: (0& \leq )\int_{a}^{b}\left\vert f\left( t\right) \right\vert dt-\left\vert

2486: \int_{a}^{b}f\left( t\right) dt\right\vert \\

2487: & \leq \frac{1}{4}\int_{a}^{b}\frac{\left[ K\left( t\right) -k\left(

2488: t\right) \right] ^{2}}{K\left( t\right) +k\left( t\right) }dt.

2489: \end{align*}%

2490: The constant $\frac{1}{4}$ is best possible in the sense that it cannot be

2491: replaced by a smaller quantity.

2492: \end{proposition}

2493: 

2494: \begin{remark}

2495: \label{RTxxr.2.4.2} One may realise that similar results can be stated if

2496: the Corollaries \ref{RTxxc2.3.2}-\ref{RTxxc2.3.5}\ obtained above are used.

2497: For the sake of brevity, we do not mention them here.

2498: \end{remark}

2499: 

2500: Let $f:\left[ a,b\right] \rightarrow \mathbb{C}$ be a Lebesgue integrable

2501: function and $M\geq 1\geq m\geq 0.$ The condition (\ref{RTxx3.2.7}) from

2502: Theorem \ref{RTxxt3.2.4}, which plays a fundamental role in the results

2503: obtained above, can be translated in this case as%

2504: \begin{equation}

2505: \func{Re}\left[ \left( Mf\left( s\right) -f\left( t\right) \right) \left(

2506: \overline{f\left( t\right) }-m\overline{f\left( s\right) }\right) \right]

2507: \geq 0  \label{RTxxe.3.4.1}

2508: \end{equation}%

2509: for a.e. $a\leq t\leq s\leq b.$

2510: 

2511: Since, obviously%

2512: \begin{multline*}

2513: \func{Re}\left[ \left( Mf\left( s\right) -f\left( t\right) \right) \left(

2514: \overline{f\left( t\right) }-m\overline{f\left( s\right) }\right) \right] \\

2515: =\left[ \left( M\func{Re}f\left( s\right) -\func{Re}f\left( t\right) \right)

2516: \left( \func{Re}f\left( t\right) -m\func{Re}f\left( s\right) \right) \right]

2517: \\

2518: +\left[ \left( M\func{Im}f\left( s\right) -\func{Im}f\left( t\right) \right)

2519: \left( \func{Im}f\left( t\right) -m\func{Im}f\left( s\right) \right) \right]

2520: \end{multline*}%

2521: hence a sufficient condition for the inequality in (\ref{RTxxe.3.4.1}) to

2522: hold is%

2523: \begin{equation}

2524: m\func{Re}f\left( s\right) \leq \func{Re}f\left( t\right) \leq M\func{Re}%

2525: f\left( s\right) \text{ }  \label{RTxxe.3.4.2}

2526: \end{equation}%

2527: and%

2528: \begin{equation*}

2529: m\func{Im}f\left( s\right) \leq \func{Im}f\left( t\right) \leq M\func{Im}%

2530: f\left( s\right)

2531: \end{equation*}%

2532: for a.e. $a\leq t\leq s\leq b.$

2533: 

2534: Utilising Theorems \ref{RTxxt3.2.4}, \ref{RTxxt3.2.5} and \ref{RTxxt3.3.1}

2535: we may state the following results incorporating quadratic reverses of the

2536: continuous triangle inequality \cite{RTCSSD3}:

2537: 

2538: \begin{proposition}

2539: \label{RTxxp.3.4.1} With the above assumptions for $f,M$ and $m,$ and if (%

2540: \ref{RTxxe.3.4.1}) holds true, then we have the inequalities%

2541: \begin{multline*}

2542: \quad \left( \int_{a}^{b}\left\vert f\left( t\right) \right\vert

2543: dt\right)

2544: ^{2}\leq \left\vert \int_{a}^{b}f\left( t\right) dt\right\vert ^{2} \\

2545: +\frac{1}{2}\cdot \frac{\left( M-m\right)

2546: ^{2}}{M+m}\int_{a}^{b}\left( s-a\right) \left\vert f\left(

2547: s\right) \right\vert ^{2}ds,\qquad

2548: \end{multline*}%

2549: \begin{equation*}

2550: \int_{a}^{b}\left\vert f\left( t\right) \right\vert dt\leq \left( \frac{M+m}{%

2551: 2\sqrt{Mm}}\right) ^{\frac{1}{2}}\left\vert \int_{a}^{b}f\left( t\right)

2552: dt\right\vert ,

2553: \end{equation*}%

2554: and%

2555: \begin{equation*}

2556: \int_{a}^{b}\left[ \left( b-s\right) +mM\left( s-a\right) \right] \left\vert

2557: f\left( s\right) \right\vert ^{2}ds\leq \frac{M+m}{2}\left\vert

2558: \int_{a}^{b}f\left( s\right) ds\right\vert ^{2}.

2559: \end{equation*}

2560: \end{proposition}

2561: 

2562: \begin{remark}

2563: \label{RTxxr.3.4.1} One may wonder if there are functions satisfying the

2564: condition (\ref{RTxxe.3.4.2}) above. It suffices to find examples of real

2565: functions $\varphi :\left[ a,b\right] \rightarrow \mathbb{R}$ verifying the

2566: following double inequality%

2567: \begin{equation}

2568: \gamma \varphi \left( s\right) \leq \varphi \left( t\right) \leq \Gamma

2569: \varphi \left( s\right)  \label{RTxxe.3.4.3}

2570: \end{equation}%

2571: for some given $\gamma ,\Gamma $ with $0\leq \gamma \leq 1\leq \Gamma

2572: <\infty $ for a.e. $a\leq t\leq s\leq b.$

2573: 

2574: For this purpose, consider $\psi :\left[ a,b\right] \rightarrow \mathbb{R}$

2575: a differentiable function on $\left( a,b\right) $, continuous on $\left[ a,b%

2576: \right] $ and with the property that there exists $\Theta \geq 0\geq \theta $

2577: such that%

2578: \begin{equation}

2579: \theta \leq \psi ^{\prime }\left( u\right) \leq \Theta \text{ for any }u\in

2580: \left( a,b\right) .  \label{RTxxe.3.4.4}

2581: \end{equation}%

2582: By Lagrange's mean value theorem, we have, for any $a\leq t\leq s\leq b$%

2583: \begin{equation*}

2584: \psi \left( s\right) -\psi \left( t\right) =\psi ^{\prime }\left( \xi

2585: \right) \left( s-t\right)

2586: \end{equation*}%

2587: with $t\leq \xi \leq s.$ Therefore, for $a\leq t\leq s\leq b,$ by (\ref%

2588: {RTxxe.3.4.4}), we have the inequality%

2589: \begin{equation*}

2590: \theta \left( b-a\right) \leq \theta \left( s-t\right) \leq \psi \left(

2591: s\right) -\psi \left( t\right) \leq \Theta \left( s-t\right) \leq \Theta

2592: \left( b-a\right) .

2593: \end{equation*}%

2594: If we choose the function $\varphi :\left[ a,b\right] \rightarrow \mathbb{R}$

2595: given by%

2596: \begin{equation*}

2597: \varphi \left( t\right) :=\exp \left[ -\psi \left( t\right) \right] ,\text{ }%

2598: t\in \left[ a,b\right] ,

2599: \end{equation*}%

2600: and $\gamma :=\exp \left[ \theta \left( b-a\right) \right] \leq 1,$ $\Gamma

2601: :=\exp \left[ \Theta \left( b-a\right) \right] ,$ then (\ref{RTxxe.3.4.3})

2602: holds true for any $a\leq t\leq s\leq b.$

2603: \end{remark}

2604: 

2605: The following reverse of the continuous triangle inequality for

2606: complex-valued functions that improves Karamata's result (\ref{RTxx0.1.1})

2607: holds \cite{RTCSSD4}.

2608: 

2609: \begin{proposition}

2610: \label{RTxxp4.4.1}Let $f\in L\left( \left[ a,b\right] ;\mathbb{C}\right) $

2611: with the property that%

2612: \begin{equation}

2613: 0\leq \varphi _{1}\leq \arg f\left( t\right) \leq \varphi _{2}<\frac{\pi }{2}

2614: \label{RTxx4.4.1}

2615: \end{equation}%

2616: for a.e. $t\in \left[ a,b\right] .$ Then we have the inequality%

2617: \begin{equation}

2618: \sqrt{\sin ^{2}\varphi _{1}+\cos ^{2}\varphi _{2}}\int_{a}^{b}\left\vert

2619: f\left( t\right) \right\vert dt\leq \left\vert \int_{a}^{b}f\left( t\right)

2620: dt\right\vert .  \label{RTxx4.4.2}

2621: \end{equation}%

2622: The equality holds in (\ref{RTxx4.4.2}) if and only if%

2623: \begin{equation}

2624: \int_{a}^{b}f\left( t\right) dt=\left( \cos \varphi _{2}+i\sin \varphi

2625: _{1}\right) \int_{a}^{b}\left\vert f\left( t\right) \right\vert dt.

2626: \label{RTxx4.4.3}

2627: \end{equation}

2628: \end{proposition}

2629: 

2630: \begin{proof}

2631: Let $f\left( t\right) =\func{Re}f\left( t\right) +i\func{Im}f\left( t\right)

2632: .$ We may assume that $\func{Re}f\left( t\right) \geq 0,$ $\func{Im}f\left(

2633: t\right) >0,$ for a.e. $t\in \left[ a,b\right] ,$ since, by (\ref{RTxx4.4.1}%

2634: ), $\frac{\func{Im}f\left( t\right) }{\func{Re}f\left( t\right) }=\tan \left[

2635: \arg f\left( t\right) \right] \in \left[ 0,\frac{\pi }{2}\right) $, for a.e.

2636: $t\in \left[ a,b\right] .$ By (\ref{RTxx4.4.1}), we obviously have%

2637: \begin{equation*}

2638: 0\leq \tan ^{2}\varphi _{1}\leq \left[ \frac{\func{Im}f\left( t\right) }{%

2639: \func{Re}f\left( t\right) }\right] ^{2}\leq \tan ^{2}\varphi _{2},

2640: \end{equation*}%

2641: for a.e. $t\in \left[ a,b\right] ,$ from where we get%

2642: \begin{equation*}

2643: \frac{\left[ \func{Im}f\left( t\right) \right] ^{2}+\left[ \func{Re}f\left(

2644: t\right) \right] ^{2}}{\left[ \func{Re}f\left( t\right) \right] ^{2}}\leq

2645: \frac{1}{\cos {}^{2}\varphi _{2}},

2646: \end{equation*}%

2647: for a.e. $t\in \left[ a,b\right] ,$ and%

2648: \begin{equation*}

2649: \frac{\left[ \func{Im}f\left( t\right) \right] ^{2}+\left[ \func{Re}f\left(

2650: t\right) \right] ^{2}}{\left[ \func{Im}f\left( t\right) \right] ^{2}}\leq

2651: \frac{1+\tan ^{2}\varphi _{1}}{\tan ^{2}\varphi _{1}}=\frac{1}{\sin \varphi

2652: _{1}},

2653: \end{equation*}%

2654: for a.e. $t\in \left[ a,b\right] ,$ giving the simpler inequalities%

2655: \begin{equation*}

2656: \left\vert f\left( t\right) \right\vert \cos \varphi _{2}\leq \func{Re}%

2657: \left( f\left( t\right) \right) ,\quad \left\vert f\left( t\right)

2658: \right\vert \sin \varphi _{1}\leq \func{Im}\left( f\left( t\right) \right)

2659: \end{equation*}%

2660: for a.e. $t\in \left[ a,b\right] .$

2661: 

2662: Now, applying Theorem \ref{RTxxt4.2.1} for the complex Hilbert space $%

2663: \mathbb{C}$ endowed with the inner product $\left\langle z,w\right\rangle

2664: =z\cdot \bar{w} $ for $k_{1}=\cos \varphi _{2},$ $k_{2}=\sin \varphi _{1}$

2665: and $e=1,$ we deduce the desired inequality (\ref{RTxx4.4.2}). The case of

2666: equality is also obvious and we omit the details.

2667: \end{proof}

2668: 

2669: Another result that has an obvious geometrical interpretation is the

2670: following one \cite{RTCSSD4}.

2671: 

2672: \begin{proposition}

2673: \label{RTxxp4.4.2}Let $e\in \mathbb{C}$ with $\left\vert e\right\vert =1$

2674: and $\rho _{1},\rho _{2}\in \left( 0,1\right) .$ If $f\left( t\right) \in

2675: L\left( \left[ a,b\right] ;\mathbb{C}\right) $ such that%

2676: \begin{equation}

2677: \left\vert f\left( t\right) -e\right\vert \leq \rho _{1},\quad \left\vert

2678: f\left( t\right) -ie\right\vert \leq \rho _{2}\qquad \text{for a.e. }t\in %

2679: \left[ a,b\right] ,  \label{RTxx4.4.5}

2680: \end{equation}%

2681: then we have the inequality%

2682: \begin{equation}

2683: \sqrt{2-\rho _{1}^{2}-\rho _{2}^{2}}\int_{a}^{b}\left\vert f\left( t\right)

2684: \right\vert dt\leq \left\vert \int_{a}^{b}f\left( t\right) dt\right\vert ,

2685: \label{RTxx4.4.6}

2686: \end{equation}%

2687: with equality if and only if%

2688: \begin{equation}

2689: \int_{a}^{b}f\left( t\right) dt=\left( \sqrt{1-\rho _{1}^{2}}+i\sqrt{1-\rho

2690: _{2}^{2}}\right) \int_{a}^{b}\left\vert f\left( t\right) \right\vert dt\cdot

2691: e.  \label{RTxx4.4.7}

2692: \end{equation}

2693: \end{proposition}

2694: 

2695: The proof is obvious by Corollary \ref{RTxxc4.2.2} applied for $H=\mathbb{C}$

2696: and we omit the details.

2697: 

2698: \begin{remark}

2699: If we choose $e=1,$ and for $\rho _{1},\rho _{2}\in \left( 0,1\right) $ we

2700: define

2701: \begin{equation*}

2702: \bar{D}\left( 1,\rho _{1}\right) :=\left\{ z\in \mathbb{C}|\left\vert

2703: z-1\right\vert \leq \rho _{1}\right\} ,\ \ \bar{D}\left( i,\rho _{2}\right)

2704: :=\left\{ z\in \mathbb{C}|\left\vert z-i\right\vert \leq \rho _{2}\right\} ,

2705: \end{equation*}%

2706: then obviously the intersection domain%

2707: \begin{equation*}

2708: S_{\rho _{1},\rho _{2}}:=\bar{D}\left( 1,\rho _{1}\right) \cap \bar{D}\left(

2709: i,\rho _{2}\right)

2710: \end{equation*}%

2711: is nonempty if and only if $\rho _{1}+\rho _{2}>\sqrt{2}.$

2712: 

2713: If $f\left( t\right) \in S_{\rho _{1},\rho _{2}}$ for a.e. $t\in \left[ a,b%

2714: \right] ,$ then (\ref{RTxx4.4.6}) holds true. The equality holds in (\ref%

2715: {RTxx4.4.6}) if and only if%

2716: \begin{equation*}

2717: \int_{a}^{b}f\left( t\right) dt=\left( \sqrt{1-\rho _{1}^{2}}+i\sqrt{1-\rho

2718: _{2}^{2}}\right) \int_{a}^{b}\left\vert f\left( t\right) \right\vert dt.

2719: \end{equation*}

2720: \end{remark}

2721: 

2722: \begin{thebibliography}{99}

2723: \bibitem{RTCDM} J.B. DIAZ and F.T. METCALF, A complementary triangle

2724: inequality in Hilbert and Banach spaces, \textit{Proceedings Amer. Math.

2725: Soc., }\textbf{17}(1) (1966), 88-97.

2726: 

2727: \bibitem{RTCSSD1} S.S. DRAGOMIR, Reverses of the continuous triangle

2728: inequality for Bochner integral of vector valued function in Hilbert spaces.

2729: Preprint, \textit{RGMIA Res. Rep. Coll., }\textbf{7}(2004)\textit{, }%

2730: Supplement\textit{, }Article 11\textit{, }\texttt{[Online

2731: http://rgmia.vu.edu.au/v7(E).html]}.

2732: 

2733: \bibitem{RTCSSD2} S.S. DRAGOMIR, Additive reverses of the continuous

2734: triangle inequality for Bochner integral of vector valued functions in

2735: Hilbert spaces. Preprint\textit{,} \textit{RGMIA Res. Rep. Coll., }\textbf{7}%

2736: (2004)\textit{, }Supplement\textit{, }Article 12\textit{, }\texttt{[Online

2737: http://rgmia.vu.edu.au/v7(E).html]}.

2738: 

2739: \bibitem{RTCSSD3} S.S. DRAGOMIR, Quadratic reverses of the continuous

2740: triangle inequality for Bochner integral of vector-valued functions in

2741: Hilbert spaces, Preprint, \textit{RGMIA Res. Rep. Coll., }\textbf{7}(2004)%

2742: \textit{, }Supplement\textit{, }Article 8\textit{, }\texttt{[Online

2743: http://rgmia.vu.edu.au/v7(E).html]}.

2744: 

2745: \bibitem{RTCSSD4} S.S. DRAGOMIR, Some reverses of the continuous triangle

2746: inequality for Bochner integral of vector-valued functions in complex

2747: Hilbert spaces, Preprint, \textit{RGMIA Res. Rep. Coll., }\textbf{7}(2004)%

2748: \textit{, }Supplement\textit{, }Article 13\textit{, }\texttt{[Online

2749: http://rgmia.vu.edu.au/v7(E).html]}.

2750: 

2751: \bibitem{RTCK} J. KARAMATA, \textit{Teorija i Praksa Stieltjesova Integrala}

2752: (Serbo-Croatian) (Stieltjes Integral, Theory and Practice), SANU, Posebna

2753: izdanja, 154, Beograd, 1949.

2754: 

2755: \bibitem{RTCM} M. MARDEN, \textit{The Geometry of the Zeros of a Polynomial

2756: in a Complex Variable}, Amer. Math. Soc. Math. Surveys, \textbf{3}, New

2757: York, 1949.

2758: 

2759: \bibitem{RTCMPF} D.S. MITRINOVI\'{C}, J.E. PE\v{C}ARI\'{C} and\ A.M. FINK,

2760: \textit{Classical and New Inequalities in Analysis, }Kluwer Academic

2761: Publishers, Dordrecht/Boston/London, 1993.

2762: 

2763: \bibitem{RTCP} M. PETROVICH, Module d'une somme, \textit{L' Ensignement Math%

2764: \'{e}matique,} \textbf{19} (1917), 53-56.

2765: 

2766: \bibitem{RTCW} H.S. WILF, Some applications of the inequality of arithmetic

2767: and geometric means to polynomial equations, \textit{Proceedings Amer. Math.

2768: Soc., }\textbf{14} (1963), 263-265.

2769: \end{thebibliography}

2770: