1: %TCIDATA{Version=4.10.0.2363}

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

3: 

4: 

5: \chapter{Reverses for the Triangle Inequality}\label{ch3}

6: 

7: \section{Introduction}

8: 

9: The following reverse of the generalised triangle inequality%

10: \begin{equation*}

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

12: \sum_{k=1}^{n}z_{k}\right\vert ,

13: \end{equation*}%

14: provided the complex numbers $z_{k},$ $k\in \left\{ 1,\dots ,n\right\} $

15: satisfy the assumption%

16: \begin{equation*}

17: a-\theta \leq \arg \left( z_{k}\right) \leq a+\theta ,\ \ \text{for any \ }%

18: k\in \left\{ 1,\dots ,n\right\} ,

19: \end{equation*}%

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

21: first discovered by M. Petrovich in 1917, \cite{RTxP} (see \cite[p. 492]%

22: {RTxMPF}) and subsequently was rediscovered by other authors, including J.

23: Karamata \cite[p. 300 -- 301]{RTxKA}, H.S. Wilf \cite{RTxW}, and in an

24: equivalent form by M. Marden \cite{RTxMA}.

25: 

26: In 1966, J.B. Diaz and F.T. Metcalf \cite{RTxDM} proved the following

27: reverse of the triangle inequality:

28: 

29: \begin{theorem}[Diaz-Metcalf, 1966]

30: \label{RT0.ta} Let $a$ be a unit vector in the inner product space $\left(

31: H;\left\langle \cdot ,\cdot \right\rangle \right) $ over the real or complex

32: number field $\mathbb{K}$. Suppose that the vectors $x_{i}\in H\backslash

33: \left\{ 0\right\} ,$ $i\in \left\{ 1,\dots ,n\right\} $ satisfy%

34: \begin{equation}

35: 0\leq r\leq \frac{\func{Re}\left\langle x_{i},a\right\rangle }{\left\Vert

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

37: \label{RT0.1.1}

38: \end{equation}%

39: Then%

40: \begin{equation}

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

42: \sum_{i=1}^{n}x_{i}\right\Vert ,  \label{RT0.1.2}

43: \end{equation}%

44: where equality holds if and only if%

45: \begin{equation}

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

47: \right) a.  \label{RT0.1.3}

48: \end{equation}

49: \end{theorem}

50: 

51: A generalisation of this result for orthonormal families is incorporated in

52: the following result \cite{RTxDM}.

53: 

54: \begin{theorem}[Diaz-Metcalf, 1966]

55: \label{RT0.tb}Let $a_{1},\dots ,a_{n}$ be orthonormal vectors in $H.$

56: Suppose the vectors $x_{1},\dots ,x_{n}\in H\backslash \left\{ 0\right\} $

57: satisfy%

58: \begin{equation}

59: 0\leq r_{k}\leq \frac{\func{Re}\left\langle x_{i},a_{k}\right\rangle }{%

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

61: k\in \left\{ 1,\dots ,m\right\} .  \label{RT0.1.4}

62: \end{equation}%

63: Then%

64: \begin{equation}

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

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

67: \label{RT0.1.5}

68: \end{equation}%

69: where equality holds if and only if%

70: \begin{equation}

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

72: \sum_{k=1}^{m}r_{k}a_{k}.  \label{RT0.1.6}

73: \end{equation}

74: \end{theorem}

75: 

76: Similar results valid for semi-inner products may be found in \cite{RTxK}

77: and \cite{RTxM}.

78: 

79: For other classical inequalities related to the triangle inequality, see

80: Chapter XVII of the book \cite{RTxMPF} and the references therein.

81: 

82: The aim of the present chapter is to provide various recent reverses for the

83: generalised triangle inequality in both its simple form that are closely

84: related to the Diaz-Metcalf results mentioned above, or in the equivalent

85: quadratic form, i.e., upper bounds for

86: 

87: \begin{equation*}

88: \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}-\left\Vert

89: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}

90: \end{equation*}%

91: and

92: 

93: \begin{equation*}

94: \frac{\left\Vert \sum_{i=1}^{n}x_{i}\right\Vert ^{2}}{\left(

95: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}}.

96: \end{equation*}

97: 

98: Applications for vector valued integral inequalities and for complex numbers

99: are given as well.

100: 

101: \section{Some Inequalities of Diaz-Metcalf Type}

102: 

103: \subsection{The Case of One Vector}

104: 

105: The following result with a natural geometrical meaning holds \cite{RTxSSD1}:

106: 

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

108: \label{RTt2.1.1}Let $a$ be a unit vector in the inner product space $\left(

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

110: 0,1\right) .$ If $x_{i}\in H,$ $i\in \left\{ 1,\dots ,n\right\} $ are such

111: that%

112: \begin{equation}

113: \left\Vert x_{i}-a\right\Vert \leq \rho \text{ \ for each \ }i\in \left\{

114: 1,\dots ,n\right\} ,  \label{RT2.1.1}

115: \end{equation}%

116: then we have the inequality%

117: \begin{equation}

118: \sqrt{1-\rho ^{2}}\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \leq \left\Vert

119: \sum_{i=1}^{n}x_{i}\right\Vert ,  \label{RT2.2.1}

120: \end{equation}%

121: with equality if and only if%

122: \begin{equation}

123: \sum_{i=1}^{n}x_{i}=\sqrt{1-\rho ^{2}}\left( \sum_{i=1}^{n}\left\Vert

124: x_{i}\right\Vert \right) a.  \label{RT2.3.1}

125: \end{equation}

126: \end{theorem}

127: 

128: \begin{proof}

129: From (\ref{RT2.1.1}) we have%

130: \begin{equation*}

131: \left\Vert x_{i}\right\Vert ^{2}-2\func{Re}\left\langle x_{i},a\right\rangle

132: +1\leq \rho ^{2},

133: \end{equation*}%

134: giving%

135: \begin{equation}

136: \left\Vert x_{i}\right\Vert ^{2}+1-\rho ^{2}\leq 2\func{Re}\left\langle

137: x_{i},a\right\rangle ,  \label{RT2.4.1}

138: \end{equation}%

139: for each $i\in \left\{ 1,\dots ,n\right\} .$

140: 

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

142: \begin{equation}

143: \frac{\left\Vert x_{i}\right\Vert ^{2}}{\sqrt{1-\rho ^{2}}}+\sqrt{1-\rho ^{2}%

144: }\leq \frac{2\func{Re}\left\langle x_{i},a\right\rangle }{\sqrt{1-\rho ^{2}}}%

145: ,  \label{RT2.5.1}

146: \end{equation}%

147: for each $i\in \left\{ 1,\dots ,n\right\} .$

148: 

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

150: \begin{equation}

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

152: \label{RT2.5.1.a}

153: \end{equation}%

154: we have%

155: \begin{equation}

156: 2\left\Vert x_{i}\right\Vert \leq \frac{\left\Vert x_{i}\right\Vert ^{2}}{%

157: \sqrt{1-\rho ^{2}}}+\sqrt{1-\rho ^{2}}  \label{RT2.6.1}

158: \end{equation}%

159: and thus, by (\ref{RT2.5.1}) and (\ref{RT2.6.1}), we deduce%

160: \begin{equation*}

161: \frac{\func{Re}\left\langle x_{i},a\right\rangle }{\left\Vert

162: x_{i}\right\Vert }\geq \sqrt{1-\rho ^{2}},

163: \end{equation*}%

164: for each $i\in \left\{ 1,\dots ,n\right\} .$ Applying Theorem \ref{RT0.ta}

165: for $r=\sqrt{1-\rho ^{2}},$ we deduce the desired inequality (\ref{RT2.2.1}).

166: \end{proof}

167: 

168: The following results may be stated as well \cite{RTxSSD1}.

169: 

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

171: \label{RTt2.3.1}Let $a$ be a unit vector in the inner product space $\left(

172: H;\left\langle \cdot ,\cdot \right\rangle \right) $ and $M\geq m>0.$ If $%

173: x_{i}\in H,$ $i\in \left\{ 1,\dots ,n\right\} $ are such that either%

174: \begin{equation}

175: \func{Re}\left\langle Ma-x_{i},x_{i}-ma\right\rangle \geq 0  \label{RT2.14.1}

176: \end{equation}%

177: or, equivalently,%

178: \begin{equation}

179: \left\Vert x_{i}-\frac{M+m}{2}\cdot a\right\Vert \leq \frac{1}{2}\left(

180: M-m\right)  \label{RT2.15.1}

181: \end{equation}%

182: holds for each $i\in \left\{ 1,\dots ,n\right\} ,$ then we have the

183: inequality%

184: \begin{equation}

185: \frac{2\sqrt{mM}}{m+M}\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \leq

186: \left\Vert \sum_{i=1}^{n}x_{i}\right\Vert ,  \label{RT2.16.1}

187: \end{equation}%

188: or, equivalently,

189: \begin{equation}

190: \left( 0\leq \right) \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\left\Vert

191: \sum_{i=1}^{n}x_{i}\right\Vert \leq \frac{\left( \sqrt{M}-\sqrt{m}\right)

192: ^{2}}{2\sqrt{mM}}\left\Vert \sum_{i=1}^{n}x_{i}\right\Vert .

193: \label{RT2.17.1}

194: \end{equation}%

195: The equality holds in (\ref{RT2.16.1}) (or in (\ref{RT2.17.1})) if and only

196: if%

197: \begin{equation}

198: \sum_{i=1}^{n}x_{i}=\frac{2\sqrt{mM}}{m+M}\left( \sum_{i=1}^{n}\left\Vert

199: x_{i}\right\Vert \right) a.  \label{RT2.19.1}

200: \end{equation}

201: \end{theorem}

202: 

203: \begin{proof}

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

205: equivalent:

206: 

207: \begin{enumerate}

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

209: 

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

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

212: \end{enumerate}

213: 

214: Using this fact, one may simply realize that (\ref{RT2.14.1}) and (\ref%

215: {RT2.15.1}) are equivalent.

216: 

217: Now, from (\ref{RT2.14.1}), we get%

218: \begin{equation*}

219: \left\Vert x_{i}\right\Vert ^{2}+mM\leq \left( M+m\right) \func{Re}%

220: \left\langle x_{i},a\right\rangle ,

221: \end{equation*}%

222: for any $i\in \left\{ 1,\dots ,n\right\} .$ Dividing this inequality by $%

223: \sqrt{mM}>0,$ we deduce the following inequality that will be used in the

224: sequel%

225: \begin{equation}

226: \frac{\left\Vert x_{i}\right\Vert ^{2}}{\sqrt{mM}}+\sqrt{mM}\leq \frac{M+m}{%

227: \sqrt{mM}}\func{Re}\left\langle x_{i},a\right\rangle ,  \label{RT2.20.1}

228: \end{equation}%

229: for each $i\in \left\{ 1,\dots ,n\right\} .$

230: 

231: Using the inequality (\ref{RT2.5.1.a}) from Theorem \ref{RTt2.1.1}, we also

232: have%

233: \begin{equation}

234: 2\left\Vert x_{i}\right\Vert \leq \frac{\left\Vert x_{i}\right\Vert ^{2}}{%

235: \sqrt{mM}}+\sqrt{mM},  \label{RT2.21.1}

236: \end{equation}%

237: for each $i\in \left\{ 1,\dots ,n\right\} .$

238: 

239: Utilizing (\ref{RT2.20.1}) and (\ref{RT2.21.1}), we may conclude with the

240: following inequality%

241: \begin{equation*}

242: \left\Vert x_{i}\right\Vert \leq \frac{M+m}{\sqrt{mM}}\func{Re}\left\langle

243: x_{i},a\right\rangle ,

244: \end{equation*}%

245: which is equivalent to%

246: \begin{equation}

247: \frac{2\sqrt{mM}}{m+M}\leq \frac{\func{Re}\left\langle x_{i},a\right\rangle

248: }{\left\Vert x_{i}\right\Vert }  \label{RT2.22.1}

249: \end{equation}%

250: for any $i\in \left\{ 1,\dots ,n\right\} .$

251: 

252: Finally, on applying the Diaz-Metcalf result in Theorem \ref{RT0.ta} for $r=%

253: \frac{2\sqrt{mM}}{m+M}$, we deduce the desired conclusion.

254: 

255: The equivalence between (\ref{RT2.16.1}) and (\ref{RT2.17.1}) follows by

256: simple calculation and we omit the details.

257: \end{proof}

258: 

259: \subsection{The Case of $m$ Vectors}

260: 

261: In a similar manner to the one used in the proof of Theorem \ref{RTt2.1.1}

262: and by the use of the Diaz-Metcalf inequality incorporated in Theorem \ref%

263: {RT0.tb}, we can also prove the following result \cite{RTxSSD1} :

264: 

265: \begin{proposition}

266: \label{RTt2.2.1}Let $a_{1},\dots ,a_{n}$ be orthonormal vectors in $H.$

267: Suppose the vectors $x_{1},\dots ,x_{n}\in H\backslash \left\{ 0\right\} $

268: satisfy%

269: \begin{equation}

270: \left\Vert x_{i}-a_{k}\right\Vert \leq \rho _{k}\text{ \ for each}\ \ i\in

271: \left\{ 1,\dots ,n\right\} ,\ k\in \left\{ 1,\dots ,m\right\} ,

272: \label{RT2.11.1}

273: \end{equation}%

274: where $\rho _{k}\in \left( 0,1\right) ,k\in \left\{ 1,\dots ,m\right\} .$

275: Then we have the following reverse of the triangle inequality%

276: \begin{equation}

277: \left( m-\sum_{k=1}^{m}\rho _{k}^{2}\right) ^{\frac{1}{2}}\sum_{i=1}^{n}%

278: \left\Vert x_{i}\right\Vert \leq \left\Vert \sum_{i=1}^{n}x_{i}\right\Vert .

279: \label{RT2.12.1}

280: \end{equation}%

281: The equality holds in (\ref{RT2.12.1}) if and only if%

282: \begin{equation}

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

284: \sum_{k=1}^{m}\left( 1-\rho _{k}^{2}\right) ^{\frac{1}{2}}a_{k}.

285: \label{RT2.13.1}

286: \end{equation}

287: \end{proposition}

288: 

289: Finally, by the use of Theorem \ref{RT0.tb} and a similar technique to that

290: employed in the proof of Theorem \ref{RTt2.3.1}, we may state the following

291: result \cite{RTxSSD1}:

292: 

293: \begin{proposition}

294: \label{RTt2.4.1}Let $a_{1},\dots ,a_{n}$ be orthonormal vectors in $H.$

295: Suppose the vectors $x_{1},\dots ,x_{n}\in H\backslash \left\{ 0\right\} $

296: satisfy%

297: \begin{equation}

298: \func{Re}\left\langle M_{k}a_{k}-x_{i},x_{i}-\mu _{k}a_{k}\right\rangle \geq

299: 0,  \label{RT2.23.1}

300: \end{equation}%

301: or, equivalently,%

302: \begin{equation}

303: \left\Vert x_{i}-\frac{M_{k}+\mu _{k}}{2}a_{k}\right\Vert \leq \frac{1}{2}%

304: \left( M_{k}-\mu _{k}\right) ,  \label{RT2.24.1}

305: \end{equation}%

306: for any $i\in \left\{ 1,\dots ,n\right\} $ and $k\in \left\{ 1,\dots

307: ,m\right\} ,$ where $M_{k}\geq \mu _{k}>0$ for each $k\in \left\{ 1,\dots

308: ,m\right\} .$

309: 

310: Then we have the inequality%

311: \begin{equation}

312: 2\left( \sum_{k=1}^{m}\frac{\mu _{k}M_{k}}{\left( \mu _{k}+M_{k}\right) ^{2}}%

313: \right) ^{\frac{1}{2}}\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \leq

314: \left\Vert \sum_{i=1}^{n}x_{i}\right\Vert .  \label{RT2.25.1}

315: \end{equation}%

316: The equality holds in (\ref{RT2.25.1}) iff%

317: \begin{equation}

318: \sum_{i=1}^{n}x_{i}=2\left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert

319: \right) \sum_{k=1}^{m}\frac{\sqrt{\mu _{k}M_{k}}}{\mu _{k}+M_{k}}a_{k}.

320: \label{RT2.26.1}

321: \end{equation}

322: \end{proposition}

323: 

324: \section{Additive Reverses for the Triangle Inequality}

325: 

326: \subsection{The Case of One Vector}

327: 

328: In this section we establish some additive reverses of the generalised

329: triangle inequality in real or complex inner product spaces.

330: 

331: The following result holds \cite{RTxSSD1}:

332: 

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

334: \label{RTt3.1.1}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

335: \right) $ be an inner product space over the real or complex number field $%

336: \mathbb{K}$ and $e,$ $x_{i}\in H,$ $i\in \left\{ 1,\dots ,n\right\} $ with $%

337: \left\Vert e\right\Vert =1.$ If $k_{i}\geq 0$, $i\in \left\{ 1,\dots

338: ,n\right\} ,$ are such that%

339: \begin{equation}

340: \left\Vert x_{i}\right\Vert -\func{Re}\left\langle e,x_{i}\right\rangle \leq

341: k_{i}\text{\ \ for each}\ \ i\in \left\{ 1,\dots ,n\right\} ,

342: \label{RT3.1.1}

343: \end{equation}%

344: then we have the inequality%

345: \begin{equation}

346: \left( 0\leq \right) \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\left\Vert

347: \sum_{i=1}^{n}x_{i}\right\Vert \leq \sum_{i=1}^{n}k_{i}.  \label{RT3.2.1}

348: \end{equation}%

349: The equality holds in (\ref{RT3.2.1}) if and only if%

350: \begin{equation}

351: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \geq \sum_{i=1}^{n}k_{i}

352: \label{RT3.3.1}

353: \end{equation}%

354: and%

355: \begin{equation}

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

357: -\sum_{i=1}^{n}k_{i}\right) e.  \label{RT3.4.1}

358: \end{equation}

359: \end{theorem}

360: 

361: \begin{proof}

362: If we sum in (\ref{RT3.1.1}) over $i$ from 1 to $n,$ then we get%

363: \begin{equation}

364: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \leq \func{Re}\left\langle

365: e,\sum_{i=1}^{n}x_{i}\right\rangle +\sum_{i=1}^{n}k_{i}.  \label{RT3.5.1}

366: \end{equation}%

367: By Schwarz's inequality for $e$ and $\sum_{i=1}^{n}x_{i},$ we have%

368: \begin{align}

369: \func{Re}\left\langle e,\sum_{i=1}^{n}x_{i}\right\rangle & \leq \left\vert

370: \func{Re}\left\langle e,\sum_{i=1}^{n}x_{i}\right\rangle \right\vert

371: \label{RT3.6.1} \\

372: & \leq \left\vert \left\langle e,\sum_{i=1}^{n}x_{i}\right\rangle

373: \right\vert \leq \left\Vert e\right\Vert \left\Vert

374: \sum_{i=1}^{n}x_{i}\right\Vert =\left\Vert \sum_{i=1}^{n}x_{i}\right\Vert .

375: \notag

376: \end{align}%

377: Making use of (\ref{RT3.5.1}) and (\ref{RT3.6.1}), we deduce the desired

378: inequality (\ref{RT3.1.1}).

379: 

380: If (\ref{RT3.3.1}) and (\ref{RT3.4.1}) hold, then%

381: \begin{equation*}

382: \left\Vert \sum_{i=1}^{n}x_{i}\right\Vert =\left\vert

383: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\sum_{i=1}^{n}k_{i}\right\vert

384: \left\Vert e\right\Vert =\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert

385: -\sum_{i=1}^{n}k_{i},

386: \end{equation*}%

387: and the equality in the second part of (\ref{RT3.2.1}) holds true.

388: 

389: Conversely, if the equality holds in (\ref{RT3.2.1}), then, obviously (\ref%

390: {RT3.3.1}) is valid and we need only to prove (\ref{RT3.4.1}).

391: 

392: Now, if the equality holds in (\ref{RT3.2.1}) then it must hold in (\ref%

393: {RT3.1.1}) for each $i\in \left\{ 1,\dots ,n\right\} $ and also must hold in

394: any of the inequalities in (\ref{RT3.6.1}).

395: 

396: It is well known that in Schwarz's inequality $\left\vert \left\langle

397: u,v\right\rangle \right\vert \leq \left\Vert u\right\Vert \left\Vert

398: v\right\Vert $ $\left( u,v\in H\right) $ the case of equality holds iff

399: there exists a $\lambda \in \mathbb{K}$ such that $u=\lambda v.$ We note

400: that in the weaker inequality $\func{Re}\left\langle u,v\right\rangle \leq

401: \left\Vert u\right\Vert \left\Vert v\right\Vert $ the case of equality holds

402: iff $\lambda \geq 0$ and $u=\lambda v.$

403: 

404: Consequently, the equality holds in all inequalities (\ref{RT3.6.1})

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

406: \begin{equation}

407: \mu e=\sum_{i=1}^{n}x_{i}.  \label{RT3.7.1}

408: \end{equation}

409: 

410: If we sum the equalities in (\ref{RT3.1.1}) over $i$ from 1 to $n,$ then we

411: deduce%

412: \begin{equation}

413: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\func{Re}\left\langle

414: e,\sum_{i=1}^{n}x_{i}\right\rangle =\sum_{i=1}^{n}k_{i}.  \label{RT3.8.1}

415: \end{equation}%

416: Replacing $\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert $ from (\ref{RT3.7.1})

417: into (\ref{RT3.8.1}), we deduce%

418: \begin{equation*}

419: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\mu \left\Vert e\right\Vert

420: ^{2}=\sum_{i=1}^{n}k_{i},

421: \end{equation*}%

422: from where we get $\mu =\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert

423: -\sum_{i=1}^{n}k_{i}.$ Using (\ref{RT3.7.1}), we deduce (\ref{RT3.4.1}) and

424: the theorem is proved.

425: \end{proof}

426: 

427: \subsection{The Case of $m$ Vectors}

428: 

429: If we turn our attention to the case of orthogonal families, then we may

430: state the following result as well \cite{RTxSSD1}.

431: 

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

433: \label{RTt3.2.1}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

434: \right) \ $be an inner product space over the real or complex number field $%

435: \mathbb{K}$, $\left\{ e_{k}\right\} _{k\in \left\{ 1,\dots ,m\right\} }$ a

436: family of orthonormal vectors in $H,$ $x_{i}\in H,$ $M_{i,k}\geq 0$ for $%

437: i\in \left\{ 1,\dots ,n\right\} $ and $k\in \left\{ 1,\dots ,m\right\} $

438: such that%

439: \begin{equation}

440: \left\Vert x_{i}\right\Vert -\func{Re}\left\langle e_{k},x_{i}\right\rangle

441: \leq M_{ik}\text{ }  \label{RT3.9.1}

442: \end{equation}%

443: \ for each$\ \ i\in \left\{ 1,\dots ,n\right\} ,\ k\in \left\{ 1,\dots

444: ,m\right\} .$ Then we have the inequality%

445: \begin{equation}

446: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \leq \frac{1}{\sqrt{m}}\left\Vert

447: \sum_{i=1}^{n}x_{i}\right\Vert +\frac{1}{m}\sum_{i=1}^{n}%

448: \sum_{k=1}^{m}M_{ik}.  \label{RT3.10.1}

449: \end{equation}%

450: The equality holds true in (\ref{RT3.10.1}) if and only if

451: \begin{equation}

452: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \geq \frac{1}{m}%

453: \sum_{i=1}^{n}\sum_{k=1}^{m}M_{ik}  \label{RT3.11.1}

454: \end{equation}%

455: and%

456: \begin{equation}

457: \sum_{i=1}^{n}x_{i}=\left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\frac{1%

458: }{m}\sum_{i=1}^{n}\sum_{k=1}^{m}M_{ik}\right) \sum_{k=1}^{m}e_{k}.

459: \label{RT3.12.1}

460: \end{equation}

461: \end{theorem}

462: 

463: \begin{proof}

464: If we sum over $i$ from 1 to $n$ in (\ref{RT3.9.1}), then we obtain%

465: \begin{equation*}

466: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \leq \func{Re}\left\langle

467: e,\sum_{i=1}^{n}x_{i}\right\rangle +\sum_{i=1}^{n}M_{ik},

468: \end{equation*}%

469: for each $k\in \left\{ 1,\dots ,m\right\} .$ Summing these inequalities over

470: $k$ from 1 to $m,$ we deduce%

471: \begin{equation}

472: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \leq \frac{1}{m}\func{Re}%

473: \left\langle \sum_{k=1}^{m}e_{k},\sum_{i=1}^{n}x_{i}\right\rangle +\frac{1}{m%

474: }\sum_{i=1}^{n}\sum_{k=1}^{m}M_{ik}.  \label{RT3.13.1}

475: \end{equation}%

476: By Schwarz's inequality for $\sum_{k=1}^{m}e_{k}$ and $\sum_{i=1}^{n}x_{i}$

477: we have%

478: \begin{align}

479: \func{Re}\left\langle \sum_{k=1}^{m}e_{k},\sum_{i=1}^{n}x_{i}\right\rangle &

480: \leq \left\vert \func{Re}\left\langle

481: \sum_{k=1}^{m}e_{k},\sum_{i=1}^{n}x_{i}\right\rangle \right\vert

482: \label{RT3.14.1} \\

483: & \leq \left\vert \left\langle

484: \sum_{k=1}^{m}e_{k},\sum_{i=1}^{n}x_{i}\right\rangle \right\vert  \notag \\

485: & \leq \left\Vert \sum_{k=1}^{m}e_{k}\right\Vert \left\Vert

486: \sum_{i=1}^{n}x_{i}\right\Vert  \notag \\

487: & =\sqrt{m}\left\Vert \sum_{i=1}^{n}x_{i}\right\Vert ,  \notag

488: \end{align}%

489: since, obviously,%

490: \begin{equation*}

491: \left\Vert \sum_{k=1}^{m}e_{k}\right\Vert =\sqrt{\left\Vert

492: \sum_{k=1}^{m}e_{k}\right\Vert ^{2}}=\sqrt{\sum_{k=1}^{m}\left\Vert

493: e_{k}\right\Vert ^{2}}=\sqrt{m}.

494: \end{equation*}%

495: Making use of (\ref{RT3.13.1}) and (\ref{RT3.14.1}), we deduce the desired

496: inequality (\ref{RT3.10.1}).

497: 

498: If (\ref{RT3.11.1}) and (\ref{RT3.12.1}) hold, then%

499: \begin{align*}

500: \frac{1}{\sqrt{m}}\left\Vert \sum_{i=1}^{n}x_{i}\right\Vert & =\left\vert

501: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\frac{1}{m}\sum_{i=1}^{n}%

502: \sum_{k=1}^{m}M_{ik}\right\vert \left\Vert \sum_{k=1}^{m}e_{k}\right\Vert \\

503: & =\frac{\sqrt{m}}{\sqrt{m}}\left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert

504: -\frac{1}{m}\sum_{i=1}^{n}\sum_{k=1}^{m}M_{ik}\right) \\

505: & =\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\frac{1}{m}\sum_{i=1}^{n}%

506: \sum_{k=1}^{m}M_{ik},

507: \end{align*}%

508: and the equality in (\ref{RT3.10.1}) holds true.

509: 

510: Conversely, if the equality holds in (\ref{RT3.10.1}), then, obviously (\ref%

511: {RT3.11.1}) is valid.

512: 

513: Now if the equality holds in (\ref{RT3.10.1}), then it must hold in (\ref%

514: {RT3.9.1}) for each $i\in \left\{ 1,\dots ,n\right\} $ and $k\in \left\{

515: 1,\dots ,m\right\} $ and also must hold in any of the inequalities in (\ref%

516: {RT3.14.1}).

517: 

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

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

520: equality occurs iff $u=\lambda v$ with $\lambda \geq 0,$ consequently, the

521: equality holds in all inequalities (\ref{RT3.14.1}) simultaneously iff there

522: exists a $\mu \geq 0$ with%

523: \begin{equation}

524: \mu \sum_{k=1}^{m}e_{k}=\sum_{i=1}^{n}x_{i}.  \label{RT3.15.1}

525: \end{equation}%

526: If we sum the equality in (\ref{RT3.9.1}) over $i$ from 1 to $n$ and $k$

527: from 1 to $m,$ then we deduce%

528: \begin{equation}

529: m\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\func{Re}\left\langle

530: \sum_{k=1}^{m}e_{k},\sum_{i=1}^{n}x_{i}\right\rangle

531: =\sum_{i=1}^{n}\sum_{k=1}^{m}M_{ik}.  \label{RT3.16.1}

532: \end{equation}%

533: Replacing $\sum_{i=1}^{n}x_{i}$ from (\ref{RT3.15.1}) into (\ref{RT3.16.1}),

534: we deduce%

535: \begin{equation*}

536: m\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\mu \sum_{k=1}^{m}\left\Vert

537: e_{k}\right\Vert ^{2}=\sum_{i=1}^{n}\sum_{k=1}^{m}M_{ik}

538: \end{equation*}%

539: giving%

540: \begin{equation*}

541: \mu =\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\frac{1}{m}%

542: \sum_{i=1}^{n}\sum_{k=1}^{m}M_{ik}.

543: \end{equation*}%

544: Using (\ref{RT3.15.1}), we deduce (\ref{RT3.12.1}) and the theorem is proved.

545: \end{proof}

546: 

547: \section{Further Additive Reverses}

548: 

549: \subsection{The Case of Small Balls}

550: 

551: In this section we point out different additive reverses of the generalised

552: triangle inequality under simpler conditions for the vectors involved.

553: 

554: The following result holds \cite{RTxSSD1}:

555: 

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

557: \label{RTt4.1.1}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

558: \right) $ be an inner product space over the real or complex number field $%

559: \mathbb{K}$ and $e,x_{i}\in H,$ $i\in \left\{ 1,\dots ,n\right\} $ with $%

560: \left\Vert e\right\Vert =1.$ If $\rho \in \left( 0,1\right) $ and $x_{i},$ $%

561: i\in \left\{ 1,\dots ,n\right\} $ are such that%

562: \begin{equation}

563: \left\Vert x_{i}-e\right\Vert \leq \rho \text{ \ for each}\ \ i\in \left\{

564: 1,\dots ,n\right\} ,  \label{RT4.1.1}

565: \end{equation}%

566: then we have the inequality%

567: \begin{align}

568: & \left( 0\leq \right) \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\left\Vert

569: \sum_{i=1}^{n}x_{i}\right\Vert  \label{RT4.2.1} \\

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

571: }\func{Re}\left\langle \sum_{i=1}^{n}x_{i},e\right\rangle  \notag \\

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

573: \right) }\left\Vert \sum_{i=1}^{n}x_{i}\right\Vert \right) .  \notag

574: \end{align}%

575: The equality holds in (\ref{RT4.2.1}) if and only if%

576: \begin{equation}

577: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \geq \frac{\rho ^{2}}{\sqrt{1-\rho

578: ^{2}}\left( 1+\sqrt{1-\rho ^{2}}\right) }\func{Re}\left\langle

579: \sum_{i=1}^{n}x_{i},e\right\rangle  \label{RT4.3.1}

580: \end{equation}%

581: and

582: \begin{multline}

583: \sum_{i=1}^{n}x_{i}  \label{RT4.4.1} \\

584: =\left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\frac{\rho ^{2}}{\sqrt{%

585: 1-\rho ^{2}}\left( 1+\sqrt{1-\rho ^{2}}\right) }\func{Re}\left\langle

586: \sum_{i=1}^{n}x_{i},e\right\rangle \right) e.

587: \end{multline}

588: \end{theorem}

589: 

590: \begin{proof}

591: We know, from the proof of Theorem \ref{RTt3.1.1}, that, if (\ref{RT4.1.1})

592: is fulfilled, then we have the inequality%

593: \begin{equation*}

594: \left\Vert x_{i}\right\Vert \leq \frac{1}{\sqrt{1-\rho ^{2}}}\func{Re}%

595: \left\langle x_{i},e\right\rangle

596: \end{equation*}%

597: for each $i\in \left\{ 1,\dots ,n\right\} ,$ implying%

598: \begin{align}

599: \left\Vert x_{i}\right\Vert -\func{Re}\left\langle x_{i},e\right\rangle &

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

601: x_{i},e\right\rangle  \label{RT4.5.1} \\

602: & =\frac{\rho ^{2}}{\sqrt{1-\rho ^{2}}\left( 1+\sqrt{1-\rho ^{2}}\right) }%

603: \func{Re}\left\langle x_{i},e\right\rangle  \notag

604: \end{align}%

605: for each $i\in \left\{ 1,\dots ,n\right\} .$

606: 

607: Now, making use of Theorem \ref{RTt2.1.1}, for

608: \begin{equation*}

609: k_{i}:=\frac{\rho ^{2}}{\sqrt{1-\rho ^{2}}\left( 1+\sqrt{1-\rho ^{2}}\right)

610: }\func{Re}\left\langle x_{i},e\right\rangle ,\ \ \ i\in \left\{ 1,\dots

611: ,n\right\} ,

612: \end{equation*}%

613: we easily deduce the conclusion of the theorem.

614: 

615: We omit the details.

616: \end{proof}

617: 

618: We may state the following result as well \cite{RTxSSD1}:

619: 

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

621: \label{RTt4.2.1}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

622: \right) $ be an inner product space and $e\in H,$ $M\geq m>0.$ If $x_{i}\in

623: H,$ $i\in \left\{ 1,\dots ,n\right\} $ are such that either%

624: \begin{equation}

625: \func{Re}\left\langle Me-x_{i},x_{i}-me\right\rangle \geq 0,  \label{RT4.6.1}

626: \end{equation}%

627: or, equivalently,%

628: \begin{equation}

629: \left\Vert x_{i}-\frac{M+m}{2}e\right\Vert \leq \frac{1}{2}\left( M-m\right)

630: \label{RT4.7.1}

631: \end{equation}%

632: holds for each $i\in \left\{ 1,\dots ,n\right\} ,$ then we have the

633: inequality%

634: \begin{align}

635: \left( 0\leq \right) \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\left\Vert

636: \sum_{i=1}^{n}x_{i}\right\Vert & \leq \frac{\left( \sqrt{M}-\sqrt{m}\right)

637: ^{2}}{2\sqrt{mM}}\func{Re}\left\langle \sum_{i=1}^{n}x_{i},e\right\rangle

638: \label{RT4.8.1} \\

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

640: \left\Vert \sum_{i=1}^{n}x_{i}\right\Vert \right) .  \notag

641: \end{align}%

642: The equality holds in (\ref{RT4.8.1}) if and only if%

643: \begin{equation}

644: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \geq \frac{\left( \sqrt{M}-\sqrt{m}%

645: \right) ^{2}}{2\sqrt{mM}}\func{Re}\left\langle

646: \sum_{i=1}^{n}x_{i},e\right\rangle  \label{RT4.9.1}

647: \end{equation}%

648: and%

649: \begin{equation}

650: \sum_{i=1}^{n}x_{i}=\left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\frac{%

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

652: \sum_{i=1}^{n}x_{i},e\right\rangle \right) e.  \label{RT4.10.1}

653: \end{equation}

654: \end{theorem}

655: 

656: \begin{proof}

657: We know, from the proof of Theorem \ref{RTt2.3.1}, that if (\ref{RT4.6.1})

658: is fulfilled, then we have the inequality%

659: \begin{equation*}

660: \left\Vert x_{i}\right\Vert \leq \frac{M+m}{2\sqrt{mM}}\func{Re}\left\langle

661: x_{i},e\right\rangle

662: \end{equation*}%

663: for each $i\in \left\{ 1,\dots ,n\right\} .$ This is equivalent to%

664: \begin{equation*}

665: \left\Vert x_{i}\right\Vert -\func{Re}\left\langle x_{i},e\right\rangle \leq

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

667: x_{i},e\right\rangle

668: \end{equation*}%

669: for each $i\in \left\{ 1,\dots ,n\right\} .$

670: 

671: Now, making use of Theorem \ref{RTt3.1.1}, we deduce the conclusion of the

672: theorem. We omit the details.

673: \end{proof}

674: 

675: \begin{remark}

676: If one uses Theorem \ref{RTt3.2.1} instead of Theorem \ref{RTt3.1.1} above,

677: then one can state the corresponding generalisation for families of

678: orthonormal vectors of the inequalities (\ref{RT4.2.1}) and (\ref{RT4.8.1})

679: respectively. We do not provide them here.

680: \end{remark}

681: 

682: \subsection{The Case of Arbitrary Balls}

683: 

684: Now, on utilising a slightly different approach, we may point out the

685: following result \cite{RTxSSD1}:

686: 

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

688: \label{RTt4.3.1}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

689: \right) $ be an inner product space over $\mathbb{K}$ and $e,$ $x_{i}\in H,$

690: $i\in \left\{ 1,\dots ,n\right\} $ with $\left\Vert e\right\Vert =1.$ If $%

691: r_{i}>0,$ $i\in \left\{ 1,\dots ,n\right\} $ are such that%

692: \begin{equation}

693: \left\Vert x_{i}-e\right\Vert \leq r_{i}\text{ \ for each}\ \ i\in \left\{

694: 1,\dots ,n\right\} ,  \label{RT4.11.1}

695: \end{equation}%

696: then we have the inequality%

697: \begin{equation}

698: 0\leq \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\left\Vert

699: \sum_{i=1}^{n}x_{i}\right\Vert \leq \frac{1}{2}\sum_{i=1}^{n}r_{i}^{2}.

700: \label{RT4.12.1}

701: \end{equation}%

702: The equality holds in (\ref{RT4.12.1}) if and only if%

703: \begin{equation}

704: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \geq \frac{1}{2}%

705: \sum_{i=1}^{n}r_{i}^{2}  \label{RT4.13.1}

706: \end{equation}%

707: and%

708: \begin{equation}

709: \sum_{i=1}^{n}x_{i}=\left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\frac{1%

710: }{2}\sum_{i=1}^{n}r_{i}^{2}\right) e.  \label{RT4.14.1}

711: \end{equation}

712: \end{theorem}

713: 

714: \begin{proof}

715: The condition (\ref{RT4.11.1}) is clearly equivalent to%

716: \begin{equation}

717: \left\Vert x_{i}\right\Vert ^{2}+1\leq \func{Re}\left\langle

718: x_{i},e\right\rangle +r_{i}^{2}  \label{RT4.15.1}

719: \end{equation}%

720: for each $i\in \left\{ 1,\dots ,n\right\} .$

721: 

722: Using the elementary inequality%

723: \begin{equation}

724: 2\left\Vert x_{i}\right\Vert \leq \left\Vert x_{i}\right\Vert ^{2}+1,

725: \label{RT4.16.1}

726: \end{equation}%

727: for each $i\in \left\{ 1,\dots ,n\right\} ,$ then, by (\ref{RT4.15.1}) and (%

728: \ref{RT4.16.1}), we deduce

729: \begin{equation*}

730: 2\left\Vert x_{i}\right\Vert \leq 2\func{Re}\left\langle

731: x_{i},e\right\rangle +r_{i}^{2},

732: \end{equation*}%

733: giving%

734: \begin{equation}

735: \left\Vert x_{i}\right\Vert -\func{Re}\left\langle x_{i},e\right\rangle \leq

736: \frac{1}{2}r_{i}^{2}  \label{RT4.17.1}

737: \end{equation}%

738: for each $i\in \left\{ 1,\dots ,n\right\} .$

739: 

740: Now, utilising Theorem \ref{RTt3.1.1} for $k_{i}=\frac{1}{2}r_{i}^{2},$ $%

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

742: details.

743: \end{proof}

744: 

745: Finally, we may state and prove the following result as well \cite{RTxSSD1}.

746: 

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

748: \label{RTt4.4.1}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

749: \right) $ be an inner product space over $\mathbb{K}$ and $e,$ $x_{i}\in H,$

750: $i\in \left\{ 1,\dots ,n\right\} $ with $\left\Vert e\right\Vert =1.$ If $%

751: M_{i}\geq m_{i}>0,$ $i\in \left\{ 1,\dots ,n\right\} ,$ are such that%

752: \begin{equation}

753: \left\Vert x_{i}-\frac{M_{i}+m_{i}}{2}e\right\Vert \leq \frac{1}{2}\left(

754: M_{i}-m_{i}\right) ,  \label{RT4.18.1}

755: \end{equation}%

756: or, equivalently,%

757: \begin{equation}

758: \func{Re}\left\langle M_{i}e-x,x-m_{i}e\right\rangle \geq 0  \label{RT4.19.1}

759: \end{equation}%

760: for each $i\in \left\{ 1,\dots ,n\right\} ,$ then we have the inequality%

761: \begin{equation}

762: \left( 0\leq \right) \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\left\Vert

763: \sum_{i=1}^{n}x_{i}\right\Vert \leq \frac{1}{4}\sum_{i=1}^{n}\frac{\left(

764: M_{i}-m_{i}\right) ^{2}}{M_{i}+m_{i}}.  \label{RT4.20.1}

765: \end{equation}%

766: The equality holds in (\ref{RT4.20.1}) if and only if%

767: \begin{equation}

768: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \geq \frac{1}{4}\sum_{i=1}^{n}%

769: \frac{\left( M_{i}-m_{i}\right) ^{2}}{M_{i}+m_{i}}  \label{RT4.21.1}

770: \end{equation}%

771: and

772: \begin{equation}

773: \sum_{i=1}^{n}x_{i}=\left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\frac{1%

774: }{4}\sum_{i=1}^{n}\frac{\left( M_{i}-m_{i}\right) ^{2}}{M_{i}+m_{i}}\right)

775: e.  \label{RT4.22.1}

776: \end{equation}

777: \end{theorem}

778: 

779: \begin{proof}

780: The condition (\ref{RT4.18.1}) is equivalent to:%

781: \begin{equation*}

782: \left\Vert x_{i}\right\Vert ^{2}+\left( \frac{M_{i}+m_{i}}{2}\right)

783: ^{2}\leq 2\func{Re}\left\langle x_{i},\frac{M_{i}+m_{i}}{2}e\right\rangle +%

784: \frac{1}{4}\left( M_{i}-m_{i}\right) ^{2}

785: \end{equation*}%

786: and since%

787: \begin{equation*}

788: 2\left( \frac{M_{i}+m_{i}}{2}\right) \left\Vert x_{i}\right\Vert \leq

789: \left\Vert x_{i}\right\Vert ^{2}+\left( \frac{M_{i}+m_{i}}{2}\right) ^{2},

790: \end{equation*}%

791: then we get%

792: \begin{equation*}

793: 2\left( \frac{M_{i}+m_{i}}{2}\right) \left\Vert x_{i}\right\Vert \leq 2\cdot

794: \frac{M_{i}+m_{i}}{2}\func{Re}\left\langle x_{i},e\right\rangle +\frac{1}{4}%

795: \left( M_{i}-m_{i}\right) ^{2},

796: \end{equation*}%

797: or, equivalently,%

798: \begin{equation*}

799: \left\Vert x_{i}\right\Vert -\func{Re}\left\langle x_{i},e\right\rangle \leq

800: \frac{1}{4}\cdot \frac{\left( M_{i}-m_{i}\right) ^{2}}{M_{i}+m_{i}}

801: \end{equation*}%

802: for each $i\in \left\{ 1,\dots ,n\right\} .$

803: 

804: Now, making use of Theorem \ref{RTt3.1.1} for $k_{i}:=\frac{1}{4}\cdot \frac{%

805: \left( M_{i}-m_{i}\right) ^{2}}{M_{i}+m_{i}},$ $i\in \left\{ 1,\dots

806: ,n\right\} ,$ we deduce the desired result.

807: \end{proof}

808: 

809: \begin{remark}

810: If one uses Theorem \ref{RTt3.2.1} instead of Theorem \ref{RTt3.1.1} above,

811: then one can state the corresponding generalisation for families of

812: orthonormal vectors of the inequalities in (\ref{RT4.12.1}) and (\ref%

813: {RT4.20.1}) respectively. We omit the details.

814: \end{remark}

815: 

816: \section{Reverses of Schwarz Inequality}

817: 

818: In this section we outline a procedure showing how some of the above results

819: for triangle inequality may be employed to obtain reverses for the

820: celebrated Schwarz inequality.

821: 

822: For $a\in H,$ $\left\Vert a\right\Vert =1$ and $r\in \left( 0,1\right) $

823: define the closed ball

824: \begin{equation*}

825: \overline{D}\left( a,r\right) :=\left\{ x\in H,\left\Vert x-a\right\Vert

826: \leq r\right\} .

827: \end{equation*}%

828: The following reverse of the Schwarz inequality holds \cite{RTxSSD1}:

829: 

830: \begin{proposition}

831: \label{RTp5.1.1} If $x,y\in \overline{D}\left( a,r\right) $ with $a\in H,$ $%

832: \left\Vert a\right\Vert =1$ and $r\in \left( 0,1\right) ,$ then we have the

833: inequality%

834: \begin{equation}

835: \left( 0\leq \right) \frac{\left\Vert x\right\Vert \left\Vert y\right\Vert -%

836: \func{Re}\left\langle x,y\right\rangle }{\left( \left\Vert x\right\Vert

837: +\left\Vert y\right\Vert \right) ^{2}}\leq \frac{1}{2}r^{2}.  \label{RT5.1.1}

838: \end{equation}%

839: The constant $\frac{1}{2}$ in $\left( \ref{RT5.1.1}\right) $ is best

840: possible in the sense that it cannot be replaced by a smaller quantity.

841: \end{proposition}

842: 

843: \begin{proof}

844: Using Theorem \ref{RTt2.1.1} for $x_{1}=x,x_{2}=y,\rho =r,$ we have

845: \begin{equation}

846: \sqrt{1-r^{2}}\left( \left\Vert x\right\Vert +\left\Vert y\right\Vert

847: \right) \leq \left\Vert x+y\right\Vert .  \label{RT5.2.1}

848: \end{equation}%

849: Taking the square in (\ref{RT5.2.1}) we deduce

850: \begin{equation*}

851: \left( 1-r^{2}\right) \left( \left\Vert x\right\Vert ^{2}+2\left\Vert

852: x\right\Vert \left\Vert y\right\Vert +\left\Vert y\right\Vert ^{2}\right)

853: \leq \left\Vert x\right\Vert ^{2}+2\func{Re}\left\langle x,y\right\rangle

854: +\left\Vert y\right\Vert ^{2}

855: \end{equation*}%

856: which is clearly equivalent to (\ref{RT5.1.1}).

857: 

858: Now, assume that (\ref{RT5.1.1}) holds with a constant $C>0$ instead of $%

859: \frac{1}{2},i.e.,$%

860: \begin{equation}

861: \frac{\left\Vert x\right\Vert \left\Vert y\right\Vert -\func{Re}\left\langle

862: x,y\right\rangle }{\left( \left\Vert x\right\Vert +\left\Vert y\right\Vert

863: \right) ^{2}}\leq Cr^{2}  \label{RT5.3.1}

864: \end{equation}%

865: provided $x,y\in \overline{D}\left( a,r\right) $ with $a\in H,$ $\left\Vert

866: a\right\Vert =1$ and $r\in \left( 0,1\right) .$

867: 

868: Let $e\in H$ with $\left\Vert e\right\Vert =1$ and $e\perp a.$ Define $%

869: x=a+re,y=a-re.$ Then

870: \begin{equation*}

871: \left\Vert x\right\Vert =\sqrt{1+r^{2}}=\left\Vert y\right\Vert ,\text{ }%

872: \func{Re}\left\langle x,y\right\rangle =1-r^{2}

873: \end{equation*}%

874: and thus, from (\ref{RT5.3.1}), we have%

875: \begin{equation*}

876: \frac{1+r^{2}-\left( 1-r^{2}\right) }{\left( 2\sqrt{1+r^{2}}\right) ^{2}}%

877: \leq Cr^{2}

878: \end{equation*}%

879: giving

880: \begin{equation*}

881: \frac{1}{2}\leq \left( 1+r^{2}\right) C

882: \end{equation*}%

883: for any $r\in \left( 0,1\right) .$ If in this inequality we let $%

884: r\rightarrow 0+,$ then we get $C\geq \frac{1}{2}$ and the proposition is

885: proved.

886: \end{proof}

887: 

888: In a similar way, by the use of Theorem \ref{RTt2.3.1}, we may prove the

889: following reverse of the Schwarz inequality as well \cite{RTxSSD1}:

890: 

891: \begin{proposition}

892: \label{RTp5.2.1} If $a\in H,$ $\left\Vert a\right\Vert =1,$ $M\geq m>0$ and $%

893: x,y\in H$ are so that either%

894: \begin{equation*}

895: \func{Re}\left\langle Ma-x,x-ma\right\rangle ,\func{Re}\left\langle

896: Ma-y,y-ma\right\rangle \geq 0

897: \end{equation*}%

898: or, equivalently,%

899: \begin{equation*}

900: \left\Vert x-\frac{m+M}{2}a\right\Vert ,\left\Vert y-\frac{m+M}{2}%

901: a\right\Vert \leq \frac{1}{2}\left( M-m\right)

902: \end{equation*}%

903: hold, then

904: \begin{equation*}

905: \left( 0\leq \right) \frac{\left\Vert x\right\Vert \left\Vert y\right\Vert -%

906: \func{Re}\left\langle x,y\right\rangle }{\left( \left\Vert x\right\Vert

907: +\left\Vert y\right\Vert \right) ^{2}}\leq \frac{1}{2}\left( \frac{M-m}{M+m}%

908: \right) ^{2}.

909: \end{equation*}%

910: The constant $\frac{1}{2}$ cannot be replaced by a smaller quantity.

911: \end{proposition}

912: 

913: \begin{remark}

914: On utilising Theorem \ref{RTt2.2.1} and Theorem \ref{RTt2.4.1}, we may

915: deduce some similar reverses of Schwarz inequality provided $x,y\in \cap

916: _{k=1}^{m}\overline{D}\left( a_{k},\rho _{k}\right) ,$ assumed not to be

917: empty, where $a_{1},...,a_{n}$ are orthonormal vectors in $H$ and $\rho

918: _{k}\in \left( 0,1\right) $ for $k\in \left\{ 1,...,m\right\} .$ We omit the

919: details.

920: \end{remark}

921: 

922: \begin{remark}

923: For various different reverses of Schwarz inequality in inner product

924: spaces, see the recent survey \cite{RTxSSD}.

925: \end{remark}

926: 

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

928: 

929: \subsection{The General Case}

930: 

931: The following lemma holds \cite{RTxSSD2}:

932: 

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

934: \label{RTl2.1.2}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

935: \right) $ be an inner product space over the real or complex number field $%

936: \mathbb{K}$, $x_{i}\in H,$ $i\in \left\{ 1,\dots ,n\right\} $ and $k_{ij}>0$

937: for $1\leq i<j\leq n$ such that%

938: \begin{equation}

939: 0\leq \left\Vert x_{i}\right\Vert \left\Vert x_{j}\right\Vert -\func{Re}%

940: \left\langle x_{i},x_{j}\right\rangle \leq k_{ij}  \label{RT2.1.2}

941: \end{equation}%

942: for $1\leq i<j\leq n.$ Then we have the following quadratic reverse of the

943: triangle inequality%

944: \begin{equation}

945: \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}\leq \left\Vert

946: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}+2\sum_{1\leq i<j\leq n}k_{ij}.

947: \label{RT2.2.2}

948: \end{equation}%

949: The case of equality holds in (\ref{RT2.2.2}) if and only if it holds in (%

950: \ref{RT2.1.2}) for each $i,j$ with $1\leq i<j\leq n.$

951: \end{lemma}

952: 

953: \begin{proof}

954: We observe that the following identity holds:%

955: \begin{align}

956: & \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}-\left\Vert

957: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}  \label{RT2.3.2} \\

958: & =\sum_{i,j=1}^{n}\left\Vert x_{i}\right\Vert \left\Vert x_{j}\right\Vert

959: -\left\langle \sum_{i=1}^{n}x_{i},\sum_{j=1}^{n}x_{j}\right\rangle  \notag \\

960: & =\sum_{i,j=1}^{n}\left\Vert x_{i}\right\Vert \left\Vert x_{j}\right\Vert

961: -\sum_{i,j=1}^{n}\func{Re}\left\langle x_{i},x_{j}\right\rangle  \notag \\

962: & =\sum_{i,j=1}^{n}\left[ \left\Vert x_{i}\right\Vert \left\Vert

963: x_{j}\right\Vert -\func{Re}\left\langle x_{i},x_{j}\right\rangle \right]

964: \notag \\

965: & =\sum_{1\leq i<j\leq n}\left[ \left\Vert x_{i}\right\Vert \left\Vert

966: x_{j}\right\Vert -\func{Re}\left\langle x_{i},x_{j}\right\rangle \right]

967: \notag \\

968: & +\sum_{1\leq j<i\leq n}\left[ \left\Vert x_{i}\right\Vert \left\Vert

969: x_{j}\right\Vert -\func{Re}\left\langle x_{i},x_{j}\right\rangle \right]

970: \notag \\

971: & =2\sum_{1\leq i<j\leq n}\left[ \left\Vert x_{i}\right\Vert \left\Vert

972: x_{j}\right\Vert -\func{Re}\left\langle x_{i},x_{j}\right\rangle \right] .

973: \notag

974: \end{align}%

975: Using the condition (\ref{RT2.1.2}), we deduce that%

976: \begin{equation*}

977: \sum_{1\leq i<j\leq n}\left[ \left\Vert x_{i}\right\Vert \left\Vert

978: x_{j}\right\Vert -\func{Re}\left\langle x_{i},x_{j}\right\rangle \right]

979: \leq \sum_{1\leq i<j\leq n}k_{ij},

980: \end{equation*}%

981: and by (\ref{RT2.3.2}), we get the desired inequality (\ref{RT2.2.2}).

982: 

983: The case of equality is obvious by the identity (\ref{RT2.3.2}) and we omit

984: the details.

985: \end{proof}

986: 

987: \begin{remark}

988: \label{RTr2.2.2}From (\ref{RT2.2.2}) one may deduce the coarser inequality

989: that might be useful in some applications:%

990: \begin{align*}

991: 0& \leq \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\left\Vert

992: \sum_{i=1}^{n}x_{i}\right\Vert  \\

993: & \leq \sqrt{2}\left( \sum_{1\leq i<j\leq n}k_{ij}\right) ^{\frac{1}{2}%

994: }\qquad \left( \leq \sqrt{2}\sum_{1\leq i<j\leq n}\sqrt{k_{ij}}\right) .

995: \end{align*}

996: \end{remark}

997: 

998: \begin{remark}

999: \label{RTr2.3.2}If the condition (\ref{RT2.1.2}) is replaced with the

1000: following refinement of Schwarz's inequality:%

1001: \begin{equation}

1002: \left( 0\leq \right) \delta _{ij}\leq \left\Vert x_{i}\right\Vert \left\Vert

1003: x_{j}\right\Vert -\func{Re}\left\langle x_{i},x_{j}\right\rangle \text{ for }%

1004: 1\leq i<j\leq n,  \label{RT2.4.2}

1005: \end{equation}%

1006: then the following refinement of the quadratic generalised triangle

1007: inequality is valid:%

1008: \begin{equation}

1009: \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}\geq \left\Vert

1010: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}+2\sum_{1\leq i<j\leq n}\delta _{ij}\quad

1011: \left( \geq \left\Vert \sum_{i=1}^{n}x_{i}\right\Vert ^{2}\right) .

1012: \label{RT2.5.2}

1013: \end{equation}%

1014: The equality holds in the first part of (\ref{RT2.5.2}) iff the case of

1015: equality holds in (\ref{RT2.4.2}) for each $1\leq i<j\leq n.$

1016: \end{remark}

1017: 

1018: The following result holds \cite{RTxSSD2}.

1019: 

1020: \begin{proposition}

1021: \label{RTp2.4.2}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

1022: \right) $ be as above, $x_{i}\in H,$ $i\in \left\{ 1,\dots ,n\right\} $ and $%

1023: r>0$ such that%

1024: \begin{equation}

1025: \left\Vert x_{i}-x_{j}\right\Vert \leq r  \label{RT2.6.2}

1026: \end{equation}%

1027: for $1\leq i<j\leq n.$ Then%

1028: \begin{equation}

1029: \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}\leq \left\Vert

1030: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}+\frac{n\left( n-1\right) }{2}r^{2}.

1031: \label{RT2.7.2}

1032: \end{equation}%

1033: The case of equality holds in (\ref{RT2.7.2}) if and only if

1034: \begin{equation}

1035: \left\Vert x_{i}\right\Vert \left\Vert x_{j}\right\Vert -\func{Re}%

1036: \left\langle x_{i},x_{j}\right\rangle =\frac{1}{2}r^{2}  \label{RT2.8.2}

1037: \end{equation}%

1038: for each $i,j$ with $1\leq i<j\leq n.$

1039: \end{proposition}

1040: 

1041: \begin{proof}

1042: The inequality (\ref{RT2.6.2}) is obviously equivalent to%

1043: \begin{equation*}

1044: \left\Vert x_{i}\right\Vert ^{2}+\left\Vert x_{j}\right\Vert ^{2}\leq 2\func{%

1045: Re}\left\langle x_{i},x_{j}\right\rangle +r^{2}

1046: \end{equation*}%

1047: for $1\leq i<j\leq n.$ Since%

1048: \begin{equation*}

1049: 2\left\Vert x_{i}\right\Vert \left\Vert x_{j}\right\Vert \leq \left\Vert

1050: x_{i}\right\Vert ^{2}+\left\Vert x_{j}\right\Vert ^{2},\ \ 1\leq i<j\leq n;

1051: \end{equation*}%

1052: hence%

1053: \begin{equation}

1054: \left\Vert x_{i}\right\Vert \left\Vert x_{j}\right\Vert -\func{Re}%

1055: \left\langle x_{i},x_{j}\right\rangle \leq \frac{1}{2}r^{2}  \label{RT2.9.2}

1056: \end{equation}%

1057: for any $i,j$ with $1\leq i<j\leq n.$

1058: 

1059: Applying Lemma \ref{RTl2.1.2} for $k_{ij}:=\frac{1}{2}r^{2}$ and taking into

1060: account that

1061: \begin{equation*}

1062: \sum_{1\leq i<j\leq n}k_{ij}=\frac{n\left( n-1\right) }{4}r^{2},

1063: \end{equation*}%

1064: we deduce the desired inequality (\ref{RT2.7.2}). The case of equality is

1065: also obvious by the above lemma and we omit the details.

1066: \end{proof}

1067: 

1068: \subsection{Inequalities in Terms of the Forward Difference}

1069: 

1070: In the same spirit, and if some information about the forward difference $%

1071: \Delta x_{k}:=x_{k+1}-x_{k}$ $\left( 1\leq k\leq n-1\right) $ are available,

1072: then the following simple quadratic reverse of the generalised triangle

1073: inequality may be stated \cite{RTxSSD2}.

1074: 

1075: \begin{corollary}

1076: \label{RTc2.5.2}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

1077: \right) $ be an inner product space and $x_{i}\in H,$ $i\in \left\{ 1,\dots

1078: ,n\right\} .$ Then we have the inequality%

1079: \begin{equation}

1080: \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}\leq \left\Vert

1081: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}+\frac{n\left( n-1\right) }{2}%

1082: \sum_{k=1}^{n-1}\left\Vert \Delta x_{k}\right\Vert .  \label{RT2.10.2}

1083: \end{equation}%

1084: The constant $\frac{1}{2}$ is best possible in the sense that it cannot be

1085: replaced in general by a smaller quantity.

1086: \end{corollary}

1087: 

1088: \begin{proof}

1089: Let $1\leq i<j\leq n.$ Then, obviously,

1090: \begin{equation*}

1091: \left\Vert x_{j}-x_{i}\right\Vert =\left\Vert \sum_{k=i}^{j-1}\Delta

1092: x_{k}\right\Vert \leq \sum_{k=i}^{j-1}\left\Vert \Delta x_{k}\right\Vert

1093: \leq \sum_{k=1}^{n-1}\left\Vert \Delta x_{k}\right\Vert .

1094: \end{equation*}%

1095: Applying Proposition \ref{RTp2.4.2} for $r:=\sum_{k=1}^{n-1}\left\Vert

1096: \Delta x_{k}\right\Vert ,$ we deduce the desired result (\ref{RT2.10.2}).

1097: 

1098: To prove the sharpness of the constant $\frac{1}{2},$ assume that the

1099: inequality (\ref{RT2.10.2}) holds with a constant $c>0,$ i.e.,

1100: \begin{equation}

1101: \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}\leq \left\Vert

1102: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}+cn\left( n-1\right)

1103: \sum_{k=1}^{n-1}\left\Vert \Delta x_{k}\right\Vert  \label{RT2.11.2}

1104: \end{equation}%

1105: for $n\geq 2,$ $x_{i}\in H,$ $i\in \left\{ 1,\dots ,n\right\} .$

1106: 

1107: If we choose in (\ref{RT2.11.2}), $n=2,$ $x_{1}=-\frac{1}{2}e,$ $x_{2}=\frac{%

1108: 1}{2}e,$ $e\in H,$ $\left\Vert e\right\Vert =1,$ then we get $1\leq 2c,$

1109: giving $c\geq \frac{1}{2}.$

1110: \end{proof}

1111: 

1112: The following result providing a reverse of the quadratic generalised

1113: triangle inequality in terms of the sup-norm of the forward differences also

1114: holds \cite{RTxSSD2}.

1115: 

1116: \begin{proposition}

1117: \label{RTp2.6.2}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

1118: \right) $ be an inner product space and $x_{i}\in H,$ $i\in \left\{ 1,\dots

1119: ,n\right\} .$ Then we have the inequality%

1120: \begin{equation}

1121: \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}\leq \left\Vert

1122: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}+\frac{n^{2}\left( n^{2}-1\right) }{12}%

1123: \max_{1\leq k\leq n-1}\left\Vert \Delta x_{k}\right\Vert ^{2}.

1124: \label{RT2.12.2}

1125: \end{equation}%

1126: The constant $\frac{1}{12}$ is best possible in (\ref{RT2.12.2}).

1127: \end{proposition}

1128: 

1129: \begin{proof}

1130: As above, we have that%

1131: \begin{equation*}

1132: \left\Vert x_{j}-x_{i}\right\Vert \leq \sum_{k=i}^{j-1}\left\Vert \Delta

1133: x_{k}\right\Vert \leq \left( j-i\right) \max_{1\leq k\leq n-1}\left\Vert

1134: \Delta x_{k}\right\Vert ,

1135: \end{equation*}%

1136: for $1\leq i<j\leq n.$

1137: 

1138: Squaring the above inequality, we get%

1139: \begin{equation*}

1140: \left\Vert x_{j}\right\Vert ^{2}+\left\Vert x_{i}\right\Vert ^{2}\leq 2\func{%

1141: Re}\left\langle x_{i},x_{j}\right\rangle +\left( j-i\right) ^{2}\max_{1\leq

1142: k\leq n-1}\left\Vert \Delta x_{k}\right\Vert ^{2}

1143: \end{equation*}%

1144: for any $i,j$ with $1\leq i<j\leq n,$ and since%

1145: \begin{equation*}

1146: 2\left\Vert x_{i}\right\Vert \left\Vert x_{j}\right\Vert \leq \left\Vert

1147: x_{j}\right\Vert ^{2}+\left\Vert x_{i}\right\Vert ^{2},

1148: \end{equation*}%

1149: hence%

1150: \begin{equation}

1151: 0\leq \left\Vert x_{i}\right\Vert \left\Vert x_{j}\right\Vert -\func{Re}%

1152: \left\langle x_{i},x_{j}\right\rangle \leq \frac{1}{2}\left( j-i\right)

1153: ^{2}\max_{1\leq k\leq n-1}\left\Vert \Delta x_{k}\right\Vert ^{2}

1154: \label{RT2.13.2}

1155: \end{equation}%

1156: for any $i,j$ with $1\leq i<j\leq n.$

1157: 

1158: Applying Lemma \ref{RTl2.1.2} for $k_{ij}:=\frac{1}{2}\left( j-i\right)

1159: ^{2}\max\limits_{1\leq k\leq n-1}\left\Vert \Delta x_{k}\right\Vert ^{2},$

1160: we can state that%

1161: \begin{equation*}

1162: \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}\leq \left\Vert

1163: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}+\sum_{1\leq i<j\leq n}\left( j-i\right)

1164: ^{2}\max_{1\leq k\leq n-1}\left\Vert \Delta x_{k}\right\Vert ^{2}.

1165: \end{equation*}%

1166: However,%

1167: \begin{align*}

1168: \sum_{1\leq i<j\leq n}\left( j-i\right) ^{2}& =\frac{1}{2}%

1169: \sum_{i,j=1}^{n}\left( j-i\right) ^{2}=n\sum_{k=1}^{n}k^{2}-\left(

1170: \sum_{k=1}^{n}k\right) ^{2} \\

1171: & =\frac{n^{2}\left( n^{2}-1\right) }{12}

1172: \end{align*}%

1173: giving the desired inequality.

1174: 

1175: To prove the sharpness of the constant, assume that (\ref{RT2.12.2}) holds

1176: with a constant $D>0,$ i.e.,%

1177: \begin{equation}

1178: \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}\leq \left\Vert

1179: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}+Dn^{2}\left( n^{2}-1\right) \max_{1\leq

1180: k\leq n-1}\left\Vert \Delta x_{k}\right\Vert ^{2}  \label{RT2.14.2}

1181: \end{equation}%

1182: for $n\geq 2,$ $x_{i}\in H,$ $i\in \left\{ 1,\dots ,n\right\} .$

1183: 

1184: If in (\ref{RT2.14.2}) we choose $n=2,$ $x_{1}=-\frac{1}{2}e,$ $x_{2}=\frac{1%

1185: }{2}e,$ $e\in H,$ $\left\Vert e\right\Vert =1,$ then we get $1\leq 12D$

1186: giving $D\geq \frac{1}{12}.$

1187: \end{proof}

1188: 

1189: The following result may be stated as well \cite{RTxSSD2}.

1190: 

1191: \begin{proposition}

1192: \label{RTp2.7.2}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

1193: \right) $ be an inner product space and $x_{i}\in H,$ $i\in \left\{ 1,\dots

1194: ,n\right\} .$ Then we have the inequality:%

1195: \begin{equation}

1196: \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}\leq \left\Vert

1197: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}+\sum_{1\leq i<j\leq n}\left( j-i\right)

1198: ^{\frac{2}{q}}\left( \sum_{k=1}^{n-1}\left\Vert \Delta x_{k}\right\Vert

1199: ^{p}\right) ^{\frac{2}{p}},  \label{RT2.15.2}

1200: \end{equation}%

1201: where $p>1,$ $\frac{1}{p}+\frac{1}{q}=1.$

1202: 

1203: The constant $E=1$ in front of the double sum cannot generally be replaced

1204: by a smaller constant.

1205: \end{proposition}

1206: 

1207: \begin{proof}

1208: Using H\"{o}lder's inequality, we have%

1209: \begin{align*}

1210: \left\Vert x_{j}-x_{i}\right\Vert & \leq \sum_{k=i}^{j-1}\left\Vert \Delta

1211: x_{k}\right\Vert \leq \left( j-i\right) ^{\frac{1}{q}}\left(

1212: \sum_{k=i}^{j-1}\left\Vert \Delta x_{k}\right\Vert ^{p}\right) ^{\frac{1}{p}}

1213: \\

1214: & \leq \left( j-i\right) ^{\frac{1}{q}}\left( \sum_{k=1}^{n-1}\left\Vert

1215: \Delta x_{k}\right\Vert ^{p}\right) ^{\frac{1}{p}},

1216: \end{align*}%

1217: for $1\leq i<j\leq n.$

1218: 

1219: Squaring the previous inequality, we get%

1220: \begin{equation*}

1221: \left\Vert x_{j}\right\Vert ^{2}+\left\Vert x_{i}\right\Vert ^{2}\leq 2\func{%

1222: Re}\left\langle x_{i},x_{j}\right\rangle +\left( j-i\right) ^{\frac{2}{q}%

1223: }\left( \sum_{k=1}^{n-1}\left\Vert \Delta x_{k}\right\Vert ^{p}\right) ^{%

1224: \frac{2}{p}},

1225: \end{equation*}%

1226: for $1\leq i<j\leq n.$

1227: 

1228: Utilising the same argument from the proof of Proposition \ref{RTp2.6.2}, we

1229: deduce the desired inequality (\ref{RT2.15.2}).

1230: 

1231: Now assume that (\ref{RT2.15.2}) holds with a constant $E>0,$ i.e.,%

1232: \begin{equation*}

1233: \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}\leq \left\Vert

1234: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}+E\sum_{1\leq i<j\leq n}\left( j-i\right)

1235: ^{\frac{2}{q}}\left( \sum_{k=1}^{n-1}\left\Vert \Delta x_{k}\right\Vert

1236: ^{p}\right) ^{\frac{2}{p}},

1237: \end{equation*}%

1238: for $n\geq 2$ and $x_{i}\in H,$ $i\in \left\{ 1,\dots ,n\right\} ,$ $p>1,$ $%

1239: \frac{1}{p}+\frac{1}{q}=1.$

1240: 

1241: For $n=2,$ $x_{1}=-\frac{1}{2}e,$ $x_{2}=\frac{1}{2}e,$ $\left\Vert

1242: e\right\Vert =1,$ we get $1\leq E,$ showing the fact that the inequality (%

1243: \ref{RT2.15.2}) is sharp.

1244: \end{proof}

1245: 

1246: The particular case $p=q=2$ is of interest \cite{RTxSSD2}.

1247: 

1248: \begin{corollary}

1249: \label{RTc2.8.2}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

1250: \right) $ be an inner product space and $x_{i}\in H,$ $i\in \left\{ 1,\dots

1251: ,n\right\} .$ Then we have the inequality:%

1252: \begin{equation}

1253: \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}\leq \left\Vert

1254: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}+\frac{\left( n^{2}-1\right) n}{6}%

1255: \sum_{k=1}^{n-1}\left\Vert \Delta x_{k}\right\Vert ^{2}.  \label{RT2.16.2}

1256: \end{equation}%

1257: The constant $\frac{1}{6}$ is best possible in (\ref{RT2.16.2}).

1258: \end{corollary}

1259: 

1260: \begin{proof}

1261: For $p=q=2,$ Proposition \ref{RTp2.7.2} provides the inequality%

1262: \begin{equation*}

1263: \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}\leq \left\Vert

1264: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}+\sum_{1\leq i<j\leq n}\left( j-i\right)

1265: \sum_{k=1}^{n-1}\left\Vert \Delta x_{k}\right\Vert ^{2},

1266: \end{equation*}%

1267: and since%

1268: \begin{align*}

1269: & \sum\limits_{1\leq i<j\leq n}\left( j-i\right)  \\

1270: & =1+\left( 1+2\right) +\left( 1+2+3\right) +\cdots +\left( 1+2+\cdots

1271: +n-1\right)  \\

1272: & =\sum\limits_{k=1}^{n-1}\left( 1+2+\cdots +k\right)

1273: =\sum\limits_{k=1}^{n-1}\frac{k\left( k+1\right) }{2}=\frac{n\left(

1274: n^{2}-1\right) }{6},

1275: \end{align*}%

1276: hence the inequality (\ref{RT2.15.2}) is proved. The best constant may be

1277: shown in the same way as above but we omit the details.

1278: \end{proof}

1279: 

1280: \subsection{A Different Quadratic Inequality}

1281: 

1282: Finally, we may state and prove the following different result \cite{RTxSSD2}%

1283: .

1284: 

1285: \begin{theorem}[Dragomir, 2004]

1286: \label{RTt2.9.2}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

1287: \right) $ be an inner product space, $y_{i}\in H,$ $i\in \left\{ 1,\dots

1288: ,n\right\} $ and $M\geq m>0$ are such that either%

1289: \begin{equation}

1290: \func{Re}\left\langle My_{j}-y_{i},y_{i}-my_{j}\right\rangle \geq 0\text{ \

1291: for }1\leq i<j\leq n,  \label{RT2.17.2}

1292: \end{equation}%

1293: or, equivalently,%

1294: \begin{equation}

1295: \left\Vert y_{i}-\frac{M+m}{2}y_{j}\right\Vert \leq \frac{1}{2}\left(

1296: M-m\right) \left\Vert y_{j}\right\Vert \text{ \ for }1\leq i<j\leq n.

1297: \label{RT2.18.2}

1298: \end{equation}%

1299: Then we have the inequality%

1300: \begin{equation}

1301: \left( \sum_{i=1}^{n}\left\Vert y_{i}\right\Vert \right) ^{2}\leq \left\Vert

1302: \sum_{i=1}^{n}y_{i}\right\Vert ^{2}+\frac{1}{2}\cdot \frac{\left( M-m\right)

1303: ^{2}}{M+m}\sum\limits_{k=1}^{n-1}k\left\Vert y_{k+1}\right\Vert ^{2}.

1304: \label{RT2.19.2}

1305: \end{equation}%

1306: The case of equality holds in (\ref{RT2.19.2}) if and only if%

1307: \begin{equation}

1308: \left\Vert y_{i}\right\Vert \left\Vert y_{j}\right\Vert -\func{Re}%

1309: \left\langle y_{i},y_{j}\right\rangle =\frac{1}{4}\cdot \frac{\left(

1310: M-m\right) ^{2}}{M+m}\left\Vert y_{j}\right\Vert ^{2}  \label{RT2.20.2}

1311: \end{equation}%

1312: for each $i,j$ with $1\leq i<j\leq n.$

1313: \end{theorem}

1314: 

1315: \begin{proof}

1316: Taking the square in (\ref{RT2.18.2}), we get%

1317: \begin{multline*}

1318: \quad \left\Vert y_{i}\right\Vert ^{2}+\frac{\left( M-m\right) ^{2}}{M+m}%

1319: \left\Vert y_{j}\right\Vert ^{2} \\

1320: \leq 2\func{Re}\left\langle y_{i},\frac{M+m}{2}y_{j}\right\rangle +\frac{1}{n%

1321: }\left( M-m\right) ^{2}\left\Vert y_{j}\right\Vert ^{2}\quad

1322: \end{multline*}%

1323: for $1\leq i<j\leq n,$ and since, obviously,%

1324: \begin{equation*}

1325: 2\left( \frac{M+m}{2}\right) \left\Vert y_{i}\right\Vert \left\Vert

1326: y_{j}\right\Vert \leq \left\Vert y_{i}\right\Vert ^{2}+\frac{\left(

1327: M-m\right) ^{2}}{M+m}\left\Vert y_{j}\right\Vert ^{2},

1328: \end{equation*}%

1329: hence%

1330: \begin{multline*}

1331: \quad  2\left( \frac{M+m}{2}\right) \left\Vert y_{i}\right\Vert

1332: \left\Vert

1333: y_{j}\right\Vert \\

1334: \leq 2\func{Re}\left\langle y_{i},\frac{M+m}{2}y_{j}\right\rangle +\frac{1}{n%

1335: }\left( M-m\right) ^{2}\left\Vert y_{j}\right\Vert ^{2},\qquad

1336: \end{multline*}%

1337: giving the much simpler inequality%

1338: \begin{equation}

1339: \left\Vert y_{i}\right\Vert \left\Vert y_{j}\right\Vert -\func{Re}%

1340: \left\langle y_{i},y_{j}\right\rangle \leq \frac{1}{4}\cdot \frac{\left(

1341: M-m\right) ^{2}}{M+m}\left\Vert y_{j}\right\Vert ^{2},  \label{RT2.21.2}

1342: \end{equation}%

1343: for $1\leq i<j\leq n.$

1344: 

1345: Applying Lemma \ref{RTl2.1.2} for $k_{ij}:=\frac{1}{4}\cdot \frac{\left(

1346: M-m\right) ^{2}}{M+m}\left\Vert y_{j}\right\Vert ^{2},$ we deduce%

1347: \begin{equation}

1348: \left( \sum_{i=1}^{n}\left\Vert y_{i}\right\Vert \right) ^{2}\leq \left\Vert

1349: \sum_{i=1}^{n}y_{i}\right\Vert ^{2}+\frac{1}{2}\frac{\left( M-m\right) ^{2}}{%

1350: M+m}\sum_{1\leq i<j\leq n}\left\Vert y_{j}\right\Vert ^{2}  \label{RT2.22.2}

1351: \end{equation}%

1352: with equality if and only if (\ref{RT2.21.2}) holds for each $i,j$ with $%

1353: 1\leq i<j\leq n.$

1354: 

1355: Since%

1356: \begin{align*}

1357: \sum_{1\leq i<j\leq n}\left\Vert y_{j}\right\Vert ^{2}& =\sum_{1<j\leq

1358: n}\left\Vert y_{j}\right\Vert ^{2}+\sum_{2<j\leq n}\left\Vert

1359: y_{j}\right\Vert ^{2}+\cdots +\sum_{n-1<j\leq n}\left\Vert y_{j}\right\Vert

1360: ^{2} \\

1361: & =\sum_{j=2}^{n}\left\Vert y_{j}\right\Vert ^{2}+\sum_{j=3}^{n}\left\Vert

1362: y_{j}\right\Vert ^{2}+\cdots +\sum_{j=n-1}^{n}\left\Vert y_{j}\right\Vert

1363: ^{2}+\left\Vert y_{n}\right\Vert ^{2} \\

1364: & =\sum_{j=2}^{n}\left( j-1\right) \left\Vert y_{j}\right\Vert

1365: ^{2}=\sum\limits_{k=1}^{n-1}k\left\Vert y_{k+1}\right\Vert ^{2},

1366: \end{align*}%

1367: hence the inequality (\ref{RT2.19.2}) is obtained.

1368: \end{proof}

1369: 

1370: \section{Further Quadratic Refinements}

1371: 

1372: \subsection{The General Case}

1373: 

1374: The following lemma is of interest in itself as well \cite{RTxSSD2}.

1375: 

1376: \begin{lemma}[Dragomir, 2004]

1377: \label{RTl3.1.2}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

1378: \right) $ be an inner product space over the real or complex number field $%

1379: \mathbb{K}$, $x_{i}\in H,$ $i\in \left\{ 1,\dots ,n\right\} $ and $k\geq 1$

1380: with the property that:

1381: \begin{equation}

1382: \left\Vert x_{i}\right\Vert \left\Vert x_{j}\right\Vert \leq k\func{Re}%

1383: \left\langle x_{i},x_{j}\right\rangle ,  \label{RT3.1.2}

1384: \end{equation}%

1385: for each $i,j$ with $1\leq i<j\leq n.$ Then%

1386: \begin{equation}

1387: \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}+\left(

1388: k-1\right) \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert ^{2}\leq k\left\Vert

1389: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}.  \label{RT3.2.2}

1390: \end{equation}%

1391: The equality holds in (\ref{RT3.2.2}) if and only if it holds in (\ref%

1392: {RT3.1.2}) for each $i,j$ with $1\leq i<j\leq n.$

1393: \end{lemma}

1394: 

1395: \begin{proof}

1396: Firstly, let us observe that the following identity holds true:%

1397: \begin{align}

1398: & \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}-k\left\Vert

1399: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}  \label{RT3.3.2} \\

1400: & =\sum_{i,j=1}^{n}\left\Vert x_{i}\right\Vert \left\Vert x_{j}\right\Vert

1401: -k\left\langle \sum_{i=1}^{n}x_{i},\sum_{j=1}^{n}x_{j}\right\rangle  \notag

1402: \\

1403: & =\sum_{i,j=1}^{n}\left[ \left\Vert x_{i}\right\Vert \left\Vert

1404: x_{j}\right\Vert -k\func{Re}\left\langle x_{i},x_{j}\right\rangle \right]

1405: \notag \\

1406: & =2\sum_{1\leq i<j\leq n}\left[ \left\Vert x_{i}\right\Vert \left\Vert

1407: x_{j}\right\Vert -k\func{Re}\left\langle x_{i},x_{j}\right\rangle \right]

1408: +\left( 1-k\right) \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert ^{2},  \notag

1409: \end{align}%

1410: since, obviously, $\func{Re}\left\langle x_{i},x_{j}\right\rangle =\func{Re}%

1411: \left\langle x_{j},x_{i}\right\rangle $ for any $i,j\in \left\{ 1,\dots

1412: ,n\right\} .$

1413: 

1414: Using the assumption (\ref{RT3.1.2}), we obtain%

1415: \begin{equation*}

1416: \sum_{1\leq i<j\leq n}\left[ \left\Vert x_{i}\right\Vert \left\Vert

1417: x_{j}\right\Vert -k\func{Re}\left\langle x_{i},x_{j}\right\rangle \right]

1418: \leq 0

1419: \end{equation*}%

1420: and thus, from (\ref{RT3.3.2}), we deduce the desired inequality (\ref%

1421: {RT3.2.2}).

1422: 

1423: The case of equality is obvious by the identity (\ref{RT3.3.2}) and we omit

1424: the details.

1425: \end{proof}

1426: 

1427: \begin{remark}

1428: \label{RTr3.4.2}The inequality (\ref{RT3.2.2}) provides the following

1429: reverse of the quadratic generalised triangle inequality:%

1430: \begin{equation}

1431: 0\leq \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right)

1432: ^{2}-\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert ^{2}\leq k\left[ \left\Vert

1433: \sum_{i=1}^{n}x_{i}\right\Vert ^{2}-\sum_{i=1}^{n}\left\Vert

1434: x_{i}\right\Vert ^{2}\right] .  \label{RT3.4.2}

1435: \end{equation}

1436: \end{remark}

1437: 

1438: \begin{remark}

1439: \label{RTr3.5.2}Since $k=1$ and $\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert

1440: ^{2}\geq 0,$ hence by (\ref{RT3.2.2}) one may deduce the following reverse

1441: of the triangle inequality%

1442: \begin{equation}

1443: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \leq \sqrt{k}\left\Vert

1444: \sum_{i=1}^{n}x_{i}\right\Vert ,  \label{RT3.5.2}

1445: \end{equation}%

1446: provided (\ref{RT3.1.2}) holds true for $1\leq i<j\leq n.$

1447: \end{remark}

1448: 

1449: The following corollary providing a better bound for $\sum_{i=1}^{n}\left%

1450: \Vert x_{i}\right\Vert ,$ holds \cite{RTxSSD2}.

1451: 

1452: \begin{corollary}

1453: \label{RTc3.6.2}With the assumptions in Lemma \ref{RTl3.1.2}, one has the

1454: inequality:%

1455: \begin{equation}

1456: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \leq \sqrt{\frac{nk}{n+k-1}}%

1457: \left\Vert \sum_{i=1}^{n}x_{i}\right\Vert .  \label{RT3.6.2}

1458: \end{equation}

1459: \end{corollary}

1460: 

1461: \begin{proof}

1462: Using the Cauchy-Bunyakovsky-Schwarz inequality%

1463: \begin{equation*}

1464: n\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert ^{2}\geq \left(

1465: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}

1466: \end{equation*}%

1467: we get%

1468: \begin{equation}

1469: \left( k-1\right) \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert ^{2}+\left(

1470: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}\geq \left( \frac{k-1}{%

1471: n}+1\right) \left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}.

1472: \label{RT3.7.2}

1473: \end{equation}%

1474: Consequently, by (\ref{RT3.7.2}) and (\ref{RT3.2.2}) we deduce%

1475: \begin{equation*}

1476: k\left\Vert \sum_{i=1}^{n}x_{i}\right\Vert ^{2}\geq \frac{n+k-1}{n}\left(

1477: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right) ^{2}

1478: \end{equation*}%

1479: giving the desired inequality (\ref{RT3.6.2}).

1480: \end{proof}

1481: 

1482: \subsection{Asymmetric Assumptions}

1483: 

1484: The following result may be stated as well \cite{RTxSSD2}.

1485: 

1486: \begin{theorem}[Dragomir, 2004]

1487: \label{RTt3.7.2}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

1488: \right) $ be an inner product space and $x_{i}\in H\backslash \left\{

1489: 0\right\} ,$ $i\in \left\{ 1,\dots ,n\right\} ,$ $\rho \in \left( 0,1\right)

1490: ,$ such that%

1491: \begin{equation}

1492: \left\Vert x_{i}-\frac{x_{j}}{\left\Vert x_{j}\right\Vert }\right\Vert \leq

1493: \rho \ \ \text{ \ for }1\leq i<j\leq n.  \label{RT3.8.2}

1494: \end{equation}%

1495: Then we have the inequality%

1496: \begin{multline}

1497: \sqrt{1-\rho ^{2}}\left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \right)

1498: ^{2}+\left( 1-\sqrt{1-\rho ^{2}}\right) \sum_{i=1}^{n}\left\Vert

1499: x_{i}\right\Vert ^{2}  \label{RT3.9.2} \\

1500: \leq \left\Vert \sum_{i=1}^{n}x_{i}\right\Vert ^{2}.

1501: \end{multline}%

1502: The case of equality holds in (\ref{RT3.9.2}) iff%

1503: \begin{equation}

1504: \left\Vert x_{i}\right\Vert \left\Vert x_{j}\right\Vert =\frac{1}{\sqrt{%

1505: 1-\rho ^{2}}}\func{Re}\left\langle x_{i},x_{j}\right\rangle  \label{RT3.10.2}

1506: \end{equation}%

1507: for any $1\leq i<j\leq n.$

1508: \end{theorem}

1509: 

1510: \begin{proof}

1511: The condition (\ref{RT3.1.2}) is obviously equivalent to%

1512: \begin{equation*}

1513: \left\Vert x_{i}\right\Vert ^{2}+1-\rho ^{2}\leq 2\func{Re}\left\langle

1514: x_{i},\frac{x_{j}}{\left\Vert x_{j}\right\Vert }\right\rangle

1515: \end{equation*}%

1516: for each $1\leq i<j\leq n.$

1517: 

1518: Dividing by $\sqrt{1-\rho ^{2}}>0,$ we deduce%

1519: \begin{equation}

1520: \frac{\left\Vert x_{i}\right\Vert ^{2}}{\sqrt{1-\rho ^{2}}}+\sqrt{1-\rho ^{2}%

1521: }\leq \frac{2}{\sqrt{1-\rho ^{2}}}\func{Re}\left\langle x_{i},\frac{x_{j}}{%

1522: \left\Vert x_{j}\right\Vert }\right\rangle ,  \label{RT3.11.2}

1523: \end{equation}%

1524: for $1\leq i<j\leq n.$

1525: 

1526: On the other hand, by the elementary inequality%

1527: \begin{equation}

1528: \frac{p}{\alpha }+q\alpha \geq 2\sqrt{pq},\ \ \ p,q\geq 0,\ \alpha >0

1529: \label{RT3.12.2}

1530: \end{equation}%

1531: we have%

1532: \begin{equation}

1533: 2\left\Vert x_{i}\right\Vert \leq \frac{\left\Vert x_{i}\right\Vert ^{2}}{%

1534: \sqrt{1-\rho ^{2}}}+\sqrt{1-\rho ^{2}}.  \label{RT3.13.2}

1535: \end{equation}%

1536: Making use of (\ref{RT3.11.2}) and (\ref{RT3.13.2}), we deduce that

1537: \begin{equation*}

1538: \left\Vert x_{i}\right\Vert \left\Vert x_{j}\right\Vert \leq \frac{1}{\sqrt{%

1539: 1-\rho ^{2}}}\func{Re}\left\langle x_{i},x_{j}\right\rangle

1540: \end{equation*}%

1541: for $1\leq i<j\leq n.$

1542: 

1543: Now, applying Lemma \ref{RTl2.1.2} for $k=\frac{1}{\sqrt{1-\rho ^{2}}},$ we

1544: deduce the desired result.

1545: \end{proof}

1546: 

1547: \begin{remark}

1548: \label{RTr3.8.2}If we assume that $\left\Vert x_{i}\right\Vert =1,$ $i\in

1549: \left\{ 1,\dots ,n\right\} ,$ satisfying the simpler condition%

1550: \begin{equation}

1551: \left\Vert x_{j}-x_{i}\right\Vert \leq \rho \ \ \text{ \ for }1\leq i<j\leq

1552: n,  \label{RT3.14.2}

1553: \end{equation}%

1554: then, from (\ref{RT3.9.2}), we deduce the following lower bound for $%

1555: \left\Vert \sum_{i=1}^{n}x_{i}\right\Vert ,$ namely%

1556: \begin{equation}

1557: \left[ n+n\left( n-1\right) \sqrt{1-\rho ^{2}}\right] ^{\frac{1}{2}}\leq

1558: \left\Vert \sum_{i=1}^{n}x_{i}\right\Vert .  \label{RT3.15.2}

1559: \end{equation}%

1560: The equality holds in (\ref{RT3.15.2}) iff $\sqrt{1-\rho ^{2}}=\func{Re}%

1561: \left\langle x_{i},x_{j}\right\rangle $ for $1\leq i<j\leq n.$

1562: \end{remark}

1563: 

1564: \begin{remark}

1565: \label{RTr3.9.2}Under the hypothesis of Proposition \ref{RTp2.7.2}, we have

1566: the coarser but simpler reverse of the triangle inequality%

1567: \begin{equation}

1568: \sqrt[4]{1-\rho ^{2}}\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \leq

1569: \left\Vert \sum_{i=1}^{n}x_{i}\right\Vert .  \label{RT3.16.2}

1570: \end{equation}%

1571: Also, applying Corollary \ref{RTc3.6.2} for $k=\frac{1}{\sqrt{1-\rho ^{2}}},$

1572: we can state that%

1573: \begin{equation}

1574: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \leq \sqrt{\frac{n}{n\sqrt{1-\rho

1575: ^{2}}+1-\sqrt{1-\rho ^{2}}}}\left\Vert \sum_{i=1}^{n}x_{i}\right\Vert ,

1576: \label{RT3.17.2}

1577: \end{equation}%

1578: provided $x_{i}\in H$ satisfy (\ref{RT3.8.2}) for $1\leq i<j\leq n.$

1579: \end{remark}

1580: 

1581: In the same manner, we can state and prove the following reverse of the

1582: quadratic generalised triangle inequality \cite{RTxSSD2}.

1583: 

1584: \begin{theorem}[Dragomir, 2004]

1585: \label{RTt3.10.2}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

1586: \right) $ be an inner product space over the real or complex number field $%

1587: \mathbb{K}$, $x_{i}\in H,$ $i\in \left\{ 1,\dots ,n\right\} $ and $M\geq m>0$

1588: such that either%

1589: \begin{equation}

1590: \func{Re}\left\langle Mx_{j}-x_{i},x_{i}-mx_{j}\right\rangle \geq 0\text{ \

1591: for }1\leq i<j\leq n,  \label{RT3.18.2}

1592: \end{equation}%

1593: or, equivalently,

1594: \begin{equation}

1595: \left\Vert x_{i}-\frac{M+m}{2}x_{j}\right\Vert \leq \frac{1}{2}\left(

1596: M-m\right) \left\Vert x_{j}\right\Vert \text{ \ for }1\leq i<j\leq n

1597: \label{RT3.19.2}

1598: \end{equation}%

1599: hold. Then%

1600: \begin{multline}

1601: \frac{2\sqrt{mM}}{M+m}\left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert

1602: \right) ^{2}+\frac{\left( \sqrt{M}-\sqrt{m}\right) ^{2}}{M+m}%

1603: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert ^{2}  \label{RT3.20.2} \\

1604: \leq \left\Vert \sum_{i=1}^{n}x_{i}\right\Vert ^{2}.

1605: \end{multline}%

1606: The case of equality holds in (\ref{RT3.20.2}) if and only if%

1607: \begin{equation}

1608: \left\Vert x_{i}\right\Vert \left\Vert x_{j}\right\Vert =\frac{M+m}{2\sqrt{mM%

1609: }}\func{Re}\left\langle x_{i},x_{j}\right\rangle \text{ \ for }1\leq i<j\leq

1610: n.  \label{RT3.21.2}

1611: \end{equation}

1612: \end{theorem}

1613: 

1614: \begin{proof}

1615: From (\ref{RT3.18.2}), observe that%

1616: \begin{equation}

1617: \left\Vert x_{i}\right\Vert ^{2}+Mm\left\Vert x_{j}\right\Vert ^{2}\leq

1618: \left( M+m\right) \func{Re}\left\langle x_{i},x_{j}\right\rangle ,

1619: \label{RT3.22.2}

1620: \end{equation}%

1621: for $1\leq i<j\leq n.$ Dividing (\ref{RT3.22.2}) by $\sqrt{mM}>0,$ we deduce%

1622: \begin{equation*}

1623: \frac{\left\Vert x_{i}\right\Vert ^{2}}{\sqrt{mM}}+\sqrt{mM}\left\Vert

1624: x_{j}\right\Vert ^{2}\leq \frac{M+m}{\sqrt{mM}}\func{Re}\left\langle

1625: x_{i},x_{j}\right\rangle ,

1626: \end{equation*}%

1627: and since, obviously%

1628: \begin{equation*}

1629: 2\left\Vert x_{i}\right\Vert \left\Vert x_{j}\right\Vert \leq \frac{%

1630: \left\Vert x_{i}\right\Vert ^{2}}{\sqrt{mM}}+\sqrt{mM}\left\Vert

1631: x_{j}\right\Vert ^{2}

1632: \end{equation*}%

1633: hence%

1634: \begin{equation*}

1635: \left\Vert x_{i}\right\Vert \left\Vert x_{j}\right\Vert \leq \frac{M+m}{2%

1636: \sqrt{mM}}\func{Re}\left\langle x_{i},x_{j}\right\rangle ,\text{ \ for }%

1637: 1\leq i<j\leq n.

1638: \end{equation*}%

1639: Applying Lemma \ref{RTl3.1.2} for $k=\frac{M+m}{2\sqrt{mM}}\geq 1,$ we

1640: deduce the desired result.

1641: \end{proof}

1642: 

1643: \begin{remark}

1644: \label{RTr3.11.2}We also must note that a simpler but coarser inequality

1645: that can be obtained from (\ref{RT3.20.2}) is

1646: \begin{equation*}

1647: \left( \frac{2\sqrt{mM}}{M+m}\right) ^{\frac{1}{2}}\sum_{i=1}^{n}\left\Vert

1648: x_{i}\right\Vert \leq \left\Vert \sum_{i=1}^{n}x_{i}\right\Vert ,

1649: \end{equation*}%

1650: provided (\ref{RT3.18.2}) holds true.

1651: \end{remark}

1652: 

1653: Finally, a different result related to the generalised triangle inequality

1654: is incorporated in the following theorem \cite{RTxSSD2}.

1655: 

1656: \begin{theorem}[Dragomir, 2004]

1657: \label{RTt3.12.2}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

1658: \right) $ be an inner product space over $\mathbb{K}$, $\eta >0$ and $%

1659: x_{i}\in H,$ $i\in \left\{ 1,\dots ,n\right\} $ with the property that%

1660: \begin{equation}

1661: \left\Vert x_{j}-x_{i}\right\Vert \leq \eta <\left\Vert x_{j}\right\Vert

1662: \text{ \ for each \ }i,j\in \left\{ 1,\dots ,n\right\} .  \label{RT3.23.2}

1663: \end{equation}%

1664: Then we have the following reverse of the triangle inequality%

1665: \begin{equation}

1666: \frac{\sum_{i=1}^{n}\sqrt{\left\Vert x_{i}\right\Vert ^{2}-\eta ^{2}}}{%

1667: \left\Vert \sum_{i=1}^{n}x_{i}\right\Vert }\leq \frac{\left\Vert

1668: \sum_{i=1}^{n}x_{i}\right\Vert }{\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert }.

1669: \label{RT3.24.2}

1670: \end{equation}%

1671: The equality holds in (\ref{RT3.24.2}) iff%

1672: \begin{equation}

1673: \left\Vert x_{i}\right\Vert \sqrt{\left\Vert x_{j}\right\Vert ^{2}-\eta ^{2}}%

1674: =\func{Re}\left\langle x_{i},x_{j}\right\rangle \text{ \ for each \ }i,j\in

1675: \left\{ 1,\dots ,n\right\} .  \label{RT3.25.2}

1676: \end{equation}

1677: \end{theorem}

1678: 

1679: \begin{proof}

1680: From (\ref{RT3.23.2}), we have%

1681: \begin{equation*}

1682: \left\Vert x_{i}\right\Vert ^{2}+\left\Vert x_{j}\right\Vert ^{2}-\eta

1683: ^{2}\leq 2\func{Re}\left\langle x_{i},x_{j}\right\rangle ,\ \ \ i,j\in

1684: \left\{ 1,\dots ,n\right\} .

1685: \end{equation*}%

1686: On the other hand,

1687: \begin{equation*}

1688: 2\left\Vert x_{i}\right\Vert \sqrt{\left\Vert x_{j}\right\Vert ^{2}-\eta ^{2}%

1689: }\leq \left\Vert x_{i}\right\Vert ^{2}+\left\Vert x_{j}\right\Vert ^{2}-\eta

1690: ^{2},\ \ \ i,j\in \left\{ 1,\dots ,n\right\}

1691: \end{equation*}%

1692: and thus%

1693: \begin{equation*}

1694: \left\Vert x_{i}\right\Vert \sqrt{\left\Vert x_{j}\right\Vert ^{2}-\eta ^{2}}%

1695: \leq \func{Re}\left\langle x_{i},x_{j}\right\rangle ,\ \ \ i,j\in \left\{

1696: 1,\dots ,n\right\} .

1697: \end{equation*}%

1698: Summing over $i,j\in \left\{ 1,\dots ,n\right\} ,$ we deduce the desired

1699: inequality (\ref{RT3.24.2}).

1700: 

1701: The case of equality is also obvious from the above, and we omit the details.

1702: \end{proof}

1703: 

1704: \section{Reverses for Complex Spaces}

1705: 

1706: \subsection{The Case of One Vector}

1707: 

1708: The following result holds \cite{RTxSSD3}.

1709: 

1710: \begin{theorem}[Dragomir, 2004]

1711: \label{RTt2.1.3}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

1712: \right) $ be a complex inner product space. Suppose that the vectors $%

1713: x_{k}\in H,$ $k\in \left\{ 1,\dots ,n\right\} $ satisfy the condition%

1714: \begin{equation}

1715: 0\leq r_{1}\left\Vert x_{k}\right\Vert \leq \func{Re}\left\langle

1716: x_{k},e\right\rangle ,\quad 0\leq r_{2}\left\Vert x_{k}\right\Vert \leq

1717: \func{Im}\left\langle x_{k},e\right\rangle  \label{RT2.1.3}

1718: \end{equation}%

1719: for each $k\in \left\{ 1,\dots ,n\right\} ,$ where $e\in H$ is such that $%

1720: \left\Vert e\right\Vert =1$ and $r_{1},r_{2}\geq 0.$ Then we have the

1721: inequality%

1722: \begin{equation}

1723: \sqrt{r_{1}^{2}+r_{2}^{2}}\sum_{k=1}^{n}\left\Vert x_{k}\right\Vert \leq

1724: \left\Vert \sum_{k=1}^{n}x_{k}\right\Vert ,  \label{RT2.2.3}

1725: \end{equation}%

1726: where equality holds if and only if%

1727: \begin{equation}

1728: \sum_{k=1}^{n}x_{k}=\left( r_{1}+ir_{2}\right) \left(

1729: \sum_{k=1}^{n}\left\Vert x_{k}\right\Vert \right) e.  \label{RT2.3.3}

1730: \end{equation}

1731: \end{theorem}

1732: 

1733: \begin{proof}

1734: In view of the Schwarz inequality in the complex inner product space $\left(

1735: H;\left\langle \cdot ,\cdot \right\rangle \right) ,$ we have%

1736: \begin{align}

1737: \left\Vert \sum_{k=1}^{n}x_{k}\right\Vert ^{2}& =\left\Vert

1738: \sum_{k=1}^{n}x_{k}\right\Vert ^{2}\left\Vert e\right\Vert ^{2}\geq

1739: \left\vert \left\langle \sum_{k=1}^{n}x_{k},e\right\rangle \right\vert ^{2}

1740: \label{RT2.4.3} \\

1741: & =\left\vert \left\langle \sum_{k=1}^{n}x_{k},e\right\rangle \right\vert

1742: ^{2}  \notag \\

1743: & =\left\vert \sum_{k=1}^{n}\func{Re}\left\langle x_{k},e\right\rangle

1744: +i\left( \sum_{k=1}^{n}\func{Im}\left\langle x_{k},e\right\rangle \right)

1745: \right\vert ^{2}  \notag \\

1746: & =\left( \sum_{k=1}^{n}\func{Re}\left\langle x_{k},e\right\rangle \right)

1747: ^{2}+\left( \sum_{k=1}^{n}\func{Im}\left\langle x_{k},e\right\rangle \right)

1748: ^{2}.  \notag

1749: \end{align}%

1750: Now, by hypothesis (\ref{RT2.1.3})%

1751: \begin{equation}

1752: \left( \sum_{k=1}^{n}\func{Re}\left\langle x_{k},e\right\rangle \right)

1753: ^{2}\geq r_{1}^{2}\left( \sum_{k=1}^{n}\left\Vert x_{k}\right\Vert \right)

1754: ^{2}  \label{RT2.5.3}

1755: \end{equation}%

1756: and%

1757: \begin{equation}

1758: \left( \sum_{k=1}^{n}\func{Im}\left\langle x_{k},e\right\rangle \right)

1759: ^{2}\geq r_{2}^{2}\left( \sum_{k=1}^{n}\left\Vert x_{k}\right\Vert \right)

1760: ^{2}.  \label{RT2.6.3}

1761: \end{equation}%

1762: If we add (\ref{RT2.5.3}) and (\ref{RT2.6.3}) and use (\ref{RT2.4.3}), then

1763: we deduce the desired inequality (\ref{RT2.2.3}).

1764: 

1765: Now, if (\ref{RT2.3.3}) holds, then%

1766: \begin{equation*}

1767: \left\Vert \sum_{k=1}^{n}x_{k}\right\Vert =\left\vert

1768: r_{1}+ir_{2}\right\vert \left( \sum_{k=1}^{n}\left\Vert x_{k}\right\Vert

1769: \right) \left\Vert e\right\Vert =\sqrt{r_{1}^{2}+r_{2}^{2}}%

1770: \sum_{k=1}^{n}\left\Vert x_{k}\right\Vert

1771: \end{equation*}%

1772: and the case of equality is valid in (\ref{RT2.2.3}).

1773: 

1774: Before we prove the reverse implication, let us observe that for $x\in H$

1775: and $e\in H,$ $\left\Vert e\right\Vert =1,$ the following identity is true%

1776: \begin{equation*}

1777: \left\Vert x-\left\langle x,e\right\rangle e\right\Vert ^{2}=\left\Vert

1778: x\right\Vert ^{2}-\left\vert \left\langle x,e\right\rangle \right\vert ^{2},

1779: \end{equation*}%

1780: therefore $\left\Vert x\right\Vert =\left\vert \left\langle x,e\right\rangle

1781: \right\vert $ if and only if $x=\left\langle x,e\right\rangle e.$

1782: 

1783: If we assume that equality holds in (\ref{RT2.2.3}), then the case of

1784: equality must hold in all the inequalities required in the argument used to

1785: prove the inequality (\ref{RT2.2.3}), and we may state that%

1786: \begin{equation}

1787: \left\Vert \sum_{k=1}^{n}x_{k}\right\Vert =\left\vert \left\langle

1788: \sum_{k=1}^{n}x_{k},e\right\rangle \right\vert ,  \label{RT2.7.3}

1789: \end{equation}%

1790: and

1791: \begin{equation}

1792: r_{1}\left\Vert x_{k}\right\Vert =\func{Re}\left\langle x_{k},e\right\rangle

1793: ,\quad r_{2}\left\Vert x_{k}\right\Vert =\func{Im}\left\langle

1794: x_{k},e\right\rangle  \label{RT2.8.3}

1795: \end{equation}%

1796: for each $k\in \left\{ 1,\dots ,n\right\} .$

1797: 

1798: From (\ref{RT2.7.3}) we deduce%

1799: \begin{equation}

1800: \sum_{k=1}^{n}x_{k}=\left\langle \sum_{k=1}^{n}x_{k},e\right\rangle e

1801: \label{RT2.9.3}

1802: \end{equation}%

1803: and from (\ref{RT2.8.3}), by multiplying the second equation with $i$ and

1804: summing both equations over $k$ from $1$ to $n,$ we deduce%

1805: \begin{equation}

1806: \left( r_{1}+ir_{2}\right) \sum_{k=1}^{n}\left\Vert x_{k}\right\Vert

1807: =\left\langle \sum_{k=1}^{n}x_{k},e\right\rangle .  \label{RT2.10.3}

1808: \end{equation}%

1809: Finally, by (\ref{RT2.10.3}) and (\ref{RT2.9.3}), we get the desired

1810: equality (\ref{RT2.3.3}).

1811: \end{proof}

1812: 

1813: The following corollary is of interest \cite{RTxSSD3}.

1814: 

1815: \begin{corollary}

1816: \label{RTc2.2.3}Let $e$ a unit vector in the complex inner product space $%

1817: \left( H;\left\langle \cdot ,\cdot \right\rangle \right) $ and $\rho

1818: _{1},\rho _{2}\in \left( 0,1\right) .$ If $x_{k}\in H,$ $k\in \left\{

1819: 1,\dots ,n\right\} $ are such that%

1820: \begin{equation}

1821: \left\Vert x_{k}-e\right\Vert \leq \rho _{1},\ \ \ \left\Vert

1822: x_{k}-ie\right\Vert \leq \rho _{2}\ \ \ \text{for each \ }k\in \left\{

1823: 1,\dots ,n\right\} ,  \label{RT2.11.3}

1824: \end{equation}%

1825: then we have the inequality%

1826: \begin{equation}

1827: \sqrt{2-\rho _{1}^{2}-\rho _{2}^{2}}\sum_{k=1}^{n}\left\Vert

1828: x_{k}\right\Vert \leq \left\Vert \sum_{k=1}^{n}x_{k}\right\Vert ,

1829: \label{RT2.12.3}

1830: \end{equation}%

1831: with equality if and only if%

1832: \begin{equation}

1833: \sum_{k=1}^{n}x_{k}=\left( \sqrt{1-\rho _{1}^{2}}+i\sqrt{1-\rho _{2}^{2}}%

1834: \right) \left( \sum_{k=1}^{n}\left\Vert x_{k}\right\Vert \right) e.

1835: \label{RT2.13.3}

1836: \end{equation}

1837: \end{corollary}

1838: 

1839: \begin{proof}

1840: From the first inequality in (\ref{RT2.11.3}) we deduce%

1841: \begin{equation}

1842: 0\leq \sqrt{1-\rho _{1}^{2}}\left\Vert x_{k}\right\Vert \leq \func{Re}%

1843: \left\langle x_{k},e\right\rangle  \label{RT2.16.3}

1844: \end{equation}%

1845: for each $k\in \left\{ 1,\dots ,n\right\} .$

1846: 

1847: From the second inequality in (\ref{RT2.11.3}) we deduce%

1848: \begin{equation*}

1849: 0\leq \sqrt{1-\rho _{2}^{2}}\left\Vert x_{k}\right\Vert \leq \func{Re}%

1850: \left\langle x_{k},ie\right\rangle

1851: \end{equation*}%

1852: for each $k\in \left\{ 1,\dots ,n\right\} .$ Since%

1853: \begin{equation*}

1854: \func{Re}\left\langle x_{k},ie\right\rangle =\func{Im}\left\langle

1855: x_{k},e\right\rangle ,

1856: \end{equation*}%

1857: hence%

1858: \begin{equation}

1859: 0\leq \sqrt{1-\rho _{2}^{2}}\left\Vert x_{k}\right\Vert \leq \func{Im}%

1860: \left\langle x_{k},e\right\rangle  \label{RT2.17.3}

1861: \end{equation}%

1862: for each $k\in \left\{ 1,\dots ,n\right\} .$

1863: 

1864: Now, observe from (\ref{RT2.16.3}) and (\ref{RT2.17.3}), that the condition (%

1865: \ref{RT2.1.3}) of Theorem \ref{RTt2.1.3} is satisfied for $r_{1}=\sqrt{%

1866: 1-\rho _{1}^{2}},$ $r_{2}=\sqrt{1-\rho _{2}^{2}}\in \left( 0,1\right) ,$ and

1867: thus the corollary is proved.

1868: \end{proof}

1869: 

1870: The following corollary may be stated as well \cite{RTxSSD3}.

1871: 

1872: \begin{corollary}

1873: \label{RTc2.3.3}Let $e$ be a unit vector in the complex inner product space $%

1874: \left( H;\left\langle \cdot ,\cdot \right\rangle \right) $ and $M_{1}\geq

1875: m_{1}>0,$ $M_{2}\geq m_{2}>0.$ If $x_{k}\in H,$ $k\in \left\{ 1,\dots

1876: ,n\right\} $ are such that either%

1877: \begin{align}

1878: \func{Re}\left\langle M_{1}e-x_{k},x_{k}-m_{1}e\right\rangle & \geq 0,\

1879: \label{RT2.18.3} \\

1880: \func{Re}\left\langle M_{2}ie-x_{k},x_{k}-m_{2}ie\right\rangle & \geq 0

1881: \notag

1882: \end{align}%

1883: or, equivalently,%

1884: \begin{align}

1885: \left\Vert x_{k}-\frac{M_{1}+m_{1}}{2}e\right\Vert & \leq \frac{1}{2}\left(

1886: M_{1}-m_{1}\right) ,  \label{RT2.19.3} \\

1887: \left\Vert x_{k}-\frac{M_{2}+m_{2}}{2}ie\right\Vert & \leq \frac{1}{2}\left(

1888: M_{2}-m_{2}\right) ,  \notag

1889: \end{align}%

1890: for each $k\in \left\{ 1,\dots ,n\right\} ,$ then we have the inequality%

1891: \begin{equation}

1892: 2\left[ \frac{m_{1}M_{1}}{\left( M_{1}+m_{1}\right) ^{2}}+\frac{m_{2}M_{2}}{%

1893: \left( M_{2}+m_{2}\right) ^{2}}\right] ^{\frac{1}{2}}\sum_{k=1}^{n}\left%

1894: \Vert x_{k}\right\Vert \leq \left\Vert \sum_{k=1}^{n}x_{k}\right\Vert .

1895: \label{RT2.20.3}

1896: \end{equation}%

1897: The equality holds in (\ref{RT2.20.3}) if and only if%

1898: \begin{equation}

1899: \sum_{k=1}^{n}x_{k}=2\left( \frac{\sqrt{m_{1}M_{1}}}{M_{1}+m_{1}}+i\frac{%

1900: \sqrt{m_{2}M_{2}}}{M_{2}+m_{2}}\right) \left( \sum_{k=1}^{n}\left\Vert

1901: x_{k}\right\Vert \right) e.  \label{RT2.21.3}

1902: \end{equation}

1903: \end{corollary}

1904: 

1905: \begin{proof}

1906: From the first inequality in (\ref{RT2.18.3})%

1907: \begin{equation}

1908: 0\leq \frac{2\sqrt{m_{1}M_{1}}}{M_{1}+m_{1}}\left\Vert x_{k}\right\Vert \leq

1909: \func{Re}\left\langle x_{k},e\right\rangle  \label{RT2.24.3}

1910: \end{equation}%

1911: for each $k\in \left\{ 1,\dots ,n\right\} .$

1912: 

1913: Now, the proof follows the same path as the one of Corollary \ref{RTc2.2.3}

1914: and we omit the details.

1915: \end{proof}

1916: 

1917: \subsection{The Case of $m$ Orthonormal Vectors}

1918: 

1919: In \cite{RTxDM}, the authors have proved the following reverse of the

1920: generalised triangle inequality in terms of orthonormal vectors \cite%

1921: {RTxSSD3}.

1922: 

1923: \begin{theorem}[Diaz-Metcalf, 1966]

1924: \label{RTt3.1.3}Let $e_{1},\dots ,e_{m}$ be orthonormal vectors in $\left(

1925: H;\left\langle \cdot ,\cdot \right\rangle \right) $, i.e., we recall that $%

1926: \left\langle e_{i},e_{j}\right\rangle =0$ if $i\neq j$ and $\left\Vert

1927: e_{i}\right\Vert =1,$ $i,j\in \left\{ 1,\dots ,m\right\} .$ Suppose that the

1928: vectors $x_{1},\dots ,x_{n}\in H$ satisfy%

1929: \begin{equation}

1930: 0\leq r_{k}\left\Vert x_{j}\right\Vert \leq \func{Re}\left\langle

1931: x_{j},e_{k}\right\rangle ,  \notag

1932: \end{equation}%

1933: $j\in \left\{ 1,\dots ,n\right\} ,\ k\in \left\{ 1,\dots ,m\right\} .$ Then%

1934: \begin{equation}

1935: \left( \sum_{k=1}^{m}r_{k}^{2}\right) ^{\frac{1}{2}}\sum_{j=1}^{n}\left\Vert

1936: x_{j}\right\Vert \leq \left\Vert \sum_{j=1}^{n}x_{j}\right\Vert ,

1937: \label{RT3.2.3}

1938: \end{equation}%

1939: where equality holds if and only if%

1940: \begin{equation}

1941: \sum_{j=1}^{n}x_{j}=\left( \sum_{j=1}^{n}\left\Vert x_{j}\right\Vert \right)

1942: \sum_{k=1}^{m}r_{k}e_{k}.  \label{RT3.3.3}

1943: \end{equation}

1944: \end{theorem}

1945: 

1946: If the space $\left( H;\left\langle \cdot ,\cdot \right\rangle \right) $ is

1947: complex and more information is available for the imaginary part, then the

1948: following result may be stated as well \cite{RTxSSD3}.

1949: 

1950: \begin{theorem}[Dragomir, 2004]

1951: \label{RTt3.2.3}Let $e_{1},\dots ,e_{m}\in H$ be an orthonormal family of

1952: vectors in the complex inner product space $H.$ If the vectors $x_{1},\dots

1953: ,x_{n}\in H$ satisfy the conditions%

1954: \begin{equation}

1955: 0\leq r_{k}\left\Vert x_{j}\right\Vert \leq \func{Re}\left\langle

1956: x_{j},e_{k}\right\rangle ,\qquad 0\leq \rho _{k}\left\Vert x_{j}\right\Vert

1957: \leq \func{Im}\left\langle x_{j},e_{k}\right\rangle  \label{RT3.4.3}

1958: \end{equation}%

1959: for each $j\in \left\{ 1,\dots ,n\right\} $ and $k\in \left\{ 1,\dots

1960: ,m\right\} ,$ then we have the following reverse of the generalised triangle

1961: inequality;%

1962: \begin{equation}

1963: \left[ \sum_{k=1}^{m}\left( r_{k}^{2}+\rho _{k}^{2}\right) \right] ^{\frac{1%

1964: }{2}}\sum_{j=1}^{n}\left\Vert x_{j}\right\Vert \leq \left\Vert

1965: \sum_{j=1}^{n}x_{j}\right\Vert .  \label{RT3.5.3}

1966: \end{equation}%

1967: The equality holds in (\ref{RT3.5.3}) if and only if%

1968: \begin{equation}

1969: \sum_{j=1}^{n}x_{j}=\left( \sum_{j=1}^{n}\left\Vert x_{j}\right\Vert \right)

1970: \sum_{k=1}^{m}\left( r_{k}+i\rho _{k}\right) e_{k}.  \label{RT3.6.3}

1971: \end{equation}

1972: \end{theorem}

1973: 

1974: \begin{proof}

1975: Before we prove the theorem, let us recall that, if $x\in H$ and $%

1976: e_{1},\dots ,e_{m}$ are orthogonal vectors, then the following identity

1977: holds true:%

1978: \begin{equation}

1979: \left\Vert x-\sum_{k=1}^{m}\left\langle x,e_{k}\right\rangle

1980: e_{k}\right\Vert ^{2}=\left\Vert x\right\Vert ^{2}-\sum_{k=1}^{n}\left\vert

1981: \left\langle x,e_{k}\right\rangle \right\vert ^{2}.  \label{RT3.7.3}

1982: \end{equation}%

1983: As a consequence of this identity, we note the \textit{Bessel inequality}%

1984: \begin{equation}

1985: \sum_{k=1}^{m}\left\vert \left\langle x,e_{k}\right\rangle \right\vert

1986: ^{2}\leq \left\Vert x\right\Vert ^{2},x\in H.  \label{RT3.8.3}

1987: \end{equation}%

1988: The case of equality holds in (\ref{RT3.8.3}) if and only if (see (\ref%

1989: {RT3.7.3}))%

1990: \begin{equation}

1991: x=\sum_{k=1}^{m}\left\langle x,e_{k}\right\rangle e_{k}.  \label{RT3.9.3}

1992: \end{equation}%

1993: Applying Bessel's inequality for $x=\sum_{j=1}^{n}x_{j},$ we have%

1994: \begin{align}

1995: \left\Vert \sum_{j=1}^{n}x_{j}\right\Vert ^{2}& \geq

1996: \sum_{k=1}^{m}\left\vert \left\langle \sum_{j=1}^{n}x_{j},e_{k}\right\rangle

1997: \right\vert ^{2}=\sum_{k=1}^{m}\left\vert \sum_{j=1}^{n}\left\langle

1998: x_{j},e_{k}\right\rangle \right\vert ^{2}  \label{RT3.10.3} \\

1999: & =\sum_{k=1}^{m}\left\vert \left( \sum_{j=1}^{n}\func{Re}\left\langle

2000: x_{j},e_{k}\right\rangle \right) +i\left( \sum_{j=1}^{n}\func{Im}%

2001: \left\langle x_{j},e_{k}\right\rangle \right) \right\vert ^{2}  \notag \\

2002: & =\sum_{k=1}^{m}\left[ \left( \sum_{j=1}^{n}\func{Re}\left\langle

2003: x_{j},e_{k}\right\rangle \right) ^{2}+\left( \sum_{j=1}^{n}\func{Im}%

2004: \left\langle x_{j},e_{k}\right\rangle \right) ^{2}\right] .  \notag

2005: \end{align}%

2006: Now, by the hypothesis (\ref{RT3.4.3}) we have%

2007: \begin{equation}

2008: \left( \sum_{j=1}^{n}\func{Re}\left\langle x_{j},e_{k}\right\rangle \right)

2009: ^{2}\geq r_{k}^{2}\left( \sum_{j=1}^{n}\left\Vert x_{j}\right\Vert \right)

2010: ^{2}  \label{RT3.11}

2011: \end{equation}%

2012: and%

2013: \begin{equation}

2014: \left( \sum_{j=1}^{n}\func{Im}\left\langle x_{j},e_{k}\right\rangle \right)

2015: ^{2}\geq \rho _{k}^{2}\left( \sum_{j=1}^{n}\left\Vert x_{j}\right\Vert

2016: \right) ^{2}.  \label{RT3.12.3}

2017: \end{equation}%

2018: Further, on making use of (\ref{RT3.10.3}) -- (\ref{RT3.12.3}), we deduce%

2019: \begin{align*}

2020: \left\Vert \sum_{j=1}^{n}x_{j}\right\Vert ^{2}& \geq \sum_{k=1}^{m}\left[

2021: r_{k}^{2}\left( \sum_{j=1}^{n}\left\Vert x_{j}\right\Vert \right) ^{2}+\rho

2022: _{k}^{2}\left( \sum_{j=1}^{n}\left\Vert x_{j}\right\Vert \right) ^{2}\right]

2023: \\

2024: & =\left( \sum_{j=1}^{n}\left\Vert x_{j}\right\Vert \right)

2025: ^{2}\sum_{k=1}^{m}\left( r_{k}^{2}+\rho _{k}^{2}\right) ,

2026: \end{align*}%

2027: which is clearly equivalent to (\ref{RT3.5.3}).

2028: 

2029: Now, if (\ref{RT3.6.3}) holds, then%

2030: \begin{align*}

2031: \left\Vert \sum_{j=1}^{n}x_{j}\right\Vert ^{2}& =\left(

2032: \sum_{j=1}^{n}\left\Vert x_{j}\right\Vert \right) ^{2}\left\Vert

2033: \sum_{k=1}^{m}\left( r_{k}+i\rho _{k}\right) e_{k}\right\Vert ^{2} \\

2034: & =\left( \sum_{j=1}^{n}\left\Vert x_{j}\right\Vert \right)

2035: ^{2}\sum_{k=1}^{m}\left\vert r_{k}+i\rho _{k}\right\vert ^{2} \\

2036: & =\left( \sum_{j=1}^{n}\left\Vert x_{j}\right\Vert \right)

2037: ^{2}\sum_{k=1}^{m}\left( r_{k}^{2}+\rho _{k}^{2}\right) ,

2038: \end{align*}%

2039: and the case of equality holds in (\ref{RT3.5.3}).

2040: 

2041: Conversely, if the equality holds in (\ref{RT3.5.3}), then it must hold in

2042: all the inequalities used to prove (\ref{RT3.5.3}) and therefore we must have%

2043: \begin{equation}

2044: \left\Vert \sum_{j=1}^{n}x_{j}\right\Vert ^{2}=\sum_{k=1}^{m}\left\vert

2045: \sum_{j=1}^{n}\left\langle x_{j},e_{k}\right\rangle \right\vert ^{2}

2046: \label{RT3.13.3}

2047: \end{equation}%

2048: and%

2049: \begin{equation}

2050: r_{k}\left\Vert x_{j}\right\Vert =\func{Re}\left\langle

2051: x_{j},e_{k}\right\rangle ,\qquad \rho _{k}\left\Vert x_{j}\right\Vert =\func{%

2052: Im}\left\langle x_{j},e_{k}\right\rangle  \label{RT3.14.3}

2053: \end{equation}%

2054: for each $j\in \left\{ 1,\dots ,n\right\} $ and $k\in \left\{ 1,\dots

2055: ,m\right\} .$

2056: 

2057: Using the identity (\ref{RT3.7.3}), we deduce from (\ref{RT3.13.3}) that%

2058: \begin{equation}

2059: \sum_{j=1}^{n}x_{j}=\sum_{k=1}^{m}\left\langle

2060: \sum_{j=1}^{n}x_{j},e_{k}\right\rangle e_{k}.  \label{RT3.15.3}

2061: \end{equation}%

2062: Multiplying the second equality in (\ref{RT3.14.3}) with the imaginary unit $%

2063: i$ and summing the equality over $j$ from $1$ to $n,$ we deduce%

2064: \begin{equation}

2065: \left( r_{k}+i\rho _{k}\right) \sum_{j=1}^{n}\left\Vert x_{j}\right\Vert

2066: =\left\langle \sum_{j=1}^{n}x_{j},e_{k}\right\rangle  \label{RT3.16.3}

2067: \end{equation}%

2068: for each $k\in \left\{ 1,\dots ,n\right\} .$

2069: 

2070: Finally, utilising (\ref{RT3.15.3}) and (\ref{RT3.16.3}), we deduce (\ref%

2071: {RT3.6.3}) and the theorem is proved.

2072: \end{proof}

2073: 

2074: The following corollaries are of interest \cite{RTxSSD3}.

2075: 

2076: \begin{corollary}

2077: \label{RTc3.2.3}Let $e_{1},\dots ,e_{m}$ be orthonormal vectors in the

2078: complex inner product space $\left( H;\left\langle \cdot ,\cdot

2079: \right\rangle \right) $ and $\rho _{k},\eta _{k}\in \left( 0,1\right) ,$ $%

2080: k\in \left\{ 1,\dots ,n\right\} .$ If $x_{1},\dots ,x_{n}\in H$ are such that%

2081: \begin{equation*}

2082: \left\Vert x_{j}-e_{k}\right\Vert \leq \rho _{k},\qquad \left\Vert

2083: x_{j}-ie_{k}\right\Vert \leq \eta _{k}

2084: \end{equation*}%

2085: for each $j\in \left\{ 1,\dots ,n\right\} $ and $k\in \left\{ 1,\dots

2086: ,m\right\} ,$ then we have the inequality%

2087: \begin{equation}

2088: \left[ \sum_{k=1}^{m}\left( 2-\rho _{k}^{2}-\eta _{k}^{2}\right) \right] ^{%

2089: \frac{1}{2}}\sum_{j=1}^{n}\left\Vert x_{j}\right\Vert \leq \left\Vert

2090: \sum_{j=1}^{n}x_{j}\right\Vert .  \label{RT3.17.3}

2091: \end{equation}%

2092: The case of equality holds in (\ref{RT3.17.3}) if and only if%

2093: \begin{equation}

2094: \sum_{j=1}^{n}x_{j}=\left( \sum_{j=1}^{n}\left\Vert x_{j}\right\Vert \right)

2095: \sum_{k=1}^{m}\left( \sqrt{1-\rho _{k}^{2}}+i\sqrt{1-\eta _{k}^{2}}\right)

2096: e_{k}.  \label{RT3.18.3}

2097: \end{equation}

2098: \end{corollary}

2099: 

2100: The proof employs Theorem \ref{RTt3.2.3} and is similar to the one from

2101: Corollary \ref{RTc2.2.3}. We omit the details.

2102: 

2103: \begin{corollary}

2104: \label{RTc3.3.3}Let $e_{1},\dots ,e_{m}$ be as in Corollary \ref{RTc3.2.3}

2105: and $M_{k}\geq m_{k}>0,$ $N_{k}\geq n_{k}>0,$ $k\in \left\{ 1,\dots

2106: ,m\right\} .$ If $x_{1},\dots ,x_{n}\in H$ are such that either%

2107: \begin{equation*}

2108: \func{Re}\left\langle M_{k}e_{k}-x_{j},x_{j}-m_{k}e_{k}\right\rangle \geq

2109: 0,\ \ \func{Re}\left\langle N_{k}ie_{k}-x_{j},x_{j}-n_{k}ie_{k}\right\rangle

2110: \geq 0

2111: \end{equation*}%

2112: or, equivalently,%

2113: \begin{align*}

2114: \left\Vert x_{j}-\frac{M_{k}+m_{k}}{2}e_{k}\right\Vert & \leq \frac{1}{2}%

2115: \left( M_{k}-m_{k}\right) ,\  \\

2116: \left\Vert x_{j}-\frac{N_{k}+n_{k}}{2}ie_{k}\right\Vert & \leq \frac{1}{2}%

2117: \left( N_{k}-n_{k}\right)

2118: \end{align*}%

2119: for each $j\in \left\{ 1,\dots ,n\right\} $ and $k\in \left\{ 1,\dots

2120: ,m\right\} ,$ then we have the inequality%

2121: \begin{equation}

2122: 2\left\{ \sum_{k=1}^{m}\left[ \frac{m_{k}M_{k}}{\left( M_{k}+m_{k}\right)

2123: ^{2}}+\frac{n_{k}N_{k}}{\left( N_{k}+n_{k}\right) ^{2}}\right] \right\} ^{%

2124: \frac{1}{2}}\sum_{j=1}^{n}\left\Vert x_{j}\right\Vert \leq \left\Vert

2125: \sum_{j=1}^{n}x_{j}\right\Vert .  \label{RT3.19.3}

2126: \end{equation}%

2127: The case of equality holds in (\ref{RT3.19.3}) if and only if%

2128: \begin{equation}

2129: \sum_{j=1}^{n}x_{j}=2\left( \sum_{j=1}^{n}\left\Vert x_{j}\right\Vert

2130: \right) \sum_{k=1}^{m}\left( \frac{\sqrt{m_{k}M_{k}}}{M_{k}+m_{k}}+i\frac{%

2131: \sqrt{n_{k}N_{k}}}{N_{k}+n_{k}}\right) e_{k}.  \label{RT3.20.3}

2132: \end{equation}

2133: \end{corollary}

2134: 

2135: The proof employs Theorem \ref{RTt3.2.3} and is similar to the one in

2136: Corollary \ref{RTc2.3.3}. We omit the details.

2137: 

2138: \section{Applications for Vector-Valued Integral Inequalities}

2139: 

2140: Let $\left( H;\left\langle \cdot ,\cdot \right\rangle \right) $ be a Hilbert

2141: space over the real or complex number field, $\left[ a,b\right] $ a compact

2142: interval in $\mathbb{R}$ and $\eta :\left[ a,b\right] \rightarrow \lbrack

2143: 0,\infty )$ a Lebesgue integrable function on $\left[ a,b\right] $ with the

2144: property that $\int_{a}^{b}\eta \left( t\right) dt=1.$ If, by $L_{\eta

2145: }\left( \left[ a,b\right] ;H\right) $ we denote the Hilbert space of all

2146: Bochner measurable functions $f:\left[ a,b\right] \rightarrow H$ with the

2147: property that $\int_{a}^{b}\eta \left( t\right) \left\Vert f\left( t\right)

2148: \right\Vert ^{2}dt<\infty ,$ then the norm $\left\Vert \cdot \right\Vert

2149: _{\eta }$ of this space is generated by the inner product $\left\langle

2150: \cdot ,\cdot \right\rangle _{\eta }:H\times H\rightarrow \mathbb{K}$ defined

2151: by

2152: \begin{equation*}

2153: \left\langle f,g\right\rangle _{\eta }:=\int_{a}^{b}\eta \left( t\right)

2154: \left\langle f\left( t\right) ,g\left( t\right) \right\rangle dt.

2155: \end{equation*}%

2156: The following proposition providing a reverse of the integral generalised

2157: triangle inequality may be stated \cite{RTxSSD1}.

2158: 

2159: \begin{proposition}

2160: \label{RTp6.1.1} Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

2161: \right) $ be a Hilbert space and $\eta :\left[ a,b\right] \rightarrow

2162: \lbrack 0,\infty )$ as above. If $g\in L_{\eta }\left( \left[ a,b\right]

2163: ;H\right) $ is so that $\int_{a}^{b}\eta \left( t\right) \left\Vert g\left(

2164: t\right) \right\Vert ^{2}dt=1$ and $f_{i}\in L_{\eta }\left( \left[ a,b%

2165: \right] ;H\right) ,i\in \left\{ 1,\dots ,n\right\} ,$ $\rho \in \left(

2166: 0,1\right) $ are so that%

2167: \begin{equation}

2168: \left\Vert f_{i}\left( t\right) -g\left( t\right) \right\Vert \leq \rho

2169: \label{RT6.1.1}

2170: \end{equation}%

2171: for a.e. $t\in \left[ a,b\right] $ and each $i\in \left\{ 1,\dots ,n\right\}

2172: ,$ then we have the inequality%

2173: \begin{multline}

2174: \quad \sqrt{1-\rho ^{2}}\sum_{i=1}^{n}\left( \int_{a}^{b}\eta

2175: \left( t\right) \left\Vert f_{i}\left( t\right) \right\Vert

2176: ^{2}dt\right) ^{\frac{1}{2}}

2177: \label{RT6.2.1} \\

2178: \leq \left( \int_{a}^{b}\eta \left( t\right) \left\Vert

2179: \sum_{i=1}^{n}f_{i}\left( t\right) \right\Vert ^{2}dt\right)

2180: ^{\frac{1}{2}}.\quad

2181: \end{multline}%

2182: The case of equality holds in (\ref{RT6.2.1}) if and only if%

2183: \begin{equation*}

2184: \sum_{i=1}^{n}f_{i}\left( t\right) =\sqrt{1-\rho ^{2}}\sum_{i=1}^{n}\left(

2185: \int_{a}^{b}\eta \left( t\right) \left\Vert f_{i}\left( t\right) \right\Vert

2186: ^{2}dt\right) ^{\frac{1}{2}}\cdot g\left( t\right)

2187: \end{equation*}%

2188: for a.e. $t\in \left[ a,b\right] .$

2189: \end{proposition}

2190: 

2191: \begin{proof}

2192: Observe, by (\ref{RT6.2.1}), that%

2193: \begin{align*}

2194: \left\Vert f_{i}-g\right\Vert _{\eta }& =\left( \int_{a}^{b}\eta \left(

2195: t\right) \left\Vert f_{i}\left( t\right) -g\left( t\right) \right\Vert

2196: ^{2}dt\right) ^{\frac{1}{2}} \\

2197: & \leq \left( \int_{a}^{b}\eta \left( t\right) \rho ^{2}dt\right) ^{\frac{1}{%

2198: 2}}=\rho

2199: \end{align*}%

2200: for each $i\in \left\{ 1,\dots ,n\right\} .$ Applying Theorem \ref{RTt2.1.1}

2201: for the Hilbert space $L_{\eta }\left( \left[ a,b\right] ;H\right) ,$ we

2202: deduce the desired result.

2203: \end{proof}

2204: 

2205: The following result may be stated as well \cite{RTxSSD1}.

2206: 

2207: \begin{proposition}

2208: \label{RTp6.2.1} Let $H,\eta ,g$ be as in Proposition \ref{RTp6.1.1}. If $%

2209: f_{i}\in L_{\eta }\left( \left[ a,b\right] ;H\right) ,i\in \left\{ 1,\dots

2210: ,n\right\} $ and $M\geq m>0$ are so that either%

2211: \begin{equation*}

2212: \func{Re}\left\langle Mg\left( t\right) -f_{i}\left( t\right) ,f_{i}\left(

2213: t\right) -mg\left( t\right) \right\rangle \geq 0

2214: \end{equation*}%

2215: or, equivalently,

2216: \begin{equation*}

2217: \left\Vert f_{i}\left( t\right) -\frac{m+M}{2}g\left( t\right) \right\Vert

2218: \leq \frac{1}{2}\left( M-m\right)

2219: \end{equation*}%

2220: for a.e. $t\in \left[ a,b\right] $ and each $i\in \left\{ 1,\dots ,n\right\}

2221: ,$ then we have the inequality%

2222: \begin{multline}

2223: \quad  \frac{2\sqrt{mM}}{m+M}\sum_{i=1}^{n}\left( \int_{a}^{b}\eta

2224: \left( t\right) \left\Vert f_{i}\left( t\right) \right\Vert

2225: ^{2}dt\right) ^{\frac{1}{2}}

2226: \label{RT6.3.1} \\

2227: \leq \left( \int_{a}^{b}\eta \left( t\right) \left\Vert

2228: \sum_{i=1}^{n}f_{i}\left( t\right) \right\Vert ^{2}dt\right)

2229: ^{\frac{1}{2}}.\quad

2230: \end{multline}%

2231: The equality holds in (\ref{RT6.3.1}) if and only if%

2232: \begin{equation*}

2233: \sum_{i=1}^{n}f_{i}\left( t\right) =\frac{2\sqrt{mM}}{m+M}%

2234: \sum_{i=1}^{n}\left( \int_{a}^{b}\eta \left( t\right) \left\Vert f_{i}\left(

2235: t\right) \right\Vert ^{2}dt\right) ^{\frac{1}{2}}\cdot g\left( t\right) ,

2236: \end{equation*}%

2237: for a.e. $t\in \left[ a,b\right] .$

2238: \end{proposition}

2239: 

2240: The following proposition providing a reverse of the integral generalised

2241: triangle inequality may be stated \cite{RTxSSD2}.

2242: 

2243: \begin{proposition}

2244: \label{RTp4.1.2} Let $\left( H;\left\langle \cdot ,\cdot \right\rangle

2245: \right) $ be a Hilbert space and $\eta :\left[ a,b\right] \rightarrow

2246: \lbrack 0,\infty )$ as above. If $g\in L_{\eta }\left( \left[ a,b\right]

2247: ;H\right) $ is so that $\int_{a}^{b}\eta \left( t\right) \left\Vert g\left(

2248: t\right) \right\Vert ^{2}dt=1$ and $f_{i}\in L_{\eta }\left( \left[ a,b%

2249: \right] ;H\right) ,i\in \left\{ 1,\dots ,n\right\} ,$ and $M\geq m>0$ are so

2250: that either%

2251: \begin{equation}

2252: \func{Re}\left\langle Mf_{j}\left( t\right) -f_{i}\left( t\right)

2253: ,f_{i}\left( t\right) -mf_{j}\left( t\right) \right\rangle \geq 0

2254: \label{RT4.1.2}

2255: \end{equation}%

2256: or, equivalently,

2257: \begin{equation*}

2258: \left\Vert f_{i}\left( t\right) -\frac{m+M}{2}f_{j}\left( t\right)

2259: \right\Vert \leq \frac{1}{2}\left( M-m\right) \left\Vert f_{j}\left(

2260: t\right) \right\Vert

2261: \end{equation*}%

2262: for a.e. $t\in \left[ a,b\right] $ and $1\leq i<j\leq n,$ then we have the

2263: inequality%

2264: \begin{multline}

2265: \left[ \sum_{i=1}^{n}\left( \int_{a}^{b}\eta \left( t\right) \left\Vert

2266: f_{i}\left( t\right) \right\Vert ^{2}dt\right) ^{\frac{1}{2}}\right] ^{2}

2267: \label{RT4.2.2} \\

2268: \leq \int_{a}^{b}\eta \left( t\right) \left\Vert \sum_{i=1}^{n}f_{i}\left(

2269: t\right) \right\Vert ^{2}dt \\

2270: +\frac{1}{2}\cdot \frac{\left( M-m\right) ^{2}}{m+M}\int_{a}^{b}\eta \left(

2271: t\right) \left( \sum_{k=1}^{n-1}k\left\Vert f_{k+1}\left( t\right)

2272: \right\Vert ^{2}\right) dt.

2273: \end{multline}%

2274: The case of equality holds in (\ref{RT4.2.2}) if and only if%

2275: \begin{multline*}

2276: \left( \int_{a}^{b}\eta \left( t\right) \left\Vert f_{i}\left( t\right)

2277: \right\Vert ^{2}dt\right) ^{\frac{1}{2}}\left( \int_{a}^{b}\eta \left(

2278: t\right) \left\Vert f_{j}\left( t\right) \right\Vert ^{2}dt\right) ^{\frac{1%

2279: }{2}} \\

2280: -\int_{a}^{b}\eta \left( t\right) \func{Re}\left\langle f_{i}\left( t\right)

2281: ,f_{j}\left( t\right) \right\rangle dt \\

2282: =\frac{1}{4}\cdot \frac{\left( M-m\right) ^{2}}{m+M}\int_{a}^{b}\eta \left(

2283: t\right) \left\Vert f_{j}\left( t\right) \right\Vert ^{2}dt

2284: \end{multline*}%

2285: for each $i,j$ with $1\leq i<j\leq n.$

2286: \end{proposition}

2287: 

2288: \begin{proof}

2289: We observe that

2290: \begin{multline*}

2291: \func{Re}\left\langle Mf_{j}-f_{i},f_{i}-mf_{j}\right\rangle _{\eta } \\

2292: =\int_{a}^{b}\eta \left( t\right) \func{Re}\left\langle Mf_{j}\left(

2293: t\right) -f_{i}\left( t\right) ,f_{i}\left( t\right) -mf_{j}\left( t\right)

2294: \right\rangle dt\geq 0

2295: \end{multline*}%

2296: for any $i,j$ with $1\leq i<j\leq n.$

2297: 

2298: Applying Theorem \ref{RTt2.9.2} for the Hilbert space $L_{\eta }\left( \left[

2299: a,b\right] ;H\right) $ and for $y_{i}=f_{i},i\in \left\{ 1,\dots ,n\right\}

2300: , $ we deduce the desired result.

2301: \end{proof}

2302: 

2303: Another integral inequality incorporated in the following proposition holds

2304: \cite{RTxSSD2}:

2305: 

2306: \begin{proposition}

2307: \label{RTp4.2.2} With the assumptions of Proposition \ref{RTp4.1.2}, we have

2308: \begin{multline}

2309: \frac{2\sqrt{mM}}{m+M}\left[ \sum_{i=1}^{n}\left( \int_{a}^{b}\eta \left(

2310: t\right) \left\Vert f_{i}\left( t\right) \right\Vert ^{2}dt\right) ^{\frac{1%

2311: }{2}}\right] ^{2}  \label{RT4.3.2} \\

2312: +\frac{\left( \sqrt{M}-\sqrt{m}\right) ^{2}}{m+M}\sum_{i=1}^{n}\int_{a}^{b}%

2313: \eta \left( t\right) \left\Vert f_{i}\left( t\right) \right\Vert ^{2}dt \\

2314: \leq \int_{a}^{b}\eta \left( t\right) \left\Vert \sum_{i=1}^{n}f_{i}\left(

2315: t\right) \right\Vert ^{2}dt.

2316: \end{multline}%

2317: The case of equality holds in (\ref{RT4.3.2}) if and only if

2318: \begin{multline*}

2319: \left( \int_{a}^{b}\eta \left( t\right) \left\Vert f_{i}\left( t\right)

2320: \right\Vert ^{2}dt\right) ^{\frac{1}{2}}\left( \int_{a}^{b}\eta \left(

2321: t\right) \left\Vert f_{j}\left( t\right) \right\Vert ^{2}dt\right) ^{\frac{1%

2322: }{2}} \\

2323: =\frac{M+m}{2\sqrt{mM}}\int_{a}^{b}\eta \left( t\right) \func{Re}%

2324: \left\langle f_{i}\left( t\right) ,f_{j}\left( t\right) \right\rangle dt

2325: \end{multline*}%

2326: for any $i,j$ with $1\leq i<j\leq n.$

2327: \end{proposition}

2328: 

2329: The proof is obvious by Theorem \ref{RTt3.10.2} and we omit the details.

2330: 

2331: \section{Applications for Complex Numbers}

2332: 

2333: The following reverse of the generalised triangle inequality with a clear

2334: geometric meaning may be stated \cite{RTxSSD3}.

2335: 

2336: \begin{proposition}

2337: \label{RTp4.1.3}Let $z_{1},\dots ,z_{n}$ be complex numbers with the

2338: property that%

2339: \begin{equation}

2340: 0\leq \varphi _{1}\leq \arg \left( z_{k}\right) \leq \varphi _{2}<\frac{\pi

2341: }{2}  \label{RT4.1.3}

2342: \end{equation}%

2343: for each $k\in \left\{ 1,\dots ,n\right\} .$ Then we have the inequality%

2344: \begin{equation}

2345: \sqrt{\sin ^{2}\varphi _{1}+\cos ^{2}\varphi _{2}}\sum_{k=1}^{n}\left\vert

2346: z_{k}\right\vert \leq \left\vert \sum_{k=1}^{n}z_{k}\right\vert .

2347: \label{RT4.3.3}

2348: \end{equation}%

2349: The equality holds in (\ref{RT4.3.3}) if and only if%

2350: \begin{equation}

2351: \sum_{k=1}^{n}z_{k}=\left( \cos \varphi _{2}+i\sin \varphi _{1}\right)

2352: \sum_{k=1}^{n}\left\vert z_{k}\right\vert .  \label{RT4.4.3}

2353: \end{equation}

2354: \end{proposition}

2355: 

2356: \begin{proof}

2357: Let $z_{k}=a_{k}+ib_{k}.$ We may assume that $b_{k}\geq 0,$ $a_{k}>0,$ $k\in

2358: \left\{ 1,\dots ,n\right\} ,$ since, by (\ref{RT4.1.3}), $\frac{b_{k}}{a_{k}}%

2359: =\tan \left[ \arg \left( z_{k}\right) \right] \in \left[ 0,\frac{\pi }{2}%

2360: \right) ,$ $k\in \left\{ 1,\dots ,n\right\} .$ By (\ref{RT4.1.3}), we

2361: obviously have%

2362: \begin{equation*}

2363: 0\leq \tan ^{2}\varphi _{1}\leq \frac{b_{k}^{2}}{a_{k}^{2}}\leq \tan

2364: ^{2}\varphi _{2},\qquad k\in \left\{ 1,\dots ,n\right\}

2365: \end{equation*}%

2366: from where we get%

2367: \begin{equation*}

2368: \frac{b_{k}^{2}+a_{k}^{2}}{a_{k}^{2}}\leq \frac{1}{\cos ^{2}\varphi _{2}}%

2369: ,\qquad k\in \left\{ 1,\dots ,n\right\} ,\ \varphi _{2}\in \left( 0,\frac{%

2370: \pi }{2}\right)

2371: \end{equation*}%

2372: and%

2373: \begin{equation*}

2374: \frac{a_{k}^{2}+b_{k}^{2}}{a_{k}^{2}}\leq \frac{1+\tan ^{2}\varphi _{1}}{%

2375: \tan ^{2}\varphi _{1}}=\frac{1}{\sin ^{2}\varphi _{1}},\qquad k\in \left\{

2376: 1,\dots ,n\right\} ,\ \varphi _{1}\in \left( 0,\frac{\pi }{2}\right)

2377: \end{equation*}%

2378: giving the inequalities%

2379: \begin{equation*}

2380: \left\vert z_{k}\right\vert \cos \varphi _{2}\leq \func{Re}\left(

2381: z_{k}\right) ,\ \ \left\vert z_{k}\right\vert \sin \varphi _{1}\leq \func{Im}%

2382: \left( z_{k}\right)

2383: \end{equation*}%

2384: for each $k\in \left\{ 1,\dots ,n\right\} .$

2385: 

2386: Now, applying Theorem \ref{RTt2.1.3} for the complex inner product $\mathbb{C%

2387: }$ endowed with the inner product $\left\langle z,w\right\rangle =z\cdot

2388: \bar{w} $ for $x_{k}=z_{k},$ $r_{1}=\cos \varphi _{2},$ $r_{2}=\sin \varphi

2389: _{1}$ and $e=1,$ we deduce the desired inequality (\ref{RT4.3.3}). The case

2390: of equality is also obvious by Theorem \ref{RTt2.1.3} and the proposition is

2391: proven.

2392: \end{proof}

2393: 

2394: Another result that has an obvious geometrical interpretation is the

2395: following one.

2396: 

2397: \begin{proposition}

2398: \label{RTp4.2.3}Let $c\in \mathbb{C}$ with $\left\vert z\right\vert =1$ and $%

2399: \rho _{1},\rho _{2}\in \left( 0,1\right) .$ If $z_{k}\in \mathbb{C}$, $k\in

2400: \left\{ 1,\dots ,n\right\} $ are such that%

2401: \begin{equation}

2402: \left\vert z_{k}-c\right\vert \leq \rho _{1},\ \ \left\vert

2403: z_{k}-ic\right\vert \leq \rho _{2}\text{ \ for each \ }k\in \left\{ 1,\dots

2404: ,n\right\} ,  \label{RT4.5.3}

2405: \end{equation}%

2406: then we have the inequality%

2407: \begin{equation}

2408: \sqrt{2-\rho _{1}^{2}-\rho _{2}^{2}}\sum_{k=1}^{n}\left\vert

2409: z_{k}\right\vert \leq \left\vert \sum_{k=1}^{n}z_{k}\right\vert ,

2410: \label{RT4.6.3}

2411: \end{equation}%

2412: with equality if and only if%

2413: \begin{equation}

2414: \sum_{k=1}^{n}z_{k}=\left( \sqrt{1-\rho _{1}^{2}}+i\sqrt{1-\rho _{2}^{2}}%

2415: \right) \left( \sum_{k=1}^{n}\left\vert z_{k}\right\vert \right) c.

2416: \label{RT4.7.3}

2417: \end{equation}

2418: \end{proposition}

2419: 

2420: The proof is obvious by Corollary \ref{RTc2.2.3} applied for $H=\mathbb{C}$.

2421: 

2422: \begin{remark}

2423: If we choose $e=1,$ and for $\rho _{1},\rho _{2}\in \left( 0,1\right) $ we

2424: define $\bar{D}\left( 1,\rho _{1}\right) :=\left\{ z\in \mathbb{C}%

2425: |\left\vert z-1\right\vert \leq \rho _{1}\right\} ,$ $\bar{D}\left( i,\rho

2426: _{2}\right) :=\left\{ z\in \mathbb{C}|\left\vert z-i\right\vert \leq \rho

2427: _{2}\right\} ,$ then obviously the intersection%

2428: \begin{equation*}

2429: S_{\rho _{1},\rho _{2}}:=\bar{D}\left( 1,\rho _{1}\right) \cap \bar{D}\left(

2430: i,\rho _{2}\right)

2431: \end{equation*}%

2432: is nonempty if and only if $\rho _{1}+\rho _{2}\geq \sqrt{2}.$

2433: 

2434: If $z_{k}\in S_{\rho _{1},\rho _{2}}$ for $k\in \left\{ 1,\dots ,n\right\} ,$

2435: then (\ref{RT4.6.3}) holds true. The equality holds in (\ref{RT4.6.3}) if

2436: and only if%

2437: \begin{equation*}

2438: \sum_{k=1}^{n}z_{k}=\left( \sqrt{1-\rho _{1}^{2}}+i\sqrt{1-\rho _{2}^{2}}%

2439: \right) \sum_{k=1}^{n}\left\vert z_{k}\right\vert .

2440: \end{equation*}

2441: \end{remark}

2442: 

2443: \begin{thebibliography}{99}

2444: \bibitem{RTxDM} J.B. DIAZ and F.T. METCALF, A complementary triangle

2445: inequality in Hilbert and Banach spaces, \textit{Proceedings Amer. Math.

2446: Soc., }\textbf{17}(1) (1966), 88-97.

2447: 

2448: \bibitem{RTxSSD} S.S. DRAGOMIR, Advances in inequalities of the Schwarz,

2449: Gruss and Bessel type in inner product spaces, Preprint, \texttt{%

2450: http://front.math.ucdavis.edu/math.FA/0309354.}

2451: 

2452: \bibitem{RTxSSD1} S.S. DRAGOMIR, Reverses of the triangle inequality in

2453: inner product spaces, \textit{RGMIA Res. Rep. Coll., }\textbf{7}(2004),

2454: Supplement, Article 7, \texttt{[ONLINE: http://rgmia.vu.edu.au/v7(E).html].}

2455: 

2456: \bibitem{RTxSSD2} S.S. DRAGOMIR, Quadratic reverses of the triangle

2457: inequality in inner product spaces, \textit{RGMIA Res. Rep. Coll., }\textbf{7%

2458: }(2004), Supplement, Article 8, \texttt{[ONLINE:

2459: http://rgmia.vu.edu.au/v7(E).html].}

2460: 

2461: \bibitem{RTxSSD3} S.S. DRAGOMIR, Some reverses of the generalised triangle

2462: inequality in complex inner product spaces, \textit{RGMIA Res. Rep. Coll., }%

2463: \textbf{7}(2004), Supplement, Article 8, \texttt{[ONLINE:

2464: http://rgmia.vu.edu.au/v7(E).html].}

2465: 

2466: \bibitem{RTxKA} J. KARAMATA, \textit{Teorija i Praksa Stieltjesova Integrala}

2467: (Serbo-Coratian) (Stieltjes Integral, Theory and Practice), SANU, Posebna

2468: izdanja, 154, Beograd, 1949.

2469: 

2470: \bibitem{RTxK} S.M. KHALEELULA, On Diaz-Metcalf's complementary triangle

2471: inequality, \textit{Kyungpook Math. J., }\textbf{15} (1975), 9-11.\textit{.}

2472: 

2473: \bibitem{RTxMA} M. MARDEN, The Geometry of the Zeros of a Polynomial in a

2474: Complex Variable, \textit{Amer. Math. Soc. Math. Surveys}, \textbf{3}, New

2475: York, 1949.

2476: 

2477: \bibitem{RTxM} P.M. MILI\v{C}I\'{C}, On a complementary inequality of the

2478: triangle inequality (French), \textit{Mat. Vesnik }\textbf{41}(1989), No. 2,

2479: 83-88.

2480: 

2481: \bibitem{RTxMPF} D.S. MITRINOVI\'{C}, J.E. PE\v{C}ARI\'{C} and\ A.M. FINK,

2482: \textit{Classical and New Inequalities in Analysis, }Kluwer Academic

2483: Publishers, Dordrecht/Boston/London, 1993.

2484: 

2485: \bibitem{RTxP} M. PETROVICH, Module d'une somme, \textit{L' Ensignement Math%

2486: \'{e}matique,} \textbf{19} (1917), 53-56.

2487: 

2488: \bibitem{RTxW} H.S. WILF, Some applications of the inequality of arithmetic

2489: and geometric means to polynomial equations, \textit{Proceedings Amer. Math.

2490: Soc., }\textbf{14} (1963), 263-265.

2491: \end{thebibliography}

2492: