1: %TCIDATA{Version=4.00.0.2321}
2: %TCIDATA{LaTeXparent=0,0,Advances2.tex}
3:
4:
5: \chapter{Schwarz Related Inequalities}
6:
7: \label{ch2}
8:
9: \section{Introduction}
10:
11: Let $H$ be a linear space over the real or complex number field $\mathbb{K}$%
12: . The functional $\left\langle \cdot ,\cdot \right\rangle :H\times
13: H\rightarrow \mathbb{K}$ is called an \textit{inner product }on $H$ if it
14: satisfies the conditions
15:
16: \begin{enumerate}
17: \item[(i)] $\left\langle x,x\right\rangle \geq 0$ for any $x\in H$ and $%
18: \left\langle x,x\right\rangle =0$ iff $x=0;$
19:
20: \item[(ii)] $\left\langle \alpha x+\beta y,z\right\rangle =\alpha
21: \left\langle x,z\right\rangle +\beta \left\langle y,z\right\rangle $ for any
22: $\alpha ,\beta \in \mathbb{K}$ and $x,y,z\in H;$
23:
24: \item[(iii)] $\left\langle y,x\right\rangle =\overline{\left\langle
25: x,y\right\rangle }$ for any $x,y\in H.$
26: \end{enumerate}
27:
28: A first fundamental consequence of the properties (i)-(iii) above, is the
29: \textit{Schwarz inequality:}%
30: \begin{equation}
31: \left\vert \left\langle x,y\right\rangle \right\vert ^{2}\leq \left\langle
32: x,x\right\rangle \left\langle y,y\right\rangle , \label{CBS}
33: \end{equation}%
34: for any $x,y\in H.$ The equality holds in (\ref{CBS}) if and only if the
35: vectors $x$ and $y$ are \textit{linearly dependent,} i.e., there exists a
36: nonzero constant $\alpha \in \mathbb{K}$ so that $x=\alpha y.$
37:
38: If we denote $\left\Vert x\right\Vert :=\sqrt{\left\langle x,x\right\rangle }%
39: ,x\in H,$ then one may state the following properties
40:
41: \begin{enumerate}
42: \item[(n)] $\left\Vert x\right\Vert \geq 0$ for any $x\in H$ and $\left\Vert
43: x\right\Vert =0$ iff $x=0;$
44:
45: \item[(nn)] $\left\Vert \alpha x\right\Vert =\left\vert \alpha \right\vert
46: \left\Vert x\right\Vert $ for any $\alpha \in \mathbb{K}$ and $x\in H;$
47:
48: \item[(nnn)] $\left\Vert x+y\right\Vert \leq \left\Vert x\right\Vert
49: +\left\Vert y\right\Vert $ for any $x,y\in H$ (the triangle inequality);
50: \end{enumerate}
51:
52: i.e., $\left\Vert \cdot \right\Vert $ is a \textit{norm} on $H.$
53:
54: In this chapter we present some classical and recent refinements and reverse
55: inequalities for the Schwarz and the triangle inequalities. More precisely,
56: we point out upper bounds or positive lower bounds for the nonnegative
57: quantities%
58: \begin{equation*}
59: \left\Vert x\right\Vert \left\Vert y\right\Vert -\left\vert \left\langle
60: x,y\right\rangle \right\vert ,\text{ \ }\left\Vert x\right\Vert
61: ^{2}\left\Vert y\right\Vert ^{2}-\left\vert \left\langle x,y\right\rangle
62: \right\vert ^{2}
63: \end{equation*}%
64: and
65: \begin{equation*}
66: \left\Vert x\right\Vert +\left\Vert y\right\Vert -\left\Vert x+y\right\Vert
67: \end{equation*}%
68: under various assumptions for the vectors $x,y\in H.$
69:
70: If the vectors $x,y\in H$ are not \textit{orthogonal}, i.e., $\left\langle
71: x,y\right\rangle \neq 0,$ then some upper and lower bounds for the
72: supra-unitary quantities%
73: \begin{equation*}
74: \frac{\left\Vert x\right\Vert \left\Vert y\right\Vert }{\left\vert
75: \left\langle x,y\right\rangle \right\vert },\ \ \frac{\left\Vert
76: x\right\Vert ^{2}\left\Vert y\right\Vert ^{2}}{\left\vert \left\langle
77: x,y\right\rangle \right\vert ^{2}}
78: \end{equation*}%
79: under appropriate restrictions for the vectors $x$ and $y$ are provided as
80: well.
81:
82: The inequalities obtained by Buzano, Richards, Precupanu and Moore and their
83: extensions and generalizations for orthonormal families of vectors in both
84: real and complex inner product spaces are presented. Recent results
85: concerning the classical refinement of Schwarz inequality due to Kurepa for
86: the complexification of real inner product spaces are also reviewed. Various
87: applications for integral inequalities including a version of Heisenberg
88: inequality for vector valued functions in Hilbert spaces are provided as
89: well.
90:
91: \section{Inequalities Related to Schwarz's One}
92:
93: \subsection{Some Refinements}
94:
95: The following result holds \cite[Theorem 1]{SSD3} (see also \cite[Theorem 2]%
96: {DS}).
97:
98: \begin{theorem}[Dragomir, 1985]
99: \label{t2.1.1}Let $\left( H,\left\langle \cdot ,\cdot \right\rangle \right) $
100: be a real or complex inner product space. Then%
101: \begin{multline}
102: \quad \left( \left\Vert x\right\Vert ^{2}\left\Vert y\right\Vert
103: ^{2}-\left\vert \left\langle x,y\right\rangle \right\vert ^{2}\right) \left(
104: \left\Vert y\right\Vert ^{2}\left\Vert z\right\Vert ^{2}-\left\vert
105: \left\langle y,z\right\rangle \right\vert ^{2}\right) \label{2.1.1} \\
106: \geq \left\vert \left\langle x,z\right\rangle \left\Vert y\right\Vert
107: ^{2}-\left\langle x,y\right\rangle \left\langle y,z\right\rangle \right\vert
108: ^{2}\quad
109: \end{multline}%
110: for any $x,y,z\in H.$
111: \end{theorem}
112:
113: \begin{proof}
114: We follow the proof in \cite{SSD3}.
115:
116: Let us consider the mapping%
117: \begin{equation*}
118: p_{y}:H\times H\rightarrow \mathbb{K},\qquad p_{y}\left( x,z\right)
119: =\left\langle x,z\right\rangle \left\Vert y\right\Vert ^{2}-\left\langle
120: x,y\right\rangle \left\langle y,z\right\rangle
121: \end{equation*}%
122: for each $y\in H\backslash \left\{ 0\right\} .$
123:
124: It is easy to see that $p_{y}\left( \cdot ,\cdot \right) $ is a nonnegative
125: Hermitian form and then on writing Schwarz's inequality%
126: \begin{equation*}
127: \left\vert p_{y}\left( x,z\right) \right\vert ^{2}\leq p_{y}\left(
128: x,x\right) p_{y}\left( z,z\right) ,\qquad x,z\in H
129: \end{equation*}%
130: we obtain the desired inequality (\ref{2.1.1}).
131: \end{proof}
132:
133: \begin{remark}
134: \label{r2.1.1}From (\ref{2.1.1}) it follows that \cite[Corollary 1]{SSD3}
135: (see also \cite[Corollary 2.1]{DS})%
136: \begin{multline}
137: \left( \left\Vert x+z\right\Vert ^{2}\left\Vert y\right\Vert ^{2}-\left\vert
138: \left\langle x+z,y\right\rangle \right\vert ^{2}\right) ^{\frac{1}{2}}
139: \label{2.1.2} \\
140: \leq \left( \left\Vert x\right\Vert ^{2}\left\Vert y\right\Vert
141: ^{2}-\left\vert \left\langle x,y\right\rangle \right\vert ^{2}\right) ^{%
142: \frac{1}{2}}+\left( \left\Vert y\right\Vert ^{2}\left\Vert z\right\Vert
143: ^{2}-\left\vert \left\langle y,z\right\rangle \right\vert ^{2}\right) ^{%
144: \frac{1}{2}}
145: \end{multline}%
146: for every $x,y,z\in H.$
147:
148: Putting $z=\lambda y$ in (\ref{2.1.2}), we get:%
149: \begin{align}
150: 0& \leq \left\Vert x+\lambda y\right\Vert ^{2}\left\Vert y\right\Vert
151: ^{2}-\left\vert \left\langle x+\lambda y,y\right\rangle \right\vert ^{2}
152: \label{2.1.3} \\
153: & \leq \left\Vert x\right\Vert ^{2}\left\Vert y\right\Vert ^{2}-\left\vert
154: \left\langle x,y\right\rangle \right\vert ^{2} \notag
155: \end{align}%
156: and, in particular,%
157: \begin{equation}
158: 0\leq \left\Vert x\pm y\right\Vert ^{2}\left\Vert y\right\Vert
159: ^{2}-\left\vert \left\langle x\pm y,y\right\rangle \right\vert ^{2}\leq
160: \left\Vert x\right\Vert ^{2}\left\Vert y\right\Vert ^{2}-\left\vert
161: \left\langle x,y\right\rangle \right\vert ^{2} \label{2.1.4}
162: \end{equation}%
163: for every $x,y\in H.$
164: \end{remark}
165:
166: Both inequalities (\ref{2.1.3}) and (\ref{2.1.4}) have been obtained in \cite%
167: {SSD3}.
168:
169: We note here that the inequality (\ref{2.1.3}) is in fact equivalent to the
170: following statement%
171: \begin{equation}
172: \sup_{\lambda \in \mathbb{K}}\left[ \left\Vert x+\lambda y\right\Vert
173: ^{2}\left\Vert y\right\Vert ^{2}-\left\vert \left\langle x+\lambda
174: y,y\right\rangle \right\vert ^{2}\right] =\left\Vert x\right\Vert
175: ^{2}\left\Vert y\right\Vert ^{2}-\left\vert \left\langle x,y\right\rangle
176: \right\vert ^{2} \label{2.1.5}
177: \end{equation}%
178: for each $x,y\in H.$
179:
180: The following corollary may be stated \cite[Corollary 2]{SSD3} (see also
181: \cite[Corollary 2.2]{DS}):
182:
183: \begin{corollary}[Dragomir, 1985]
184: \label{c2.1.1}For any $x,y,z\in H\backslash \left\{ 0\right\} $ we have the
185: inequality%
186: \begin{equation}
187: \left\vert \frac{\left\langle x,y\right\rangle }{\left\Vert x\right\Vert
188: \left\Vert y\right\Vert }\right\vert ^{2}+\left\vert \frac{\left\langle
189: y,z\right\rangle }{\left\Vert y\right\Vert \left\Vert z\right\Vert }%
190: \right\vert ^{2}+\left\vert \frac{\left\langle z,x\right\rangle }{\left\Vert
191: z\right\Vert \left\Vert x\right\Vert }\right\vert ^{2}\leq 1+2\left\vert
192: \frac{\left\langle x,y\right\rangle \left\langle y,z\right\rangle
193: \left\langle z,x\right\rangle }{\left\Vert x\right\Vert ^{2}\left\Vert
194: y\right\Vert ^{2}\left\Vert z\right\Vert ^{2}}\right\vert . \label{2.1.6}
195: \end{equation}
196: \end{corollary}
197:
198: \begin{proof}
199: By the modulus properties we obviously have%
200: \begin{equation*}
201: \left\vert \left\langle x,z\right\rangle \left\Vert y\right\Vert
202: ^{2}-\left\langle x,y\right\rangle \left\langle y,z\right\rangle \right\vert
203: \geq \left\vert \left\vert \left\langle x,z\right\rangle \right\vert
204: \left\Vert y\right\Vert ^{2}-\left\vert \left\langle x,y\right\rangle
205: \right\vert \left\vert \left\langle y,z\right\rangle \right\vert \right\vert
206: .
207: \end{equation*}%
208: Therefore, by (\ref{2.1.1}) we may state that%
209: \begin{multline*}
210: \left( \left\Vert x\right\Vert ^{2}\left\Vert y\right\Vert ^{2}-\left\vert
211: \left\langle x,y\right\rangle \right\vert ^{2}\right) \left( \left\Vert
212: y\right\Vert ^{2}\left\Vert z\right\Vert ^{2}-\left\vert \left\langle
213: y,z\right\rangle \right\vert ^{2}\right) \\
214: \geq \left\vert \left\langle x,z\right\rangle \right\vert ^{2}\left\Vert
215: y\right\Vert ^{4}-2\left\vert \left\langle x,y\right\rangle \left\langle
216: y,z\right\rangle \left\langle z,x\right\rangle \right\vert \left\Vert
217: y\right\Vert ^{2}+\left\vert \left\langle x,y\right\rangle \right\vert
218: ^{2}\left\vert \left\langle y,z\right\rangle \right\vert ^{2},
219: \end{multline*}%
220: which, upon elementary calculation, is equivalent to (\ref{2.1.6}).
221: \end{proof}
222:
223: \begin{remark}
224: \label{r2.1.2}If we utilise the elementary inequality $a^{2}+b^{2}+c^{2}\geq
225: 3abc$ when $a,b,c\geq 0,$ then one can state the following inequality%
226: \begin{equation}
227: 3\left\vert \frac{\left\langle x,y\right\rangle \left\langle
228: y,z\right\rangle \left\langle z,x\right\rangle }{\left\Vert x\right\Vert
229: ^{2}\left\Vert y\right\Vert ^{2}\left\Vert z\right\Vert ^{2}}\right\vert
230: \leq \left\vert \frac{\left\langle x,y\right\rangle }{\left\Vert
231: x\right\Vert \left\Vert y\right\Vert }\right\vert ^{2}+\left\vert \frac{%
232: \left\langle y,z\right\rangle }{\left\Vert y\right\Vert \left\Vert
233: z\right\Vert }\right\vert ^{2}+\left\vert \frac{\left\langle
234: z,x\right\rangle }{\left\Vert z\right\Vert \left\Vert x\right\Vert }%
235: \right\vert ^{2} \label{2.1.7}
236: \end{equation}%
237: for any $x,y,z\in H\backslash \left\{ 0\right\} .$ Therefore, the inequality
238: (\ref{2.1.6}) may be regarded as a reverse inequality of (\ref{2.1.7}).
239: \end{remark}
240:
241: The following refinement of the Schwarz inequality holds \cite[Theorem 2]%
242: {SSD3} (see also \cite[Corollary 1.1]{DS}):
243:
244: \begin{theorem}[Dragomir, 1985]
245: \label{t2.1.2}For any $x,y\in H$ and $e\in H$ with $\left\Vert e\right\Vert
246: =1,$ the following refinement of the Schwarz inequality holds:%
247: \begin{equation}
248: \left\Vert x\right\Vert \left\Vert y\right\Vert \geq \left\vert \left\langle
249: x,y\right\rangle -\left\langle x,e\right\rangle \left\langle
250: e,y\right\rangle \right\vert +\left\vert \left\langle x,e\right\rangle
251: \left\langle e,y\right\rangle \right\vert \geq \left\vert \left\langle
252: x,y\right\rangle \right\vert . \label{2.1.8}
253: \end{equation}
254: \end{theorem}
255:
256: \begin{proof}
257: We follow the proof in \cite{SSD3}.
258:
259: Applying the inequality (\ref{2.1.1}), we can state that%
260: \begin{equation}
261: \left( \left\Vert x\right\Vert ^{2}-\left\vert \left\langle x,e\right\rangle
262: \right\vert ^{2}\right) \left( \left\Vert y\right\Vert ^{2}-\left\vert
263: \left\langle y,e\right\rangle \right\vert ^{2}\right) \geq \left\vert
264: \left\langle x,y\right\rangle -\left\langle x,e\right\rangle \left\langle
265: e,y\right\rangle \right\vert ^{2}. \label{2.1.9}
266: \end{equation}%
267: Utilising the elementary inequality for real numbers%
268: \begin{equation}
269: \left( m^{2}-n^{2}\right) \left( p^{2}-q^{2}\right) \leq \left( mp-nq\right)
270: ^{2}, \label{2.1.10}
271: \end{equation}%
272: we can easily see that%
273: \begin{multline}
274: \quad \left( \left\Vert x\right\Vert \left\Vert y\right\Vert -\left\vert
275: \left\langle x,e\right\rangle \left\langle e,y\right\rangle \right\vert
276: \right) ^{2} \label{2.1.11} \\
277: \geq \left( \left\Vert x\right\Vert ^{2}-\left\vert \left\langle
278: x,e\right\rangle \right\vert ^{2}\right) \left( \left\Vert y\right\Vert
279: ^{2}-\left\vert \left\langle y,e\right\rangle \right\vert ^{2}\right) \quad
280: \end{multline}%
281: for any $x,y,e\in H$ with $\left\Vert e\right\Vert =1.$
282:
283: Since, by Schwarz's inequality%
284: \begin{equation}
285: \left\Vert x\right\Vert \left\Vert y\right\Vert \geq \left\vert \left\langle
286: x,e\right\rangle \left\langle e,y\right\rangle \right\vert \label{2.1.12}
287: \end{equation}%
288: hence, by (\ref{2.1.9}) and (\ref{2.1.11}) we deduce the first part of (\ref%
289: {2.1.9}).
290:
291: The second part of (\ref{2.1.9}) is obvious.
292: \end{proof}
293:
294: \begin{corollary}[Dragomir, 1985]
295: \label{c2.1.2}If $x,y,e\in H$ are such that $\left\Vert e\right\Vert =1$ and
296: $x\perp y,$ then%
297: \begin{equation}
298: \left\Vert x\right\Vert \left\Vert y\right\Vert \geq 2\left\vert
299: \left\langle x,e\right\rangle \left\langle e,y\right\rangle \right\vert .
300: \label{2.1.13}
301: \end{equation}
302: \end{corollary}
303:
304: \begin{remark}
305: \label{r2.1.3}Assume that $A:H\rightarrow H$ is a bounded linear operator on
306: $H.$ For $x,e\in H$ with $\left\Vert x\right\Vert =\left\Vert e\right\Vert
307: =1,$ we have by (\ref{2.1.8}) that%
308: \begin{equation}
309: \left\Vert Ay\right\Vert \geq \left\vert \left\langle x,Ay\right\rangle
310: -\left\langle x,e\right\rangle \left\langle e,Ay\right\rangle \right\vert
311: +\left\vert \left\langle x,e\right\rangle \left\langle e,Ay\right\rangle
312: \right\vert \geq \left\vert \left\langle x,Ay\right\rangle \right\vert
313: \label{2.1.14}
314: \end{equation}%
315: for any $y\in H.$
316:
317: Taking the supremum over $x\in H,$ $\left\Vert x\right\Vert =1$ in (\ref%
318: {2.1.14}) and noting that $\left\Vert Ay\right\Vert =\sup\limits_{\left\Vert
319: x\right\Vert =1}\left\vert \left\langle x,Ay\right\rangle \right\vert ,$ we
320: deduce the representation%
321: \begin{equation}
322: \left\Vert Ay\right\Vert =\sup_{\left\Vert x\right\Vert =1}\left\{
323: \left\vert \left\langle x,Ay\right\rangle -\left\langle x,e\right\rangle
324: \left\langle e,Ay\right\rangle \right\vert +\left\vert \left\langle
325: x,e\right\rangle \left\langle e,Ay\right\rangle \right\vert \right\}
326: \label{2.1.15}
327: \end{equation}%
328: for any $y\in H.$ Finally, on taking the supremum over $y\in H,$ $\left\Vert
329: y\right\Vert =1$ in (\ref{2.1.15}) we get%
330: \begin{equation}
331: \left\Vert A\right\Vert =\sup_{\left\Vert y\right\Vert =1,\left\Vert
332: x\right\Vert =1}\left\{ \left\vert \left\langle x,Ay\right\rangle
333: -\left\langle x,e\right\rangle \left\langle e,Ay\right\rangle \right\vert
334: +\left\vert \left\langle x,e\right\rangle \left\langle e,Ay\right\rangle
335: \right\vert \right\} \label{2.1.16}
336: \end{equation}%
337: for any $e\in H,$ $\left\Vert e\right\Vert =1,$ a representation that has
338: been obtained in \cite[Eq. 9]{SSD3}.
339: \end{remark}
340:
341: \begin{remark}
342: \label{r2.1.4}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle \right) $
343: be a Hilbert space. Then for any continuous linear functional $%
344: f:H\rightarrow \mathbb{K}$, $f\neq 0,$ there exists, by the Riesz
345: representation theorem a unique vector $e\in H\backslash \left\{ 0\right\} $
346: such that $f\left( x\right) =\left\langle x,e\right\rangle $ for $x\in H$
347: and $\left\Vert f\right\Vert =\left\Vert e\right\Vert .$
348:
349: If $E$ is a nonzero linear subspace of $H$ and if we denote by $E^{\perp }$
350: its \ orthogonal complement, i.e., we recall that $E^{\perp }:=\left\{ y\in
351: H|y\perp x\right\} $ then for any $x\in E$ and $y\in E^{\perp },$ by (\ref%
352: {2.1.13}) we may state that%
353: \begin{equation*}
354: \left\Vert x\right\Vert \left\Vert y\right\Vert \geq 2\left\vert
355: \left\langle x,\frac{e}{\left\Vert x\right\Vert }\right\rangle \left\langle
356: y,\frac{e}{\left\Vert y\right\Vert }\right\rangle \right\vert ,
357: \end{equation*}%
358: giving, for $x,y\neq 0,$ that%
359: \begin{equation}
360: \left\Vert f\right\Vert ^{2}\geq 2\left\vert \left\langle x,e\right\rangle
361: \left\langle y,e\right\rangle \right\vert =2\left\vert f\left( x\right)
362: \right\vert \left\vert f\left( y\right) \right\vert \label{2.1.17}
363: \end{equation}%
364: for any $x\in E$ and $y\in E^{\perp }.$
365:
366: If by $\left\Vert f\right\Vert _{E}$ we denote the norm of the functional $f$
367: restricted to $E,$ i.e., $\left\Vert f\right\Vert _{E}=\sup_{x\in
368: E\backslash \left\{ 0\right\} }\frac{\left\vert f\left( x\right) \right\vert
369: }{\left\Vert x\right\Vert },$ then, on taking the supremum over $x\in E$ and
370: $y\in E^{\perp }$ in (\ref{2.1.17}) we deduce%
371: \begin{equation}
372: \left\Vert f\right\Vert ^{2}\geq 2\left\Vert f\right\Vert _{E}\cdot
373: \left\Vert f\right\Vert _{E^{\perp }} \label{2.1.18}
374: \end{equation}%
375: for any $E$ a nonzero linear subspace of the Hilbert space $H$ and a given
376: functional $f\in H^{\ast }\backslash \left\{ 0\right\} .$
377:
378: We note that the inequality (\ref{2.1.18}) has been obtained in \cite[Eq. 10]%
379: {SSD3}.
380: \end{remark}
381:
382: \subsection{A Conditional Inequality}
383:
384: The following result providing a lower bound for the norm product under
385: suitable conditions holds \cite{DS1} (see also \cite[Theorem 1]{DS}):
386:
387: \begin{theorem}[Dragomir-S\'{a}ndor, 1986]
388: \label{t2.2.3}Let $x,y,a,b\in H,$ where $\left( H;\left\langle \cdot ,\cdot
389: \right\rangle \right) $ is an inner product space, be such that%
390: \begin{equation}
391: \left\Vert a\right\Vert ^{2}\leq 2\func{Re}\left\langle x,a\right\rangle
392: \qquad \text{and\qquad }\left\Vert y\right\Vert ^{2}\leq 2\func{Re}%
393: \left\langle y,b\right\rangle \label{2.2.1}
394: \end{equation}%
395: holds true. Then%
396: \begin{multline}
397: \left\Vert x\right\Vert \left\Vert y\right\Vert \geq \left( 2\func{Re}%
398: \left\langle x,a\right\rangle -\left\Vert a\right\Vert ^{2}\right) ^{\frac{1%
399: }{2}}\left( 2\func{Re}\left\langle y,b\right\rangle -\left\Vert b\right\Vert
400: ^{2}\right) ^{\frac{1}{2}} \label{2.2.2} \\
401: +\left\vert \left\langle x,y\right\rangle -\left\langle x,b\right\rangle
402: -\left\langle a,y\right\rangle +\left\langle a,b\right\rangle \right\vert .
403: \end{multline}
404: \end{theorem}
405:
406: \begin{proof}
407: We follow the proof in \cite{DS1}.
408:
409: Observe that%
410: \begin{align}
411: & \left\vert \left\langle x,y\right\rangle -\left\langle x,b\right\rangle
412: -\left\langle a,y\right\rangle +\left\langle a,b\right\rangle \right\vert
413: \label{2.2.3} \\
414: & =\left\vert \left\langle x-a,y-b\right\rangle \right\vert ^{2} \notag \\
415: & \leq \left\Vert x-a\right\Vert ^{2}\left\Vert y-b\right\Vert ^{2} \notag
416: \\
417: & =\left[ \left\Vert x\right\Vert ^{2}-\left( 2\func{Re}\left\langle
418: x,a\right\rangle -\left\Vert a\right\Vert ^{2}\right) \right] \left[
419: \left\Vert y\right\Vert ^{2}-\left( 2\func{Re}\left\langle y,b\right\rangle
420: -\left\Vert b\right\Vert ^{2}\right) \right] . \notag
421: \end{align}%
422: Applying the elementary inequality (\ref{2.1.10}) we have%
423: \begin{multline}
424: \left\{ \left\Vert x\right\Vert ^{2}-\left[ \left( 2\func{Re}\left\langle
425: x,a\right\rangle -\left\Vert a\right\Vert ^{2}\right) ^{\frac{1}{2}}\right]
426: ^{2}\right\} \label{2.2.4} \\
427: \times \left\{ \left\Vert y\right\Vert ^{2}-\left[ \left( 2\func{Re}%
428: \left\langle y,b\right\rangle -\left\Vert b\right\Vert ^{2}\right) ^{\frac{1%
429: }{2}}\right] ^{2}\right\} \\
430: \leq \left[ \left\Vert x\right\Vert \left\Vert y\right\Vert -\left( 2\func{Re%
431: }\left\langle x,a\right\rangle -\left\Vert a\right\Vert ^{2}\right) ^{\frac{1%
432: }{2}}\left( 2\func{Re}\left\langle y,b\right\rangle -\left\Vert b\right\Vert
433: ^{2}\right) ^{\frac{1}{2}}\right] .
434: \end{multline}%
435: Since%
436: \begin{align*}
437: 0& \leq \left( 2\func{Re}\left\langle x,a\right\rangle -\left\Vert
438: a\right\Vert ^{2}\right) ^{\frac{1}{2}}\leq \left\Vert x\right\Vert \quad
439: \text{and} \\
440: 0& \leq \left( 2\func{Re}\left\langle y,b\right\rangle -\left\Vert
441: b\right\Vert ^{2}\right) ^{\frac{1}{2}}\leq \left\Vert y\right\Vert
442: \end{align*}%
443: hence%
444: \begin{equation*}
445: \left\Vert x\right\Vert \left\Vert y\right\Vert \geq \left( 2\func{Re}%
446: \left\langle x,a\right\rangle -\left\Vert a\right\Vert ^{2}\right) ^{\frac{1%
447: }{2}}\left( 2\func{Re}\left\langle y,b\right\rangle -\left\Vert b\right\Vert
448: ^{2}\right) ^{\frac{1}{2}}
449: \end{equation*}%
450: and by (\ref{2.2.3}) and (\ref{2.2.4}) we deduce the desired result (\ref%
451: {2.2.2}).
452: \end{proof}
453:
454: \begin{remark}
455: \label{r2.2.4}As pointed out in \cite{DS1}, if we consider $a=\left\langle
456: x,e\right\rangle e,$ $b=\left\langle y,e\right\rangle e$ with $e\in H,$ $%
457: \left\Vert e\right\Vert =1$, then the condition (\ref{2.2.1}) is obviously
458: satisfied and the inequality (\ref{2.2.2}) becomes%
459: \begin{align}
460: \left\Vert x\right\Vert \left\Vert y\right\Vert & \geq \left\vert
461: \left\langle x,e\right\rangle \left\langle e,y\right\rangle \right\vert
462: +\left\vert \left\langle x,y\right\rangle -\left\langle x,e\right\rangle
463: \left\langle e,y\right\rangle \right\vert \label{2.2.5} \\
464: (& \geq \left\vert \left\langle x,y\right\rangle \right\vert ), \notag
465: \end{align}%
466: which is the refinement of the Schwarz inequality incorporated in (\ref%
467: {2.1.8}).
468: \end{remark}
469:
470: For vectors located in a closed ball centered at $0$ and of radius $\sqrt{2}%
471: , $ one can state the following corollary as well \cite[Corollary 1.2]{DS}.
472:
473: \begin{corollary}
474: \label{c2.2.3}Let $x,y\in H$ such that $\left\Vert x\right\Vert ,\left\Vert
475: y\right\Vert \leq \sqrt{2}.$ Then%
476: \begin{multline}
477: \left\Vert x\right\Vert \left\Vert y\right\Vert \geq \left\vert \left\langle
478: x,y\right\rangle \right\vert ^{2}\left( 2-\left\Vert x\right\Vert
479: ^{2}\right) ^{\frac{1}{2}}\left( 2-\left\Vert y\right\Vert ^{2}\right) ^{%
480: \frac{1}{2}} \label{2.2.6} \\
481: +\left\vert \left\langle x,y\right\rangle \right\vert \left\vert
482: 1-\left\Vert x\right\Vert ^{2}-\left\Vert y\right\Vert ^{2}+\left\vert
483: \left\langle x,y\right\rangle \right\vert ^{2}\right\vert .
484: \end{multline}
485: \end{corollary}
486:
487: \begin{proof}
488: Follows by Theorem \ref{t2.2.3} on choosing $a=\left\langle x,y\right\rangle
489: y,$ $b=\left\langle y,x\right\rangle x.$ We omit the details.
490: \end{proof}
491:
492: \subsection{A Refinement for Orthonormal Families}
493:
494: The following result provides a generalisation for a refinement of the
495: Schwarz inequality incorporated in (\ref{2.1.8}) \cite[Theorem 3]{SSD3} (see
496: also \cite[Theorem]{SSD4} or \cite[Theorem 3]{DS}):
497:
498: \begin{theorem}[Dragomir, 1985]
499: \label{t2.3.4}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle \right) $
500: be an inner product space over the real or complex number field $\mathbb{K}$
501: and $\left\{ e_{i}\right\} _{i\in I}$ an orthonormal family in $I.$ For any $%
502: F$ a nonempty finite part of $I$ we have the following refinement of
503: Schwarz's inequality:%
504: \begin{align}
505: \left\Vert x\right\Vert \left\Vert y\right\Vert & \geq \left\vert
506: \left\langle x,y\right\rangle -\sum_{i\in F}\left\langle
507: x,e_{i}\right\rangle \left\langle e_{i},y\right\rangle \right\vert
508: +\sum_{i\in F}\left\vert \left\langle x,e_{i}\right\rangle \left\langle
509: e_{i},y\right\rangle \right\vert \label{2.3.1} \\
510: & \geq \left\vert \left\langle x,y\right\rangle -\sum_{i\in F}\left\langle
511: x,e_{i}\right\rangle \left\langle e_{i},y\right\rangle \right\vert
512: +\left\vert \sum_{i\in F}\left\langle x,e_{i}\right\rangle \left\langle
513: e_{i},y\right\rangle \right\vert \notag \\
514: & \geq \left\vert \left\langle x,y\right\rangle \right\vert , \notag
515: \end{align}%
516: where $x,y\in H.$
517: \end{theorem}
518:
519: \begin{proof}
520: We follow the proof in \cite{SSD3}.
521:
522: We apply the Schwarz inequality to obtain%
523: \begin{multline}
524: \left\vert \left\langle x-\sum_{i\in F}\left\langle x,e_{i}\right\rangle
525: e_{i},\ \ y-\sum_{i\in F}\left\langle y,e_{i}\right\rangle
526: e_{i}\right\rangle \right\vert ^{2} \label{2.3.2} \\
527: \leq \left\Vert x-\sum_{i\in F}\left\langle x,e_{i}\right\rangle
528: e_{i}\right\Vert ^{2}\left\Vert y-\sum_{i\in F}\left\langle
529: y,e_{i}\right\rangle e_{i}\right\Vert ^{2}.
530: \end{multline}%
531: Since a simple calculation with orthonormal vectors shows that%
532: \begin{align*}
533: \left\Vert x-\sum_{i\in F}\left\langle x,e_{i}\right\rangle e_{i}\right\Vert
534: ^{2}& =\left\Vert x\right\Vert ^{2}-\sum_{i\in F}\left\vert \left\langle
535: x,e_{i}\right\rangle \right\vert ^{2}, \\
536: \left\Vert y-\sum_{i\in F}\left\langle y,e_{i}\right\rangle e_{i}\right\Vert
537: ^{2}& =\left\Vert y\right\Vert ^{2}-\sum_{i\in F}\left\vert \left\langle
538: y,e_{i}\right\rangle \right\vert ^{2},
539: \end{align*}%
540: and%
541: \begin{equation*}
542: \left\langle x-\sum_{i\in F}\left\langle x,e_{i}\right\rangle e_{i},\ \
543: y-\sum_{i\in F}\left\langle y,e_{i}\right\rangle e_{i}\right\rangle
544: =\left\langle x,y\right\rangle -\sum_{i\in F}\left\langle
545: x,e_{i}\right\rangle \left\langle e_{i},y\right\rangle ,
546: \end{equation*}%
547: hence (\ref{2.3.2}) is equivalent to%
548: \begin{multline}
549: \left\vert \left\langle x,y\right\rangle -\sum_{i\in F}\left\langle
550: x,e_{i}\right\rangle \left\langle e_{i},y\right\rangle \right\vert ^{2}
551: \label{2.3.3} \\
552: \leq \left( \left\Vert x\right\Vert ^{2}-\sum_{i\in F}\left\vert
553: \left\langle x,e_{i}\right\rangle \right\vert ^{2}\right) \left( \left\Vert
554: y\right\Vert ^{2}-\sum_{i\in F}\left\vert \left\langle y,e_{i}\right\rangle
555: \right\vert ^{2}\right)
556: \end{multline}%
557: for any $x,y\in H.$
558:
559: Further, we need the following Acz\'{e}l type inequality%
560: \begin{equation}
561: \left( \alpha ^{2}-\sum_{i\in F}\alpha _{i}^{2}\right) \left( \beta
562: ^{2}-\sum_{i\in F}\beta _{i}^{2}\right) \leq \left( \alpha \beta -\sum_{i\in
563: F}\alpha _{i}\beta _{i}\right) ^{2}, \label{2.3.4}
564: \end{equation}%
565: provided that $\alpha ^{2}\geq \sum_{i\in F}\alpha _{i}^{2}$ and $\beta
566: ^{2}\geq \sum_{i\in F}\beta _{i}^{2}$, where $\alpha ,\beta ,\alpha
567: _{i},\beta _{i}\in \mathbb{R}$, $i\in F.$
568:
569: For an Acz\'{e}l inequality that holds under slightly weaker conditions and
570: a different proof based on polynomials, see \cite[p. 57]{MIT}.
571:
572: For the sake of completeness, we give here a direct proof of (\ref{2.3.4}).
573:
574: Utilising the elementary inequality (\ref{2.1.10}), we can write%
575: \begin{multline}
576: \left( \alpha ^{2}-\left[ \left( \sum_{i\in F}\alpha _{i}^{2}\right) ^{\frac{%
577: 1}{2}}\right] ^{2}\right) \left( \beta ^{2}-\left[ \left( \sum_{i\in F}\beta
578: _{i}^{2}\right) ^{\frac{1}{2}}\right] ^{2}\right) \label{2.3.5} \\
579: \leq \left[ \left\vert \alpha \beta \right\vert -\left( \sum_{i\in F}\alpha
580: _{i}^{2}\right) ^{\frac{1}{2}}\left( \sum_{i\in F}\beta _{i}^{2}\right) ^{%
581: \frac{1}{2}}\right] ^{2}.
582: \end{multline}%
583: Since $\left\vert \alpha \right\vert \geq \left( \sum_{i\in F}\alpha
584: _{i}^{2}\right) ^{\frac{1}{2}}$ and $\left\vert \beta \right\vert \geq
585: \left( \sum_{i\in F}\beta _{i}^{2}\right) ^{\frac{1}{2}},$ then%
586: \begin{equation*}
587: \left\vert \alpha \beta \right\vert \geq \left( \sum_{i\in F}\alpha
588: _{i}^{2}\right) ^{\frac{1}{2}}\left( \sum_{i\in F}\beta _{i}^{2}\right) ^{%
589: \frac{1}{2}}.
590: \end{equation*}%
591: Therefore, by the Cauchy-Bunyakovsky-Schwarz inequality, we have that%
592: \begin{align*}
593: \left\vert \left\vert \alpha \beta \right\vert -\left( \sum_{i\in F}\alpha
594: _{i}^{2}\right) ^{\frac{1}{2}}\left( \sum_{i\in F}\beta _{i}^{2}\right) ^{%
595: \frac{1}{2}}\right\vert & =\left\vert \alpha \beta \right\vert -\left(
596: \sum_{i\in F}\alpha _{i}^{2}\right) ^{\frac{1}{2}}\left( \sum_{i\in F}\beta
597: _{i}^{2}\right) ^{\frac{1}{2}} \\
598: & \leq \left\vert \alpha \beta \right\vert -\left\vert \sum_{i\in F}\alpha
599: _{i}\beta _{i}\right\vert \\
600: & =\left\vert \left\vert \alpha \beta \right\vert -\left\vert \sum_{i\in
601: F}\alpha _{i}\beta _{i}\right\vert \right\vert \\
602: & \leq \left\vert \alpha \beta -\sum_{i\in F}\alpha _{i}\beta
603: _{i}\right\vert ,
604: \end{align*}%
605: showing that%
606: \begin{equation}
607: \left[ \left\vert \alpha \beta \right\vert -\left( \sum_{i\in F}\alpha
608: _{i}^{2}\right) ^{\frac{1}{2}}\left( \sum_{i\in F}\beta _{i}^{2}\right) ^{%
609: \frac{1}{2}}\right] ^{2}\leq \left( \alpha \beta -\sum_{i\in F}\alpha
610: _{i}\beta _{i}\right) ^{2} \label{2.3.6}
611: \end{equation}%
612: and then, by (\ref{2.3.5}) and (\ref{2.3.6}) we deduce the desired result (%
613: \ref{2.3.4}).
614:
615: By Bessel's inequality we obviously have that%
616: \begin{equation*}
617: \left\Vert x\right\Vert ^{2}\geq \sum_{i\in F}\left\vert \left\langle
618: x,e_{i}\right\rangle \right\vert ^{2}\qquad \text{and\qquad }\left\Vert
619: y\right\Vert ^{2}\geq \sum_{i\in F}\left\vert \left\langle
620: y,e_{i}\right\rangle \right\vert ^{2},
621: \end{equation*}%
622: therefore, on applying the inequality (\ref{2.3.4}) we deduce that%
623: \begin{multline}
624: \left( \left\Vert x\right\Vert ^{2}-\sum_{i\in F}\left\vert \left\langle
625: x,e_{i}\right\rangle \right\vert ^{2}\right) \left( \left\Vert y\right\Vert
626: ^{2}-\sum_{i\in F}\left\vert \left\langle y,e_{i}\right\rangle \right\vert
627: ^{2}\right) \label{2.3.7} \\
628: \leq \left( \left\Vert x\right\Vert \left\Vert y\right\Vert -\sum_{i\in
629: F}\left\vert \left\langle x,e_{i}\right\rangle \left\langle
630: e_{i},y\right\rangle \right\vert \right) ^{2}.
631: \end{multline}%
632: Since $\left\Vert x\right\Vert \left\Vert y\right\Vert -\sum_{i\in
633: F}\left\vert \left\langle x,e_{i}\right\rangle \left\langle
634: e_{i},y\right\rangle \right\vert \geq 0,$ hence by (\ref{2.3.3}) and (\ref%
635: {2.3.7}) we deduce the first part of (\ref{2.3.1}).
636:
637: The second and third parts are obvious.
638: \end{proof}
639:
640: When the vectors are orthogonal, the following result may be stated \cite%
641: {SSD4} (see also \cite[Corollary 3.1]{DS}).
642:
643: \begin{corollary}
644: \label{c2.3.3}If $\left\{ e_{i}\right\} _{i\in I}$ is an orthonormal family
645: in $\left( H,\left\langle \cdot ,\cdot \right\rangle \right) $ and $x,y\in H$
646: with $x\perp y,$ then we have the inequality:%
647: \begin{align}
648: \left\Vert x\right\Vert \left\Vert y\right\Vert & \geq \left\vert \sum_{i\in
649: F}\left\langle x,e_{i}\right\rangle \left\langle e_{i},y\right\rangle
650: \right\vert +\sum_{i\in F}\left\vert \left\langle x,e_{i}\right\rangle
651: \left\langle e_{i},y\right\rangle \right\vert \label{2.3.8} \\
652: & \geq 2\left\vert \sum_{i\in F}\left\langle x,e_{i}\right\rangle
653: \left\langle e_{i},y\right\rangle \right\vert , \notag
654: \end{align}%
655: for any nonempty finite part of $I.$
656: \end{corollary}
657:
658: \section{Kurepa Type Refinements for the Schwarz Inequality}
659:
660: \subsection{Kurepa's Inequality}
661:
662: In 1960, N.G. de Bruijn proved the following refinement of the celebrated
663: Cauchy-Bunyakovsky-Schwarz (CBS) inequality for a sequence of real numbers
664: and the second of complex numbers, see \cite{BR} or \cite[p. 48]{SSD1}:
665:
666: \begin{theorem}[de Bruijn, 1960]
667: \label{t3.1.1}Let $\left( a_{1},\dots ,a_{n}\right) $ be an $n-$tuple of
668: real numbers and $\left( z_{1},\dots ,z_{n}\right) $ an $n-$tuple of complex
669: numbers. Then%
670: \begin{align}
671: \left\vert \sum_{k=1}^{n}a_{k}z_{k}\right\vert ^{2}& \leq \frac{1}{2}%
672: \sum_{k=1}^{n}a_{k}^{2}\left[ \sum_{k=1}^{n}\left\vert z_{k}\right\vert
673: ^{2}+\left\vert \sum_{k=1}^{n}z_{k}^{2}\right\vert \right] \label{3.1.1} \\
674: & \left( \leq \sum_{k=1}^{n}a_{k}^{2}\cdot \sum_{k=1}^{n}\left\vert
675: z_{k}\right\vert ^{2}\right) . \notag
676: \end{align}%
677: Equality holds in (\ref{3.1.1}) if and only if, for $k\in \left\{ 1,\dots
678: ,n\right\} ,$ $a_{k}=\func{Re}\left( \lambda z_{k}\right) ,$ where $\lambda $
679: is a complex number such that $\lambda ^{2}\sum_{k=1}^{n}z_{n}^{2}$ is a
680: nonnegative real number.
681: \end{theorem}
682:
683: In 1966, in an effort to extend this result to inner products, Kurepa \cite%
684: {KU} obtained the following refinement for the complexification of a real
685: inner product space $\left( H;\left\langle \cdot ,\cdot \right\rangle
686: \right) :$
687:
688: \begin{theorem}[Kurepa, 1966]
689: \label{t3.1.2}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle \right) $
690: be a real inner product space and $\left( H_{\mathbb{C}},\left\langle \cdot
691: ,\cdot \right\rangle _{\mathbb{C}}\right) $ its complexification. For any $%
692: a\in H$ and $z\in H_{\mathbb{C}}$ we have the inequality:%
693: \begin{align}
694: \left\vert \left\langle z,a\right\rangle _{\mathbb{C}}\right\vert ^{2}& \leq
695: \frac{1}{2}\left\Vert a\right\Vert ^{2}\left[ \left\Vert z\right\Vert _{%
696: \mathbb{C}}^{2}+\left\vert \left\langle z,\bar{z}\right\rangle _{\mathbb{C}%
697: }\right\vert \right] \label{3.1.2} \\
698: & \left( \leq \left\Vert a\right\Vert ^{2}\left\Vert z\right\Vert _{\mathbb{C%
699: }}^{2}\right) . \notag
700: \end{align}
701: \end{theorem}
702:
703: To be comprehensive, we define in the following the concept of
704: complexification for a real inner product space.
705:
706: Let $H$ be a real inner product space with the scalar product $\left\langle
707: \cdot ,\cdot \right\rangle $ and the norm $\left\Vert \cdot \right\Vert .$
708: The \textit{complexification }$H_{\mathbb{C}}$ of $H$ is defined as a
709: complex linear space $H\times H$ of all ordered pairs $\left( x,y\right) $ $%
710: \left( x,y\in H\right) $ endowed with the operations%
711: \begin{align*}
712: \left( x,y\right) +\left( x^{\prime },y^{\prime }\right) & :=\left(
713: x+x^{\prime },y+y^{\prime }\right) ,\qquad x,x^{\prime },y,y^{\prime }\in H;
714: \\
715: \left( \sigma +i\tau \right) \cdot \left( x,y\right) & :=\left( \sigma
716: x-\tau y,\tau x+\sigma y\right) ,\qquad x,y\in H\text{ \ and \ }\sigma ,\tau
717: \in \mathbb{R}.
718: \end{align*}%
719: On $H_{\mathbb{C}}$ one can canonically consider the \textit{scalar product}
720: $\left\langle \cdot ,\cdot \right\rangle _{\mathbb{C}}$ defined by:%
721: \begin{equation*}
722: \left\langle z,z^{\prime }\right\rangle _{\mathbb{C}}:=\left\langle
723: x,x^{\prime }\right\rangle +\left\langle y,y^{\prime }\right\rangle +i\left[
724: \left\langle y,x^{\prime }\right\rangle -\left\langle x,y^{\prime
725: }\right\rangle \right]
726: \end{equation*}%
727: where $z=\left( x,y\right) ,$ $z^{\prime }=\left( x^{\prime },y^{\prime
728: }\right) \in H_{\mathbb{C}}.$ Obviously,%
729: \begin{equation*}
730: \left\Vert z\right\Vert _{\mathbb{C}}^{2}=\left\Vert x\right\Vert
731: ^{2}+\left\Vert y\right\Vert ^{2},
732: \end{equation*}%
733: where $z=\left( x,y\right) .$
734:
735: The conjugate of a vector $z=\left( x,y\right) \in H_{\mathbb{C}}$ is
736: defined by $\bar{z}:=\left( x,-y\right) .$
737:
738: It is easy to see that the elements of $H_{\mathbb{C}}$ under defined
739: operations behave as formal \textquotedblleft complex\textquotedblright\
740: combinations $x+iy$ with $x,y\in H.$ Because of this, we may write $z=x+iy$
741: instead of $z=\left( x,y\right) .$ Thus, $\bar{z}=x-iy.$
742:
743: \subsection{A Generalisation of Kurepa's Inequality}
744:
745: The following lemma is of interest \cite{SSD2}.
746:
747: \begin{lemma}
748: \label{l3.1.1}Let $f:\left[ 0,2\pi \right] \rightarrow \mathbb{R}$ given by%
749: \begin{equation}
750: f\left( \alpha \right) =\lambda \sin ^{2}\alpha +2\beta \sin \alpha \cos
751: \alpha +\alpha \cos ^{2}\alpha , \label{3.1.3}
752: \end{equation}%
753: where $\lambda ,\beta ,\gamma \in \mathbb{R}$. Then%
754: \begin{equation}
755: \sup_{\alpha \in \left[ 0,2\pi \right] }f\left( \alpha \right) =\frac{1}{2}%
756: \left( \lambda +\gamma \right) +\frac{1}{2}\left[ \left( \gamma -\lambda
757: \right) ^{2}+4\beta ^{2}\right] ^{\frac{1}{2}}. \label{3.1.4}
758: \end{equation}
759: \end{lemma}
760:
761: \begin{proof}
762: Since%
763: \begin{equation*}
764: \sin ^{2}\alpha =\frac{1-\cos 2\alpha }{2},\quad \cos ^{2}\alpha =\frac{%
765: 1+\cos 2\alpha }{2},\quad 2\sin \alpha \cos \alpha =\sin 2\alpha ,
766: \end{equation*}%
767: hence $f$ may be written as%
768: \begin{equation}
769: f\left( \alpha \right) =\frac{1}{2}\left( \lambda +\gamma \right) +\frac{1}{2%
770: }\left( \gamma -\lambda \right) \cos 2\alpha +\beta \sin 2\alpha .
771: \label{3.1.5}
772: \end{equation}%
773: If $\beta =0,$ then (\ref{3.1.5}) becomes%
774: \begin{equation*}
775: f\left( \alpha \right) =\frac{1}{2}\left( \lambda +\gamma \right) +\frac{1}{2%
776: }\left( \gamma -\lambda \right) \cos 2\alpha .
777: \end{equation*}%
778: Obviously, in this case%
779: \begin{equation*}
780: \sup_{\alpha \in \left[ 0,2\pi \right] }f\left( \alpha \right) =\frac{1}{2}%
781: \left( \lambda +\gamma \right) +\frac{1}{2}\left\vert \gamma -\lambda
782: \right\vert =\max \left\{ \gamma ,\lambda \right\} .
783: \end{equation*}%
784: If $\beta \neq 0,$ then (\ref{3.1.5}) becomes%
785: \begin{equation*}
786: f\left( \alpha \right) =\frac{1}{2}\left( \lambda +\gamma \right) +\beta %
787: \left[ \sin 2\alpha +\frac{\left( \gamma -\lambda \right) }{\beta }\cos
788: 2\alpha \right] .
789: \end{equation*}%
790: Let $\varphi \in \left( -\frac{\pi }{2},\frac{\pi }{2}\right) $ for which $%
791: \tan \varphi =\frac{\gamma -\lambda }{2\beta }.$ Then $f$ can be written as%
792: \begin{equation*}
793: f\left( \alpha \right) =\frac{1}{2}\left( \lambda +\gamma \right) +\frac{%
794: \beta }{\cos \varphi }\sin \left( 2\alpha +\varphi \right) .
795: \end{equation*}%
796: For this function, obviously%
797: \begin{equation}
798: \sup_{\alpha \in \left[ 0,2\pi \right] }f\left( \alpha \right) =\frac{1}{2}%
799: \left( \lambda +\gamma \right) +\frac{\left\vert \beta \right\vert }{%
800: \left\vert \cos \varphi \right\vert }. \label{3.1.6}
801: \end{equation}%
802: Since%
803: \begin{equation*}
804: \frac{\sin ^{2}\varphi }{\cos ^{2}\varphi }=\frac{\left( \gamma -\lambda
805: \right) ^{2}}{4\beta ^{2}},
806: \end{equation*}%
807: hence,%
808: \begin{equation*}
809: \frac{1}{\left\vert \cos \varphi \right\vert }=\frac{\left[ \left( \gamma
810: -\lambda \right) ^{2}+4\beta ^{2}\right] ^{\frac{1}{2}}}{2\left\vert \beta
811: \right\vert },
812: \end{equation*}%
813: and from (\ref{3.1.6}) we deduce the desired result (\ref{3.1.4}).
814: \end{proof}
815:
816: The following result holds \cite{SSD2}.
817:
818: \begin{theorem}[Dragomir, 2004]
819: \label{t3.1.3}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle \right) $
820: be a complex inner product space. If $x,y,z\in H$ are such that%
821: \begin{equation}
822: \func{Im}\left\langle x,z\right\rangle =\func{Im}\left\langle
823: y,z\right\rangle =0, \label{3.1.7}
824: \end{equation}%
825: then we have the inequality:%
826: \begin{align}
827: & \func{Re}^{2}\left\langle x,z\right\rangle +\func{Re}^{2}\left\langle
828: y,z\right\rangle \label{3.1.8} \\
829: & =\left\vert \left\langle x+iy,z\right\rangle \right\vert ^{2} \notag \\
830: & \leq \frac{1}{2}\left\{ \left\Vert x\right\Vert ^{2}+\left\Vert
831: y\right\Vert ^{2}+\left[ \left( \left\Vert x\right\Vert ^{2}-\left\Vert
832: y\right\Vert ^{2}\right) ^{2}-4\func{Re}^{2}\left\langle x,y\right\rangle %
833: \right] ^{\frac{1}{2}}\right\} \left\Vert z\right\Vert ^{2} \notag \\
834: & \leq \left( \left\Vert x\right\Vert ^{2}+\left\Vert y\right\Vert
835: ^{2}\right) \left\Vert z\right\Vert ^{2}. \notag
836: \end{align}
837: \end{theorem}
838:
839: \begin{proof}
840: Obviously, by (\ref{3.1.7}), we have%
841: \begin{equation*}
842: \left\langle x+iy,z\right\rangle =\func{Re}\left\langle x,z\right\rangle +i%
843: \func{Re}\left\langle y,z\right\rangle
844: \end{equation*}%
845: and the first part of (\ref{3.1.8}) holds true.
846:
847: Now, let $\varphi \in \left[ 0,2\pi \right] $ be such that $\left\langle
848: x+iy,z\right\rangle =e^{i\varphi }\left\vert \left\langle
849: x+iy,z\right\rangle \right\vert .$ Then%
850: \begin{equation*}
851: \left\vert \left\langle x+iy,z\right\rangle \right\vert =e^{-i\varphi
852: }\left\langle x+iy,z\right\rangle =\left\langle e^{-i\varphi }\left(
853: x+iy\right) ,z\right\rangle .
854: \end{equation*}%
855: Utilising the above identity, we can write:%
856: \begin{align*}
857: \left\vert \left\langle x+iy,z\right\rangle \right\vert & =\func{Re}%
858: \left\langle e^{-i\varphi }\left( x+iy\right) ,z\right\rangle \\
859: & =\func{Re}\left\langle \left( \cos \varphi -i\sin \varphi \right) \left(
860: x+iy\right) ,z\right\rangle \\
861: & =\func{Re}\left\langle \cos \varphi \cdot x+\sin \varphi \cdot y-i\sin
862: \varphi \cdot x+i\cos \varphi \cdot y,z\right\rangle \\
863: & =\func{Re}\left\langle \cos \varphi \cdot x+\sin \varphi \cdot
864: y,z\right\rangle +\func{Im}\left\langle \sin \varphi \cdot x-\cos \varphi
865: \cdot y,z\right\rangle \\
866: & =\func{Re}\left\langle \cos \varphi \cdot x+\sin \varphi \cdot
867: y,z\right\rangle +\sin \varphi \func{Im}\left\langle x,z\right\rangle -\cos
868: \varphi \func{Im}\left\langle y,z\right\rangle \\
869: & =\func{Re}\left\langle \cos \varphi \cdot x+\sin \varphi \cdot
870: y,z\right\rangle ,
871: \end{align*}%
872: and for the last equality we have used the assumption (\ref{3.1.7}).
873:
874: Taking the square and using the Schwarz inequality for the inner product $%
875: \left\langle \cdot ,\cdot \right\rangle ,$ we have%
876: \begin{align}
877: \left\vert \left\langle x+iy,z\right\rangle \right\vert ^{2}& =\left[ \func{%
878: Re}\left\langle \cos \varphi \cdot x+\sin \varphi \cdot y,z\right\rangle %
879: \right] ^{2} \label{3.1.9} \\
880: & \leq \left\Vert \cos \varphi \cdot x+\sin \varphi \cdot y\right\Vert
881: ^{2}\left\Vert z\right\Vert ^{2}. \notag
882: \end{align}%
883: On making use of Lemma \ref{l3.1.1}, we have%
884: \begin{align*}
885: & \sup_{\alpha \in \left[ 0,2\pi \right] }\left\Vert \cos \varphi \cdot
886: x+\sin \varphi \cdot y\right\Vert ^{2} \\
887: & =\sup_{\alpha \in \left[ 0,2\pi \right] }\left[ \left\Vert x\right\Vert
888: ^{2}\cos ^{2}\varphi +2\func{Re}\left\langle x,y\right\rangle \sin \varphi
889: \cos \varphi +\left\Vert y\right\Vert ^{2}\sin {}^{2}\varphi \right] \\
890: & =\frac{1}{2}\left\{ \left\Vert x\right\Vert ^{2}+\left\Vert y\right\Vert
891: ^{2}+\left[ \left( \left\Vert x\right\Vert ^{2}-\left\Vert y\right\Vert
892: ^{2}\right) ^{2}+4\func{Re}^{2}\left\langle x,y\right\rangle \right] ^{\frac{%
893: 1}{2}}\right\}
894: \end{align*}%
895: and the first inequality in (\ref{3.1.8}) is proved.
896:
897: Observe that%
898: \begin{align*}
899: & \left( \left\Vert x\right\Vert ^{2}-\left\Vert y\right\Vert ^{2}\right)
900: ^{2}+4\func{Re}^{2}\left\langle x,y\right\rangle \\
901: & =\left( \left\Vert x\right\Vert ^{2}+\left\Vert y\right\Vert ^{2}\right)
902: ^{2}-4\left[ \left\Vert x\right\Vert ^{2}\left\Vert y\right\Vert ^{2}-\func{%
903: Re}^{2}\left\langle x,y\right\rangle \right] \\
904: & \leq \left( \left\Vert x\right\Vert ^{2}+\left\Vert y\right\Vert
905: ^{2}\right) ^{2}
906: \end{align*}%
907: and the last part of (\ref{3.1.8}) is proved.
908: \end{proof}
909:
910: \begin{remark}
911: \label{r3.1.4}Observe that if $\left( H,\left\langle \cdot ,\cdot
912: \right\rangle \right) $ is a real inner product space, then for any $%
913: x,y,z\in H$ one has:%
914: \begin{align}
915: & \left\langle x,z\right\rangle ^{2}+\left\langle y,z\right\rangle ^{2}
916: \label{3.1.10} \\
917: & \leq \frac{1}{2}\left\{ \left\Vert x\right\Vert ^{2}+\left\Vert
918: y\right\Vert ^{2}+\left[ \left( \left\Vert x\right\Vert ^{2}-\left\Vert
919: y\right\Vert ^{2}\right) ^{2}+4\left\langle x,y\right\rangle ^{2}\right]
920: \right\} ^{\frac{1}{2}}\left\Vert z\right\Vert ^{2} \notag \\
921: & \leq \left( \left\Vert x\right\Vert ^{2}+\left\Vert y\right\Vert
922: ^{2}\right) \left\Vert z\right\Vert ^{2}. \notag
923: \end{align}
924: \end{remark}
925:
926: \begin{remark}
927: \label{r3.1.5}If $H$ is a real space, $\left\langle \cdot ,\cdot
928: \right\rangle $ the real inner product, $H_{\mathbb{C}}$ its
929: complexification and $\left\langle \cdot ,\cdot \right\rangle _{\mathbb{C}}$
930: the corresponding complexification for $\left\langle \cdot ,\cdot
931: \right\rangle $, then for $x,y\in H$ and $w:=x+iy\in H_{\mathbb{C}}$ and for
932: $e\in H$ we have%
933: \begin{equation*}
934: \func{Im}\left\langle x,e\right\rangle _{\mathbb{C}}=\func{Im}\left\langle
935: y,e\right\rangle _{\mathbb{C}}=0,
936: \end{equation*}%
937: \begin{equation*}
938: \left\Vert w\right\Vert _{\mathbb{C}}^{2}=\left\Vert x\right\Vert
939: ^{2}+\left\Vert y\right\Vert ^{2},\qquad \left\vert \left\langle w,\bar{w}%
940: \right\rangle _{\mathbb{C}}\right\vert =\left( \left\Vert x\right\Vert
941: ^{2}-\left\Vert y\right\Vert ^{2}\right) ^{2}+4\left\langle x,y\right\rangle
942: ^{2},
943: \end{equation*}%
944: where $\bar{w}=x-iy\in H_{\mathbb{C}}.$
945:
946: Applying Theorem \ref{t3.1.3} for the complex space $H_{\mathbb{C}}$ and
947: complex inner product $\left\langle \cdot ,\cdot \right\rangle _{\mathbb{C}%
948: }, $ we deduce%
949: \begin{equation}
950: \left\vert \left\langle w,e\right\rangle _{\mathbb{C}}\right\vert ^{2}\leq
951: \frac{1}{2}\left\Vert e\right\Vert ^{2}\left[ \left\Vert w\right\Vert _{%
952: \mathbb{C}}^{2}+\left\vert \left\langle w,\bar{w}\right\rangle _{\mathbb{C}%
953: }\right\vert \right] \leq \left\Vert e\right\Vert ^{2}\left\Vert
954: w\right\Vert _{\mathbb{C}}^{2}, \label{3.1.11}
955: \end{equation}%
956: which is Kurepa's inequality (\ref{3.1.2}).
957: \end{remark}
958:
959: \begin{corollary}
960: \label{c3.1.1}Let $x,y,z$ be as in Theorem \ref{t3.1.3}. In addition, if $%
961: \func{Re}\left\langle x,y\right\rangle =0,$ then%
962: \begin{equation}
963: \left[ \func{Re}^{2}\left\langle x,z\right\rangle +\func{Re}^{2}\left\langle
964: y,z\right\rangle \right] ^{\frac{1}{2}}\leq \left\Vert z\right\Vert \cdot
965: \max \left\{ \left\Vert x\right\Vert ,\left\Vert y\right\Vert \right\} .
966: \label{3.1.12}
967: \end{equation}
968: \end{corollary}
969:
970: \begin{remark}
971: \label{r3.1.6}If $H$ is a real space and $\left\langle \cdot ,\cdot
972: \right\rangle $ a real inner product on $H,$ then for any $x,y,z\in H$ with $%
973: \left\langle x,y\right\rangle =0$ we have%
974: \begin{equation}
975: \left[ \left\langle x,z\right\rangle ^{2}+\left\langle y,z\right\rangle ^{2}%
976: \right] ^{\frac{1}{2}}\leq \left\Vert z\right\Vert \cdot \max \left\{
977: \left\Vert x\right\Vert ,\left\Vert y\right\Vert \right\} . \label{3.1.13}
978: \end{equation}
979: \end{remark}
980:
981: \subsection{A Related Result}
982:
983: Utilising Lemma \ref{l3.1.1}, we may state and prove the following result as
984: well.
985:
986: \begin{theorem}[Dragomir, 2004]
987: \label{t3.1.4}Let $\left( H,\left\langle \cdot ,\cdot \right\rangle \right) $
988: \ be a real or complex inner product space. Then we have the inequalities:%
989: \begin{align}
990: & \frac{1}{2}\bigg\{\left\vert \left\langle v,t\right\rangle \right\vert
991: ^{2}+\left\vert \left\langle w,t\right\rangle \right\vert ^{2}+\left[ \left(
992: \left\vert \left\langle v,t\right\rangle \right\vert ^{2}-\left\vert
993: \left\langle w,t\right\rangle \right\vert ^{2}\right) ^{2}\right.
994: \label{3.1.14} \\
995: & \quad \left. +\left. 4\left( \func{Re}\left\langle v,t\right\rangle \func{%
996: Re}\left\langle w,t\right\rangle +\func{Im}\left\langle v,t\right\rangle
997: \func{Im}\left\langle w,t\right\rangle \right) ^{2}\right] ^{\frac{1}{2}%
998: }\right\} \notag \\
999: & \leq \frac{1}{2}\left\Vert t\right\Vert ^{2}\left\{ \left\Vert
1000: v\right\Vert ^{2}+\left\Vert w\right\Vert ^{2}+\left[ \left( \left\Vert
1001: v\right\Vert ^{2}-\left\Vert w\right\Vert ^{2}\right) ^{2}+4\func{Re}%
1002: ^{2}\left( v,w\right) \right] ^{\frac{1}{2}}\right\} \notag \\
1003: & \leq \left( \left\Vert v\right\Vert ^{2}+\left\Vert w\right\Vert
1004: ^{2}\right) \left\Vert t\right\Vert ^{2}, \notag
1005: \end{align}%
1006: for all $v,w,t\in H.$
1007: \end{theorem}
1008:
1009: \begin{proof}
1010: Observe that, by Schwarz's inequality%
1011: \begin{equation}
1012: \left\vert \left( \cos \varphi \cdot v+\sin \varphi \cdot w,z\right)
1013: \right\vert ^{2}\leq \left\Vert \cos \varphi \cdot v+\sin \varphi \cdot
1014: w\right\Vert ^{2}\left\Vert z\right\Vert ^{2} \label{3.1.15}
1015: \end{equation}%
1016: for any $\varphi \in \left[ 0,2\pi \right] .$
1017:
1018: Since%
1019: \begin{align*}
1020: I\left( \varphi \right) & :=\left\Vert \cos \varphi \cdot v+\sin \varphi
1021: \cdot w\right\Vert ^{2} \\
1022: & =\cos ^{2}\varphi \left\Vert v\right\Vert ^{2}+2\func{Re}\left( v,w\right)
1023: \sin \varphi \cos \varphi +\left\Vert w\right\Vert ^{2}\sin ^{2}\varphi ,
1024: \end{align*}%
1025: hence, as in Theorem \ref{t3.1.3},%
1026: \begin{equation*}
1027: \sup_{\varphi \in \left[ 0,2\pi \right] }I\left( \varphi \right) =\frac{1}{2}%
1028: \left\{ \left\Vert v\right\Vert ^{2}+\left\Vert w\right\Vert ^{2}+\left[
1029: \left( \left\Vert v\right\Vert ^{2}-\left\Vert w\right\Vert ^{2}\right)
1030: ^{2}+4\func{Re}^{2}\left( v,w\right) \right] ^{\frac{1}{2}}\right\} .
1031: \end{equation*}%
1032: Also, denoting%
1033: \begin{align*}
1034: J\left( \varphi \right) & :=\left\vert \cos \varphi \left\langle
1035: v,z\right\rangle +\sin \varphi \left\langle w,z\right\rangle \right\vert \\
1036: & =\cos ^{2}\varphi \left\vert \left\langle v,z\right\rangle \right\vert
1037: ^{2}+2\sin \varphi \cos \varphi \func{Re}\left[ \left\langle
1038: v,z\right\rangle \overline{\left\langle w,z\right\rangle }\right] +\sin
1039: ^{2}\varphi \left\vert \left\langle w,z\right\rangle \right\vert ^{2},
1040: \end{align*}%
1041: then, on applying Lemma \ref{l3.1.1}, we deduce that%
1042: \begin{multline*}
1043: \sup_{\varphi \in \left[ 0,2\pi \right] }J\left( \varphi \right) =\frac{1}{2}%
1044: \bigg\{\left\vert \left\langle v,t\right\rangle \right\vert ^{2}+\left\vert
1045: \left\langle w,t\right\rangle \right\vert ^{2} \\
1046: +\left. \left[ \left( \left\vert \left\langle v,t\right\rangle \right\vert
1047: ^{2}-\left\vert \left\langle w,t\right\rangle \right\vert ^{2}\right) ^{2}+4%
1048: \func{Re}^{2}\left[ \left\langle v,z\right\rangle \overline{\left\langle
1049: w,z\right\rangle }\right] \right] ^{\frac{1}{2}}\right\}
1050: \end{multline*}%
1051: and, since%
1052: \begin{equation*}
1053: \func{Re}\left[ \left\langle v,z\right\rangle \overline{\left\langle
1054: w,z\right\rangle }\right] =\func{Re}\left\langle v,t\right\rangle \func{Re}%
1055: \left\langle w,t\right\rangle +\func{Im}\left\langle v,t\right\rangle \func{%
1056: Im}\left\langle w,t\right\rangle ,
1057: \end{equation*}%
1058: hence, on taking the supremum in the inequality (\ref{3.1.15}), we deduce
1059: the desired inequality (\ref{3.1.14}).
1060: \end{proof}
1061:
1062: \begin{remark}
1063: \label{r3.1.7}In the real case, (\ref{3.1.14}) provides the same inequality
1064: we obtained in (\ref{3.1.10}).
1065:
1066: In the complex case, if we assume that \ $v,w,t\in H$ are such that%
1067: \begin{equation*}
1068: \func{Re}\left\langle v,t\right\rangle \func{Re}\left\langle
1069: w,t\right\rangle =-\func{Im}\left\langle v,t\right\rangle \func{Im}%
1070: \left\langle w,t\right\rangle ,
1071: \end{equation*}%
1072: then (\ref{3.1.14}) becomes:%
1073: \begin{multline}
1074: \max \left\{ \left\vert \left\langle v,t\right\rangle \right\vert
1075: ^{2},\left\vert \left\langle w,t\right\rangle \right\vert ^{2}\right\}
1076: \label{3.1.1.6} \\
1077: \leq \frac{1}{2}\left\Vert t\right\Vert ^{2}\left\{ \left\Vert v\right\Vert
1078: ^{2}+\left\Vert w\right\Vert ^{2}+\left[ \left( \left\Vert v\right\Vert
1079: ^{2}-\left\Vert w\right\Vert ^{2}\right) ^{2}+4\func{Re}^{2}\left(
1080: v,w\right) \right] ^{\frac{1}{2}}\right\} .
1081: \end{multline}
1082: \end{remark}
1083:
1084: \section{Refinements of Buzano's and Kurepa's Inequalities}
1085:
1086: \subsection{Introduction}
1087:
1088: In \cite{B}, M.L. Buzano obtained the following extension of the celebrated
1089: Schwarz's inequality in a real or complex inner product space $\left(
1090: H;\left\langle \cdot ,\cdot \right\rangle \right) :$%
1091: \begin{equation}
1092: \left\vert \left\langle a,x\right\rangle \left\langle x,b\right\rangle
1093: \right\vert \leq \frac{1}{2}\left[ \left\Vert a\right\Vert \cdot \left\Vert
1094: b\right\Vert +\left\vert \left\langle a,b\right\rangle \right\vert \right]
1095: \left\Vert x\right\Vert ^{2}, \label{ch3.1.1}
1096: \end{equation}%
1097: for any $a,b,x\in H.$
1098:
1099: It is clear that for $a=b,$ the above inequality becomes the standard
1100: Schwarz inequality%
1101: \begin{equation}
1102: \left\vert \left\langle a,x\right\rangle \right\vert ^{2}\leq \left\Vert
1103: a\right\Vert ^{2}\left\Vert x\right\Vert ^{2},\qquad a,x\in H;
1104: \label{ch3.1.2}
1105: \end{equation}%
1106: with equality if and only if there exists a scalar $\lambda \in \mathbb{K}$ $%
1107: \left( \mathbb{K}=\mathbb{R}\text{ \ or }\mathbb{C}\right) $ such that $%
1108: x=\lambda a.$
1109:
1110: As noted by M. Fujii and F. Kubo in \cite{FK}, where they provided a simple
1111: proof of (\ref{ch3.1.1}) by utilising orthogonal projection arguments, the
1112: case of equality holds in (\ref{ch3.1.1}) if%
1113: \begin{equation*}
1114: x=\left\{
1115: \begin{array}{ll}
1116: \alpha \left( \frac{a}{\left\Vert a\right\Vert }+\frac{\left\langle
1117: a,b\right\rangle }{\left\vert \left\langle a,b\right\rangle \right\vert }%
1118: \cdot \frac{b}{\left\Vert b\right\Vert }\right) , & \text{when \ }%
1119: \left\langle a,b\right\rangle \neq 0 \\
1120: & \\
1121: \alpha \left( \frac{a}{\left\Vert a\right\Vert }+\beta \cdot \frac{b}{%
1122: \left\Vert b\right\Vert }\right) , & \text{when\ }\left\langle
1123: a,b\right\rangle =0,%
1124: \end{array}%
1125: \right.
1126: \end{equation*}%
1127: where $\alpha ,\beta \in \mathbb{K}$.
1128:
1129: It might be useful to observe that, out of (\ref{ch3.1.1}), one may get the
1130: following discrete inequality:%
1131: \begin{multline}
1132: \left\vert \sum_{i=1}^{n}p_{i}a_{i}\overline{x_{i}}\sum_{i=1}^{n}p_{i}x_{i}%
1133: \overline{b_{i}}\right\vert \label{ch3.1.3} \\
1134: \leq \frac{1}{2}\left[ \left( \sum_{i=1}^{n}p_{i}\left\vert a_{i}\right\vert
1135: ^{2}\sum_{i=1}^{n}p_{i}\left\vert b_{i}\right\vert ^{2}\right) ^{\frac{1}{2}%
1136: }+\left\vert \sum_{i=1}^{n}p_{i}a_{i}\overline{b_{i}}\right\vert \right]
1137: \sum_{i=1}^{n}p_{i}\left\vert x_{i}\right\vert ^{2},
1138: \end{multline}%
1139: where $p_{i}\geq 0,$ $a_{i},x_{i},b_{i}\in \mathbb{C}$, $i\in \left\{
1140: 1,\dots ,n\right\} .$
1141:
1142: If one takes in (\ref{ch3.1.3}) $b_{i}=\overline{a_{i}}$ for $i\in \left\{
1143: 1,\dots ,n\right\} ,$ then one obtains%
1144: \begin{equation}
1145: \left\vert \sum_{i=1}^{n}p_{i}a_{i}\overline{x_{i}}%
1146: \sum_{i=1}^{n}p_{i}a_{i}x_{i}\right\vert \leq \frac{1}{2}\left[
1147: \sum_{i=1}^{n}p_{i}\left\vert a_{i}\right\vert ^{2}+\left\vert
1148: \sum_{i=1}^{n}p_{i}a_{i}^{2}\right\vert \right] \sum_{i=1}^{n}p_{i}\left%
1149: \vert x_{i}\right\vert ^{2}, \label{ch3.1.4}
1150: \end{equation}%
1151: for any $p_{i}\geq 0,$ $a_{i},x_{i},b_{i}\in \mathbb{C}$, $i\in \left\{
1152: 1,\dots ,n\right\} .$
1153:
1154: Note that, if $x_{i},$ $i\in \left\{ 1,\dots ,n\right\} $ are real numbers,
1155: then out of (\ref{ch3.1.4}), we may deduce the de Bruijn refinement of the
1156: celebrated Cauchy-Bunyakovsky-Schwarz inequality \cite{BR}%
1157: \begin{equation}
1158: \left\vert \sum_{i=1}^{n}p_{i}x_{i}z_{i}\right\vert ^{2}\leq \frac{1}{2}%
1159: \sum_{i=1}^{n}p_{i}x_{i}^{2}\left[ \sum_{i=1}^{n}p_{i}\left\vert
1160: z_{i}\right\vert ^{2}+\left\vert \sum_{i=1}^{n}p_{i}z_{i}^{2}\right\vert %
1161: \right] , \label{ch3.1.5}
1162: \end{equation}%
1163: where $z_{i}\in \mathbb{C}$, $i\in \left\{ 1,\dots ,n\right\} .$ In this
1164: way, Buzano's result may be regarded as a generalisation of de Bruijn's
1165: inequality.
1166:
1167: Similar comments obviously apply for integrals, but, for the sake of brevity
1168: we do not mention them here.
1169:
1170: The aim of the present section is to establish some related results as well
1171: as a refinement of Buzano's inequality for real or complex inner product
1172: spaces. An improvement of Kurepa's inequality for the complexification of a
1173: real inner product and the corresponding applications for discrete and
1174: integral inequalities are also provided.
1175:
1176: \subsection{Some Buzano Type Inequalities}
1177:
1178: The following result may be stated \cite{DRAG1}.
1179:
1180: \begin{theorem}[Dragomir, 2004]
1181: \label{ch3.t2.1}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
1182: \right) $ be an inner product space over the real or complex number field $%
1183: \mathbb{K}$. For all $\alpha \in \mathbb{K}\backslash \left\{ 0\right\} $
1184: and $x,a,b\in H,$ $\alpha \neq 0,$ one has the inequality%
1185: \begin{multline}
1186: \left\vert \frac{\left\langle a,x\right\rangle \left\langle x,b\right\rangle
1187: }{\left\Vert x\right\Vert ^{2}}-\frac{\left\langle a,b\right\rangle }{\alpha
1188: }\right\vert \label{ch3.2.1} \\
1189: \leq \frac{\left\Vert b\right\Vert }{\left\vert \alpha \right\vert
1190: \left\Vert x\right\Vert }\left[ \left\vert \alpha -1\right\vert
1191: ^{2}\left\vert \left\langle a,x\right\rangle \right\vert ^{2}+\left\Vert
1192: x\right\Vert ^{2}\left\Vert a\right\Vert ^{2}-\left\vert \left\langle
1193: a,x\right\rangle \right\vert ^{2}\right] .
1194: \end{multline}%
1195: The case of equality holds in (\ref{ch3.2.1}) if and only if there exists a
1196: scalar $\lambda \in \mathbb{K}$ so that%
1197: \begin{equation}
1198: \alpha \cdot \frac{\left\langle a,x\right\rangle }{\left\Vert x\right\Vert
1199: ^{2}}x=a+\lambda b. \label{ch3.2.1.a}
1200: \end{equation}
1201: \end{theorem}
1202:
1203: \begin{proof}
1204: We follow the proof in \cite{DRAG1}.
1205:
1206: Using Schwarz's inequality, we have that%
1207: \begin{equation}
1208: \left\vert \left\langle \alpha \cdot \frac{\left\langle a,x\right\rangle }{%
1209: \left\Vert x\right\Vert ^{2}}x-a,b\right\rangle \right\vert ^{2}\leq
1210: \left\Vert \alpha \cdot \frac{\left\langle a,x\right\rangle }{\left\Vert
1211: x\right\Vert ^{2}}x-a\right\Vert ^{2}\left\Vert b\right\Vert ^{2}
1212: \label{ch3.2.2}
1213: \end{equation}%
1214: and since%
1215: \begin{align*}
1216: \left\Vert \alpha \cdot \frac{\left\langle a,x\right\rangle }{\left\Vert
1217: x\right\Vert ^{2}}x-a\right\Vert ^{2}& =\left\vert \alpha \right\vert ^{2}%
1218: \frac{\left\vert \left\langle a,x\right\rangle \right\vert ^{2}}{\left\Vert
1219: x\right\Vert ^{2}}-2\frac{\left\vert \left\langle a,x\right\rangle
1220: \right\vert ^{2}}{\left\Vert x\right\Vert ^{2}}\func{Re}\alpha +\left\Vert
1221: a\right\Vert ^{2} \\
1222: & =\frac{\left\vert \alpha -1\right\vert ^{2}\left\vert \left\langle
1223: a,x\right\rangle \right\vert ^{2}+\left\Vert x\right\Vert ^{2}\left\Vert
1224: a\right\Vert ^{2}-\left\vert \left\langle a,x\right\rangle \right\vert ^{2}}{%
1225: \left\Vert x\right\Vert ^{2}}
1226: \end{align*}%
1227: and%
1228: \begin{equation*}
1229: \left\langle \alpha \cdot \frac{\left\langle a,x\right\rangle }{\left\Vert
1230: x\right\Vert ^{2}}x-a,b\right\rangle =\alpha \left[ \frac{\left\langle
1231: a,x\right\rangle \left\langle x,b\right\rangle }{\left\Vert x\right\Vert ^{2}%
1232: }-\frac{\left\langle a,b\right\rangle }{\alpha }\right] ,
1233: \end{equation*}%
1234: hence by (\ref{ch3.2.1}) we deduce the desired inequality (\ref{ch3.2.1}).
1235:
1236: The case of equality is obvious from the above considerations related to the
1237: Schwarz's inequality (\ref{ch3.1.2}).
1238: \end{proof}
1239:
1240: \begin{remark}
1241: \label{ch3.r2.2}Using the continuity property of the modulus, i.e., $%
1242: \left\vert \left\vert z\right\vert -\left\vert u\right\vert \right\vert \leq
1243: \left\vert z-u\right\vert ,$ $z,u\in \mathbb{K}$, we have:%
1244: \begin{equation}
1245: \left\vert \frac{\left\vert \left\langle a,x\right\rangle \left\langle
1246: x,b\right\rangle \right\vert }{\left\Vert x\right\Vert ^{2}}-\frac{%
1247: \left\vert \left\langle a,b\right\rangle \right\vert }{\left\vert \alpha
1248: \right\vert }\right\vert \leq \left\vert \frac{\left\langle a,x\right\rangle
1249: \left\langle x,b\right\rangle }{\left\Vert x\right\Vert ^{2}}-\frac{%
1250: \left\langle a,b\right\rangle }{\alpha }\right\vert . \label{ch3.2.3}
1251: \end{equation}%
1252: Therefore, by (\ref{ch3.2.1}) and (\ref{ch3.2.3}), one may deduce the
1253: following double inequality:%
1254: \begin{align}
1255: & \frac{1}{\left\vert \alpha \right\vert }\left[ \left\vert \left\langle
1256: a,b\right\rangle \right\vert -\frac{\left\Vert b\right\Vert }{\left\Vert
1257: x\right\Vert }\right. \label{ch3.2.4} \\
1258: & \qquad \times \left. \left[ \left( \left\vert \alpha -1\right\vert
1259: ^{2}\left\vert \left\langle x,a\right\rangle \right\vert ^{2}+\left\Vert
1260: x\right\Vert ^{2}\left\Vert a\right\Vert ^{2}-\left\vert \left\langle
1261: a,x\right\rangle \right\vert ^{2}\right) ^{\frac{1}{2}}\right] \right]
1262: \notag \\
1263: & \leq \frac{\left\vert \left\langle a,x\right\rangle \left\langle
1264: x,b\right\rangle \right\vert }{\left\Vert x\right\Vert ^{2}} \notag \\
1265: & \leq \frac{1}{\left\vert \alpha \right\vert }\left[ \left\vert
1266: \left\langle a,b\right\rangle \right\vert +\frac{\left\Vert b\right\Vert }{%
1267: \left\Vert x\right\Vert }\right] \notag \\
1268: & \qquad \times \left[ \left( \left\vert \alpha -1\right\vert ^{2}\left\vert
1269: \left\langle x,a\right\rangle \right\vert ^{2}+\left\Vert x\right\Vert
1270: ^{2}\left\Vert a\right\Vert ^{2}-\left\vert \left\langle x,a\right\rangle
1271: \right\vert ^{2}\right) ^{\frac{1}{2}}\right] , \notag
1272: \end{align}%
1273: for each $\alpha \in \mathbb{K}\backslash \left\{ 0\right\} ,$ $a,b,x\in H$
1274: and $x\neq 0.$
1275: \end{remark}
1276:
1277: It is obvious that, out of (\ref{ch3.2.1}), we can obtain various particular
1278: inequalities. We mention in the following a class of these which is related
1279: to Buzano's result (\ref{ch3.1.1}) \cite{DRAG1}.
1280:
1281: \begin{corollary}[Dragomir, 2004]
1282: \label{ch3.c2.3}Let $a,b,x\in H,$ $x\neq 0$ and $\eta \in \mathbb{K}$ with $%
1283: \left\vert \eta \right\vert =1,$ $\func{Re}\eta \neq -1.$ Then we have the
1284: inequality:%
1285: \begin{equation}
1286: \left\vert \frac{\left\langle a,x\right\rangle \left\langle x,b\right\rangle
1287: }{\left\Vert x\right\Vert ^{2}}-\frac{\left\langle a,b\right\rangle }{1+\eta
1288: }\right\vert \leq \frac{\left\Vert a\right\Vert \left\Vert b\right\Vert }{%
1289: \sqrt{2}\sqrt{1+\func{Re}\eta }}, \label{ch3.2.5}
1290: \end{equation}%
1291: and, in particular, for $\eta =1,$ the inequality:%
1292: \begin{equation}
1293: \left\vert \frac{\left\langle a,x\right\rangle \left\langle x,b\right\rangle
1294: }{\left\Vert x\right\Vert ^{2}}-\frac{\left\langle a,b\right\rangle }{2}%
1295: \right\vert \leq \frac{\left\Vert a\right\Vert \left\Vert b\right\Vert }{2}.
1296: \label{ch3.2.6}
1297: \end{equation}
1298: \end{corollary}
1299:
1300: \begin{proof}
1301: It follows by Theorem \ref{ch3.t2.1} on choosing $\alpha =1+\eta $ and we
1302: omit the details.
1303: \end{proof}
1304:
1305: \begin{remark}
1306: Using the continuity property of modulus, we get from (\ref{ch3.2.5}) that:%
1307: \begin{equation*}
1308: \frac{\left\vert \left\langle a,x\right\rangle \left\langle x,b\right\rangle
1309: \right\vert }{\left\Vert x\right\Vert ^{2}}\leq \frac{\left\vert
1310: \left\langle a,b\right\rangle \right\vert +\left\Vert a\right\Vert
1311: \left\Vert b\right\Vert }{\sqrt{2}\sqrt{1+\func{Re}\eta }},\qquad \left\vert
1312: \eta \right\vert =1,\ \ \func{Re}\eta \neq -1,
1313: \end{equation*}%
1314: which provides, as the best possible inequality, the above result due to
1315: Buzano (\ref{ch3.1.1}).
1316: \end{remark}
1317:
1318: \begin{remark}
1319: \label{ch3.r2.3}If the space is real, then the inequality (\ref{ch3.2.1}) is
1320: obviously equivalent to:%
1321: \begin{align}
1322: & \frac{\left\langle a,b\right\rangle }{\alpha }-\frac{\left\Vert
1323: b\right\Vert }{\left\vert \alpha \right\vert \left\Vert x\right\Vert }\left[
1324: \left( \alpha -1\right) ^{2}\left\langle a,x\right\rangle ^{2}+\left\Vert
1325: x\right\Vert ^{2}\left\Vert a\right\Vert ^{2}-\left\langle a,x\right\rangle
1326: ^{2}\right] ^{\frac{1}{2}} \label{ch3.2.7} \\
1327: & \leq \frac{\left\langle a,x\right\rangle \left\langle x,b\right\rangle }{%
1328: \left\Vert x\right\Vert ^{2}} \notag \\
1329: & \leq \frac{\left\langle a,b\right\rangle }{\alpha }+\frac{\left\Vert
1330: b\right\Vert }{\left\vert \alpha \right\vert \left\Vert x\right\Vert }\left[
1331: \left( \alpha -1\right) ^{2}\left\langle a,x\right\rangle ^{2}+\left\Vert
1332: x\right\Vert ^{2}\left\Vert a\right\Vert ^{2}-\left\langle a,x\right\rangle
1333: ^{2}\right] ^{\frac{1}{2}} \notag
1334: \end{align}%
1335: for any $\alpha \in \mathbb{R}\backslash \left\{ 0\right\} $ and $a,b,x\in
1336: H, $ $x\neq 0.$
1337:
1338: If in (\ref{ch3.2.7}) we take $\alpha =2,$ then we get%
1339: \begin{align}
1340: \frac{1}{2}\left[ \left\langle a,b\right\rangle -\left\Vert a\right\Vert
1341: \left\Vert b\right\Vert \right] \left\Vert x\right\Vert ^{2}& \leq
1342: \left\langle a,x\right\rangle \left\langle x,b\right\rangle \label{ch3.2.8}
1343: \\
1344: & \leq \frac{1}{2}\left[ \left\langle a,b\right\rangle +\left\Vert
1345: a\right\Vert \left\Vert y\right\Vert \right] \left\Vert x\right\Vert ^{2},
1346: \notag
1347: \end{align}%
1348: which apparently, as mentioned by T. Precupanu in \cite{P}, has been
1349: obtained independently of Buzano, by U. Richard in \cite{R}.
1350:
1351: In \cite{PE}, Pe\v{c}ari\'{c} gave a simple direct proof of (\ref{ch3.2.8})
1352: without mentioning the work of either Buzano or Richard, but tracked down
1353: the result, in a particular form, to an earlier paper due to C. Blatter \cite%
1354: {BL}.
1355: \end{remark}
1356:
1357: Obviously, the following refinement of Buzano's result may be stated \cite%
1358: {DRAG1}.
1359:
1360: \begin{corollary}[Dragomir, 2004]
1361: \label{ch3.c2.4}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
1362: \right) $ be a real or complex inner product space and $a,b,x\in H.$ Then%
1363: \begin{align}
1364: \left\vert \left\langle a,x\right\rangle \left\langle x,b\right\rangle
1365: \right\vert & \leq \left\vert \left\langle a,x\right\rangle \left\langle
1366: x,b\right\rangle -\frac{1}{2}\left\langle a,b\right\rangle \left\Vert
1367: x\right\Vert ^{2}\right\vert +\frac{1}{2}\left\vert \left\langle
1368: a,b\right\rangle \right\vert \left\Vert x\right\Vert ^{2} \label{ch3.2.9} \\
1369: & \leq \frac{1}{2}\left[ \left\Vert a\right\Vert \left\Vert b\right\Vert
1370: +\left\vert \left\langle a,b\right\rangle \right\vert \right] \left\Vert
1371: x\right\Vert ^{2}. \notag
1372: \end{align}
1373: \end{corollary}
1374:
1375: \begin{proof}
1376: The first inequality in (\ref{ch3.2.9}) follows by the triangle inequality
1377: for the modulus $\left\vert \cdot \right\vert .$ The second inequality is
1378: merely (\ref{ch3.2.6}) in which we added the same quantity to both sides.
1379: \end{proof}
1380:
1381: \begin{remark}
1382: \label{ch3.r2.5}For $\alpha =1,$ we deduce from (\ref{ch3.2.1}) the
1383: following inequality:%
1384: \begin{equation}
1385: \left\vert \frac{\left\langle a,x\right\rangle \left\langle x,b\right\rangle
1386: }{\left\Vert x\right\Vert ^{2}}-\left\langle a,b\right\rangle \right\vert
1387: \leq \frac{\left\Vert b\right\Vert }{\left\Vert x\right\Vert }\left[
1388: \left\Vert x\right\Vert ^{2}\left\Vert a\right\Vert ^{2}-\left\vert
1389: \left\langle a,x\right\rangle \right\vert ^{2}\right] ^{\frac{1}{2}}
1390: \label{ch3.2.10}
1391: \end{equation}%
1392: for any $a,b,x\in H$ with $x\neq 0.$
1393:
1394: If the space is real, then (\ref{ch3.2.10}) is equivalent to%
1395: \begin{align}
1396: & \left\langle a,b\right\rangle -\frac{\left\Vert b\right\Vert }{\left\Vert
1397: x\right\Vert }\left[ \left\Vert x\right\Vert ^{2}\left\Vert a\right\Vert
1398: ^{2}-\left\vert \left\langle a,x\right\rangle \right\vert ^{2}\right] ^{%
1399: \frac{1}{2}} \label{ch3.2.11} \\
1400: & \leq \frac{\left\langle a,x\right\rangle \left\langle x,b\right\rangle }{%
1401: \left\Vert x\right\Vert ^{2}} \notag \\
1402: & \leq \frac{\left\Vert b\right\Vert }{\left\Vert x\right\Vert }\left[
1403: \left\Vert x\right\Vert ^{2}\left\Vert a\right\Vert ^{2}-\left\vert
1404: \left\langle a,x\right\rangle \right\vert ^{2}\right] ^{\frac{1}{2}%
1405: }+\left\langle a,b\right\rangle , \notag
1406: \end{align}%
1407: which is similar to Richard's inequality (\ref{ch3.2.8}).
1408: \end{remark}
1409:
1410: \subsection{Applications to Kurepa's Inequality}
1411:
1412: In 1960, N.G. de Bruijn \cite{BR} obtained the following refinement of the
1413: Cauchy-Bunyakovsky-Schwarz inequality:%
1414: \begin{equation}
1415: \left\vert \sum_{i=1}^{n}a_{i}z_{i}\right\vert ^{2}\leq \frac{1}{2}%
1416: \sum_{i=1}^{n}a_{i}^{2}\left[ \sum_{i=1}^{n}\left\vert z_{i}\right\vert
1417: ^{2}+\left\vert \sum_{i=1}^{n}z_{i}^{2}\right\vert \right] , \label{ch3.3.1}
1418: \end{equation}%
1419: provided that $a_{i}$ are real numbers while $z_{i}$ are complex for each $%
1420: i\in \left\{ 1,...,n\right\} .$
1421:
1422: In \cite{KU}, S. Kurepa proved the following generalisation of the de Bruijn
1423: result:
1424:
1425: \begin{theorem}[Kurepa, 1966]
1426: \label{ch3.t3.1}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
1427: \right) $ be a real inner product space and $\left( H_{\mathbb{C}%
1428: },\left\langle \cdot ,\cdot \right\rangle _{\mathbb{C}}\right) $ its
1429: complexification. Then for any $a\in H$ and $z\in H_{\mathbb{C}},$ one has
1430: the following refinement of Schwarz's inequality%
1431: \begin{equation}
1432: \left\vert \left\langle a,z\right\rangle _{\mathbb{C}}\right\vert ^{2}\leq
1433: \frac{1}{2}\left\Vert a\right\Vert ^{2}\left[ \left\Vert z\right\Vert _{%
1434: \mathbb{C}}^{2}+\left\vert \left\langle z,\bar{z}\right\rangle _{\mathbb{C}%
1435: }\right\vert \right] \leq \left\Vert a\right\Vert ^{2}\left\Vert
1436: z\right\Vert _{\mathbb{C}}^{2}, \label{ch3.3.2}
1437: \end{equation}%
1438: where $\bar{z}$ denotes the conjugate of $z\in H_{\mathbb{C}}.$
1439: \end{theorem}
1440:
1441: As consequences of this general result, Kurepa noted the following integral,
1442: respectively, discrete inequality:
1443:
1444: \begin{corollary}[Kurepa, 1966]
1445: \label{ch3.c3.2}Let $\left( S,\Sigma ,\mu \right) $ be a positive measure
1446: space and $a,z\in L_{2}\left( S,\Sigma ,\mu \right) ,$ the Hilbert space of
1447: complex-valued $2-\mu -$integrable functions defined on $S.$ If $a$ is a
1448: real function and $z$ is a complex function, then
1449: \begin{multline}
1450: \left\vert \int_{S}a\left( t\right) z\left( t\right) d\mu \left( t\right)
1451: \right\vert ^{2} \label{ch3.3.3} \\
1452: \leq \frac{1}{2}\cdot \int_{S}a^{2}\left( t\right) d\mu \left( t\right) %
1453: \left[ \int_{S}\left\vert z\left( t\right) \right\vert ^{2}d\mu \left(
1454: t\right) +\left\vert \int_{S}z^{2}\left( t\right) d\mu \left( t\right)
1455: \right\vert \right] .
1456: \end{multline}
1457: \end{corollary}
1458:
1459: \begin{corollary}[Kurepa, 1966]
1460: \label{ch3.c3.3}If $a_{1},\dots ,a_{n}$ are real numbers, $z_{1},\dots
1461: ,z_{n} $ are complex numbers and $\left( A_{ij}\right) $ is a positive
1462: definite real matrix of dimension $n\times n$, then%
1463: \begin{equation}
1464: \left\vert \sum_{i,j=1}^{n}A_{ij}a_{i}z_{j}\right\vert ^{2}\leq \frac{1}{2}%
1465: \sum_{i,j=1}^{n}A_{ij}a_{i}a_{j}\left[ \sum_{i,j=1}^{n}A_{ij}z_{i}\overline{%
1466: z_{j}}+\left\vert \sum_{i,j=1}^{n}A_{ij}z_{i}\overline{z_{j}}\right\vert %
1467: \right] . \label{ch3.3.4}
1468: \end{equation}
1469: \end{corollary}
1470:
1471: The following refinement of Kurepa's result may be stated \cite{DRAG1}.
1472:
1473: \begin{theorem}[Dragomir, 2004]
1474: \label{ch3.t3.3}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
1475: \right) $ be a real inner product space and $\left( H_{\mathbb{C}%
1476: },\left\langle \cdot ,\cdot \right\rangle _{\mathbb{C}}\right) $ its
1477: complexification. Then for any $e\in H$ and $w\in H_{\mathbb{C}},$ one has
1478: the inequality:%
1479: \begin{align}
1480: \left\vert \left\langle w,e\right\rangle _{\mathbb{C}}\right\vert ^{2}& \leq
1481: \left\vert \left\langle w,e\right\rangle _{\mathbb{C}}^{2}-\frac{1}{2}%
1482: \left\langle w,\bar{w}\right\rangle _{\mathbb{C}}\left\Vert e\right\Vert
1483: ^{2}\right\vert +\frac{1}{2}\left\vert \left\langle w,\bar{w}\right\rangle _{%
1484: \mathbb{C}}\right\vert \left\Vert e\right\Vert ^{2} \label{ch3.3.5} \\
1485: & \leq \frac{1}{2}\left\Vert e\right\Vert ^{2}\left[ \left\Vert w\right\Vert
1486: _{\mathbb{C}}^{2}+\left\vert \left\langle w,\bar{w}\right\rangle _{\mathbb{C}%
1487: }\right\vert \right] . \notag
1488: \end{align}
1489: \end{theorem}
1490:
1491: \begin{proof}
1492: We follow the proof in \cite{DRAG1}.
1493:
1494: If we apply Corollary \ref{ch3.c3.3} for $\left( H_{\mathbb{C}},\left\langle
1495: \cdot ,\cdot \right\rangle _{\mathbb{C}}\right) $ and $x=e\in H,$ $a=w$ and $%
1496: b=\bar{w},$ then we have%
1497: \begin{align}
1498: & \left\vert \left\langle w,e\right\rangle _{\mathbb{C}}\left\langle e,\bar{w%
1499: }\right\rangle _{\mathbb{C}}\right\vert \label{ch3.3.6} \\
1500: & \leq \left\vert \left\langle w,e\right\rangle _{\mathbb{C}}\left\langle e,%
1501: \bar{w}\right\rangle _{\mathbb{C}}-\frac{1}{2}\left\langle w,\bar{w}%
1502: \right\rangle _{\mathbb{C}}\left\Vert e\right\Vert ^{2}\right\vert +\frac{1}{%
1503: 2}\left\vert \left\langle w,\bar{w}\right\rangle _{\mathbb{C}}\right\vert
1504: \left\Vert e\right\Vert ^{2} \notag \\
1505: & \leq \frac{1}{2}\left\Vert e\right\Vert ^{2}\left[ \left\Vert w\right\Vert
1506: _{\mathbb{C}}\left\Vert \bar{w}\right\Vert _{\mathbb{C}}+\left\vert
1507: \left\langle w,\bar{w}\right\rangle _{\mathbb{C}}\right\vert \right] .
1508: \notag
1509: \end{align}%
1510: Now, if we assume that $w=\left( x,y\right) \in H_{\mathbb{C}},$ then, by
1511: the definition of $\left\langle \cdot ,\cdot \right\rangle _{\mathbb{C}},$
1512: we have%
1513: \begin{align*}
1514: \left\langle w,e\right\rangle _{\mathbb{C}}& =\left\langle \left( x,y\right)
1515: ,\left( e,0\right) \right\rangle _{\mathbb{C}} \\
1516: & =\left\langle x,e\right\rangle +\left\langle y,0\right\rangle +i\left[
1517: \left\langle y,e\right\rangle -\left\langle x,0\right\rangle \right] \\
1518: & =\left\langle e,x\right\rangle +i\left\langle e,y\right\rangle ,
1519: \end{align*}%
1520: \begin{align*}
1521: \left\langle e,\bar{w}\right\rangle _{\mathbb{C}}& =\left\langle \left(
1522: e,0\right) ,\left( x,-y\right) \right\rangle _{\mathbb{C}} \\
1523: & =\left\langle e,x\right\rangle +\left\langle 0,-y\right\rangle +i\left[
1524: \left\langle 0,x\right\rangle -\left\langle e,-y\right\rangle \right] \\
1525: & =\left\langle e,x\right\rangle +i\left\langle e,y\right\rangle
1526: =\left\langle w,e\right\rangle _{\mathbb{C}}
1527: \end{align*}%
1528: and%
1529: \begin{equation*}
1530: \left\Vert \bar{w}\right\Vert _{\mathbb{C}}^{2}=\left\Vert x\right\Vert
1531: ^{2}+\left\Vert y\right\Vert ^{2}=\left\Vert w\right\Vert _{\mathbb{C}}^{2}.
1532: \end{equation*}%
1533: Therefore, by (\ref{ch3.3.6}), we deduce the desired result (\ref{ch3.3.5}).
1534: \end{proof}
1535:
1536: Denote by $\ell _{\rho }^{2}\left( \mathbb{C}\right) $ the Hilbert space of
1537: all complex sequences $z=\left( z_{i}\right) _{i\in \mathbb{N}}$ with the
1538: property that for $\rho _{i}\geq 0$ with $\sum_{i=1}^{\infty }\rho _{i}=1$
1539: we have $\sum_{i=1}^{\infty }\rho _{i}\left\vert z_{i}\right\vert
1540: ^{2}<\infty .$ If $a=\left( a_{i}\right) _{i\in \mathbb{N}}$ is a sequence
1541: of real numbers such that $a\in \ell _{\rho }^{2}\left( \mathbb{C}\right) ,$
1542: then for any $z\in \ell _{\rho }^{2}\left( \mathbb{C}\right) $ we have the
1543: inequality:%
1544: \begin{align}
1545: & \left\vert \sum_{i=1}^{\infty }\rho _{i}a_{i}z_{i}\right\vert ^{2}
1546: \label{ch3.3.7} \\
1547: & \leq \left\vert \left( \sum_{i=1}^{\infty }\rho _{i}a_{i}z_{i}\right) ^{2}-%
1548: \frac{1}{2}\sum_{i=1}^{\infty }\rho _{i}a_{i}^{2}\sum_{i=1}^{\infty }\rho
1549: _{i}z_{i}^{2}\right\vert +\frac{1}{2}\sum_{i=1}^{\infty }\rho
1550: _{i}a_{i}^{2}\left\vert \sum_{i=1}^{\infty }\rho _{i}z_{i}^{2}\right\vert
1551: \notag \\
1552: & \leq \frac{1}{2}\sum_{i=1}^{\infty }\rho _{i}a_{i}^{2}\left[
1553: \sum_{i=1}^{\infty }\rho _{i}\left\vert z_{i}\right\vert ^{2}+\left\vert
1554: \sum_{i=1}^{\infty }\rho _{i}z_{i}^{2}\right\vert \right] . \notag
1555: \end{align}
1556:
1557: Similarly, if by $L_{\rho }^{2}\left( S,\Sigma ,\mu \right) $ we understand
1558: the Hilbert space of all complex-valued functions $f:S\rightarrow \mathbb{C}$
1559: with the property that for the $\mu -$measurable function $\rho \geq 0$ with
1560: $\int_{S}\rho \left( t\right) d\mu \left( t\right) =1$ we have%
1561: \begin{equation*}
1562: \int_{S}\rho \left( t\right) \left\vert f\left( t\right) \right\vert
1563: ^{2}d\mu \left( t\right) <\infty ,
1564: \end{equation*}%
1565: then for a real function $a\in L_{\rho }^{2}\left( S,\Sigma ,\mu \right) $
1566: and any $f\in L_{\rho }^{2}\left( S,\Sigma ,\mu \right) ,$ we have the
1567: inequalities%
1568: \begin{align}
1569: & \left\vert \int_{S}\rho \left( t\right) a\left( t\right) f\left( t\right)
1570: d\mu \left( t\right) \right\vert ^{2} \label{ch3.3.8} \\
1571: & \leq \left\vert \left( \int_{S}\rho \left( t\right) a\left( t\right)
1572: f\left( t\right) d\mu \left( t\right) \right) ^{2}\right. \notag \\
1573: & \qquad \qquad -\frac{1}{2}\left. \int_{S}\rho \left( t\right) f^{2}\left(
1574: t\right) d\mu \left( t\right) \int_{S}\rho \left( t\right) a^{2}\left(
1575: t\right) d\mu \left( t\right) \right\vert \notag \\
1576: & \qquad \qquad +\frac{1}{2}\left\vert \int_{S}\rho \left( t\right)
1577: f^{2}\left( t\right) d\mu \left( t\right) \right\vert \int_{S}\rho \left(
1578: t\right) a^{2}\left( t\right) d\mu \left( t\right) \notag \\
1579: & \leq \frac{1}{2}\int_{S}\rho \left( t\right) a^{2}\left( t\right) d\mu
1580: \left( t\right) \notag \\
1581: & \qquad \qquad \times \left[ \int_{S}\rho \left( t\right) \left\vert
1582: f\left( t\right) \right\vert ^{2}d\mu \left( t\right) +\left\vert
1583: \int_{S}\rho \left( t\right) f^{2}\left( t\right) d\mu \left( t\right)
1584: \right\vert \right] . \notag
1585: \end{align}
1586:
1587: \section{Inequalities for Orthornormal Families}
1588:
1589: \subsection{\label{ch4.s1}Introduction}
1590:
1591: In \cite{B}, M.L. Buzano obtained the following extension of the celebrated
1592: Schwarz's inequality in a real or complex inner product space $\left(
1593: H;\left\langle \cdot ,\cdot \right\rangle \right) :$%
1594: \begin{equation}
1595: \left\vert \left\langle a,x\right\rangle \left\langle x,b\right\rangle
1596: \right\vert \leq \frac{1}{2}\left[ \left\Vert a\right\Vert \left\Vert
1597: b\right\Vert +\left\vert \left\langle a,b\right\rangle \right\vert \right]
1598: \left\Vert x\right\Vert ^{2}, \label{ch4.1.1}
1599: \end{equation}%
1600: for any $a,b,x\in H.$
1601:
1602: It is clear that the above inequality becomes, for $a=b,$ the Schwarz's
1603: inequality%
1604: \begin{equation}
1605: \left\vert \left\langle a,x\right\rangle \right\vert ^{2}\leq \left\Vert
1606: a\right\Vert ^{2}\left\Vert x\right\Vert ^{2},\quad a,x\in H;
1607: \label{ch4.1.2}
1608: \end{equation}%
1609: in which the equality holds if and only if there exists a scalar $\lambda
1610: \in \mathbb{K}$ $\left( \mathbb{R},\mathbb{C}\right) $ so that $x=\lambda a.$
1611:
1612: As noted by T. Precupanu in \cite{P}, independently of Buzano, U. Richard
1613: \cite{R} obtained the following similar inequality holding in real inner
1614: product spaces:%
1615: \begin{align}
1616: \frac{1}{2}\left\Vert x\right\Vert ^{2}\left[ \left\langle a,b\right\rangle
1617: -\left\Vert a\right\Vert \left\Vert b\right\Vert \right] & \leq \left\langle
1618: a,x\right\rangle \left\langle x,b\right\rangle \label{ch4.1.4} \\
1619: & \leq \frac{1}{2}\left\Vert x\right\Vert ^{2}\left[ \left\langle
1620: a,b\right\rangle +\left\Vert a\right\Vert \left\Vert b\right\Vert \right] .
1621: \notag
1622: \end{align}
1623:
1624: The main aim of the present section is to obtain similar results for
1625: families of orthonormal vectors in $\left( H;\left\langle \cdot ,\cdot
1626: \right\rangle \right) ,$ real or complex space, that are naturally connected
1627: with the celebrated Bessel inequality and improve the results of Busano,
1628: Richard and Kurepa.
1629:
1630: \subsection{A Generalisation for Orthonormal Families\label{ch4.s2}}
1631:
1632: We say that the finite family $\left\{ e_{i}\right\} _{i\in I}$ \ ($I$ is
1633: finite) of vectors is \textit{orthonormal} if $\left\langle
1634: e_{i},e_{j}\right\rangle =0$ if $i,j\in I$ with $i\neq j$ and $\left\Vert
1635: e_{i}\right\Vert =1$ for each $i\in I.$ The following result may be stated
1636: \cite{DRAG2}:
1637:
1638: \begin{theorem}[Dragomir, 2004]
1639: \label{ch4.t2.1}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
1640: \right) $ be an inner product space over the real or complex number field $%
1641: \mathbb{K}$ and $\left\{ e_{i}\right\} _{i\in I}$ a finite orthonormal
1642: family in $H.$ Then for any $a,b\in H,$ one has the inequality:%
1643: \begin{equation}
1644: \left\vert \sum_{i\in I}\left\langle a,e_{i}\right\rangle \left\langle
1645: e_{i},b\right\rangle -\frac{1}{2}\left\langle a,b\right\rangle \right\vert
1646: \leq \frac{1}{2}\left\Vert a\right\Vert \left\Vert b\right\Vert .
1647: \label{ch4.2.1}
1648: \end{equation}%
1649: The case of equality holds in (\ref{ch4.2.1}) if and only if%
1650: \begin{equation}
1651: \sum_{i\in I}\left\langle a,e_{i}\right\rangle e_{i}=\frac{1}{2}a+\left(
1652: \sum_{i\in I}\left\langle a,e_{i}\right\rangle \left\langle
1653: e_{i},b\right\rangle -\frac{1}{2}\left\langle a,b\right\rangle \right) \cdot
1654: \frac{b}{\left\Vert b\right\Vert ^{2}}. \label{ch4.2.2}
1655: \end{equation}
1656: \end{theorem}
1657:
1658: \begin{proof}
1659: We follow the proof in \cite{DRAG2}.
1660:
1661: It is well known that, for $e\neq 0$ and $f\in H,$ the following identity
1662: holds:%
1663: \begin{equation}
1664: \frac{\left\Vert f\right\Vert ^{2}\left\Vert e\right\Vert ^{2}-\left\vert
1665: \left\langle f,e\right\rangle \right\vert ^{2}}{\left\Vert e\right\Vert ^{2}}%
1666: =\left\Vert f-\frac{\left\langle f,e\right\rangle e}{\left\Vert e\right\Vert
1667: ^{2}}\right\Vert ^{2}. \label{ch4.2.3}
1668: \end{equation}%
1669: Therefore, in Schwarz's inequality%
1670: \begin{equation}
1671: \left\vert \left\langle f,e\right\rangle \right\vert ^{2}\leq \left\Vert
1672: f\right\Vert ^{2}\left\Vert e\right\Vert ^{2},\quad f,e\in H;
1673: \label{ch4.2.4}
1674: \end{equation}%
1675: the case of equality, for $e\neq 0,$ holds if and only if%
1676: \begin{equation*}
1677: f=\frac{\left\langle f,e\right\rangle e}{\left\Vert e\right\Vert ^{2}}.
1678: \end{equation*}%
1679: Let $f:=2\sum_{i\in I}\left\langle a,e_{i}\right\rangle e_{i}-a$ and $e:=b.$
1680: Then, by Schwarz's inequality (\ref{ch4.2.4}), we may state that%
1681: \begin{equation}
1682: \left\vert \left\langle 2\sum_{i\in I}\left\langle a,e_{i}\right\rangle
1683: e_{i}-a,b\right\rangle \right\vert ^{2}\leq \left\Vert 2\sum_{i\in
1684: I}\left\langle a,e_{i}\right\rangle e_{i}-a\right\Vert ^{2}\left\Vert
1685: b\right\Vert ^{2} \label{ch4.2.5}
1686: \end{equation}%
1687: with equality, for $b\neq 0,$ if and only if%
1688: \begin{equation}
1689: 2\sum_{i\in I}\left\langle a,e_{i}\right\rangle e_{i}-a=\left\langle
1690: 2\sum_{i\in I}\left\langle a,e_{i}\right\rangle e_{i}-a,b\right\rangle \frac{%
1691: b}{\left\Vert b\right\Vert ^{2}}. \label{ch4.2.6}
1692: \end{equation}%
1693: Since%
1694: \begin{equation*}
1695: \left\langle 2\sum_{i\in I}\left\langle a,e_{i}\right\rangle
1696: e_{i}-a,b\right\rangle =2\sum_{i\in I}\left\langle a,e_{i}\right\rangle
1697: \left\langle e_{i},b\right\rangle -\left\langle a,b\right\rangle
1698: \end{equation*}%
1699: and%
1700: \begin{align*}
1701: & \left\Vert 2\sum_{i\in I}\left\langle a,e_{i}\right\rangle
1702: e_{i}-a\right\Vert ^{2} \\
1703: & =4\left\Vert \sum_{i\in I}\left\langle a,e_{i}\right\rangle
1704: e_{i}\right\Vert ^{2}-4\func{Re}\left\langle \sum_{i\in I}\left\langle
1705: a,e_{i}\right\rangle e_{i},a\right\rangle +\left\Vert a\right\Vert ^{2} \\
1706: & =4\sum_{i\in I}\left\vert \left\langle a,e_{i}\right\rangle \right\vert
1707: ^{2}-4\sum_{i\in I}\left\vert \left\langle a,e_{i}\right\rangle \right\vert
1708: ^{2}+\left\Vert a\right\Vert ^{2} \\
1709: & =\left\Vert a\right\Vert ^{2},
1710: \end{align*}%
1711: hence by (\ref{ch4.2.5}) we deduce the desired inequality (\ref{ch4.2.1}).
1712:
1713: Finally, as (\ref{ch4.2.2}) is equivalent to%
1714: \begin{equation*}
1715: \sum_{i\in I}\left\langle a,e_{i}\right\rangle e_{i}-\frac{a}{2}=\left(
1716: \sum_{i\in I}\left\langle a,e_{i}\right\rangle \left\langle
1717: e_{i},b\right\rangle -\frac{1}{2}\left\langle a,b\right\rangle \right) \frac{%
1718: b}{\left\Vert b\right\Vert ^{2}},
1719: \end{equation*}%
1720: hence the equality holds in (\ref{ch4.2.1}) if and only if (\ref{ch4.2.2})
1721: is valid.
1722: \end{proof}
1723:
1724: The following result is well known in the literature as Bessel's inequality%
1725: \begin{equation}
1726: \sum_{i\in I}\left\vert \left\langle x,e_{i}\right\rangle \right\vert
1727: ^{2}\leq \left\Vert x\right\Vert ^{2},\quad x\in H, \label{ch4.2.7}
1728: \end{equation}%
1729: where, as above, $\left\{ e_{i}\right\} _{i\in I}$ is a finite orthonormal
1730: family in the inner product space $\left( H;\left\langle \cdot ,\cdot
1731: \right\rangle \right) .$
1732:
1733: If one chooses $a=b=x$ in (\ref{ch4.2.1}), then one gets the inequality%
1734: \begin{equation*}
1735: \left\vert \sum_{i\in I}\left\vert \left\langle x,e_{i}\right\rangle
1736: \right\vert ^{2}-\frac{1}{2}\left\Vert x\right\Vert ^{2}\right\vert \leq
1737: \frac{1}{2}\left\Vert x\right\Vert ^{2},
1738: \end{equation*}%
1739: which is obviously equivalent to Bessel's inequality (\ref{ch4.2.7}).
1740: Therefore, the inequality (\ref{ch4.2.1}) may be regarded as a
1741: generalisation of Bessel's inequality as well.
1742:
1743: Utilising the Bessel and Cauchy-Bunyakovsky-Schwarz inequalities, one may
1744: state that%
1745: \begin{equation}
1746: \left\vert \sum_{i\in I}\left\langle a,e_{i}\right\rangle \left\langle
1747: e_{i},b\right\rangle \right\vert \leq \left[ \sum_{i\in I}\left\vert
1748: \left\langle a,e_{i}\right\rangle \right\vert ^{2}\sum_{i\in I}\left\vert
1749: \left\langle b,e_{i}\right\rangle \right\vert ^{2}\right] ^{\frac{1}{2}}\leq
1750: \left\Vert a\right\Vert \left\Vert b\right\Vert \label{ch4.2.8}
1751: \end{equation}
1752:
1753: A different refinement of the inequality between the first and the last term
1754: in (\ref{ch4.2.8}) is incorporated in the following \cite{DRAG2}:
1755:
1756: \begin{corollary}[Dragomir, 2004]
1757: \label{ch4.c2.2}With the assumption of Theorem \ref{ch4.t2.1}, we have%
1758: \begin{align}
1759: \left\vert \sum_{i\in I}\left\langle a,e_{i}\right\rangle \left\langle
1760: e_{i},b\right\rangle \right\vert & \leq \left\vert \sum_{i\in I}\left\langle
1761: a,e_{i}\right\rangle \left\langle e_{i},b\right\rangle -\frac{1}{2}%
1762: \left\langle a,b\right\rangle \right\vert +\frac{1}{2}\left\vert
1763: \left\langle a,b\right\rangle \right\vert \label{ch4.2.9} \\
1764: & \leq \frac{1}{2}\left[ \left\Vert a\right\Vert \left\Vert b\right\Vert
1765: +\left\vert \left\langle a,b\right\rangle \right\vert \right] \notag \\
1766: & \leq \left\Vert a\right\Vert \left\Vert b\right\Vert . \notag
1767: \end{align}
1768: \end{corollary}
1769:
1770: \begin{remark}
1771: \label{ch4.r2.3}If the space $\left( H;\left\langle \cdot ,\cdot
1772: \right\rangle \right) $ is real, then, obviously, (\ref{ch4.2.1}) is
1773: equivalent to:%
1774: \begin{equation}
1775: \frac{1}{2}\left( \left\langle a,b\right\rangle -\left\Vert a\right\Vert
1776: \left\Vert b\right\Vert \right) \leq \sum_{i\in I}\left\langle
1777: a,e_{i}\right\rangle \left\langle e_{i},b\right\rangle \leq \frac{1}{2}\left[
1778: \left\Vert a\right\Vert \left\Vert b\right\Vert +\left\langle
1779: a,b\right\rangle \right] . \label{ch4.2.10}
1780: \end{equation}
1781: \end{remark}
1782:
1783: \begin{remark}
1784: \label{ch4.r2.4}It is obvious that if the family comprises of only a single
1785: element $e=\frac{x}{\left\Vert x\right\Vert },$ $x\in H,$ $x\neq 0,$ then
1786: from (\ref{ch4.2.9}) we recapture the refinement of Buzano's inequality
1787: incorporated in (\ref{ch4.1.1}) while from (\ref{ch4.2.10}) we deduce
1788: Richard's result from (\ref{ch4.1.4}).
1789: \end{remark}
1790:
1791: The following corollary of Theorem \ref{ch4.t2.1} is of interest as well
1792: \cite{DRAG2}:
1793:
1794: \begin{corollary}[Dragomir, 2004]
1795: \label{ch4.c2.3}Let $\left\{ e_{i}\right\} _{i\in I}$ be a finite
1796: orthonormal family in $\left( H;\left\langle \cdot ,\cdot \right\rangle
1797: \right) .$ If $x,y\in H\backslash \left\{ 0\right\} $ are such that there
1798: exists the constants $m_{i},n_{i},$ $M_{i},$ $N_{i}\in \mathbb{R}$, $i\in I$
1799: such that:%
1800: \begin{equation}
1801: -1\leq m_{i}\leq \frac{\func{Re}\left\langle x,e_{i}\right\rangle }{%
1802: \left\Vert x\right\Vert }\cdot \frac{\func{Re}\left\langle
1803: y,e_{i}\right\rangle }{\left\Vert y\right\Vert }\leq M_{i}\leq 1,\quad i\in I
1804: \label{ch4.2.11}
1805: \end{equation}%
1806: and%
1807: \begin{equation}
1808: -1\leq n_{i}\leq \frac{\func{Im}\left\langle x,e_{i}\right\rangle }{%
1809: \left\Vert x\right\Vert }\cdot \frac{\func{Im}\left\langle
1810: y,e_{i}\right\rangle }{\left\Vert y\right\Vert }\leq N_{i}\leq 1,\quad i\in I
1811: \label{ch4.2.12}
1812: \end{equation}%
1813: then%
1814: \begin{equation}
1815: 2\sum_{i\in I}\left( m_{i}+n_{i}\right) -1\leq \frac{\func{Re}\left\langle
1816: x,y\right\rangle }{\left\Vert x\right\Vert \left\Vert y\right\Vert }\leq
1817: 1+2\sum_{i\in I}\left( M_{i}+N_{i}\right) . \label{ch4.2.13}
1818: \end{equation}
1819: \end{corollary}
1820:
1821: \begin{proof}
1822: We follow the proof in \cite{DRAG2}.
1823:
1824: Using Theorem \ref{ch4.t2.1} and the fact that for any complex number $z,$ $%
1825: \left\vert z\right\vert \geq \left\vert \func{Re}z\right\vert ,$ we have%
1826: \begin{align}
1827: & \left\vert \sum_{i\in I}\func{Re}\left[ \left\langle x,e_{i}\right\rangle
1828: \left\langle e_{i},y\right\rangle \right] -\frac{1}{2}\func{Re}\left\langle
1829: x,y\right\rangle \right\vert \label{ch4.2.14} \\
1830: & \leq \left\vert \sum_{i\in I}\left\langle x,e_{i}\right\rangle
1831: \left\langle e_{i},y\right\rangle -\frac{1}{2}\left\langle x,y\right\rangle
1832: \right\vert \notag \\
1833: & \leq \frac{1}{2}\left\Vert x\right\Vert \left\Vert y\right\Vert . \notag
1834: \end{align}%
1835: Since%
1836: \begin{equation*}
1837: \func{Re}\left[ \left\langle x,e_{i}\right\rangle \left\langle
1838: e_{i},y\right\rangle \right] =\func{Re}\left\langle x,e_{i}\right\rangle
1839: \func{Re}\left\langle y,e_{i}\right\rangle +\func{Im}\left\langle
1840: x,e_{i}\right\rangle \func{Im}\left\langle y,e_{i}\right\rangle ,
1841: \end{equation*}%
1842: hence by (\ref{ch4.2.14}) we have:%
1843: \begin{align}
1844: & -\frac{1}{2}\left\Vert x\right\Vert \left\Vert y\right\Vert +\frac{1}{2}%
1845: \func{Re}\left\langle x,y\right\rangle \label{ch4.2.15} \\
1846: & \leq \sum_{i\in I}\func{Re}\left\langle x,e_{i}\right\rangle \func{Re}%
1847: \left\langle y,e_{i}\right\rangle +\sum_{i\in I}\func{Im}\left\langle
1848: x,e_{i}\right\rangle \func{Im}\left\langle y,e_{i}\right\rangle \notag \\
1849: & \leq \frac{1}{2}\left\Vert x\right\Vert \left\Vert y\right\Vert +\frac{1}{2%
1850: }\func{Re}\left\langle x,y\right\rangle . \notag
1851: \end{align}%
1852: Utilising the assumptions (\ref{ch4.2.11}) and (\ref{ch4.2.12}), we have%
1853: \begin{equation}
1854: \sum_{i\in I}m_{i}\leq \sum_{i\in I}\frac{\func{Re}\left\langle
1855: x,e_{i}\right\rangle \func{Re}\left\langle y,e_{i}\right\rangle }{\left\Vert
1856: x\right\Vert \left\Vert y\right\Vert }\leq \sum_{i\in I}M_{i}
1857: \label{ch4.2.16}
1858: \end{equation}%
1859: and
1860: \begin{equation}
1861: \sum_{i\in I}n_{i}\leq \sum_{i\in I}\frac{\func{Im}\left\langle
1862: x,e_{i}\right\rangle \func{Im}\left\langle y,e_{i}\right\rangle }{\left\Vert
1863: x\right\Vert \left\Vert y\right\Vert }\leq \sum_{i\in I}N_{i}.
1864: \label{ch4.2.17}
1865: \end{equation}%
1866: Finally, on making use of (\ref{ch4.2.15}) -- (\ref{ch4.2.17}), we deduce
1867: the desired result (\ref{ch4.2.13}).
1868: \end{proof}
1869:
1870: \begin{remark}
1871: By Schwarz's inequality, is it obvious that, in general,%
1872: \begin{equation*}
1873: -1\leq \frac{\func{Re}\left\langle x,y\right\rangle }{\left\Vert
1874: x\right\Vert \left\Vert y\right\Vert }\leq 1.
1875: \end{equation*}%
1876: Consequently, the left inequality in (\ref{ch4.2.13}) is of interest when $%
1877: \sum_{i\in I}\left( m_{i}+n_{i}\right) >0,$ while the right inequality in (%
1878: \ref{ch4.2.13}) is of interest when $\sum_{i\in I}\left( M_{i}+N_{i}\right)
1879: <0.$
1880: \end{remark}
1881:
1882: \subsection{Refinements of Kurepa's Inequality\label{ch4.s3}}
1883:
1884: The following result holds \cite{DRAG2}.
1885:
1886: \begin{theorem}[Dragomir, 2004]
1887: \label{ch4.t3.1}Let $\left\{ e_{j}\right\} _{j\in I}$ be a finite
1888: orthonormal family in the real inner product space $\left( H;\left\langle
1889: \cdot ,\cdot \right\rangle \right) .$ Then for any $w\in H_{\mathbb{C}},$
1890: where $\left( H_{\mathbb{C}};\left\langle \cdot ,\cdot \right\rangle _{%
1891: \mathbb{C}}\right) $ is the complexification of $\left( H;\left\langle \cdot
1892: ,\cdot \right\rangle \right) ,$ one has the following Bessel's type
1893: inequality:%
1894: \begin{align}
1895: \left\vert \sum_{j\in I}\left\langle w,e_{j}\right\rangle _{\mathbb{C}%
1896: }^{2}\right\vert & \leq \left\vert \sum_{j\in I}\left\langle
1897: w,e_{j}\right\rangle _{\mathbb{C}}^{2}-\frac{1}{2}\left\langle w,\bar{w}%
1898: \right\rangle _{\mathbb{C}}\right\vert +\frac{1}{2}\left\vert \left\langle w,%
1899: \bar{w}\right\rangle _{\mathbb{C}}\right\vert \label{ch4.3.3} \\
1900: & \leq \frac{1}{2}\left[ \left\Vert w\right\Vert _{\mathbb{C}%
1901: }^{2}+\left\vert \left\langle w,\bar{w}\right\rangle _{\mathbb{C}%
1902: }\right\vert \right] \leq \left\Vert w\right\Vert _{\mathbb{C}}^{2}. \notag
1903: \end{align}
1904: \end{theorem}
1905:
1906: \begin{proof}
1907: We follow the proof in \cite{DRAG2}.
1908:
1909: Define $f_{j}\in H_{\mathbb{C}},$ $f_{j}:=\left( e_{j},0\right) ,$ $j\in I.$
1910: For any $k,j\in I$ we have%
1911: \begin{equation*}
1912: \left\langle f_{i},f_{j}\right\rangle _{\mathbb{C}}=\left\langle \left(
1913: e_{k},0\right) ,\left( e_{j},0\right) \right\rangle _{\mathbb{C}%
1914: }=\left\langle e_{k},e_{j}\right\rangle =\delta _{kj},
1915: \end{equation*}%
1916: therefore $\left\{ f_{j}\right\} _{j\in I}$ is an orthonormal family in $%
1917: \left( H_{\mathbb{C}};\left\langle \cdot ,\cdot \right\rangle _{\mathbb{C}%
1918: }\right) .$
1919:
1920: If we apply Theorem \ref{ch4.t2.1} for $\left( H_{\mathbb{C}};\left\langle
1921: \cdot ,\cdot \right\rangle _{\mathbb{C}}\right) ,$ $a=w,$ $b=\bar{w},$ we
1922: may write:%
1923: \begin{equation}
1924: \left\vert \sum_{j\in I}\left\langle w,e_{j}\right\rangle _{\mathbb{C}%
1925: }\left\langle e_{j},\bar{w}\right\rangle _{\mathbb{C}}-\frac{1}{2}%
1926: \left\langle w,\bar{w}\right\rangle _{\mathbb{C}}\right\vert \leq \frac{1}{2}%
1927: \left\Vert w\right\Vert _{\mathbb{C}}\left\Vert \bar{w}\right\Vert _{\mathbb{%
1928: C}}. \label{ch4.3.4}
1929: \end{equation}%
1930: However, for $w:=\left( x,y\right) \in H_{\mathbb{C}}$, we have $\bar{w}%
1931: =\left( x,-y\right) $ and%
1932: \begin{equation*}
1933: \left\langle e_{j},\bar{w}\right\rangle _{\mathbb{C}}=\left\langle \left(
1934: e_{j},0\right) ,\left( x,-y\right) \right\rangle _{\mathbb{C}}=\left\langle
1935: e_{j},x\right\rangle -i\left\langle e_{j},-y\right\rangle =\left\langle
1936: e_{j},x\right\rangle +i\left\langle e_{j},y\right\rangle
1937: \end{equation*}%
1938: and%
1939: \begin{equation*}
1940: \left\langle w,e_{j}\right\rangle _{\mathbb{C}}=\left\langle \left(
1941: x,y\right) ,\left( e_{j},0\right) \right\rangle _{\mathbb{C}}=\left\langle
1942: e_{j},x\right\rangle -i\left\langle e_{j},-y\right\rangle =\left\langle
1943: x,e_{j}\right\rangle +i\left\langle e_{j},y\right\rangle
1944: \end{equation*}%
1945: for any $j\in I.$ Thus $\left\langle e_{j},\bar{w}\right\rangle
1946: =\left\langle w,e_{j}\right\rangle $ for each $j\in I$ and since%
1947: \begin{equation*}
1948: \left\Vert w\right\Vert _{\mathbb{C}}=\left\Vert \bar{w}\right\Vert _{%
1949: \mathbb{C}}=\left( \left\Vert x\right\Vert ^{2}+\left\Vert y\right\Vert
1950: ^{2}\right) ^{\frac{1}{2}},
1951: \end{equation*}%
1952: we get from (\ref{ch4.3.4}) that%
1953: \begin{equation}
1954: \left\vert \sum_{j\in I}\left\langle w,e_{j}\right\rangle _{\mathbb{C}}^{2}-%
1955: \frac{1}{2}\left\langle w,\bar{w}\right\rangle _{\mathbb{C}}\right\vert \leq
1956: \frac{1}{2}\left\Vert w\right\Vert _{\mathbb{C}}^{2}. \label{ch4.3.5}
1957: \end{equation}%
1958: Now, observe that the first inequality in (\ref{ch4.3.3}) follows by the
1959: triangle inequality, the second is an obvious consequence of (\ref{ch4.3.5})
1960: and the last one is derived from Schwarz's result.
1961: \end{proof}
1962:
1963: \begin{remark}
1964: \label{ch4.r3.2}If the family $\left\{ e_{j}\right\} _{j\in I}$ contains
1965: only a single element $e=\frac{x}{\left\Vert x\right\Vert },$ $x\in H,$ $%
1966: x\neq 0,$ then from (\ref{ch4.3.3}) we deduce (\ref{ch3.3.6}), which, in its
1967: turn, provides a refinement of Kurepa's inequality (\ref{ch3.3.2}).
1968: \end{remark}
1969:
1970: \subsection{\label{ch4.s4}An Application for $L_{2}\left[ -\protect\pi ,%
1971: \protect\pi \right] $}
1972:
1973: It is well known that in the Hilbert space $L_{2}\left[ -\pi ,\pi \right] $
1974: of all functions $f:\left[ -\pi ,\pi \right] \rightarrow \mathbb{C}$ with
1975: the property that $f$ is Lebesgue measurable on $\left[ -\pi ,\pi \right] $
1976: and $\int_{-\pi }^{\pi }\left\vert f\left( t\right) \right\vert
1977: ^{2}dt<\infty ,$ the set of functions%
1978: \begin{equation*}
1979: \left\{ \frac{1}{\sqrt{2\pi }},\frac{1}{\sqrt{\pi }}\cos t,\frac{1}{\sqrt{%
1980: \pi }}\sin t,\dots ,\frac{1}{\sqrt{\pi }}\cos nt,\frac{1}{\sqrt{\pi }}\sin
1981: nt,\dots \right\}
1982: \end{equation*}%
1983: is orthonormal.
1984:
1985: If by $\limfunc{trig}t,$ we denote either $\sin t$ or $\cos t$, $t\in \left[
1986: -\pi ,\pi \right] ,$ then on using the results from Sections \ref{ch4.s2}
1987: and \ref{ch4.s3}, we may state the following inequality:%
1988: \begin{multline}
1989: \left\vert \frac{1}{\pi }\sum_{k=1}^{n}\int_{-\pi }^{\pi }f\left( t\right)
1990: \limfunc{trig}\left( kt\right) dt\cdot \int_{-\pi }^{\pi }\overline{g\left(
1991: t\right) }\limfunc{trig}\left( kt\right) dt\right. \label{ch4.4.1} \\
1992: \left. -\frac{1}{2}\int_{-\pi }^{\pi }f\left( t\right) \overline{g\left(
1993: t\right) }dt\right\vert ^{2} \\
1994: \leq \frac{1}{4}\int_{-\pi }^{\pi }\left\vert f\left( t\right) \right\vert
1995: ^{2}dt\int_{-\pi }^{\pi }\left\vert g\left( t\right) \right\vert ^{2}dt,
1996: \end{multline}%
1997: where all $\limfunc{trig}\left( kt\right) $ is either $\sin kt$ or $\cos kt,$
1998: $k\in \left\{ 1,\dots ,n\right\} $ and $f\in L_{2}\left[ -\pi ,\pi \right] .$
1999:
2000: This follows by Theorem \ref{ch4.t2.1}.
2001:
2002: If one uses Corollary \ref{ch4.c2.2}, then one can state the following chain
2003: of inequalities%
2004: \begin{align}
2005: & \left\vert \frac{1}{\pi }\sum_{k=1}^{n}\int_{-\pi }^{\pi }f\left( t\right)
2006: \limfunc{trig}\left( kt\right) dt\cdot \int_{-\pi }^{\pi }\overline{g\left(
2007: t\right) }\limfunc{trig}\left( kt\right) dt\right\vert \label{ch4.4.2} \\
2008: & \leq \left\vert \frac{1}{\pi }\sum_{k=1}^{n}\int_{-\pi }^{\pi }f\left(
2009: t\right) \limfunc{trig}\left( kt\right) dt\cdot \int_{-\pi }^{\pi }\overline{%
2010: g\left( t\right) }\limfunc{trig}\left( kt\right) dt\right. \notag \\
2011: & \qquad \qquad \left. -\frac{1}{2}\int_{-\pi }^{\pi }f\left( t\right)
2012: \overline{g\left( t\right) }dt\right\vert +\frac{1}{2}\left\vert \int_{-\pi
2013: }^{\pi }f\left( t\right) \overline{g\left( t\right) }dt\right\vert \notag \\
2014: & \leq \frac{1}{2}\left[ \left( \int_{-\pi }^{\pi }\left\vert f\left(
2015: t\right) \right\vert ^{2}dt\int_{-\pi }^{\pi }\left\vert g\left( t\right)
2016: \right\vert ^{2}dt\right) ^{\frac{1}{2}}+\left\vert \int_{-\pi }^{\pi
2017: }f\left( t\right) \overline{g\left( t\right) }dt\right\vert \right] \notag
2018: \\
2019: & \leq \left( \int_{-\pi }^{\pi }\left\vert f\left( t\right) \right\vert
2020: ^{2}dt\int_{-\pi }^{\pi }\left\vert g\left( t\right) \right\vert
2021: ^{2}dt\right) ^{\frac{1}{2}}, \notag
2022: \end{align}%
2023: where $f\in L_{2}\left[ -\pi ,\pi \right] .$
2024:
2025: Finally, by employing Theorem \ref{ch4.t3.1}, we may state:%
2026: \begin{align*}
2027: & \frac{1}{\pi }\left\vert \sum_{k=1}^{n}\left[ \int_{-\pi }^{\pi }f\left(
2028: t\right) \limfunc{trig}\left( kt\right) dt\right] ^{2}\right\vert \\
2029: & \leq \left\vert \frac{1}{\pi }\sum_{k=1}^{n}\left[ \int_{-\pi }^{\pi
2030: }f\left( t\right) \limfunc{trig}\left( kt\right) dt\right] ^{2}-\frac{1}{2}%
2031: \int_{-\pi }^{\pi }f^{2}\left( t\right) dt\right\vert +\frac{1}{2}\left\vert
2032: \int_{-\pi }^{\pi }f^{2}\left( t\right) dt\right\vert \\
2033: & \leq \frac{1}{2}\left[ \int_{-\pi }^{\pi }\left\vert f\left( t\right)
2034: \right\vert ^{2}dt+\left\vert \int_{-\pi }^{\pi }f^{2}\left( t\right)
2035: dt\right\vert \right] \leq \int_{-\pi }^{\pi }\left\vert f\left( t\right)
2036: \right\vert ^{2}dt,
2037: \end{align*}%
2038: where $f\in L_{2}\left[ -\pi ,\pi \right] .$
2039:
2040: \section{Generalizations of Precupanu's Inequality}
2041:
2042: \subsection{\label{ch5.s1}Introduction}
2043:
2044: In 1976, T. Precupanu \cite{P} obtained the following result related to the
2045: Schwarz inequality in a real inner product space $\left( H;\left\langle
2046: \cdot ,\cdot \right\rangle \right) :$
2047:
2048: \begin{theorem}[Precupanu, 1976]
2049: \label{ch5.t1.1}For any $a\in H,$ $x,y\in H\backslash \left\{ 0\right\} ,$
2050: we have the inequality:%
2051: \begin{align}
2052: & \frac{-\left\Vert a\right\Vert \left\Vert b\right\Vert +\left\langle
2053: a,b\right\rangle }{2} \label{ch5.1.1} \\
2054: & \leq \frac{\left\langle x,a\right\rangle \left\langle x,b\right\rangle }{%
2055: \left\Vert x\right\Vert ^{2}}+\frac{\left\langle y,a\right\rangle
2056: \left\langle y,b\right\rangle }{\left\Vert y\right\Vert ^{2}}-2\cdot \frac{%
2057: \left\langle x,a\right\rangle \left\langle y,b\right\rangle \left\langle
2058: x,y\right\rangle }{\left\Vert x\right\Vert ^{2}\left\Vert y\right\Vert ^{2}}
2059: \notag \\
2060: & \leq \frac{\left\Vert a\right\Vert \left\Vert b\right\Vert +\left\langle
2061: a,b\right\rangle }{2}. \notag
2062: \end{align}%
2063: In the right-hand side or in the left-hand side of (\ref{ch5.1.1}) we have
2064: equality if and only if there are $\lambda ,\mu \in \mathbb{R}$ such that%
2065: \begin{equation}
2066: \lambda \frac{\left\langle x,a\right\rangle }{\left\Vert x\right\Vert ^{2}}%
2067: \cdot x+\mu \frac{\left\langle y,b\right\rangle }{\left\Vert y\right\Vert
2068: ^{2}}\cdot y=\frac{1}{2}\left( \lambda a+\mu b\right) . \label{ch5.1.2}
2069: \end{equation}
2070: \end{theorem}
2071:
2072: Note for instance that \cite{P}, if $y\perp b,$ i.e., $\left\langle
2073: y,b\right\rangle =0,$ then by (\ref{ch5.1.1}) one may deduce:%
2074: \begin{equation}
2075: \frac{-\left\Vert a\right\Vert \left\Vert b\right\Vert +\left\langle
2076: a,b\right\rangle }{2}\left\Vert x\right\Vert ^{2}\leq \left\langle
2077: x,a\right\rangle \left\langle x,b\right\rangle \leq \frac{\left\Vert
2078: a\right\Vert \left\Vert b\right\Vert +\left\langle a,b\right\rangle }{2}%
2079: \left\Vert x\right\Vert ^{2} \label{ch5.1.3}
2080: \end{equation}%
2081: for any $a,b,x\in H,$ an inequality that has been obtained previously by U.
2082: Richard \cite{R}. The case of equality in the right-hand side or in the
2083: left-hand side of (\ref{ch5.1.3}) holds if and only if there are $\lambda
2084: ,\mu \in \mathbb{R}$ with%
2085: \begin{equation}
2086: 2\lambda \left\langle x,a\right\rangle x=\left( \lambda a+\mu b\right)
2087: \left\Vert x\right\Vert ^{2}. \label{ch5.1.4}
2088: \end{equation}%
2089: For $a=b,$ we may obtain from (\ref{ch5.1.1}) the following inequality \cite%
2090: {P}%
2091: \begin{equation}
2092: 0\leq \frac{\left\langle x,a\right\rangle ^{2}}{\left\Vert x\right\Vert ^{2}}%
2093: +\frac{\left\langle y,a\right\rangle ^{2}}{\left\Vert y\right\Vert ^{2}}%
2094: -2\cdot \frac{\left\langle x,a\right\rangle \left\langle y,a\right\rangle
2095: \left\langle x,y\right\rangle }{\left\Vert x\right\Vert ^{2}\left\Vert
2096: y\right\Vert ^{2}}\leq \left\Vert a\right\Vert ^{2}. \label{ch5.1.5}
2097: \end{equation}%
2098: This inequality implies \cite{P}:%
2099: \begin{equation}
2100: \frac{\left\langle x,y\right\rangle }{\left\Vert x\right\Vert \left\Vert
2101: y\right\Vert }\geq \frac{1}{2}\left[ \frac{\left\langle x,a\right\rangle }{%
2102: \left\Vert x\right\Vert \left\Vert a\right\Vert }+\frac{\left\langle
2103: y,a\right\rangle }{\left\Vert y\right\Vert \left\Vert a\right\Vert }\right]
2104: ^{2}-\frac{3}{2}. \label{ch5.1.6}
2105: \end{equation}
2106:
2107: In \cite{M}, M.H. Moore pointed out the following reverse of the Schwarz
2108: inequality%
2109: \begin{equation}
2110: \left\vert \left\langle y,z\right\rangle \right\vert \leq \left\Vert
2111: y\right\Vert \left\Vert z\right\Vert ,\qquad y,z\in H, \label{ch5.1.7}
2112: \end{equation}%
2113: where some information about a third vector $x$ is known:
2114:
2115: \begin{theorem}[Moore, 1973]
2116: \label{ch5.t1.2}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
2117: \right) $ be an inner product space over the real field $\mathbb{R}$ and $%
2118: x,y,z\in H$ such that:%
2119: \begin{equation}
2120: \left\vert \left\langle x,y\right\rangle \right\vert \geq \left(
2121: 1-\varepsilon \right) \left\Vert x\right\Vert \left\Vert y\right\Vert
2122: ,\qquad \left\vert \left\langle x,z\right\rangle \right\vert \geq \left(
2123: 1-\varepsilon \right) \left\Vert x\right\Vert \left\Vert z\right\Vert ,
2124: \label{ch5.1.8}
2125: \end{equation}%
2126: where $\varepsilon $ is a positive real number, reasonably small. Then%
2127: \begin{equation}
2128: \left\vert \left\langle y,z\right\rangle \right\vert \geq \max \left\{
2129: 1-\varepsilon -\sqrt{2\varepsilon },1-4\varepsilon ,0\right\} \left\Vert
2130: y\right\Vert \left\Vert z\right\Vert . \label{ch5.1.9}
2131: \end{equation}
2132: \end{theorem}
2133:
2134: Utilising Richard's inequality (\ref{ch5.1.3}) written in the following
2135: equivalent form:%
2136: \begin{equation}
2137: 2\cdot \frac{\left\langle x,a\right\rangle \left\langle x,b\right\rangle }{%
2138: \left\Vert x\right\Vert ^{2}}-\left\Vert a\right\Vert \left\Vert
2139: b\right\Vert \leq \left\langle a,b\right\rangle \leq 2\cdot \frac{%
2140: \left\langle x,a\right\rangle \left\langle x,b\right\rangle }{\left\Vert
2141: x\right\Vert ^{2}}+\left\Vert a\right\Vert \left\Vert b\right\Vert
2142: \label{ch5.1.10}
2143: \end{equation}%
2144: for any $a,b\in H$ and $a\in H\backslash \left\{ 0\right\} ,$ Precupanu has
2145: obtained the following Moore's type result:
2146:
2147: \begin{theorem}[Precupanu, 1976]
2148: \label{ch5.t1.3}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
2149: \right) $ be a real inner product space. If $a,b,x\in H$ and $0<\varepsilon
2150: _{1}<\varepsilon _{2}$ are such that:%
2151: \begin{align}
2152: \varepsilon _{1}\left\Vert x\right\Vert \left\Vert a\right\Vert & \leq
2153: \left\langle x,a\right\rangle \leq \varepsilon _{2}\left\Vert x\right\Vert
2154: \left\Vert a\right\Vert , \label{ch5.1.11} \\
2155: \varepsilon _{1}\left\Vert x\right\Vert \left\Vert b\right\Vert & \leq
2156: \left\langle x,b\right\rangle \leq \varepsilon _{2}\left\Vert x\right\Vert
2157: \left\Vert b\right\Vert , \notag
2158: \end{align}%
2159: then%
2160: \begin{equation}
2161: \left( 2\varepsilon _{1}^{2}-1\right) \left\Vert a\right\Vert \left\Vert
2162: b\right\Vert \leq \left\langle a,b\right\rangle \leq \left( 2\varepsilon
2163: _{1}^{2}+1\right) \left\Vert a\right\Vert \left\Vert b\right\Vert .
2164: \label{ch5.1.12}
2165: \end{equation}
2166: \end{theorem}
2167:
2168: Remark that the right inequality is always satisfied, since by Schwarz's
2169: inequality, we have $\left\langle a,b\right\rangle \leq \left\Vert
2170: a\right\Vert \left\Vert b\right\Vert $. The left inequality may be useful
2171: when one assumes that $\varepsilon _{1}\in (0,1].$ In that case, from (\ref%
2172: {ch5.1.12}), we obtain%
2173: \begin{equation}
2174: -\left\Vert a\right\Vert \left\Vert b\right\Vert \leq \left( 2\varepsilon
2175: _{1}^{2}-1\right) \left\Vert a\right\Vert \left\Vert b\right\Vert \leq
2176: \left\langle a,b\right\rangle \label{ch5.1.13}
2177: \end{equation}%
2178: provided $\varepsilon _{1}\left\Vert x\right\Vert \left\Vert a\right\Vert
2179: \leq \left\langle x,a\right\rangle $ and $\varepsilon _{1}\left\Vert
2180: x\right\Vert \left\Vert b\right\Vert \leq \left\langle x,b\right\rangle ,$
2181: which is a refinement of Schwarz's inequality%
2182: \begin{equation*}
2183: -\left\Vert a\right\Vert \left\Vert b\right\Vert \leq \left\langle
2184: a,b\right\rangle .
2185: \end{equation*}
2186:
2187: In the complex case, apparently independent of Richard, M.L. Buzano obtained
2188: in \cite{B} the following inequality%
2189: \begin{equation}
2190: \left\vert \left\langle x,a\right\rangle \left\langle x,b\right\rangle
2191: \right\vert \leq \frac{\left\Vert a\right\Vert \left\Vert b\right\Vert
2192: +\left\vert \left\langle a,b\right\rangle \right\vert }{2}\cdot \left\Vert
2193: x\right\Vert ^{2}, \label{ch5.1.14}
2194: \end{equation}%
2195: provided $x,a,b$ are vectors in the complex inner product space $\left(
2196: H;\left\langle \cdot ,\cdot \right\rangle \right) .$
2197:
2198: In the same paper \cite{P}, Precupanu, without mentioning Buzano's name in
2199: relation to the inequality (\ref{ch5.1.14}), observed that, on utilising (%
2200: \ref{ch5.1.14}), one may obtain the following result of Moore type:
2201:
2202: \begin{theorem}[Precupanu, 1976]
2203: \label{ch5.t1.4}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
2204: \right) $ be a (real or) complex inner product space. If $x,a,b\in H$ are
2205: such that%
2206: \begin{equation}
2207: \left\vert \left\langle x,a\right\rangle \right\vert \geq \left(
2208: 1-\varepsilon \right) \left\Vert x\right\Vert \left\Vert a\right\Vert
2209: ,\qquad \left\vert \left\langle x,b\right\rangle \right\vert \geq \left(
2210: 1-\varepsilon \right) \left\Vert x\right\Vert \left\Vert b\right\Vert ,
2211: \label{ch5.1.15}
2212: \end{equation}%
2213: then%
2214: \begin{equation}
2215: \left\vert \left\langle a,b\right\rangle \right\vert \geq \left(
2216: 1-4\varepsilon +2\varepsilon ^{2}\right) \left\Vert a\right\Vert \left\Vert
2217: b\right\Vert . \label{ch5.1.16}
2218: \end{equation}
2219: \end{theorem}
2220:
2221: Note that the above theorem is useful when, for $\varepsilon \in (0,1],$ the
2222: quantity $1-4\varepsilon +2\varepsilon ^{2}>0,$ i.e., $\varepsilon \in
2223: \left( 0,1-\frac{\sqrt{2}}{2}\right] .$
2224:
2225: \begin{remark}
2226: When the space is real, the inequality (\ref{ch5.1.16}) provides a better
2227: lower bound for $\left\vert \left\langle a,b\right\rangle \right\vert $ than
2228: the second bound in Moore's result (\ref{ch5.1.9}). However, it is not known
2229: if the first bound in (\ref{ch5.1.9}) remains valid for the case of complex
2230: spaces. From Moore's original proof, apparently, the fact that the space $%
2231: \left( H;\left\langle \cdot ,\cdot \right\rangle \right) $ is real plays an
2232: essential role.
2233: \end{remark}
2234:
2235: Before we point out some new results for orthonormal families of vectors in
2236: real or complex inner product spaces, we state the following result that
2237: complements the Moore type results outlined above for real spaces \cite%
2238: {DRAG3}:
2239:
2240: \begin{theorem}[Dragomir, 2004]
2241: \label{ch5.t1.5}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
2242: \right) $ be a real inner product space and $a,b,x,y\in H\backslash \left\{
2243: 0\right\} .$
2244:
2245: \begin{enumerate}
2246: \item[(i)] If there exist $\delta _{1},\delta _{2}\in (0,1]$ such that%
2247: \begin{equation*}
2248: \frac{\left\langle x,a\right\rangle }{\left\Vert x\right\Vert \left\Vert
2249: a\right\Vert }\geq \delta _{1},\qquad \frac{\left\langle y,a\right\rangle }{%
2250: \left\Vert y\right\Vert \left\Vert a\right\Vert }\geq \delta _{2}
2251: \end{equation*}%
2252: and $\delta _{1}+\delta _{2}\geq 1,$ then%
2253: \begin{equation}
2254: \frac{\left\langle x,y\right\rangle }{\left\Vert x\right\Vert \left\Vert
2255: y\right\Vert }\geq \frac{1}{2}\left( \delta _{1}+\delta _{2}\right) ^{2}-%
2256: \frac{3}{2}\qquad \left( \geq -1\right) . \label{ch5.1.17}
2257: \end{equation}
2258:
2259: \item[(ii)] If there exist $\mu _{1}\left( \mu _{2}\right) \in \mathbb{R}$
2260: such that%
2261: \begin{equation*}
2262: \mu _{1}\left\Vert a\right\Vert \left\Vert b\right\Vert \leq \frac{%
2263: \left\langle x,a\right\rangle \left\langle x,b\right\rangle }{\left\Vert
2264: x\right\Vert ^{2}}\left( \leq \mu _{2}\left\Vert a\right\Vert \left\Vert
2265: b\right\Vert \right)
2266: \end{equation*}%
2267: and $1\geq \mu _{1}\geq 0$ $\left( -1\leq \mu _{2}\leq 0\right) ,$ then%
2268: \begin{equation}
2269: \left[ -1\leq \right] 2\mu _{1}-1\leq \frac{\left\langle a,b\right\rangle }{%
2270: \left\Vert a\right\Vert \left\Vert b\right\Vert }\left( \leq 2\mu _{2}+1%
2271: \left[ \leq 1\right] \right) . \label{ch5.1.18}
2272: \end{equation}
2273: \end{enumerate}
2274: \end{theorem}
2275:
2276: The proof is obvious by the inequalities (\ref{ch5.1.6}) and (\ref{ch5.1.10}%
2277: ). We omit the details.
2278:
2279: \subsection{Inequalities for Orthonormal Families}
2280:
2281: The following result may be stated \cite{DRAG3}.
2282:
2283: \begin{theorem}[Dragomir, 2004]
2284: \label{ch5.t2.1}Let $\left\{ e_{i}\right\} _{i\in I}$ and $\left\{
2285: f_{j}\right\} _{j\in J}$ be two finite families of orthonormal vectors in $%
2286: \left( H;\left\langle \cdot ,\cdot \right\rangle \right) .$ For any $x,y\in
2287: H\backslash \left\{ 0\right\} $ one has the inequality%
2288: \begin{multline}
2289: \left\vert \sum_{i\in I}\left\langle x,e_{i}\right\rangle \left\langle
2290: e_{i},y\right\rangle +\sum_{j\in J}\left\langle x,f_{j}\right\rangle
2291: \left\langle f_{j},y\right\rangle \right. \label{ch5.2.1} \\
2292: -\left. 2\sum_{i\in I,j\in J}\left\langle x,e_{i}\right\rangle \left\langle
2293: f_{j},y\right\rangle \left\langle e_{i},f_{j}\right\rangle -\frac{1}{2}%
2294: \left\langle x,y\right\rangle \right\vert \leq \frac{1}{2}\left\Vert
2295: x\right\Vert \left\Vert y\right\Vert .
2296: \end{multline}%
2297: The case of equality holds in (\ref{ch5.2.1}) if and only if there exists a $%
2298: \lambda \in \mathbb{K}$ such that%
2299: \begin{equation}
2300: x-\lambda y=2\left( \sum_{i\in I}\left\langle x,e_{i}\right\rangle
2301: e_{i}-\lambda \sum_{j\in J}\left\langle y,f_{j}\right\rangle f_{j}\right) .
2302: \label{ch5.2.2}
2303: \end{equation}
2304: \end{theorem}
2305:
2306: \begin{proof}
2307: We follow the proof in \cite{DRAG3}.
2308:
2309: We know that, if $u,v\in H,$ $v\neq 0,$ then%
2310: \begin{equation}
2311: \left\Vert u-\frac{\left\langle u,v\right\rangle }{\left\Vert v\right\Vert
2312: ^{2}}\cdot v\right\Vert ^{2}=\frac{\left\Vert u\right\Vert ^{2}\left\Vert
2313: v\right\Vert ^{2}-\left\vert \left\langle u,v\right\rangle \right\vert ^{2}}{%
2314: \left\Vert v\right\Vert ^{2}} \label{ch5.2.3}
2315: \end{equation}%
2316: showing that, in Schwarz's inequality%
2317: \begin{equation}
2318: \left\vert \left\langle u,v\right\rangle \right\vert ^{2}\leq \left\Vert
2319: u\right\Vert ^{2}\left\Vert v\right\Vert ^{2}, \label{ch5.2.4}
2320: \end{equation}%
2321: the case of equality, for $v\neq 0,$ holds if and only if%
2322: \begin{equation}
2323: u=\frac{\left\langle u,v\right\rangle }{\left\Vert v\right\Vert ^{2}}\cdot
2324: v,\ \label{ch5.2.5}
2325: \end{equation}%
2326: i.e. there exists a $\lambda \in \mathbb{R}$ \ such that $u=\lambda v.$
2327:
2328: Now, let $u:=2\sum_{i\in I}\left\langle x,e_{i}\right\rangle e_{i}-x$ and $%
2329: v:=2\sum_{j\in J}\left\langle y,f_{j}\right\rangle f_{j}-y.$
2330:
2331: Observe that%
2332: \begin{align*}
2333: \left\Vert u\right\Vert ^{2}& =\left\Vert 2\sum_{i\in I}\left\langle
2334: x,e_{i}\right\rangle e_{i}\right\Vert ^{2}-4\func{Re}\left\langle \sum_{i\in
2335: I}\left\langle x,e_{i}\right\rangle e_{i},x\right\rangle +\left\Vert
2336: x\right\Vert ^{2} \\
2337: & =4\sum_{i\in I}\left\vert \left\langle x,e_{i}\right\rangle \right\vert
2338: ^{2}-4\sum_{i\in I}\left\vert \left\langle x,e_{i}\right\rangle \right\vert
2339: ^{2}+\left\Vert x\right\Vert ^{2}=\left\Vert x\right\Vert ^{2},
2340: \end{align*}%
2341: and, similarly%
2342: \begin{equation*}
2343: \left\Vert v\right\Vert ^{2}=\left\Vert y\right\Vert ^{2}.
2344: \end{equation*}%
2345: Also,%
2346: \begin{multline*}
2347: \left\langle u,v\right\rangle =4\sum_{i\in I,j\in J}\left\langle
2348: x,e_{i}\right\rangle \left\langle f_{j},y\right\rangle \left\langle
2349: e_{i},f_{j}\right\rangle +\left\langle x,y\right\rangle \\
2350: -2\sum_{i\in I}\left\langle x,e_{i}\right\rangle \left\langle
2351: e_{i},y\right\rangle -2\sum_{j\in J}\left\langle x,f_{j}\right\rangle
2352: \left\langle f_{j},y\right\rangle .
2353: \end{multline*}%
2354: Therefore, by Schwarz's inequality (\ref{ch5.2.4}) we deduce the desired
2355: inequality (\ref{ch5.2.1}). By (\ref{ch5.2.5}), the case of equality holds
2356: in (\ref{ch5.2.1}) if and only if there exists a $\lambda \in \mathbb{K}$
2357: such that%
2358: \begin{equation*}
2359: 2\sum_{i\in I}\left\langle x,e_{i}\right\rangle e_{i}-x=\lambda \left(
2360: 2\sum_{j\in J}\left\langle y,f_{j}\right\rangle f_{j}-y\right) ,
2361: \end{equation*}%
2362: which is equivalent to (\ref{ch5.2.2}).
2363: \end{proof}
2364:
2365: \begin{remark}
2366: If in (\ref{ch5.2.2}) we choose $x=y,$ then we get the inequality:%
2367: \begin{multline}
2368: \left\vert \sum_{i\in I}\left\vert \left\langle x,e_{i}\right\rangle
2369: \right\vert ^{2}+\sum_{j\in J}\left\vert \left\langle x,f_{j}\right\rangle
2370: \right\vert ^{2}\right. \label{ch5.2.6} \\
2371: -2\left. \sum_{i\in I,j\in J}\left\langle x,e_{i}\right\rangle \left\langle
2372: f_{j},x\right\rangle \left\langle e_{i},f_{j}\right\rangle -\frac{1}{2}%
2373: \left\Vert x\right\Vert ^{2}\right\vert \leq \frac{1}{2}\left\Vert
2374: x\right\Vert ^{2}
2375: \end{multline}%
2376: for any $x\in H.$
2377:
2378: If in the above theorem we assume that $I=J$ and $f_{i}=e_{i},$ $i\in I,$
2379: then we get from (\ref{ch5.2.1}) the Schwarz inequality $\left\vert
2380: \left\langle x,y\right\rangle \right\vert \leq \left\Vert x\right\Vert
2381: \left\Vert y\right\Vert .$
2382:
2383: If $I\cap J=\varnothing ,$ $I\cup J=K,$ $g_{k}=e_{k},$ $k\in I,$ $%
2384: g_{k}=f_{k},$ $k\in J$ and $\left\{ g_{k}\right\} _{k\in K}$ is orthonormal,
2385: then from (\ref{ch5.2.1}) we get:%
2386: \begin{equation}
2387: \left\vert \sum_{k\in K}\left\langle x,g_{k}\right\rangle \left\langle
2388: g_{k},y\right\rangle -\frac{1}{2}\left\langle x,y\right\rangle \right\vert
2389: \leq \frac{1}{2}\left\Vert x\right\Vert \left\Vert y\right\Vert ,\qquad
2390: x,y\in H \label{ch5.2.7}
2391: \end{equation}%
2392: which has been obtained earlier by the author in \cite{DRAG1}.
2393: \end{remark}
2394:
2395: If $I$ and $J$ reduce to one element, namely $e_{1}=\frac{e}{\left\Vert
2396: e\right\Vert },$ $f_{1}=\frac{f}{\left\Vert f\right\Vert }$ with $e,f\neq 0,$
2397: then from (\ref{ch5.2.1}) we get%
2398: \begin{multline}
2399: \left\vert \frac{\left\langle x,e\right\rangle \left\langle e,y\right\rangle
2400: }{\left\Vert e\right\Vert ^{2}}+\frac{\left\langle x,f\right\rangle
2401: \left\langle f,y\right\rangle }{\left\Vert f\right\Vert ^{2}}-2\cdot \frac{%
2402: \left\langle x,e\right\rangle \left\langle f,y\right\rangle \left\langle
2403: e,f\right\rangle }{\left\Vert e\right\Vert ^{2}\left\Vert f\right\Vert ^{2}}-%
2404: \frac{1}{2}\left\langle x,y\right\rangle \right\vert \label{ch5.2.8} \\
2405: \leq \frac{1}{2}\left\Vert x\right\Vert \left\Vert y\right\Vert ,\qquad
2406: x,y\in H
2407: \end{multline}%
2408: which is the corresponding complex version of Precupanu's inequality (\ref%
2409: {ch5.1.1}).
2410:
2411: If in (\ref{ch5.2.8}) we assume that $x=y,$ then we get%
2412: \begin{multline}
2413: \left\vert \frac{\left\vert \left\langle x,e\right\rangle \right\vert ^{2}}{%
2414: \left\Vert e\right\Vert ^{2}}+\frac{\left\vert \left\langle x,f\right\rangle
2415: \right\vert ^{2}}{\left\Vert f\right\Vert ^{2}}-2\cdot \frac{\left\langle
2416: x,e\right\rangle \left\langle f,e\right\rangle \left\langle e,f\right\rangle
2417: }{\left\Vert e\right\Vert ^{2}\left\Vert f\right\Vert ^{2}}-\frac{1}{2}%
2418: \left\Vert x\right\Vert ^{2}\right\vert \label{ch5.2.9} \\
2419: \leq \frac{1}{2}\left\Vert x\right\Vert ^{2}.
2420: \end{multline}
2421:
2422: The following corollary may be stated \cite{DRAG3}:
2423:
2424: \begin{corollary}[Dragomir, 2004]
2425: \label{ch5.c2.3}With the assumptions of Theorem \ref{ch5.t2.1}, we have:%
2426: \begin{align}
2427: & \left\vert \sum_{i\in I}\left\langle x,e_{i}\right\rangle \left\langle
2428: e_{i},y\right\rangle +\sum_{j\in J}\left\langle x,f_{j}\right\rangle
2429: \left\langle f_{j},y\right\rangle \right. \label{ch5.2.10} \\
2430: & \qquad \qquad -\left. 2\sum_{i\in I,j\in J}\left\langle
2431: x,e_{i}\right\rangle \left\langle f_{j},y\right\rangle \left\langle
2432: e_{i},f_{j}\right\rangle \right\vert \notag \\
2433: & \leq \frac{1}{2}\left\vert \left\langle x,y\right\rangle \right\vert
2434: +\left\vert \sum_{i\in I}\left\langle x,e_{i}\right\rangle \left\langle
2435: e_{i},y\right\rangle +\sum_{j\in J}\left\langle x,f_{j}\right\rangle
2436: \left\langle f_{j},y\right\rangle \right. \notag \\
2437: & \qquad \qquad \qquad -\left. 2\sum_{i\in I,j\in J}\left\langle
2438: x,e_{i}\right\rangle \left\langle f_{j},y\right\rangle \left\langle
2439: e_{i},f_{j}\right\rangle -\frac{1}{2}\left\vert \left\langle
2440: x,y\right\rangle \right\vert \right\vert \notag \\
2441: & \leq \frac{1}{2}\left[ \left\vert \left\langle x,y\right\rangle
2442: \right\vert +\left\Vert x\right\Vert \left\Vert y\right\Vert \right] .
2443: \notag
2444: \end{align}
2445: \end{corollary}
2446:
2447: \begin{proof}
2448: The first inequality follows by the triangle inequality for the modulus. The
2449: second inequality follows by (\ref{ch5.2.1}) on adding the quantity $\frac{1%
2450: }{2}\left\vert \left\langle x,y\right\rangle \right\vert $ on both sides.
2451: \end{proof}
2452:
2453: \begin{remark}
2454:
2455: \begin{enumerate}
2456: \item \label{ch5.r2.4}If we choose in (\ref{ch5.2.10}), $x=y,$ then we get:%
2457: \begin{align}
2458: & \left\vert \sum_{i\in I}\left\vert \left\langle x,e_{i}\right\rangle
2459: \right\vert ^{2}+\sum_{j\in J}\left\vert \left\langle x,f_{j}\right\rangle
2460: \right\vert ^{2}\right. \label{ch5.2.11} \\
2461: & \qquad -\left. 2\sum_{i\in I,j\in J}\left\langle x,e_{i}\right\rangle
2462: \left\langle f_{j},x\right\rangle \left\langle e_{i},f_{j}\right\rangle
2463: \right\vert \notag \\
2464: & \leq \left\vert \sum_{i\in I}\left\vert \left\langle x,e_{i}\right\rangle
2465: \right\vert ^{2}+\sum_{j\in J}\left\vert \left\langle x,f_{j}\right\rangle
2466: \right\vert ^{2}\right. \notag \\
2467: & \qquad -\left. 2\sum_{i\in I,j\in J}\left\langle x,e_{i}\right\rangle
2468: \left\langle f_{j},x\right\rangle \left\langle e_{i},f_{j}\right\rangle -%
2469: \frac{1}{2}\left\Vert x\right\Vert ^{2}\right\vert +\frac{1}{2}\left\Vert
2470: x\right\Vert ^{2} \notag \\
2471: & \leq \left\Vert x\right\Vert ^{2}. \notag
2472: \end{align}%
2473: We observe that (\ref{ch5.2.11}) will generate Bessel's inequality if $%
2474: \left\{ e_{i}\right\} _{i\in I},$ $\left\{ f_{j}\right\} _{j\in J}$ are
2475: disjoint parts of a larger orthonormal family.
2476:
2477: \item From (\ref{ch5.2.8}) one can obtain:%
2478: \begin{multline}
2479: \left\vert \frac{\left\langle x,e\right\rangle \left\langle e,y\right\rangle
2480: }{\left\Vert e\right\Vert ^{2}}+\frac{\left\langle x,f\right\rangle
2481: \left\langle f,y\right\rangle }{\left\Vert f\right\Vert ^{2}}-2\cdot \frac{%
2482: \left\langle x,e\right\rangle \left\langle f,y\right\rangle \left\langle
2483: e,f\right\rangle }{\left\Vert e\right\Vert ^{2}\left\Vert f\right\Vert ^{2}}%
2484: \right\vert \label{ch5.2.12} \\
2485: \leq \frac{1}{2}\left[ \left\Vert x\right\Vert \left\Vert y\right\Vert
2486: +\left\vert \left\langle x,y\right\rangle \right\vert \right]
2487: \end{multline}%
2488: and in particular
2489: \begin{equation}
2490: \left\vert \frac{\left\vert \left\langle x,e\right\rangle \right\vert ^{2}}{%
2491: \left\Vert e\right\Vert ^{2}}+\frac{\left\vert \left\langle x,f\right\rangle
2492: \right\vert ^{2}}{\left\Vert f\right\Vert ^{2}}-2\cdot \frac{\left\langle
2493: x,e\right\rangle \left\langle f,e\right\rangle \left\langle e,f\right\rangle
2494: }{\left\Vert e\right\Vert ^{2}\left\Vert f\right\Vert ^{2}}\right\vert \leq
2495: \left\Vert x\right\Vert ^{2}, \label{ch5.2.13}
2496: \end{equation}%
2497: for any $x,y\in H.$
2498: \end{enumerate}
2499: \end{remark}
2500:
2501: The case of real inner products will provide a natural genearlization for
2502: Precupanu's inequality (\ref{ch5.1.1}) \cite{DRAG3}:
2503:
2504: \begin{corollary}[Dragomir, 2004]
2505: \label{ch5.c2.5}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
2506: \right) $ be a real inner product space and $\left\{ e_{i}\right\} _{i\in
2507: I}, $ $\left\{ f_{j}\right\} _{j\in J}$ two finite families of orthonormal
2508: vectors in $\left( H;\left\langle \cdot ,\cdot \right\rangle \right) .$ For
2509: any $x,y\in H\backslash \left\{ 0\right\} $ one has the double inequality:%
2510: \begin{align}
2511: \frac{1}{2}\left[ \left\vert \left\langle x,y\right\rangle \right\vert
2512: -\left\Vert x\right\Vert \left\Vert y\right\Vert \right] & \leq \sum_{i\in
2513: I}\left\langle x,e_{i}\right\rangle \left\langle y,e_{i}\right\rangle
2514: +\sum_{j\in J}\left\langle x,f_{j}\right\rangle \left\langle
2515: y,f_{j}\right\rangle \label{ch5.2.14} \\
2516: & \qquad \qquad -2\sum_{i\in I,j\in J}\left\langle x,e_{i}\right\rangle
2517: \left\langle y,f_{j}\right\rangle \left\langle e_{i},f_{j}\right\rangle
2518: \notag \\
2519: & \leq \frac{1}{2}\left[ \left\Vert x\right\Vert \left\Vert y\right\Vert
2520: +\left\vert \left\langle x,y\right\rangle \right\vert \right] . \notag
2521: \end{align}%
2522: In particular, we have%
2523: \begin{align}
2524: 0& \leq \sum_{i\in I}\left\langle x,e_{i}\right\rangle ^{2}+\sum_{j\in
2525: J}\left\langle x,f_{j}\right\rangle ^{2}-2\sum_{i\in I,j\in J}\left\langle
2526: x,e_{i}\right\rangle \left\langle x,f_{j}\right\rangle \left\langle
2527: e_{i},f_{j}\right\rangle \label{ch5.2.15} \\
2528: & \leq \left\Vert x\right\Vert ^{2}, \notag
2529: \end{align}%
2530: for any $x\in H.$
2531: \end{corollary}
2532:
2533: \begin{remark}
2534: Similar particular inequalities to those incorporated in (\ref{ch5.2.7}) -- (%
2535: \ref{ch5.2.13}) may be stated, but we omit them.
2536: \end{remark}
2537:
2538: \subsection{Refinements of Kurepa's Inequality\label{ch5.s3}}
2539:
2540: The following result may be stated \cite{DRAG3}.
2541:
2542: \begin{theorem}[Dragomir, 2004]
2543: \label{ch5.t3.1}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
2544: \right) $ be a real inner product space and $\left\{ e_{i}\right\} _{i\in
2545: I},\left\{ f_{j}\right\} _{j\in J}$ two finite families in $H.$ If $\left(
2546: H_{\mathbb{C}};\left\langle \cdot ,\cdot \right\rangle _{\mathbb{C}}\right) $
2547: is the complexification of $\left( H;\left\langle \cdot ,\cdot \right\rangle
2548: \right) ,$ then for any $w\in H_{\mathbb{C}},$ we have the inequalities%
2549: \begin{align}
2550: & \left\vert \sum_{i\in I}\left\langle w,e_{i}\right\rangle _{\mathbb{C}%
2551: }^{2}+\sum_{j\in J}\left\langle w,f_{j}\right\rangle _{\mathbb{C}%
2552: }^{2}-2\sum_{i\in I,j\in J}\left\langle w,e_{i}\right\rangle _{\mathbb{C}%
2553: }\left\langle w,f_{j}\right\rangle _{\mathbb{C}}\left\langle
2554: e_{i},f_{j}\right\rangle \right\vert \label{ch5.3.3} \\
2555: & \leq \frac{1}{2}\left\vert \left\langle w,\bar{w}\right\rangle _{\mathbb{C}%
2556: }\right\vert +\left\vert \sum_{i\in I}\left\langle w,e_{i}\right\rangle _{%
2557: \mathbb{C}}^{2}+\sum_{j\in J}\left\langle w,f_{j}\right\rangle _{\mathbb{C}%
2558: }^{2}\right. \notag \\
2559: & \qquad \qquad \qquad -\left. 2\sum_{i\in I,j\in J}\left\langle
2560: w,e_{i}\right\rangle _{\mathbb{C}}\left\langle w,f_{j}\right\rangle _{%
2561: \mathbb{C}}\left\langle e_{i},f_{j}\right\rangle -\frac{1}{2}\left\langle w,%
2562: \bar{w}\right\rangle _{\mathbb{C}}\right\vert \notag \\
2563: & \leq \frac{1}{2}\left[ \left\Vert w\right\Vert _{\mathbb{C}%
2564: }^{2}+\left\vert \left\langle w,\bar{w}\right\rangle _{\mathbb{C}%
2565: }\right\vert \right] \leq \left\Vert w\right\Vert _{\mathbb{C}}^{2}. \notag
2566: \end{align}
2567: \end{theorem}
2568:
2569: \begin{proof}
2570: Define $g_{j}\in H_{\mathbb{C}},$ $g_{j}:=\left( e_{j},0\right) ,$ $j\in I.$
2571: For any $k,j\in I$ we have%
2572: \begin{equation*}
2573: \left\langle g_{k},g_{j}\right\rangle _{\mathbb{C}}=\left\langle \left(
2574: e_{k},0\right) ,\left( e_{j},0\right) \right\rangle _{\mathbb{C}%
2575: }=\left\langle e_{k},e_{j}\right\rangle =\delta _{kj},
2576: \end{equation*}%
2577: therefore $\left\{ g_{j}\right\} _{j\in I}$ is an orthonormal family in $%
2578: \left( H_{\mathbb{C}};\left\langle \cdot ,\cdot \right\rangle _{\mathbb{C}%
2579: }\right) .$
2580:
2581: If we apply Corollary \ref{ch5.c2.3} for $\left( H_{\mathbb{C}};\left\langle
2582: \cdot ,\cdot \right\rangle _{\mathbb{C}}\right) ,$ $x=w,$ $y=\bar{w},$ we
2583: may write:%
2584: \begin{align}
2585: & \left\vert \sum_{i\in I}\left\langle w,e_{i}\right\rangle _{\mathbb{C}%
2586: }\left\langle e_{i},\bar{w}\right\rangle _{\mathbb{C}}+\sum_{j\in
2587: J}\left\langle w,f_{j}\right\rangle \left\langle f_{j},\bar{w}\right\rangle
2588: \right. \label{ch5.3.4} \\
2589: & \qquad \qquad \qquad -\left. 2\sum_{i\in I,j\in J}\left\langle
2590: w,e_{i}\right\rangle _{\mathbb{C}}\left\langle f_{j},\overline{w}%
2591: \right\rangle _{\mathbb{C}}\left\langle e_{i},f_{j}\right\rangle \right\vert
2592: \notag \\
2593: & \leq \frac{1}{2}\left\Vert w\right\Vert _{\mathbb{C}}\left\Vert \bar{w}%
2594: \right\Vert _{\mathbb{C}}+\left\vert \sum_{i\in I}\left\langle
2595: w,e_{i}\right\rangle _{\mathbb{C}}\left\langle e_{i},\bar{w}\right\rangle _{%
2596: \mathbb{C}}+\sum_{j\in J}\left\langle w,f_{j}\right\rangle \left\langle
2597: f_{j},\bar{w}\right\rangle \right. \notag \\
2598: & \qquad \qquad \qquad -\left. 2\sum_{i\in I,j\in J}\left\langle
2599: w,e_{i}\right\rangle _{\mathbb{C}}\left\langle f_{j},\overline{w}%
2600: \right\rangle _{\mathbb{C}}\left\langle e_{i},f_{j}\right\rangle -\frac{1}{2}%
2601: \left\langle w,\bar{w}\right\rangle _{\mathbb{C}}\right\vert \notag \\
2602: & \leq \frac{1}{2}\left[ \left\vert \left\langle w,\bar{w}\right\rangle _{%
2603: \mathbb{C}}\right\vert +\left\Vert w\right\Vert _{\mathbb{C}}\left\Vert \bar{%
2604: w}\right\Vert _{\mathbb{C}}\right] . \notag
2605: \end{align}%
2606: However, for $w:=\left( x,y\right) \in H_{\mathbb{C}}$, we have $\bar{w}%
2607: =\left( x,-y\right) $ and%
2608: \begin{equation*}
2609: \left\langle e_{j},\bar{w}\right\rangle _{\mathbb{C}}=\left\langle \left(
2610: e_{j},0\right) ,\left( x,-y\right) \right\rangle _{\mathbb{C}}=\left\langle
2611: e_{j},x\right\rangle +i\left\langle e_{j},y\right\rangle
2612: \end{equation*}%
2613: and%
2614: \begin{equation*}
2615: \left\langle w,e_{j}\right\rangle _{\mathbb{C}}=\left\langle \left(
2616: x,y\right) ,\left( e_{j},0\right) \right\rangle _{\mathbb{C}}=\left\langle
2617: x,e_{j}\right\rangle +i\left\langle e_{j},y\right\rangle
2618: \end{equation*}%
2619: showing that $\left\langle e_{j},\bar{w}\right\rangle =\left\langle
2620: w,e_{j}\right\rangle $ for any $j\in I$. A similar relation is true for $%
2621: f_{j}$ and since%
2622: \begin{equation*}
2623: \left\Vert w\right\Vert _{\mathbb{C}}=\left\Vert \bar{w}\right\Vert _{%
2624: \mathbb{C}}=\left( \left\Vert x\right\Vert ^{2}+\left\Vert y\right\Vert
2625: ^{2}\right) ^{\frac{1}{2}},
2626: \end{equation*}%
2627: hence from (\ref{ch5.3.4}) we deduce the desired inequality (\ref{ch5.3.3}).
2628: \end{proof}
2629:
2630: \begin{remark}
2631: It is obvious that, if one family, say $\left\{ f_{j}\right\} _{j\in J}$ is
2632: empty, then, on observing that all sums $\sum_{j\in J}$ should be zero, from
2633: (\ref{ch5.3.3}) one would get \cite{DRAG1}%
2634: \begin{align}
2635: & \left\vert \sum_{i\in I}\left\langle w,e_{i}\right\rangle _{\mathbb{C}%
2636: }^{2}\right\vert \label{ch5.3.5} \\
2637: & \leq \frac{1}{2}\left\vert \left\langle w,\bar{w}\right\rangle _{\mathbb{C}%
2638: }\right\vert +\left\vert \sum_{i\in I}\left\langle w,e_{i}\right\rangle _{%
2639: \mathbb{C}}^{2}-\frac{1}{2}\left\langle w,\bar{w}\right\rangle _{\mathbb{C}%
2640: }\right\vert \notag \\
2641: & \leq \frac{1}{2}\left[ \left\Vert w\right\Vert _{\mathbb{C}%
2642: }^{2}+\left\vert \left\langle w,\bar{w}\right\rangle _{\mathbb{C}%
2643: }\right\vert \right] \leq \left\Vert w\right\Vert _{\mathbb{C}}^{2}. \notag
2644: \end{align}%
2645: If in (\ref{ch5.3.5}) one assumes that the family $\left\{ e_{i}\right\}
2646: _{i\in I}$ contains only one element $e=\frac{a}{\left\Vert a\right\Vert }%
2647: ,a\neq 0,$ then by selecting $w=z,$ one would deduce (\ref{ch3.3.5}), which
2648: is a refinement for Kurepa's inequality.
2649: \end{remark}
2650:
2651: \section{Some New Refinements of the Schwarz Inequality}
2652:
2653: \subsection{Refinements}
2654:
2655: The following result holds \cite{DRAG01}.
2656:
2657: \begin{theorem}[Dragomir, 2004]
2658: \label{ch2.t2.1}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
2659: \right) $ be an inner product space over the real or complex number field $%
2660: \mathbb{K}$ and $r_{1},r_{2}>0.$ If $x,y\in H$ are with the property that
2661: \begin{equation}
2662: \left\Vert x-y\right\Vert \geq r_{2}\geq r_{1}\geq \left\vert \left\Vert
2663: x\right\Vert -\left\Vert y\right\Vert \right\vert , \label{ch2.2.1}
2664: \end{equation}%
2665: then we have the following refinement of Schwarz's inequality
2666: \begin{equation}
2667: \left\Vert x\right\Vert \left\Vert y\right\Vert -\func{Re}\left\langle
2668: x,y\right\rangle \geq \frac{1}{2}\left( r_{2}^{2}-r_{1}^{2}\right) \left(
2669: \geq 0\right) . \label{ch2.2.2}
2670: \end{equation}%
2671: The constant $\frac{1}{2}$ \ is best possible in the sense that it cannot be
2672: replaced by a larger quantity.
2673: \end{theorem}
2674:
2675: \begin{proof}
2676: From the first inequality in (\ref{ch2.2.1}) we have
2677: \begin{equation}
2678: \left\| x\right\| ^{2}+\left\| y\right\| ^{2}\geq r_{2}^{2}+2\func{Re}%
2679: \left\langle x,y\right\rangle . \label{ch2.2.3}
2680: \end{equation}
2681: Subtracting in (\ref{ch2.2.3}) the quantity $2\left\| x\right\| \left\|
2682: y\right\| ,$ we get
2683: \begin{equation}
2684: \left( \left\| x\right\| -\left\| y\right\| \right) ^{2}\geq
2685: r_{2}^{2}-2\left( \left\| x\right\| \left\| y\right\| -\func{Re}\left\langle
2686: x,y\right\rangle \right) . \label{ch2.2.4}
2687: \end{equation}
2688: Since, by the second inequality in (\ref{ch2.2.1}) we have
2689: \begin{equation}
2690: r_{1}^{2}\geq \left( \left\| x\right\| -\left\| y\right\| \right) ^{2},
2691: \label{ch2.2.5}
2692: \end{equation}
2693: hence from (\ref{ch2.2.4}) and (\ref{ch2.2.5}) we deduce the desired
2694: inequality (\ref{ch2.2.2}).
2695:
2696: To prove the sharpness of the constant $\frac{1}{2}$ in (\ref{ch2.2.2}), let
2697: us assume that there is a constant $C>0$ such that
2698: \begin{equation}
2699: \left\| x\right\| \left\| y\right\| -\func{Re}\left\langle x,y\right\rangle
2700: \geq C\left( r_{2}^{2}-r_{1}^{2}\right) , \label{ch2.2.6}
2701: \end{equation}
2702: provided that $x$ and $y$ satisfy (\ref{ch2.2.1}).
2703:
2704: Let $e\in H$ with $\left\Vert e\right\Vert =1$ and for $r_{2}>r_{1}>0,$
2705: define
2706: \begin{equation}
2707: x=\frac{r_{2}+r_{1}}{2}\cdot e\text{ \ and \ }y=\frac{r_{1}-r_{2}}{2}\cdot e.
2708: \label{ch2.2.7}
2709: \end{equation}%
2710: Then
2711: \begin{equation*}
2712: \left\Vert x-y\right\Vert =r_{2}\text{ and }\left\vert \left\Vert
2713: x\right\Vert -\left\Vert y\right\Vert \right\vert =r_{1},
2714: \end{equation*}%
2715: showing that the condition (\ref{ch2.2.1}) is fulfilled with equality.
2716:
2717: If we replace $x$ and $y$ as defined in (\ref{ch2.2.7}) into the inequality (%
2718: \ref{ch2.2.6}), then we get
2719: \begin{equation*}
2720: \frac{r_{2}^{2}-r_{1}^{2}}{2}\geq C\left( r_{2}^{2}-r_{1}^{2}\right) ,
2721: \end{equation*}%
2722: which implies that $C\leq \frac{1}{2},$ and the theorem is completely proved.
2723: \end{proof}
2724:
2725: The following corollary holds.
2726:
2727: \begin{corollary}
2728: \label{ch2.c2.2}With the assumptions of Theorem \ref{ch2.t2.1}, we have the
2729: inequality:
2730: \begin{equation}
2731: \left\Vert x\right\Vert +\left\Vert y\right\Vert -\frac{\sqrt{2}}{2}%
2732: \left\Vert x+y\right\Vert \geq \frac{\sqrt{2}}{2}\sqrt{r_{2}^{2}-r_{1}^{2}}.
2733: \label{ch2.2.8}
2734: \end{equation}
2735: \end{corollary}
2736:
2737: \begin{proof}
2738: We have, by (\ref{ch2.2.2}), that
2739: \begin{equation*}
2740: \left( \left\| x\right\| +\left\| y\right\| \right) ^{2}-\left\| x+y\right\|
2741: ^{2}=2\left( \left\| x\right\| \left\| y\right\| -\func{Re}\left\langle
2742: x,y\right\rangle \right) \geq r_{2}^{2}-r_{1}^{2}\geq 0
2743: \end{equation*}
2744: which gives
2745: \begin{equation}
2746: \left( \left\| x\right\| +\left\| y\right\| \right) ^{2}\geq \left\|
2747: x+y\right\| ^{2}+\left( \sqrt{r_{2}^{2}-r_{1}^{2}}\right) ^{2}.
2748: \label{ch2.2.9}
2749: \end{equation}
2750: By making use of the elementary inequality
2751: \begin{equation*}
2752: 2\left( \alpha ^{2}+\beta ^{2}\right) \geq \left( \alpha +\beta \right)
2753: ^{2},\qquad \alpha ,\beta \geq 0;
2754: \end{equation*}
2755: we get
2756: \begin{equation}
2757: \left\| x+y\right\| ^{2}+\left( \sqrt{r_{2}^{2}-r_{1}^{2}}\right) ^{2}\geq
2758: \frac{1}{2}\left( \left\| x+y\right\| +\sqrt{r_{2}^{2}-r_{1}^{2}}\right)
2759: ^{2}. \label{ch2.2.10}
2760: \end{equation}
2761: Utilising (\ref{ch2.2.9}) and (\ref{ch2.2.10}), we deduce the desired
2762: inequality (\ref{ch2.2.8}).
2763: \end{proof}
2764:
2765: If $\left( H;\left\langle \cdot ,\cdot \right\rangle \right) $ is a Hilbert
2766: space and $\left\{ e_{i}\right\} _{i\in I}$ is an orthornormal family in $H,$
2767: i.e., we recall that $\left\langle e_{i},e_{j}\right\rangle =\delta _{ij}$
2768: for any $i,j\in I,$ where $\delta _{ij}$ is Kronecker's delta, then we have
2769: the following inequality which is well known in the literature as \textit{%
2770: Bessel's inequality}
2771: \begin{equation}
2772: \sum_{i\in I}\left\vert \left\langle x,e_{i}\right\rangle \right\vert
2773: ^{2}\leq \left\Vert x\right\Vert ^{2}\text{ \ \ for each \ }x\in H.
2774: \label{ch2.2.11}
2775: \end{equation}%
2776: Here, the meaning of the sum is
2777: \begin{equation*}
2778: \sum_{i\in I}\left\vert \left\langle x,e_{i}\right\rangle \right\vert
2779: ^{2}=\sup_{F\subset I}\left\{ \sum_{i\in F}\left\vert \left\langle
2780: x,e_{i}\right\rangle \right\vert ^{2},\ F\text{ is a finite part of }%
2781: I\right\} .
2782: \end{equation*}
2783:
2784: The following result providing a refinement of the Bessel inequality (\ref%
2785: {ch2.2.11}) holds \cite{DRAG01}.
2786:
2787: \begin{theorem}[Dragomir, 2004]
2788: \label{ch2.t2.3}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
2789: \right) $ be a Hilbert space and $\left\{ e_{i}\right\} _{i\in I}$ an
2790: orthonormal family in $H.$ If $x\in H,$ $x\neq 0,$ and $r_{2},r_{1}>0$ are
2791: such that:
2792: \begin{multline}
2793: \left\Vert x-\sum_{i\in I}\left\langle x,e_{i}\right\rangle e_{i}\right\Vert
2794: \\
2795: \geq r_{2}\geq r_{1}\geq \left\Vert x\right\Vert -\left( \sum_{i\in
2796: I}\left\vert \left\langle x,e_{i}\right\rangle \right\vert ^{2}\right) ^{%
2797: \frac{1}{2}}\left( \geq 0\right) , \label{ch2.2.12}
2798: \end{multline}%
2799: then we have the inequality
2800: \begin{equation}
2801: \left\Vert x\right\Vert -\left( \sum_{i\in I}\left\vert \left\langle
2802: x,e_{i}\right\rangle \right\vert ^{2}\right) ^{\frac{1}{2}}\geq \frac{1}{2}%
2803: \cdot \frac{r_{2}^{2}-r_{1}^{2}}{\left( \sum_{i\in I}\left\vert \left\langle
2804: x,e_{i}\right\rangle \right\vert ^{2}\right) ^{\frac{1}{2}}}\left( \geq
2805: 0\right) . \label{ch2.2.13}
2806: \end{equation}%
2807: The constant $\frac{1}{2}$ is best possible.
2808: \end{theorem}
2809:
2810: \begin{proof}
2811: Consider $y:=\sum_{i\in I}\left\langle x,e_{i}\right\rangle e_{i}.$
2812: Obviously, since $H$ is a Hilbert space, $y\in H.$ We also note that
2813: \begin{equation*}
2814: \left\Vert y\right\Vert =\left\Vert \sum_{i\in I}\left\langle
2815: x,e_{i}\right\rangle e_{i}\right\Vert =\sqrt{\left\Vert \sum_{i\in
2816: I}\left\langle x,e_{i}\right\rangle e_{i}\right\Vert ^{2}}=\sqrt{\sum_{i\in
2817: I}\left\vert \left\langle x,e_{i}\right\rangle \right\vert ^{2}},
2818: \end{equation*}
2819: and thus (\ref{ch2.2.12}) is in fact (\ref{ch2.2.1}) of Theorem \ref%
2820: {ch2.t2.1}.
2821:
2822: Since
2823: \begin{align*}
2824: \left\Vert x\right\Vert \left\Vert y\right\Vert -\func{Re}\left\langle
2825: x,y\right\rangle & =\left\Vert x\right\Vert \left( \sum_{i\in I}\left\vert
2826: \left\langle x,e_{i}\right\rangle \right\vert ^{2}\right) ^{\frac{1}{2}}-%
2827: \func{Re}\left\langle x,\sum_{i\in I}\left\langle x,e_{i}\right\rangle
2828: e_{i}\right\rangle \\
2829: & =\left( \sum_{i\in I}\left\vert \left\langle x,e_{i}\right\rangle
2830: \right\vert ^{2}\right) ^{\frac{1}{2}}\left[ \left\Vert x\right\Vert -\left(
2831: \sum_{i\in I}\left\vert \left\langle x,e_{i}\right\rangle \right\vert
2832: ^{2}\right) ^{\frac{1}{2}}\right] ,
2833: \end{align*}%
2834: hence, by (\ref{ch2.2.2}), we deduce the desired result (\ref{ch2.2.13}).
2835:
2836: We will prove the sharpness of the constant for the case of one element,
2837: i.e., $I=\left\{ 1\right\} ,$ $e_{1}=e\in H,$ $\left\Vert e\right\Vert =1.$
2838: For this, assume that there exists a constant $D>0$ such that
2839: \begin{equation}
2840: \left\Vert x\right\Vert -\left\vert \left\langle x,e\right\rangle
2841: \right\vert \geq D\cdot \frac{r_{2}^{2}-r_{1}^{2}}{\left\vert \left\langle
2842: x,e\right\rangle \right\vert } \label{ch2.2.14}
2843: \end{equation}%
2844: provided $x\in H\backslash \left\{ 0\right\} $ satisfies the condition
2845: \begin{equation}
2846: \left\Vert x-\left\langle x,e\right\rangle e\right\Vert \geq r_{2}\geq
2847: r_{1}\geq \left\Vert x\right\Vert -\left\vert \left\langle x,e\right\rangle
2848: \right\vert . \label{ch2.2.15}
2849: \end{equation}%
2850: Assume that $x=\lambda e+\mu f$ with $e,f\in H,$ $\left\Vert e\right\Vert
2851: =\left\Vert f\right\Vert =1$ and $e\perp f.$ We wish to see if there exists
2852: positive numbers $\lambda ,\mu $ such that
2853: \begin{equation}
2854: \left\Vert x-\left\langle x,e\right\rangle e\right\Vert
2855: =r_{2}>r_{1}=\left\Vert x\right\Vert -\left\vert \left\langle
2856: x,e\right\rangle \right\vert . \label{ch2.2.16}
2857: \end{equation}%
2858: Since (for $\lambda ,\mu >0$)
2859: \begin{equation*}
2860: \left\Vert x-\left\langle x,e\right\rangle e\right\Vert =\mu
2861: \end{equation*}%
2862: and
2863: \begin{equation*}
2864: \left\Vert x\right\Vert -\left\vert \left\langle x,e\right\rangle
2865: \right\vert =\sqrt{\lambda ^{2}+\mu ^{2}}-\lambda
2866: \end{equation*}%
2867: hence, by (\ref{ch2.2.16}), we get $\mu =r_{2}$ and
2868: \begin{equation*}
2869: \sqrt{\lambda ^{2}+r_{2}^{2}}-\lambda =r_{1}
2870: \end{equation*}%
2871: giving
2872: \begin{equation*}
2873: \lambda ^{2}+r_{2}^{2}=\lambda ^{2}+2\lambda r_{1}+r_{1}^{2}
2874: \end{equation*}%
2875: from where we get
2876: \begin{equation*}
2877: \lambda =\frac{r_{2}^{2}-r_{1}^{2}}{2r_{1}}>0.
2878: \end{equation*}%
2879: With these values for $\lambda $ and $\mu ,$ we have
2880: \begin{equation*}
2881: \left\Vert x\right\Vert -\left\vert \left\langle x,e\right\rangle
2882: \right\vert =r_{1},\qquad \left\vert \left\langle x,e\right\rangle
2883: \right\vert =\frac{r_{2}^{2}-r_{1}^{2}}{2r_{1}}
2884: \end{equation*}%
2885: and thus, from (\ref{ch2.2.14}), we deduce
2886: \begin{equation*}
2887: r_{1}\geq D\cdot \frac{r_{2}^{2}-r_{1}^{2}}{\frac{r_{2}^{2}-r_{1}^{2}}{2r_{1}%
2888: }},
2889: \end{equation*}%
2890: giving $D\leq \frac{1}{2}.$ This proves the theorem.
2891: \end{proof}
2892:
2893: The following corollary is obvious.
2894:
2895: \begin{corollary}
2896: \label{ch2.c2.4}Let $x,y\in H$ with $\left\langle x,y\right\rangle \neq 0$
2897: and $r_{2}\geq r_{1}>0$ such that
2898: \begin{align}
2899: \left\Vert \left\Vert y\right\Vert x-\frac{\left\langle x,y\right\rangle }{%
2900: \left\Vert y\right\Vert }\cdot y\right\Vert & \geq r_{2}\left\Vert
2901: y\right\Vert \geq r_{1}\left\Vert y\right\Vert \label{ch2.2.17} \\
2902: & \geq \left\Vert x\right\Vert \left\Vert y\right\Vert -\left\vert
2903: \left\langle x,y\right\rangle \right\vert \left( \geq 0\right) . \notag
2904: \end{align}%
2905: Then we have the following refinement of the Schwarz's inequality:
2906: \begin{equation}
2907: \left\Vert x\right\Vert \left\Vert y\right\Vert -\left\vert \left\langle
2908: x,y\right\rangle \right\vert \geq \frac{1}{2}\left(
2909: r_{2}^{2}-r_{1}^{2}\right) \frac{\left\Vert y\right\Vert ^{2}}{\left\vert
2910: \left\langle x,y\right\rangle \right\vert }\left( \geq 0\right) .
2911: \label{ch2.2.18}
2912: \end{equation}%
2913: The constant $\frac{1}{2}$ is best possible.
2914: \end{corollary}
2915:
2916: The following lemma holds \cite{DRAG01}.
2917:
2918: \begin{lemma}[Dragomir, 2004]
2919: \label{ch2.l2.5}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
2920: \right) $ be an inner product space and $R\geq 1.$ For $x,y\in H,$ the
2921: subsequent statements are equivalent:
2922:
2923: \begin{enumerate}
2924: \item[(i)] The following refinement of the triangle inequality holds:
2925: \begin{equation}
2926: \left\Vert x\right\Vert +\left\Vert y\right\Vert \geq R\left\Vert
2927: x+y\right\Vert ; \label{ch2.2.19}
2928: \end{equation}
2929:
2930: \item[(ii)] The following refinement of the Schwarz inequality holds:
2931: \begin{equation}
2932: \left\Vert x\right\Vert \left\Vert y\right\Vert -\func{Re}\left\langle
2933: x,y\right\rangle \geq \frac{1}{2}\left( R^{2}-1\right) \left\Vert
2934: x+y\right\Vert ^{2}. \label{ch2.2.20}
2935: \end{equation}
2936: \end{enumerate}
2937: \end{lemma}
2938:
2939: \begin{proof}
2940: Taking the square in (\ref{ch2.2.19}), we have
2941: \begin{equation}
2942: 2\left\Vert x\right\Vert \left\Vert y\right\Vert \geq \left( R^{2}-1\right)
2943: \left\Vert x\right\Vert ^{2}+2R^{2}\func{Re}\left\langle x,y\right\rangle
2944: +\left( R^{2}-1\right) \left\Vert y\right\Vert ^{2}. \label{ch2.2.21}
2945: \end{equation}
2946: Subtracting from both sides of (\ref{ch2.2.21}) the quantity $2\func{Re}%
2947: \left\langle x,y\right\rangle ,$ we obtain
2948: \begin{align*}
2949: 2\left( \left\Vert x\right\Vert \left\Vert y\right\Vert -\func{Re}%
2950: \left\langle x,y\right\rangle \right) & \geq \left( R^{2}-1\right) \left[
2951: \left\Vert x\right\Vert ^{2}+2\func{Re}\left\langle x,y\right\rangle
2952: +\left\Vert y\right\Vert ^{2}\right] \\
2953: & =\left( R^{2}-1\right) \left\Vert x+y\right\Vert ^{2},
2954: \end{align*}
2955: which is clearly equivalent to (\ref{ch2.2.20}).
2956: \end{proof}
2957:
2958: By the use of the above lemma, we may now state the following theorem
2959: concerning another refinement of the Schwarz inequality \cite{DRAG01}.
2960:
2961: \begin{theorem}[Dragomir, 2004]
2962: \label{ch2.t2.6}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
2963: \right) $ be an inner product space over the real or complex number field
2964: and $R\geq 1,$ $r\geq 0.$ If $x,y\in H$ are such that
2965: \begin{equation}
2966: \frac{1}{R}\left( \left\Vert x\right\Vert +\left\Vert y\right\Vert \right)
2967: \geq \left\Vert x+y\right\Vert \geq r, \label{ch2.2.22}
2968: \end{equation}%
2969: then we have the following refinement of the Schwarz inequality
2970: \begin{equation}
2971: \left\Vert x\right\Vert \left\Vert y\right\Vert -\func{Re}\left\langle
2972: x,y\right\rangle \geq \frac{1}{2}\left( R^{2}-1\right) r^{2}.
2973: \label{ch2.2.23}
2974: \end{equation}%
2975: The constant $\frac{1}{2}$ is best possible in the sense that it cannot be
2976: replaced by a larger quantity.
2977: \end{theorem}
2978:
2979: \begin{proof}
2980: The inequality (\ref{ch2.2.23}) follows easily from Lemma \ref{ch2.l2.5}. We
2981: need only prove that $\frac{1}{2}$ is the best possible constant in (\ref%
2982: {ch2.2.23}).
2983:
2984: Assume that there exists a $C>0$ such that
2985: \begin{equation}
2986: \left\Vert x\right\Vert \left\Vert y\right\Vert -\func{Re}\left\langle
2987: x,y\right\rangle \geq C\left( R^{2}-1\right) r^{2} \label{ch2.2.24}
2988: \end{equation}%
2989: provided $x,y,$ $R$ and $r$ satisfy (\ref{ch2.2.22}).
2990:
2991: Consider $r=1,$ $R>1$ and choose $x=\frac{1-R}{2}e,$ $y=\frac{1+R}{2}e$ with
2992: $e\in H,$ $\left\Vert e\right\Vert =1.$ Then
2993: \begin{equation*}
2994: x+y=e,\ \ \ \frac{\left\Vert x\right\Vert +\left\Vert y\right\Vert }{R}=1
2995: \end{equation*}%
2996: and thus (\ref{ch2.2.22}) holds with equality on both sides.
2997:
2998: From (\ref{ch2.2.24}), for the above choices, we have $\frac{1}{2}\left(
2999: R^{2}-1\right) \geq C\left( R^{2}-1\right) ,$ which shows that $C\leq \frac{1%
3000: }{2}.$
3001: \end{proof}
3002:
3003: Finally, the following result also holds \cite{DRAG01}.
3004:
3005: \begin{theorem}[Dragomir, 2004]
3006: \label{ch2.t2.7}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
3007: \right) $ be an inner product space over the real or complex number field $%
3008: \mathbb{K}$ and $r\in (0,1].$ For $x,y\in H,$ the following statements are
3009: equivalent:
3010:
3011: \begin{enumerate}
3012: \item[(i)] We have the inequality
3013: \begin{equation}
3014: \left\vert \left\Vert x\right\Vert -\left\Vert y\right\Vert \right\vert \leq
3015: r\left\Vert x-y\right\Vert ; \label{ch2.2.25}
3016: \end{equation}
3017:
3018: \item[(ii)] We have the following refinement of the Schwarz inequality
3019: \begin{equation}
3020: \left\Vert x\right\Vert \left\Vert y\right\Vert -\func{Re}\left\langle
3021: x,y\right\rangle \geq \frac{1}{2}\left( 1-r^{2}\right) \left\Vert
3022: x-y\right\Vert ^{2}. \label{ch2.2.26}
3023: \end{equation}
3024: The constant $\frac{1}{2}$ in (\ref{ch2.2.26}) is best possible.
3025: \end{enumerate}
3026: \end{theorem}
3027:
3028: \begin{proof}
3029: Taking the square in (\ref{ch2.2.25}), we have
3030: \begin{equation*}
3031: \left\Vert x\right\Vert ^{2}-2\left\Vert x\right\Vert \left\Vert
3032: y\right\Vert +\left\Vert y\right\Vert ^{2}\leq r^{2}\left( \left\Vert
3033: x\right\Vert ^{2}-2\func{Re}\left\langle x,y\right\rangle +\left\Vert
3034: y\right\Vert ^{2}\right)
3035: \end{equation*}
3036: which is clearly equivalent to
3037: \begin{equation*}
3038: \left( 1-r^{2}\right) \left[ \left\Vert x\right\Vert ^{2}-2\func{Re}%
3039: \left\langle x,y\right\rangle +\left\Vert y\right\Vert ^{2}\right] \leq
3040: 2\left( \left\Vert x\right\Vert \left\Vert y\right\Vert -\func{Re}%
3041: \left\langle x,y\right\rangle \right)
3042: \end{equation*}
3043: or with (\ref{ch2.2.26}).
3044:
3045: Now, assume that (\ref{ch2.2.26}) holds with a constant $E>0,$ i.e.,
3046: \begin{equation}
3047: \left\| x\right\| \left\| y\right\| -\func{Re}\left\langle x,y\right\rangle
3048: \geq E\left( 1-r^{2}\right) \left\| x-y\right\| ^{2}, \label{ch2.2.27}
3049: \end{equation}
3050: provided (\ref{ch2.2.25}) holds.
3051:
3052: Define $x=\frac{r+1}{2}e,$ $y=\frac{r-1}{2}e$ with $e\in H,$ $\left\Vert
3053: e\right\Vert =1.$ Then
3054: \begin{equation*}
3055: \left\vert \left\Vert x\right\Vert -\left\Vert y\right\Vert \right\vert
3056: =r,\quad \left\Vert x-y\right\Vert =1
3057: \end{equation*}%
3058: showing that (\ref{ch2.2.25}) holds with equality.
3059:
3060: If we replace $x$ and $y$ in (\ref{ch2.2.27}), then we get $E\left(
3061: 1-r^{2}\right) \leq \frac{1}{2}\left( 1-r^{2}\right) ,$ implying that $E\leq
3062: \frac{1}{2}.$
3063: \end{proof}
3064:
3065: \subsection{Discrete Inequalities}
3066:
3067: Assume that $\left( K;\left( \cdot ,\cdot \right) \right) $ is a Hilbert
3068: space over the real or complex number field. Assume also that $p_{i}\geq 0$,
3069: $i\in H$ with $\sum_{i=1}^{\infty }p_{i}=1$ and define
3070: \begin{equation*}
3071: \ell _{p}^{2}\left( K\right) :=\left\{ \mathbf{x}:=\left. \left(
3072: x_{i}\right) _{i\in \mathbb{N}}\right| x_{i}\in \mathbb{K},\ i\in \mathbb{N}%
3073: \text{ \ and \ }\sum_{i=1}^{\infty }p_{i}\left\| x_{i}\right\| ^{2}<\infty
3074: \right\} .
3075: \end{equation*}
3076: It is well known that $\ell _{p}^{2}\left( K\right) $ endowed with the inner
3077: product $\left\langle \cdot ,\cdot \right\rangle _{p}$ defined by
3078: \begin{equation*}
3079: \left\langle \mathbf{x},\mathbf{y}\right\rangle _{p}:=\sum_{i=1}^{\infty
3080: }p_{i}\left( x_{i},y_{i}\right)
3081: \end{equation*}
3082: and generating the norm
3083: \begin{equation*}
3084: \left\| \mathbf{x}\right\| _{p}:=\left( \sum_{i=1}^{\infty }p_{i}\left\|
3085: x_{i}\right\| ^{2}\right) ^{\frac{1}{2}}
3086: \end{equation*}
3087: is a Hilbert space over $\mathbb{K}$.
3088:
3089: We may state the following discrete inequality improving the
3090: Cauchy-Bunyakovsky-Schwarz classical result \cite{DRAG01}.
3091:
3092: \begin{proposition}
3093: \label{ch2.p3.1}Let $\left( K;\left( \cdot ,\cdot \right) \right) $ be a
3094: Hilbert space and $p_{i}\geq 0$ $\left( i\in \mathbb{N}\right) $ with $%
3095: \sum_{i=1}^{\infty }p_{i}=1.$ Assume that $\mathbf{x},\mathbf{y}\in \ell
3096: _{p}^{2}\left( K\right) $ and $r_{1},r_{2}>0$ satisfy the condition
3097: \begin{equation}
3098: \left\Vert x_{i}-y_{i}\right\Vert \geq r_{2}\geq r_{1}\geq \left\vert
3099: \left\Vert x_{i}\right\Vert -\left\Vert y_{i}\right\Vert \right\vert
3100: \label{ch2.3.1}
3101: \end{equation}%
3102: for each $i\in \mathbb{N}$. Then we have the following refinement of the
3103: Cauchy-Bunyakovsky-Schwarz inequality
3104: \begin{multline}
3105: \left( \sum_{i=1}^{\infty }p_{i}\left\Vert x_{i}\right\Vert
3106: ^{2}\sum_{i=1}^{\infty }p_{i}\left\Vert y_{i}\right\Vert ^{2}\right) ^{\frac{%
3107: 1}{2}}-\sum_{i=1}^{\infty }p_{i}\func{Re}\left( x_{i},y_{i}\right)
3108: \label{ch2.3.2} \\
3109: \geq \frac{1}{2}\left( r_{2}^{2}-r_{1}^{2}\right) \geq 0.
3110: \end{multline}%
3111: The constant $\frac{1}{2}$ is best possible.
3112: \end{proposition}
3113:
3114: \begin{proof}
3115: From the condition (\ref{ch2.3.1}) we simply deduce
3116: \begin{align}
3117: \sum_{i=1}^{\infty }p_{i}\left\Vert x_{i}-y_{i}\right\Vert ^{2}& \geq
3118: r_{2}^{2}\geq r_{1}^{2}\geq \sum_{i=1}^{\infty }p_{i}\left( \left\Vert
3119: x_{i}\right\Vert -\left\Vert y_{i}\right\Vert \right) ^{2} \label{ch2.3.3}
3120: \\
3121: & \geq \left[ \left( \sum_{i=1}^{\infty }p_{i}\left\Vert x_{i}\right\Vert
3122: ^{2}\right) ^{\frac{1}{2}}-\left( \sum_{i=1}^{\infty }p_{i}\left\Vert
3123: y_{i}\right\Vert ^{2}\right) ^{\frac{1}{2}}\right] ^{2}. \notag
3124: \end{align}
3125: In terms of the norm $\left\Vert \cdot \right\Vert _{p},$ the inequality (%
3126: \ref{ch2.3.3}) may be written as
3127: \begin{equation}
3128: \left\Vert \mathbf{x}-\mathbf{y}\right\Vert _{p}\geq r_{2}\geq r_{1}\geq
3129: \left\vert \left\Vert \mathbf{x}\right\Vert _{p}-\left\Vert \mathbf{y}%
3130: \right\Vert _{p}\right\vert . \label{ch2.3.4}
3131: \end{equation}%
3132: Utilising Theorem \ref{ch2.t2.1} for the Hilbert space $\left( \ell
3133: _{p}^{2}\left( K\right) ,\left\langle \cdot ,\cdot \right\rangle _{p}\right)
3134: ,$ we deduce the desired inequality (\ref{ch2.3.2}).
3135:
3136: For $n=1$ $\left( p_{1}=1\right) ,$ the inequality (\ref{ch2.3.2}) reduces
3137: to (\ref{ch2.2.2}) for which we have shown that $\frac{1}{2}$ is the best
3138: possible constant.
3139: \end{proof}
3140:
3141: By the use of Corollary \ref{ch2.c2.2}, we may state the following result as
3142: well.
3143:
3144: \begin{corollary}
3145: \label{ch2.c3.2}With the assumptions of Proposition \ref{ch2.p3.1}, we have
3146: the inequality
3147: \begin{multline}
3148: \left( \sum_{i=1}^{\infty }p_{i}\left\Vert x_{i}\right\Vert ^{2}\right) ^{%
3149: \frac{1}{2}}+\left( \sum_{i=1}^{\infty }p_{i}\left\Vert y_{i}\right\Vert
3150: ^{2}\right) ^{\frac{1}{2}} \label{ch2.3.5} \\
3151: -\frac{\sqrt{2}}{2}\left( \sum_{i=1}^{\infty }p_{i}\left\Vert
3152: x_{i}+y_{i}\right\Vert ^{2}\right) ^{\frac{1}{2}}\geq \frac{\sqrt{2}}{2}%
3153: \sqrt{r_{2}^{2}-r_{1}^{2}}.
3154: \end{multline}
3155: \end{corollary}
3156:
3157: The following proposition also holds \cite{DRAG01}.
3158:
3159: \begin{proposition}
3160: \label{ch2.p3.3}Let $\left( K;\left( \cdot ,\cdot \right) \right) $ be a
3161: Hilbert space and $p_{i}\geq 0$ $\left( i\in \mathbb{N}\right) $ with $%
3162: \sum_{i=1}^{\infty }p_{i}=1.$ Assume that $\mathbf{x},\mathbf{y}\in \ell
3163: _{p}^{2}\left( K\right) $ and $R\geq 1,$ $r\geq 0$ satisfy the condition
3164: \begin{equation}
3165: \frac{1}{R}\left( \left\Vert x_{i}\right\Vert +\left\Vert y_{i}\right\Vert
3166: \right) \geq \left\Vert x_{i}+y_{i}\right\Vert \geq r \label{ch2.3.6}
3167: \end{equation}%
3168: for each $i\in \mathbb{N}$. Then we have the following refinement of the
3169: Schwarz inequality
3170: \begin{multline}
3171: \left( \sum_{i=1}^{\infty }p_{i}\left\Vert x_{i}\right\Vert
3172: ^{2}\sum_{i=1}^{\infty }p_{i}\left\Vert y_{i}\right\Vert ^{2}\right) ^{\frac{%
3173: 1}{2}}-\sum_{i=1}^{\infty }p_{i}\func{Re}\left( x_{i},y_{i}\right)
3174: \label{ch2.3.7} \\
3175: \geq \frac{1}{2}\left( R^{2}-1\right) r^{2}.
3176: \end{multline}%
3177: The constant $\frac{1}{2}$ is best possible in the sense that it cannot be
3178: replaced by a larger quantity.
3179: \end{proposition}
3180:
3181: \begin{proof}
3182: By (\ref{ch2.3.6}) we deduce
3183: \begin{equation}
3184: \frac{1}{R}\left[ \sum_{i=1}^{\infty }p_{i}\left( \left\Vert
3185: x_{i}\right\Vert +\left\Vert y_{i}\right\Vert \right) ^{2}\right] ^{\frac{1}{%
3186: 2}}\geq \left( \sum_{i=1}^{\infty }p_{i}\left\Vert x_{i}+y_{i}\right\Vert
3187: ^{2}\right) ^{\frac{1}{2}}\geq r. \label{ch2.3.8}
3188: \end{equation}%
3189: By the classical Minkowsky inequality for nonnegative numbers, we have
3190: \begin{multline}
3191: \quad \left( \sum_{i=1}^{\infty }p_{i}\left\Vert x_{i}\right\Vert
3192: ^{2}\right) ^{\frac{1}{2}}+\left( \sum_{i=1}^{\infty }p_{i}\left\Vert
3193: y_{i}\right\Vert ^{2}\right) ^{\frac{1}{2}} \label{ch2.3.9} \\
3194: \geq \left[ \sum_{i=1}^{\infty }p_{i}\left( \left\Vert x_{i}\right\Vert
3195: +\left\Vert y_{i}\right\Vert \right) ^{2}\right] ^{\frac{1}{2}},\quad
3196: \end{multline}%
3197: and thus, by utilising (\ref{ch2.3.8}) and (\ref{ch2.3.9}), we may state in
3198: terms of $\left\Vert \cdot \right\Vert _{p}$ the following inequality
3199: \begin{equation}
3200: \frac{1}{R}\left( \left\Vert \mathbf{x}\right\Vert _{p}+\left\Vert \mathbf{y}%
3201: \right\Vert _{p}\right) \geq \left\Vert \mathbf{x}+\mathbf{y}\right\Vert
3202: _{p}\geq r. \label{ch2.3.10}
3203: \end{equation}%
3204: Employing Theorem \ref{ch2.t2.6} for the Hilbert space $\ell _{p}^{2}\left(
3205: K\right) $ and the inequality (\ref{ch2.3.10}), we deduce the desired result
3206: (\ref{ch2.3.7}).
3207:
3208: Since, for $p=1,$ $n=1,$ (\ref{ch2.3.7}) reduced to (\ref{ch2.2.23}) for
3209: which we have shown that $\frac{1}{2}$ is the best constant, we conclude
3210: that $\frac{1}{2}$ is the best constant in (\ref{ch2.3.7}) as well.
3211: \end{proof}
3212:
3213: Finally, we may state and prove the following result \cite{DRAG01}
3214: incorporated in
3215:
3216: \begin{proposition}
3217: \label{ch2.p3.4}Let $\left( K;\left( \cdot ,\cdot \right) \right) $ be a
3218: Hilbert space and $p_{i}\geq 0$ $\left( i\in \mathbb{N}\right) $ with $%
3219: \sum_{i=1}^{\infty }p_{i}=1.$ Assume that $\mathbf{x},\mathbf{y}\in \ell
3220: _{p}^{2}\left( K\right) $ and $r\in (0,1]$ such that
3221: \begin{equation}
3222: \left\vert \left\Vert x_{i}\right\Vert -\left\Vert y_{i}\right\Vert
3223: \right\vert \leq r\left\Vert x_{i}-y_{i}\right\Vert \text{ \ for each }i\in
3224: \mathbb{N}, \label{ch2.3.11}
3225: \end{equation}
3226: holds true. Then we have the following refinement of the Schwarz inequality
3227: \begin{multline}
3228: \left( \sum_{i=1}^{\infty }p_{i}\left\Vert x_{i}\right\Vert
3229: ^{2}\sum_{i=1}^{\infty }p_{i}\left\Vert y_{i}\right\Vert ^{2}\right) ^{\frac{%
3230: 1}{2}}-\sum_{i=1}^{\infty }p_{i}\func{Re}\left( x_{i},y_{i}\right)
3231: \label{ch2.3.12} \\
3232: \geq \frac{1}{2}\left( 1-r^{2}\right) \sum_{i=1}^{\infty }p_{i}\left\Vert
3233: x_{i}-y_{i}\right\Vert ^{2}.
3234: \end{multline}
3235: The constant $\frac{1}{2}$ is best possible in (\ref{ch2.3.12}).
3236: \end{proposition}
3237:
3238: \begin{proof}
3239: From (\ref{ch2.3.11}) we have
3240: \begin{equation*}
3241: \left[ \sum_{i=1}^{\infty }p_{i}\left( \left\Vert x_{i}\right\Vert
3242: -\left\Vert y_{i}\right\Vert \right) ^{2}\right] ^{\frac{1}{2}}\leq r\left[
3243: \sum_{i=1}^{\infty }p_{i}\left\Vert x_{i}-y_{i}\right\Vert ^{2}\right] ^{%
3244: \frac{1}{2}}.
3245: \end{equation*}%
3246: Utilising the following elementary result
3247: \begin{equation*}
3248: \left\vert \left( \sum_{i=1}^{\infty }p_{i}\left\Vert x_{i}\right\Vert
3249: ^{2}\right) ^{\frac{1}{2}}-\left( \sum_{i=1}^{\infty }p_{i}\left\Vert
3250: y_{i}\right\Vert ^{2}\right) ^{\frac{1}{2}}\right\vert \leq \left(
3251: \sum_{i=1}^{\infty }p_{i}\left( \left\Vert x_{i}\right\Vert -\left\Vert
3252: y_{i}\right\Vert \right) ^{2}\right) ^{\frac{1}{2}},
3253: \end{equation*}%
3254: we may state that
3255: \begin{equation*}
3256: \left\vert \left\Vert \mathbf{x}\right\Vert _{p}-\left\Vert \mathbf{y}%
3257: \right\Vert _{p}\right\vert \leq r\left\Vert \mathbf{x}-\mathbf{y}%
3258: \right\Vert _{p}.
3259: \end{equation*}%
3260: Now, by making use of Theorem \ref{ch2.t2.7}, we deduce the desired
3261: inequality (\ref{ch2.3.12}) and the fact that $\frac{1}{2}$ is the best
3262: possible constant. We omit the details.
3263: \end{proof}
3264:
3265: \subsection{Integral Inequalities}
3266:
3267: Assume that $\left( K;\left( \cdot ,\cdot \right) \right) $ is a Hilbert
3268: space over the real or complex number field $\mathbb{K}$. If $\rho :\left[
3269: a,b\right] \subset \mathbb{R}\rightarrow \lbrack 0,\infty )$ is a Lebesgue
3270: integrable function with $\int_{a}^{b}\rho \left( t\right) dt=1,$ then we
3271: may consider the space $L_{\rho }^{2}\left( \left[ a,b\right] ;K\right) $ of
3272: all functions $f:\left[ a,b\right] \rightarrow K,$ that are Bochner
3273: measurable and $\int_{a}^{b}\rho \left( t\right) \left\Vert f\left( t\right)
3274: \right\Vert ^{2}dt<\infty $. It is known that $L_{\rho }^{2}\left( \left[ a,b%
3275: \right] ;K\right) $ endowed with the inner product $\left\langle \cdot
3276: ,\cdot \right\rangle _{\rho }$ defined by
3277: \begin{equation*}
3278: \left\langle f,g\right\rangle _{\rho }:=\int_{a}^{b}\rho \left( t\right)
3279: \left( f\left( t\right) ,g\left( t\right) \right) dt
3280: \end{equation*}%
3281: and generating the norm
3282: \begin{equation*}
3283: \left\Vert f\right\Vert _{\rho }:=\left( \int_{a}^{b}\rho \left( t\right)
3284: \left\Vert f\left( t\right) \right\Vert ^{2}dt\right) ^{\frac{1}{2}}
3285: \end{equation*}%
3286: is a Hilbert space over $\mathbb{K}$.
3287:
3288: Now we may state and prove the first refinement of the
3289: Cauchy-Bunyakovsky-Schwarz integral inequality \cite{DRAG01}.
3290:
3291: \begin{proposition}
3292: \label{ch2.p4.1}Assume that $f,g\in L_{\rho }^{2}\left( \left[ a,b\right]
3293: ;K\right) $ and $r_{2},r_{1}>0$ satisfy the condition
3294: \begin{equation}
3295: \left\Vert f\left( t\right) -g\left( t\right) \right\Vert \geq r_{2}\geq
3296: r_{1}\geq \left\vert \left\Vert f\left( t\right) \right\Vert -\left\Vert
3297: g\left( t\right) \right\Vert \right\vert \label{ch2.4.1}
3298: \end{equation}%
3299: for a.e. $t\in \left[ a,b\right] .$ Then we have the inequality
3300: \begin{multline}
3301: \left( \int_{a}^{b}\rho \left( t\right) \left\Vert f\left( t\right)
3302: \right\Vert ^{2}dt\int_{a}^{b}\rho \left( t\right) \left\Vert g\left(
3303: t\right) \right\Vert ^{2}dt\right) ^{\frac{1}{2}} \label{ch2.4.2} \\
3304: -\int_{a}^{b}\rho \left( t\right) \func{Re}\left( f\left( t\right) ,g\left(
3305: t\right) \right) dt\geq \frac{1}{2}\left( r_{2}^{2}-r_{1}^{2}\right) \left(
3306: \geq 0\right) .
3307: \end{multline}%
3308: The constant $\frac{1}{2}$ is best possible in (\ref{ch2.4.2}).
3309: \end{proposition}
3310:
3311: \begin{proof}
3312: Integrating (\ref{ch2.4.1}), we get
3313: \begin{multline}
3314: \left( \int_{a}^{b}\rho \left( t\right) \left( \left\Vert f\left( t\right)
3315: -g\left( t\right) \right\Vert \right) ^{2}dt\right) ^{\frac{1}{2}}
3316: \label{ch2.4.3} \\
3317: \geq r_{2}\geq r_{1}\geq \left( \int_{a}^{b}\rho \left( t\right) \left(
3318: \left\Vert f\left( t\right) \right\Vert -\left\Vert g\left( t\right)
3319: \right\Vert \right) ^{2}dt\right) ^{\frac{1}{2}}.
3320: \end{multline}%
3321: Utilising the obvious fact
3322: \begin{multline}
3323: \left[ \int_{a}^{b}\rho \left( t\right) \left( \left\Vert f\left( t\right)
3324: \right\Vert -\left\Vert g\left( t\right) \right\Vert \right) ^{2}dt\right] ^{%
3325: \frac{1}{2}} \label{ch2.4.4} \\
3326: \geq \left\vert \left( \int_{a}^{b}\rho \left( t\right) \left\Vert f\left(
3327: t\right) \right\Vert ^{2}dt\right) ^{\frac{1}{2}}-\left( \int_{a}^{b}\rho
3328: \left( t\right) \left\Vert g\left( t\right) \right\Vert ^{2}dt\right) ^{%
3329: \frac{1}{2}}\right\vert ,
3330: \end{multline}
3331: we can state the following inequality in terms of the $\left\Vert \cdot
3332: \right\Vert _{\rho }$ norm:
3333: \begin{equation}
3334: \left\Vert f-g\right\Vert _{\rho }\geq r_{2}\geq r_{1}\geq \left\vert
3335: \left\Vert f\right\Vert _{\rho }-\left\Vert g\right\Vert _{\rho }\right\vert
3336: . \label{ch2.4.5}
3337: \end{equation}%
3338: Employing Theorem \ref{ch2.t2.1} for the Hilbert space $L_{\rho }^{2}\left( %
3339: \left[ a,b\right] ;K\right) ,$ we deduce the desired inequality (\ref%
3340: {ch2.4.2}).
3341:
3342: To prove the sharpness of $\frac{1}{2}$ in (\ref{ch2.4.2}), we choose $a=0,$
3343: $b=1,$ $f\left( t\right) =1,$ $t\in \left[ 0,1\right] $ and $f\left(
3344: t\right) =x,$ $g\left( t\right) =y,$ $t\in \left[ a,b\right] ,$ $x,y\in K.$
3345: Then (\ref{ch2.4.2}) becomes
3346: \begin{equation*}
3347: \left\| x\right\| \left\| y\right\| -\func{Re}\left\langle x,y\right\rangle
3348: \geq \frac{1}{2}\left( r_{2}^{2}-r_{1}^{2}\right)
3349: \end{equation*}
3350: provided
3351: \begin{equation*}
3352: \left\| x-y\right\| \geq r_{2}\geq r_{1}\geq \left| \left\| x\right\|
3353: -\left\| y\right\| \right| ,
3354: \end{equation*}
3355: which, by Theorem \ref{ch2.t2.1} has the quantity $\frac{1}{2}$ as the best
3356: possible constant.
3357: \end{proof}
3358:
3359: The following corollary holds.
3360:
3361: \begin{corollary}
3362: \label{ch2.c4.2}With the assumptions of Proposition \ref{ch2.p4.1}, we have
3363: the inequality
3364: \begin{multline}
3365: \left( \int_{a}^{b}\rho \left( t\right) \left\Vert f\left( t\right)
3366: \right\Vert ^{2}dt\right) ^{\frac{1}{2}}+\left( \int_{a}^{b}\rho \left(
3367: t\right) \left\Vert g\left( t\right) \right\Vert ^{2}dt\right) ^{\frac{1}{2}}
3368: \label{ch2.4.6} \\
3369: -\frac{\sqrt{2}}{2}\left( \int_{a}^{b}\rho \left( t\right) \left\Vert
3370: f\left( t\right) +g\left( t\right) \right\Vert ^{2}dt\right) ^{\frac{1}{2}%
3371: }\geq \frac{\sqrt{2}}{2}\sqrt{r_{2}^{2}-r_{1}^{2}}.
3372: \end{multline}
3373: \end{corollary}
3374:
3375: The following two refinements of the Cauchy-Bunyakovsky-Schwarz (CBS)
3376: integral inequality also hold.
3377:
3378: \begin{proposition}
3379: \label{ch2.p4.3}If $f,g\in L_{\rho }^{2}\left( \left[ a,b\right] ;K\right) $
3380: and $R\geq 1,r\geq 0$ satisfy the condition
3381: \begin{equation}
3382: \frac{1}{R}\left( \left\Vert f\left( t\right) \right\Vert +\left\Vert
3383: g\left( t\right) \right\Vert \right) \geq \left\Vert f\left( t\right)
3384: +g\left( t\right) \right\Vert \geq r \label{ch2.4.7}
3385: \end{equation}%
3386: for a.e. $t\in \left[ a,b\right] ,$ then we have the inequality
3387: \begin{multline}
3388: \left( \int_{a}^{b}\rho \left( t\right) \left\Vert f\left( t\right)
3389: \right\Vert ^{2}dt\int_{a}^{b}\rho \left( t\right) \left\Vert g\left(
3390: t\right) \right\Vert ^{2}dt\right) ^{\frac{1}{2}} \label{ch2.4.8} \\
3391: -\int_{a}^{b}\rho \left( t\right) \func{Re}\left( f\left( t\right) ,g\left(
3392: t\right) \right) dt\geq \frac{1}{2}\left( R^{2}-1\right) r^{2}.
3393: \end{multline}%
3394: The constant $\frac{1}{2}$ is best possible in (\ref{ch2.4.8}).
3395: \end{proposition}
3396:
3397: The proof follows by Theorem \ref{ch2.t2.6} and we omit the details.
3398:
3399: \begin{proposition}
3400: \label{ch2.p4.4}If $f,g\in L_{\rho }^{2}\left( \left[ a,b\right] ;K\right) $
3401: and $\zeta \in (0,1]$ satisfy the condition
3402: \begin{equation}
3403: \left\vert \left\Vert f\left( t\right) \right\Vert -\left\Vert g\left(
3404: t\right) \right\Vert \right\vert \leq \zeta \left\Vert f\left( t\right)
3405: -g\left( t\right) \right\Vert \label{ch2.4.9}
3406: \end{equation}%
3407: for a.e. $t\in \left[ a,b\right] ,$ then we have the inequality
3408: \begin{multline}
3409: \left( \int_{a}^{b}\rho \left( t\right) \left\Vert f\left( t\right)
3410: \right\Vert ^{2}dt\int_{a}^{b}\rho \left( t\right) \left\Vert g\left(
3411: t\right) \right\Vert ^{2}dt\right) ^{\frac{1}{2}} \label{ch2.4.10} \\
3412: -\int_{a}^{b}\rho \left( t\right) \func{Re}\left( f\left( t\right) ,g\left(
3413: t\right) \right) dt \\
3414: \geq \frac{1}{2}\left( 1-\zeta ^{2}\right) \int_{a}^{b}\rho \left( t\right)
3415: \left\Vert f\left( t\right) -g\left( t\right) \right\Vert ^{2}dt.
3416: \end{multline}%
3417: The constant $\frac{1}{2}$ is best possible in (\ref{ch2.4.10}).
3418: \end{proposition}
3419:
3420: The proof follows by Theorem \ref{ch2.t2.7} and we omit the details.
3421:
3422: \subsection{Refinements of the Heisenberg Inequality}
3423:
3424: It is well known that if $\left( H;\left\langle \cdot ,\cdot \right\rangle
3425: \right) $ is a real or complex Hilbert space and $f:\left[ a,b\right]
3426: \subset \mathbb{R\rightarrow }H$ is an \textit{absolutely continuous
3427: vector-valued }function, then $f$ is differentiable almost everywhere on $%
3428: \left[ a,b\right] ,$ the derivative $f^{\prime }:\left[ a,b\right]
3429: \rightarrow H$ is Bochner integrable on $\left[ a,b\right] $ and
3430: \begin{equation}
3431: f\left( t\right) =\int_{a}^{t}f^{\prime }\left( s\right) ds\qquad \text{for
3432: any \ }t\in \left[ a,b\right] . \label{ch2.5.1}
3433: \end{equation}
3434:
3435: The following theorem provides a version of the Heisenberg inequalities in
3436: the general setting of Hilbert spaces \cite{DRAG01}.
3437:
3438: \begin{theorem}[Dragomir, 2004]
3439: \label{ch2.t5.1}Let $\varphi :\left[ a,b\right] \rightarrow H$ be an
3440: absolutely continuous function with the property that $b\left\Vert \varphi
3441: \left( b\right) \right\Vert ^{2}=a\left\Vert \varphi \left( a\right)
3442: \right\Vert ^{2}.$ Then we have the inequality:
3443: \begin{equation}
3444: \left( \int_{a}^{b}\left\Vert \varphi \left( t\right) \right\Vert
3445: ^{2}dt\right) ^{2}\leq 4\int_{a}^{b}t^{2}\left\Vert \varphi \left( t\right)
3446: \right\Vert ^{2}dt\cdot \int_{a}^{b}\left\Vert \varphi ^{\prime }\left(
3447: t\right) \right\Vert ^{2}dt. \label{ch2.5.2}
3448: \end{equation}%
3449: The constant $4$ is best possible in the sense that it cannot be replaced by
3450: a smaller constant.
3451: \end{theorem}
3452:
3453: \begin{proof}
3454: Integrating by parts, we have successively
3455: \begin{align}
3456: & \int_{a}^{b}\left\Vert \varphi \left( t\right) \right\Vert ^{2}dt
3457: \label{ch2.5.3} \\
3458: & =t\left\Vert \varphi \left( t\right) \right\Vert ^{2}\bigg|%
3459: _{a}^{b}-\int_{a}^{b}t\frac{d}{dt}\left( \left\Vert \varphi \left( t\right)
3460: \right\Vert ^{2}\right) dt \notag \\
3461: & =b\left\Vert \varphi \left( b\right) \right\Vert ^{2}-a\left\Vert \varphi
3462: \left( a\right) \right\Vert ^{2}-\int_{a}^{b}t\frac{d}{dt}\left\langle
3463: \varphi \left( t\right) ,\varphi \left( t\right) \right\rangle dt \notag \\
3464: & =-\int_{a}^{b}t\left[ \left\langle \varphi ^{\prime }\left( t\right)
3465: ,\varphi \left( t\right) \right\rangle +\left\langle \varphi \left( t\right)
3466: ,\varphi ^{\prime }\left( t\right) \right\rangle \right] dt \notag \\
3467: & =-2\int_{a}^{b}t\func{Re}\left\langle \varphi ^{\prime }\left( t\right)
3468: ,\varphi \left( t\right) \right\rangle dt \notag \\
3469: & =2\int_{a}^{b}\func{Re}\left\langle \varphi ^{\prime }\left( t\right)
3470: ,\left( -t\right) \varphi \left( t\right) \right\rangle dt. \notag
3471: \end{align}%
3472: If we apply the Cauchy-Bunyakovsky-Schwarz integral inequality
3473: \begin{equation*}
3474: \int_{a}^{b}\func{Re}\left\langle g\left( t\right) ,h\left( t\right)
3475: \right\rangle dt\leq \left( \int_{a}^{b}\left\Vert g\left( t\right)
3476: \right\Vert ^{2}dt\int_{a}^{b}\left\Vert h\left( t\right) \right\Vert
3477: ^{2}dt\right) ^{\frac{1}{2}}
3478: \end{equation*}%
3479: for $g\left( t\right) =\varphi ^{\prime }\left( t\right) ,$ $h\left(
3480: t\right) =-t\varphi \left( t\right) ,$ $t\in \left[ a,b\right] ,$ then we
3481: deduce the desired inequality (\ref{ch2.4.2}).
3482:
3483: The fact that $4$ is the best constant in (\ref{ch2.4.2}) follows from the
3484: fact that in the (CBS) inequality, the case of equality holds iff $g\left(
3485: t\right) =\lambda h\left( t\right) $ for a.e. $t\in \left[ a,b\right] $ and $%
3486: \lambda $ a given scalar in $\mathbb{K}$. We omit the details.
3487: \end{proof}
3488:
3489: For details on the classical Heisenberg inequality, see, for instance, \cite%
3490: {HLP}.
3491:
3492: Utilising Proposition \ref{ch2.p4.1}, we can state the following refinement
3493: \cite{DRAG01} of the Heisenberg inequality obtained above in (\ref{ch2.5.2}):
3494:
3495: \begin{proposition}
3496: \label{ch2.p5.1}Assume that $\varphi :\left[ a,b\right] \rightarrow H$ is as
3497: in the hypothesis of Theorem \ref{ch2.t5.1}. In addition, if there exists $%
3498: r_{2},r_{1}>0$ so that
3499: \begin{equation*}
3500: \left\Vert \varphi ^{\prime }\left( t\right) +t\varphi \left( t\right)
3501: \right\Vert \geq r_{2}\geq r_{1}\geq \left\vert \left\Vert \varphi ^{\prime
3502: }\left( t\right) \right\Vert -\left\vert t\right\vert \left\Vert \varphi
3503: \left( t\right) \right\Vert \right\vert
3504: \end{equation*}%
3505: for a.e. $t\in \left[ a,b\right] ,$ then we have the inequality
3506: \begin{multline*}
3507: \left( \int_{a}^{b}t^{2}\left\Vert \varphi \left( t\right) \right\Vert
3508: ^{2}dt\cdot \int_{a}^{b}\left\Vert \varphi ^{\prime }\left( t\right)
3509: \right\Vert ^{2}dt\right) ^{\frac{1}{2}}-\frac{1}{2}\int_{a}^{b}\left\Vert
3510: \varphi \left( t\right) \right\Vert ^{2}dt \\
3511: \geq \frac{1}{2}\left( b-a\right) \left( r_{2}^{2}-r_{1}^{2}\right) \left(
3512: \geq 0\right) .
3513: \end{multline*}
3514: \end{proposition}
3515:
3516: The proof follows by Proposition \ref{ch2.p4.1} on choosing $f\left(
3517: t\right) =\varphi ^{\prime }\left( t\right) ,g\left( t\right) =-t\varphi
3518: \left( t\right) $ and $\rho \left( t\right) =\frac{1}{b-a},t\in \left[ a,b%
3519: \right] .$
3520:
3521: On utilising the Proposition \ref{ch2.p4.3} for the same choices of $f,g$
3522: and $\rho ,$ we may state the following results as well \cite{DRAG01}:
3523:
3524: \begin{proposition}
3525: \label{ch2.p5.2}Assume that $\varphi :\left[ a,b\right] \rightarrow H$ is as
3526: in the hypothesis of Theorem \ref{ch2.t5.1}. In addition, if there exists $%
3527: R\geq 1$ and $r>0$ so that
3528: \begin{equation*}
3529: \frac{1}{R}\left( \left\Vert \varphi ^{\prime }\left( t\right) \right\Vert
3530: +\left\vert t\right\vert \left\Vert \varphi \left( t\right) \right\Vert
3531: \right) \geq \left\Vert \varphi ^{\prime }\left( t\right) -t\varphi \left(
3532: t\right) \right\Vert \geq r
3533: \end{equation*}%
3534: for a.e. $t\in \left[ a,b\right] ,$ then we have the inequality
3535: \begin{multline*}
3536: \left( \int_{a}^{b}t^{2}\left\Vert \varphi \left( t\right) \right\Vert
3537: ^{2}dt\cdot \int_{a}^{b}\left\Vert \varphi ^{\prime }\left( t\right)
3538: \right\Vert ^{2}dt\right) ^{\frac{1}{2}}-\frac{1}{2}\int_{a}^{b}\left\Vert
3539: \varphi \left( t\right) \right\Vert ^{2}dt \\
3540: \geq \frac{1}{2}\left( b-a\right) \left( R^{2}-1\right) r^{2}\left( \geq
3541: 0\right) .
3542: \end{multline*}
3543: \end{proposition}
3544:
3545: Finally, we can state
3546:
3547: \begin{proposition}
3548: \label{ch2.p5.3}Let $\varphi :\left[ a,b\right] \rightarrow H$ be as in the
3549: hypothesis of Theorem \ref{ch2.t5.1}. In addition, if there exists $\zeta
3550: \in (0,1]$ so that
3551: \begin{equation*}
3552: \left\vert \left\Vert \varphi ^{\prime }\left( t\right) \right\Vert
3553: -\left\vert t\right\vert \left\Vert \varphi \left( t\right) \right\Vert
3554: \right\vert \leq \zeta \left\Vert \varphi ^{\prime }\left( t\right)
3555: +t\varphi \left( t\right) \right\Vert
3556: \end{equation*}%
3557: for a.e. $t\in \left[ a,b\right] ,$ then we have the inequality
3558: \begin{multline*}
3559: \left( \int_{a}^{b}t^{2}\left\Vert \varphi \left( t\right) \right\Vert
3560: ^{2}dt\cdot \int_{a}^{b}\left\Vert \varphi ^{\prime }\left( t\right)
3561: \right\Vert ^{2}dt\right) ^{\frac{1}{2}}-\frac{1}{2}\int_{a}^{b}\left\Vert
3562: \varphi \left( t\right) \right\Vert ^{2}dt \\
3563: \geq \frac{1}{2}\left( 1-\zeta ^{2}\right) \int_{a}^{b}\left\Vert \varphi
3564: ^{\prime }\left( t\right) +t\varphi \left( t\right) \right\Vert ^{2}dt\left(
3565: \geq 0\right) .
3566: \end{multline*}
3567: \end{proposition}
3568:
3569: This follows by Proposition \ref{ch2.p4.4} and we omit the details.
3570:
3571: \section{More Schwarz Related Inequalities}
3572:
3573: \subsection{Introduction}
3574:
3575: In practice, one may need reverses of the Schwarz inequality, namely, upper
3576: bounds for the quantities%
3577: \begin{equation*}
3578: \left\Vert x\right\Vert \left\Vert y\right\Vert -\func{Re}\left\langle
3579: x,y\right\rangle ,\qquad \left\Vert x\right\Vert ^{2}\left\Vert y\right\Vert
3580: ^{2}-\left( \func{Re}\left\langle x,y\right\rangle \right) ^{2}
3581: \end{equation*}%
3582: and%
3583: \begin{equation*}
3584: \frac{\left\Vert x\right\Vert \left\Vert y\right\Vert }{\func{Re}%
3585: \left\langle x,y\right\rangle }
3586: \end{equation*}%
3587: or the corresponding expressions where $\func{Re}\left\langle
3588: x,y\right\rangle $ is replaced by either $\left\vert \func{Re}\left\langle
3589: x,y\right\rangle \right\vert $ or $\left\vert \left\langle x,y\right\rangle
3590: \right\vert ,$ under suitable assumptions for the vectors $x,y$ in an inner
3591: product space $\left( H;\left\langle \cdot ,\cdot \right\rangle \right) $
3592: over the real or complex number field $\mathbb{K}$.
3593:
3594: In this class of results, we mention the following recent reverses of the
3595: Schwarz inequality due to the present author, that can be found, for
3596: instance, in the survey work \cite{SSD3a}, where more specific references
3597: are provided:
3598:
3599: \begin{theorem}[Dragomir, 2004]
3600: \label{ch2x.t1.4}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
3601: \right) $ be an inner product space over $\mathbb{K}$ $\left( \mathbb{K}=%
3602: \mathbb{C},\mathbb{R}\right) .$ If $a,A\in \mathbb{K}$ and $x,y\in H$ are
3603: such that either%
3604: \begin{equation}
3605: \func{Re}\left\langle Ay-x,x-ay\right\rangle \geq 0, \label{ch2x.1.8}
3606: \end{equation}%
3607: or, equivalently,%
3608: \begin{equation}
3609: \left\Vert x-\frac{A+a}{2}y\right\Vert \leq \frac{1}{2}\left\vert
3610: A-a\right\vert \left\Vert y\right\Vert , \label{ch2x.1.9}
3611: \end{equation}%
3612: then the following reverse for the quadratic form of the Schwarz inequality%
3613: \begin{align}
3614: (0& \leq )\left\Vert x\right\Vert ^{2}\left\Vert y\right\Vert
3615: ^{2}-\left\vert \left\langle x,y\right\rangle \right\vert ^{2}
3616: \label{ch2x.1.10} \\
3617: & \leq \left\{
3618: \begin{array}{l}
3619: \frac{1}{4}\left\vert A-a\right\vert ^{2}\left\Vert y\right\Vert
3620: ^{4}-\left\vert \frac{A+a}{2}\left\Vert y\right\Vert ^{2}-\left\langle
3621: x,y\right\rangle \right\vert ^{2} \\
3622: \\
3623: \frac{1}{4}\left\vert A-a\right\vert ^{2}\left\Vert y\right\Vert
3624: ^{4}-\left\Vert y\right\Vert ^{2}\func{Re}\left\langle Ay-x,x-ay\right\rangle%
3625: \end{array}%
3626: \right. \notag \\
3627: & \leq \frac{1}{4}\left\vert A-a\right\vert ^{2}\left\Vert y\right\Vert ^{4}
3628: \notag
3629: \end{align}%
3630: holds.
3631:
3632: If in addition, we have $\func{Re}\left( A\bar{a}\right) >0,$ then%
3633: \begin{equation}
3634: \left\Vert x\right\Vert \left\Vert y\right\Vert \leq \frac{1}{2}\cdot \frac{%
3635: \func{Re}\left[ \left( \bar{A}+\bar{a}\right) \left\langle x,y\right\rangle %
3636: \right] }{\sqrt{\func{Re}\left( A\bar{a}\right) }}\leq \frac{1}{2}\cdot
3637: \frac{\left\vert A+a\right\vert }{\sqrt{\func{Re}\left( A\bar{a}\right) }}%
3638: \left\vert \left\langle x,y\right\rangle \right\vert , \label{ch2x.1.11}
3639: \end{equation}%
3640: and%
3641: \begin{equation}
3642: (0\leq )\left\Vert x\right\Vert ^{2}\left\Vert y\right\Vert ^{2}-\left\vert
3643: \left\langle x,y\right\rangle \right\vert ^{2}\leq \frac{1}{4}\cdot \frac{%
3644: \left\vert A-a\right\vert ^{2}}{\func{Re}\left( A\bar{a}\right) }\left\vert
3645: \left\langle x,y\right\rangle \right\vert ^{2}. \label{ch2x.1.12}
3646: \end{equation}%
3647: Also, if (\ref{ch2x.1.8}) or (\ref{ch2x.1.9}) are valid and $A\neq -a,$ then
3648: we have the reverse for the simple form of Schwarz inequality%
3649: \begin{align}
3650: (0& \leq )\left\Vert x\right\Vert \left\Vert y\right\Vert -\left\vert
3651: \left\langle x,y\right\rangle \right\vert \leq \left\Vert x\right\Vert
3652: \left\Vert y\right\Vert -\left\vert \func{Re}\left[ \frac{\bar{A}+\bar{a}}{%
3653: \left\vert A+a\right\vert }\left\langle x,y\right\rangle \right] \right\vert
3654: \label{ch2x.1.13} \\
3655: & \leq \left\Vert x\right\Vert \left\Vert y\right\Vert -\func{Re}\left[
3656: \frac{\bar{A}+\bar{a}}{\left\vert A+a\right\vert }\left\langle
3657: x,y\right\rangle \right] \leq \frac{1}{4}\cdot \frac{\left\vert
3658: A-a\right\vert ^{2}}{\left\vert A+a\right\vert }\left\Vert y\right\Vert ^{2}.
3659: \notag
3660: \end{align}%
3661: The multiplicative constants $\frac{1}{4}$ and $\frac{1}{2}$ above are best
3662: possible in the sense that they cannot be replaced by a smaller quantity.
3663: \end{theorem}
3664:
3665: For some classical results related to Schwarz inequality, see \cite{B}, \cite%
3666: {FK}, \cite{PE}, \cite{P}, \cite{R} and the references therein.
3667:
3668: The main aim of the present section is to point out other results in
3669: connection with both the quadratic and simple forms of the Schwarz
3670: inequality. As applications, some reverse results for the generalised
3671: triangle inequality, i.e., upper bounds for the quantity%
3672: \begin{equation*}
3673: (0\leq )\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\left\Vert
3674: \sum_{i=1}^{n}x_{i}\right\Vert
3675: \end{equation*}%
3676: under various assumptions for the vectors $x_{i}\in H,$ $i\in \left\{
3677: 1,\dots ,n\right\} ,$ are established.
3678:
3679: \subsection{Refinements and Reverses\label{ch2x.s2}}
3680:
3681: The following result holds \cite{DRAG02}.
3682:
3683: \begin{proposition}
3684: \label{ch2x.p2.1}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
3685: \right) $ be an inner product space over the real or complex number field $%
3686: \mathbb{K}$. The subsequent statements are equivalent.
3687:
3688: \begin{enumerate}
3689: \item[(i)] The following inequality holds%
3690: \begin{equation}
3691: \left\Vert \frac{x}{\left\Vert x\right\Vert }-\frac{y}{\left\Vert
3692: y\right\Vert }\right\Vert \leq \left( \geq \right) r; \label{ch2x.2.1}
3693: \end{equation}
3694:
3695: \item[(ii)] The following reverse (improvement) of Schwarz's inequality holds%
3696: \begin{equation}
3697: \left\Vert x\right\Vert \left\Vert y\right\Vert -\func{Re}\left\langle
3698: x,y\right\rangle \leq \left( \geq \right) \frac{1}{2}r^{2}\left\Vert
3699: x\right\Vert \left\Vert y\right\Vert . \label{ch2x.2.2}
3700: \end{equation}%
3701: The constant $\frac{1}{2}$ is best possible in (\ref{ch2x.2.2}) in the sense
3702: that it cannot be replaced by a larger (smaller) quantity.
3703: \end{enumerate}
3704: \end{proposition}
3705:
3706: \begin{remark}
3707: \label{ch2x.r2.2}Since%
3708: \begin{align*}
3709: \left\Vert \left\Vert y\right\Vert x-\left\Vert x\right\Vert y\right\Vert &
3710: =\left\Vert \left\Vert y\right\Vert \left( x-y\right) +\left( \left\Vert
3711: y\right\Vert -\left\Vert x\right\Vert \right) y\right\Vert \\
3712: & \leq \left\Vert y\right\Vert \left\Vert x-y\right\Vert +\left\vert
3713: \left\Vert y\right\Vert -\left\Vert x\right\Vert \right\vert \left\Vert
3714: y\right\Vert \\
3715: & \leq 2\left\Vert y\right\Vert \left\Vert x-y\right\Vert
3716: \end{align*}%
3717: hence a sufficient condition for (\ref{ch2x.2.1}) to hold is%
3718: \begin{equation}
3719: \left\Vert x-y\right\Vert \leq \frac{r}{2}\left\Vert x\right\Vert .
3720: \label{ch2x.2.3}
3721: \end{equation}
3722: \end{remark}
3723:
3724: \begin{remark}
3725: \label{ch2x.r2.3}Utilising the Dunkl-Williams inequality \cite{DW}%
3726: \begin{equation}
3727: \left\Vert a-b\right\Vert \geq \frac{1}{2}\left( \left\Vert a\right\Vert
3728: +\left\Vert b\right\Vert \right) \left\Vert \frac{a}{\left\Vert a\right\Vert
3729: }-\frac{b}{\left\Vert b\right\Vert }\right\Vert ,\ \ a,b\in H\backslash
3730: \left\{ 0\right\} \label{ch2x.2.4}
3731: \end{equation}%
3732: with equality if and only if either $\left\Vert a\right\Vert =\left\Vert
3733: b\right\Vert $ or $\left\Vert a\right\Vert +\left\Vert b\right\Vert
3734: =\left\Vert a-b\right\Vert ,$ we can state the following inequality%
3735: \begin{equation}
3736: \frac{\left\Vert x\right\Vert \left\Vert y\right\Vert -\func{Re}\left\langle
3737: x,y\right\rangle }{\left\Vert x\right\Vert \left\Vert y\right\Vert }\leq
3738: 2\left( \frac{\left\Vert x-y\right\Vert }{\left\Vert x\right\Vert
3739: +\left\Vert y\right\Vert }\right) ^{2},\ \ x,y\in H\backslash \left\{
3740: 0\right\} . \label{ch2x.2.5}
3741: \end{equation}%
3742: Obviously, if $x,y\in H\backslash \left\{ 0\right\} $ are such that%
3743: \begin{equation}
3744: \left\Vert x-y\right\Vert \leq \eta \left( \left\Vert x\right\Vert
3745: +\left\Vert y\right\Vert \right) , \label{ch2x.2.6}
3746: \end{equation}%
3747: with $\eta \in (0,1],$ then one has the following reverse of the Schwarz
3748: inequality%
3749: \begin{equation}
3750: \left\Vert x\right\Vert \left\Vert y\right\Vert -\func{Re}\left\langle
3751: x,y\right\rangle \leq 2\eta ^{2}\left\Vert x\right\Vert \left\Vert
3752: y\right\Vert \label{ch2x.2.7}
3753: \end{equation}%
3754: that is similar to (\ref{ch2x.2.2}).
3755: \end{remark}
3756:
3757: The following result may be stated as well \cite{DRAG02}.
3758:
3759: \begin{proposition}
3760: \label{ch2x.t2.4}If $x,y\in H\backslash \left\{ 0\right\} $ and $\rho >0$
3761: are such that%
3762: \begin{equation}
3763: \left\Vert \frac{x}{\left\Vert y\right\Vert }-\frac{y}{\left\Vert
3764: x\right\Vert }\right\Vert \leq \rho , \label{ch2x.2.8}
3765: \end{equation}%
3766: then we have the following reverse of Schwarz's inequality%
3767: \begin{align}
3768: \left( 0\leq \right) \left\Vert x\right\Vert \left\Vert y\right\Vert
3769: -\left\vert \left\langle x,y\right\rangle \right\vert & \leq \left\Vert
3770: x\right\Vert \left\Vert y\right\Vert -\func{Re}\left\langle x,y\right\rangle
3771: \label{ch2x.2.9} \\
3772: & \leq \frac{1}{2}\rho ^{2}\left\Vert x\right\Vert \left\Vert y\right\Vert .
3773: \notag
3774: \end{align}%
3775: The case of equality holds in the last inequality in (\ref{ch2x.2.9}) if and
3776: only if%
3777: \begin{equation}
3778: \left\Vert x\right\Vert =\left\Vert y\right\Vert \qquad \text{and}\qquad
3779: \left\Vert x-y\right\Vert =\rho . \label{ch2x.2.10}
3780: \end{equation}%
3781: The constant $\frac{1}{2}$ in (\ref{ch2x.2.9}) cannot be replaced by a
3782: smaller quantity.
3783: \end{proposition}
3784:
3785: \begin{proof}
3786: Taking the square in (\ref{ch2x.2.8}), we get%
3787: \begin{equation}
3788: \frac{\left\Vert x\right\Vert ^{2}}{\left\Vert y\right\Vert ^{2}}-\frac{2%
3789: \func{Re}\left\langle x,y\right\rangle }{\left\Vert x\right\Vert \left\Vert
3790: y\right\Vert }+\frac{\left\Vert y\right\Vert ^{2}}{\left\Vert x\right\Vert
3791: ^{2}}\leq \rho ^{2}. \label{ch2x.2.11}
3792: \end{equation}%
3793: Since, obviously%
3794: \begin{equation}
3795: 2\leq \frac{\left\Vert x\right\Vert ^{2}}{\left\Vert y\right\Vert ^{2}}+%
3796: \frac{\left\Vert y\right\Vert ^{2}}{\left\Vert x\right\Vert ^{2}}
3797: \label{ch2x.2.12}
3798: \end{equation}%
3799: with equality iff $\left\Vert x\right\Vert =\left\Vert y\right\Vert ,$ hence
3800: by (\ref{ch2x.2.11}) we deduce the second inequality in (\ref{ch2x.2.9}).
3801:
3802: The case of equality and the best constant are obvious and we omit the
3803: details.
3804: \end{proof}
3805:
3806: \begin{remark}
3807: \label{ch2x.r2.5}In \cite{H}, Hile obtained the following inequality%
3808: \begin{equation}
3809: \left\Vert \left\Vert x\right\Vert ^{v}x-\left\Vert y\right\Vert
3810: ^{v}y\right\Vert \leq \frac{\left\Vert x\right\Vert ^{v+1}-\left\Vert
3811: y\right\Vert ^{v+1}}{\left\Vert x\right\Vert -\left\Vert y\right\Vert }%
3812: \left\Vert x-y\right\Vert \label{ch2x.2.12a}
3813: \end{equation}%
3814: provided $v>0$ and $\left\Vert x\right\Vert \neq \left\Vert y\right\Vert .$
3815:
3816: If in (\ref{ch2x.2.12a}) we choose $v=1$ and take the square, then we get%
3817: \begin{equation}
3818: \left\Vert x\right\Vert ^{4}-2\left\Vert x\right\Vert \left\Vert
3819: y\right\Vert \func{Re}\left\langle x,y\right\rangle +\left\Vert y\right\Vert
3820: ^{4}\leq \left( \left\Vert x\right\Vert +\left\Vert y\right\Vert \right)
3821: ^{2}\left\Vert x-y\right\Vert ^{2}. \label{ch2x.2.13}
3822: \end{equation}%
3823: Since,%
3824: \begin{equation*}
3825: \left\Vert x\right\Vert ^{4}+\left\Vert y\right\Vert ^{4}\geq 2\left\Vert
3826: x\right\Vert ^{2}\left\Vert y\right\Vert ^{2},
3827: \end{equation*}%
3828: hence, by (\ref{ch2x.2.13}) we deduce%
3829: \begin{equation}
3830: \left( 0\leq \right) \left\Vert x\right\Vert \left\Vert y\right\Vert -\func{%
3831: Re}\left\langle x,y\right\rangle \leq \frac{1}{2}\cdot \frac{\left(
3832: \left\Vert x\right\Vert +\left\Vert y\right\Vert \right) ^{2}\left\Vert
3833: x-y\right\Vert ^{2}}{\left\Vert x\right\Vert \left\Vert y\right\Vert },
3834: \label{ch2x.2.13a}
3835: \end{equation}%
3836: provided $x,y\in H\backslash \left\{ 0\right\} .$
3837: \end{remark}
3838:
3839: The following inequality is due to Goldstein, Ryff and Clarke \cite[p. 309]%
3840: {GRC}:%
3841: \begin{multline}
3842: \left\Vert x\right\Vert ^{2r}+\left\Vert y\right\Vert ^{2r}-2\left\Vert
3843: x\right\Vert ^{r}\left\Vert y\right\Vert ^{r}\cdot \frac{\func{Re}%
3844: \left\langle x,y\right\rangle }{\left\Vert x\right\Vert \left\Vert
3845: y\right\Vert } \label{ch2x.2.13.a} \\
3846: \leq \left\{
3847: \begin{array}{ll}
3848: r^{2}\left\Vert x\right\Vert ^{2r-2}\left\Vert x-y\right\Vert ^{2} & \text{%
3849: if \ }r\geq 1 \\
3850: & \\
3851: \left\Vert y\right\Vert ^{2r-2}\left\Vert x-y\right\Vert ^{2} & \text{if \ }%
3852: r<1%
3853: \end{array}%
3854: \right.
3855: \end{multline}%
3856: provided $r\in \mathbb{R}$ and $x,y\in H$ with $\left\Vert x\right\Vert \geq
3857: \left\Vert y\right\Vert .$
3858:
3859: Utilising (\ref{ch2x.2.13.a}) we may state the following proposition
3860: containing a different reverse of the Schwarz inequality in inner product
3861: spaces \cite{DRAG02}.
3862:
3863: \begin{proposition}
3864: \label{ch2x.p2.5.a}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
3865: \right) $ be an inner product space over the real or complex number field $%
3866: \mathbb{K}$. If $x,y\in H\backslash \left\{ 0\right\} $ and $\left\Vert
3867: x\right\Vert \geq \left\Vert y\right\Vert ,$ then we have%
3868: \begin{align}
3869: 0& \leq \left\Vert x\right\Vert \left\Vert y\right\Vert -\left\vert
3870: \left\langle x,y\right\rangle \right\vert \leq \left\Vert x\right\Vert
3871: \left\Vert y\right\Vert -\func{Re}\left\langle x,y\right\rangle
3872: \label{ch2x.2.13.b} \\
3873: & \leq \left\{
3874: \begin{array}{ll}
3875: \frac{1}{2}r^{2}\left( \frac{\left\Vert x\right\Vert }{\left\Vert
3876: y\right\Vert }\right) ^{r-1}\left\Vert x-y\right\Vert ^{2} & \text{if \ }%
3877: r\geq 1, \\
3878: & \\
3879: \frac{1}{2}\left( \frac{\left\Vert x\right\Vert }{\left\Vert y\right\Vert }%
3880: \right) ^{1-r}\left\Vert x-y\right\Vert ^{2} & \text{if \ }r<1.%
3881: \end{array}%
3882: \right. \notag
3883: \end{align}
3884: \end{proposition}
3885:
3886: \begin{proof}
3887: It follows from (\ref{ch2x.2.13.a}), on dividing by $\left\Vert x\right\Vert
3888: ^{r}\left\Vert y\right\Vert ^{r},$ that%
3889: \begin{multline}
3890: \left( \frac{\left\Vert x\right\Vert }{\left\Vert y\right\Vert }\right)
3891: ^{r}+\left( \frac{\left\Vert y\right\Vert }{\left\Vert x\right\Vert }\right)
3892: ^{r}-2\cdot \frac{\func{Re}\left\langle x,y\right\rangle }{\left\Vert
3893: x\right\Vert \left\Vert y\right\Vert } \label{ch2x.2.13.c} \\
3894: \leq \left\{
3895: \begin{array}{ll}
3896: r^{2}\cdot \frac{\left\Vert x\right\Vert ^{r-2}}{\left\Vert y\right\Vert ^{r}%
3897: }\left\Vert x-y\right\Vert ^{2} & \text{if \ }r\geq 1, \\
3898: & \\
3899: \frac{\left\Vert y\right\Vert ^{r-2}}{\left\Vert x\right\Vert ^{r}}%
3900: \left\Vert x-y\right\Vert ^{2} & \text{if \ }r<1.%
3901: \end{array}%
3902: \right.
3903: \end{multline}%
3904: Since%
3905: \begin{equation*}
3906: \left( \frac{\left\Vert x\right\Vert }{\left\Vert y\right\Vert }\right)
3907: ^{r}+\left( \frac{\left\Vert y\right\Vert }{\left\Vert x\right\Vert }\right)
3908: ^{r}\geq 2,
3909: \end{equation*}%
3910: hence, by (\ref{ch2x.2.13.c}) one has%
3911: \begin{equation*}
3912: 2-2\cdot \frac{\func{Re}\left\langle x,y\right\rangle }{\left\Vert
3913: x\right\Vert \left\Vert y\right\Vert }\leq \left\{
3914: \begin{array}{ll}
3915: r^{2}\frac{\left\Vert x\right\Vert ^{r-2}}{\left\Vert y\right\Vert ^{r}}%
3916: \left\Vert x-y\right\Vert ^{2} & \text{if \ }r\geq 1, \\
3917: & \\
3918: \frac{\left\Vert y\right\Vert ^{r-2}}{\left\Vert x\right\Vert ^{r}}%
3919: \left\Vert x-y\right\Vert ^{2} & \text{if \ }r<1.%
3920: \end{array}%
3921: \right.
3922: \end{equation*}%
3923: Dividing this inequality by 2 and multiplying with $\left\Vert x\right\Vert
3924: \left\Vert y\right\Vert ,$ we deduce the desired result in (\ref{ch2x.2.13.b}%
3925: ).
3926: \end{proof}
3927:
3928: Another result providing a different additive reverse (refinement) of the
3929: Schwarz inequality may be stated \cite{DRAG02}.
3930:
3931: \begin{proposition}
3932: \label{ch2x.p2.6}Let $x,y\in H$ with $y\neq 0$ and $r>0.$ The subsequent
3933: statements are equivalent:
3934:
3935: \begin{enumerate}
3936: \item[(i)] The following inequality holds:%
3937: \begin{equation}
3938: \left\Vert x-\frac{\left\langle x,y\right\rangle }{\left\Vert y\right\Vert
3939: ^{2}}\cdot y\right\Vert \leq \left( \geq \right) r; \label{ch2x.2.14}
3940: \end{equation}
3941:
3942: \item[(ii)] The following reverse (refinement) of the quadratic Schwarz
3943: inequality holds:%
3944: \begin{equation}
3945: \left\Vert x\right\Vert ^{2}\left\Vert y\right\Vert ^{2}-\left\vert
3946: \left\langle x,y\right\rangle \right\vert ^{2}\leq \left( \geq \right)
3947: r^{2}\left\Vert y\right\Vert ^{2}. \label{ch2x.2.15}
3948: \end{equation}
3949: \end{enumerate}
3950: \end{proposition}
3951:
3952: The proof is obvious on taking the square in (\ref{ch2x.2.14}) and
3953: performing the calculation.
3954:
3955: \begin{remark}
3956: \label{ch2x.r2.7}Since%
3957: \begin{align*}
3958: \left\Vert \left\Vert y\right\Vert ^{2}x-\left\langle x,y\right\rangle
3959: y\right\Vert & =\left\Vert \left\Vert y\right\Vert ^{2}\left( x-y\right)
3960: -\left\langle x-y,y\right\rangle y\right\Vert \\
3961: & \leq \left\Vert y\right\Vert ^{2}\left\Vert x-y\right\Vert +\left\vert
3962: \left\langle x-y,y\right\rangle \right\vert \left\Vert y\right\Vert \\
3963: & \leq 2\left\Vert x-y\right\Vert \left\Vert y\right\Vert ^{2},
3964: \end{align*}%
3965: hence a sufficient condition for the inequality (\ref{ch2x.2.14}) to hold is
3966: that%
3967: \begin{equation}
3968: \left\Vert x-y\right\Vert \leq \frac{r}{2}. \label{ch2x.2.16}
3969: \end{equation}
3970: \end{remark}
3971:
3972: The following proposition may give a complementary approach \cite{DRAG02}:
3973:
3974: \begin{proposition}
3975: \label{ch2x.p2.6.a}Let $x,y\in H$ with $\left\langle x,y\right\rangle \neq 0$
3976: and $\rho >0.$ If
3977: \begin{equation}
3978: \left\Vert x-\frac{\left\langle x,y\right\rangle }{\left\vert \left\langle
3979: x,y\right\rangle \right\vert }\cdot y\right\Vert \leq \rho ,
3980: \label{ch2x.2.16.a}
3981: \end{equation}%
3982: then%
3983: \begin{equation}
3984: (0\leq )\left\Vert x\right\Vert \left\Vert y\right\Vert -\left\vert
3985: \left\langle x,y\right\rangle \right\vert \leq \frac{1}{2}\rho ^{2}.
3986: \label{ch2x.2.16.b}
3987: \end{equation}%
3988: The multiplicative constant $\frac{1}{2}$ is best possible in (\ref%
3989: {ch2x.2.16.b}).
3990: \end{proposition}
3991:
3992: The proof is similar to the ones outlined above and we omit it.
3993:
3994: For the case of complex inner product spaces, we may state the following
3995: result \cite{DRAG02}.
3996:
3997: \begin{proposition}
3998: \label{ch2x.t2.8}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
3999: \right) $ be a complex inner product space and $\alpha \in \mathbb{C}$ a
4000: given complex number with $\func{Re}\alpha ,$ $\func{Im}\alpha >0.$ If $%
4001: x,y\in H$ are such that%
4002: \begin{equation}
4003: \left\Vert x-\frac{\func{Im}\alpha }{\func{Re}\alpha }\cdot y\right\Vert
4004: \leq r, \label{ch2x.2.17}
4005: \end{equation}%
4006: then we have the inequality%
4007: \begin{align}
4008: (0& \leq )\left\Vert x\right\Vert \left\Vert y\right\Vert -\left\vert
4009: \left\langle x,y\right\rangle \right\vert \leq \left\Vert x\right\Vert
4010: \left\Vert y\right\Vert -\func{Re}\left\langle x,y\right\rangle
4011: \label{ch2x.2.18} \\
4012: & \leq \frac{1}{2}\cdot \frac{\func{Re}\alpha }{\func{Im}\alpha }\cdot r^{2}.
4013: \notag
4014: \end{align}%
4015: The equality holds in the second inequality in (\ref{ch2x.2.18}) if and only
4016: if the case of equality holds in (\ref{ch2x.2.17}) and $\func{Re}\alpha
4017: \cdot \left\Vert x\right\Vert =\func{Im}\alpha \cdot \left\Vert y\right\Vert
4018: .$
4019: \end{proposition}
4020:
4021: \begin{proof}
4022: Observe that the condition (\ref{ch2x.2.17}) is equivalent to%
4023: \begin{equation}
4024: \left[ \func{Re}\alpha \right] ^{2}\left\Vert x\right\Vert ^{2}+\left[ \func{%
4025: Im}\alpha \right] ^{2}\left\Vert y\right\Vert ^{2}\leq 2\func{Re}\alpha
4026: \func{Im}\alpha \func{Re}\left\langle x,y\right\rangle +\left[ \func{Re}%
4027: \alpha \right] ^{2}r^{2}. \label{ch2x.2.19}
4028: \end{equation}%
4029: On the other hand, on utilising the elementary inequality%
4030: \begin{equation}
4031: 2\func{Re}\alpha \func{Im}\alpha \left\Vert x\right\Vert \left\Vert
4032: y\right\Vert \leq \left[ \func{Re}\alpha \right] ^{2}\left\Vert x\right\Vert
4033: ^{2}+\left[ \func{Im}\alpha \right] ^{2}\left\Vert y\right\Vert ^{2},
4034: \label{ch2x.2.20}
4035: \end{equation}%
4036: with equality if and only if $\func{Re}\alpha \cdot \left\Vert x\right\Vert =%
4037: \func{Im}\alpha \cdot \left\Vert y\right\Vert ,$ we deduce from (\ref%
4038: {ch2x.2.19}) that%
4039: \begin{equation}
4040: 2\func{Re}\alpha \func{Im}\alpha \left\Vert x\right\Vert \left\Vert
4041: y\right\Vert \leq 2\func{Re}\alpha \func{Im}\alpha \func{Re}\left\langle
4042: x,y\right\rangle +r^{2}\left[ \func{Re}\alpha \right] ^{2} \label{ch2x.2.21}
4043: \end{equation}%
4044: giving the desired inequality (\ref{ch2x.2.18}).
4045:
4046: The case of equality follows from the above and we omit the details.
4047: \end{proof}
4048:
4049: The following different reverse for the Schwarz inequality that holds for
4050: both real and complex inner product spaces may be stated as well \cite%
4051: {DRAG02}.
4052:
4053: \begin{theorem}[Dragomir, 2004]
4054: \label{ch2x.t2.9}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
4055: \right) $ be an inner product space over $\mathbb{K}$, $\mathbb{K}=\mathbb{C}%
4056: ,\mathbb{R}.$ If $\alpha \in \mathbb{K}\backslash \left\{ 0\right\} ,$ then%
4057: \begin{align}
4058: 0& \leq \left\Vert x\right\Vert \left\Vert y\right\Vert -\left\vert
4059: \left\langle x,y\right\rangle \right\vert \leq \left\Vert x\right\Vert
4060: \left\Vert y\right\Vert -\func{Re}\left[ \frac{\alpha ^{2}}{\left\vert
4061: \alpha \right\vert ^{2}}\left\langle x,y\right\rangle \right]
4062: \label{ch2x.2.22} \\
4063: & \leq \frac{1}{2}\cdot \frac{\left[ \left\vert \func{Re}\alpha \right\vert
4064: \left\Vert x-y\right\Vert +\left\vert \func{Im}\alpha \right\vert \left\Vert
4065: x+y\right\Vert \right] ^{2}}{\left\vert \alpha \right\vert ^{2}}\leq \frac{1%
4066: }{2}\cdot I^{2}, \notag
4067: \end{align}%
4068: where%
4069: \begin{equation}
4070: I:=\left\{
4071: \begin{array}{l}
4072: \max \left\{ \left\vert \func{Re}\alpha \right\vert ,\left\vert \func{Im}%
4073: \alpha \right\vert \right\} \left( \left\Vert x-y\right\Vert +\left\Vert
4074: x+y\right\Vert \right) ; \\
4075: \\
4076: \left( \left\vert \func{Re}\alpha \right\vert ^{p}+\left\vert \func{Im}%
4077: \alpha \right\vert ^{p}\right) ^{\frac{1}{p}}\left( \left\Vert
4078: x-y\right\Vert ^{q}+\left\Vert x+y\right\Vert ^{q}\right) ^{\frac{1}{q}}, \\
4079: \hfill p>1,\ \frac{1}{p}+\frac{1}{q}=1; \\
4080: \max \left\{ \left\Vert x-y\right\Vert ,\left\Vert x+y\right\Vert \right\}
4081: \left( \left\vert \func{Re}\alpha \right\vert +\left\vert \func{Im}\alpha
4082: \right\vert \right) .%
4083: \end{array}%
4084: \right. \label{ch2x.2.23}
4085: \end{equation}
4086: \end{theorem}
4087:
4088: \begin{proof}
4089: Observe, for $\alpha \in \mathbb{K}\backslash \left\{ 0\right\} ,$ that%
4090: \begin{align*}
4091: \left\Vert \alpha x-\bar{\alpha}y\right\Vert ^{2}& =\left\vert \alpha
4092: \right\vert ^{2}\left\Vert x\right\Vert ^{2}-2\func{Re}\left\langle \alpha x,%
4093: \bar{\alpha}y\right\rangle +\left\vert \alpha \right\vert ^{2}\left\Vert
4094: y\right\Vert ^{2} \\
4095: & =\left\vert \alpha \right\vert ^{2}\left( \left\Vert x\right\Vert
4096: ^{2}+\left\Vert y\right\Vert ^{2}\right) -2\func{Re}\left[ \alpha
4097: ^{2}\left\langle x,y\right\rangle \right] .
4098: \end{align*}%
4099: Since $\left\Vert x\right\Vert ^{2}+\left\Vert y\right\Vert ^{2}\geq
4100: 2\left\Vert x\right\Vert \left\Vert y\right\Vert ,$ hence%
4101: \begin{equation}
4102: \left\Vert \alpha x-\bar{\alpha}y\right\Vert ^{2}\geq 2\left\vert \alpha
4103: \right\vert ^{2}\left\{ \left\Vert x\right\Vert \left\Vert y\right\Vert -%
4104: \func{Re}\left[ \frac{\alpha ^{2}}{\left\vert \alpha \right\vert ^{2}}%
4105: \left\langle x,y\right\rangle \right] \right\} . \label{ch2x.2.24}
4106: \end{equation}%
4107: On the other hand, we have%
4108: \begin{align}
4109: \left\Vert \alpha x-\bar{\alpha}y\right\Vert & =\left\Vert \left( \func{Re}%
4110: \alpha +i\func{Im}\alpha \right) x-\left( \func{Re}\alpha -i\func{Im}\alpha
4111: \right) y\right\Vert \label{ch2x.2.25} \\
4112: & =\left\Vert \func{Re}\alpha \left( x-y\right) +i\func{Im}\alpha \left(
4113: x+y\right) \right\Vert \notag \\
4114: & \leq \left\vert \func{Re}\alpha \right\vert \left\Vert x-y\right\Vert
4115: +\left\vert \func{Im}\alpha \right\vert \left\Vert x+y\right\Vert . \notag
4116: \end{align}%
4117: Utilising (\ref{ch2x.2.24}) and (\ref{ch2x.2.25}) we deduce the third
4118: inequality in (\ref{ch2x.2.22}).
4119:
4120: For the last inequality we use the following elementary inequality%
4121: \begin{equation}
4122: \alpha a+\beta b\leq \left\{
4123: \begin{array}{ll}
4124: \max \left\{ \alpha ,\beta \right\} \left( a+b\right) & \\
4125: & \\
4126: \left( \alpha ^{p}+\beta ^{p}\right) ^{\frac{1}{p}}\left( a^{q}+b^{q}\right)
4127: ^{\frac{1}{q}}, & p>1,\ \frac{1}{p}+\frac{1}{q}=1,%
4128: \end{array}%
4129: \right. \label{ch2x.2.26}
4130: \end{equation}%
4131: provided $\alpha ,\beta ,a,b\geq 0.$
4132: \end{proof}
4133:
4134: The following result may be stated \cite{DRAG02}.
4135:
4136: \begin{proposition}
4137: \label{ch2x.p2.11}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
4138: \right) $ be an inner product over$\mathbb{\ K}$ and $e\in H,$ $\left\Vert
4139: e\right\Vert =1.$ If $\lambda \in \left( 0,1\right) ,$ then%
4140: \begin{multline}
4141: \func{Re}\left[ \left\langle x,y\right\rangle -\left\langle x,e\right\rangle
4142: \left\langle e,y\right\rangle \right] \label{ch2x.2.28} \\
4143: \leq \frac{1}{4}\cdot \frac{1}{\lambda \left( 1-\lambda \right) }\left[
4144: \left\Vert \lambda x+\left( 1-\lambda \right) y\right\Vert ^{2}-\left\vert
4145: \left\langle \lambda x+\left( 1-\lambda \right) y,e\right\rangle \right\vert
4146: ^{2}\right] .
4147: \end{multline}%
4148: The constant $\frac{1}{4}$ is best possible.
4149: \end{proposition}
4150:
4151: \begin{proof}
4152: Firstly, note that the following equality holds true%
4153: \begin{equation*}
4154: \left\langle x-\left\langle x,e\right\rangle e,y-\left\langle
4155: y,e\right\rangle e\right\rangle =\left\langle x,y\right\rangle -\left\langle
4156: x,e\right\rangle \left\langle e,y\right\rangle .
4157: \end{equation*}%
4158: Utilising the elementary inequality%
4159: \begin{equation*}
4160: \func{Re}\left\langle z,w\right\rangle \leq \frac{1}{4}\left\Vert
4161: z+w\right\Vert ^{2},\qquad z,w\in H
4162: \end{equation*}%
4163: we have%
4164: \begin{align*}
4165: & \func{Re}\left\langle x-\left\langle x,e\right\rangle e,y-\left\langle
4166: y,e\right\rangle e\right\rangle \\
4167: & =\frac{1}{\lambda \left( 1-\lambda \right) }\func{Re}\left\langle \lambda
4168: x-\left\langle \lambda x,e\right\rangle e,\left( 1-\lambda \right)
4169: y-\left\langle \left( 1-\lambda \right) y,e\right\rangle e\right\rangle \\
4170: & \leq \frac{1}{4}\cdot \frac{1}{\lambda \left( 1-\lambda \right) }\left[
4171: \left\Vert \lambda x+\left( 1-\lambda \right) y\right\Vert ^{2}-\left\vert
4172: \left\langle \lambda x+\left( 1-\lambda \right) y,e\right\rangle \right\vert
4173: ^{2}\right] ,
4174: \end{align*}%
4175: proving the desired inequality (\ref{ch2x.2.28}).
4176: \end{proof}
4177:
4178: \begin{remark}
4179: \label{ch2x.r2.12}For $\lambda =\frac{1}{2},$ we get the simpler inequality:%
4180: \begin{equation}
4181: \func{Re}\left[ \left\langle x,y\right\rangle -\left\langle x,e\right\rangle
4182: \left\langle e,y\right\rangle \right] \leq \left\Vert \frac{x+y}{2}%
4183: \right\Vert ^{2}-\left\vert \left\langle \frac{x+y}{2},e\right\rangle
4184: \right\vert ^{2}, \label{ch2x.2.29}
4185: \end{equation}%
4186: that has been obtained in \cite[p. 46]{SSD3a}, for which the sharpness of
4187: the inequality was established.
4188: \end{remark}
4189:
4190: The following result may be stated as well \cite{DRAG02}.
4191:
4192: \begin{theorem}[Dragomir, 2004]
4193: \label{ch2x.t2.13}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
4194: \right) $ be an inner product space over $\mathbb{K}$ and $p\geq 1.$ Then
4195: for any $x,y\in H$ we have%
4196: \begin{align}
4197: 0& \leq \left\Vert x\right\Vert \left\Vert y\right\Vert -\left\vert
4198: \left\langle x,y\right\rangle \right\vert \leq \left\Vert x\right\Vert
4199: \left\Vert y\right\Vert -\func{Re}\left\langle x,y\right\rangle
4200: \label{ch2x.2.30} \\
4201: & \leq \frac{1}{2}\times \left\{
4202: \begin{array}{l}
4203: \left[ \left( \left\Vert x\right\Vert +\left\Vert y\right\Vert \right)
4204: ^{2p}-\left\Vert x+y\right\Vert ^{2p}\right] ^{\frac{1}{p}}, \\
4205: \\
4206: \left[ \left\Vert x-y\right\Vert ^{2p}-\left\vert \left\Vert x\right\Vert
4207: -\left\Vert y\right\Vert \right\vert ^{2p}\right] ^{\frac{1}{p}}.%
4208: \end{array}%
4209: \right. \notag
4210: \end{align}
4211: \end{theorem}
4212:
4213: \begin{proof}
4214: Firstly, observe that%
4215: \begin{equation*}
4216: 2\left( \left\Vert x\right\Vert \left\Vert y\right\Vert -\func{Re}%
4217: \left\langle x,y\right\rangle \right) =\left( \left\Vert x\right\Vert
4218: +\left\Vert y\right\Vert \right) ^{2}-\left\Vert x+y\right\Vert ^{2}.
4219: \end{equation*}%
4220: Denoting $D:=\left\Vert x\right\Vert \left\Vert y\right\Vert -\func{Re}%
4221: \left\langle x,y\right\rangle ,$ then we have%
4222: \begin{equation}
4223: 2D+\left\Vert x+y\right\Vert ^{2}=\left( \left\Vert x\right\Vert +\left\Vert
4224: y\right\Vert \right) ^{2}. \label{ch2x.2.31}
4225: \end{equation}%
4226: Taking in (\ref{ch2x.2.31}) the power $p\geq 1$ and using the elementary
4227: inequality
4228: \begin{equation}
4229: \left( a+b\right) ^{p}\geq a^{p}+b^{p};a,b\geq 0, \label{ch2x.2.31.a}
4230: \end{equation}%
4231: we have%
4232: \begin{equation*}
4233: \left( \left\Vert x\right\Vert +\left\Vert y\right\Vert \right) ^{2p}=\left(
4234: 2D+\left\Vert x+y\right\Vert ^{2}\right) ^{p}\geq 2^{p}D^{p}+\left\Vert
4235: x+y\right\Vert ^{2p}
4236: \end{equation*}%
4237: giving%
4238: \begin{equation*}
4239: D^{p}\leq \frac{1}{2^{p}}\left[ \left( \left\Vert x\right\Vert +\left\Vert
4240: y\right\Vert \right) ^{2p}-\left\Vert x+y\right\Vert ^{2p}\right] ,
4241: \end{equation*}%
4242: which is clearly equivalent to the first branch of the third inequality in (%
4243: \ref{ch2x.2.30}).
4244:
4245: With the above notation, we also have%
4246: \begin{equation}
4247: 2D+\left( \left\Vert x\right\Vert -\left\Vert y\right\Vert \right)
4248: ^{2}=\left\Vert x-y\right\Vert ^{2}. \label{ch2x.2.32}
4249: \end{equation}%
4250: Taking the power $p\geq 1$ in (\ref{ch2x.2.32}) and using the inequality (%
4251: \ref{ch2x.2.31.a}) we deduce%
4252: \begin{equation*}
4253: \left\Vert x-y\right\Vert ^{2p}\geq 2^{p}D^{p}+\left\vert \left\Vert
4254: x\right\Vert -\left\Vert y\right\Vert \right\vert ^{2p},
4255: \end{equation*}%
4256: from where we get the last part of (\ref{ch2x.2.30}).
4257: \end{proof}
4258:
4259: \subsection{More Schwarz Related Inequalities}
4260:
4261: Before we point out other inequalities related to the Schwarz inequality, we
4262: need the following identity that is interesting in itself \cite{DRAG02}.
4263:
4264: \begin{lemma}[Dragomir, 2004]
4265: \label{ch2x.l2.14}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
4266: \right) $ be an inner product space over the real or complex number field $%
4267: \mathbb{K}$, $e\in H,$ $\left\Vert e\right\Vert =1,$ $\alpha \in H$ and $%
4268: \gamma ,\Gamma \in \mathbb{K}$. Then we have the identity:%
4269: \begin{multline}
4270: \left\Vert x\right\Vert ^{2}-\left\vert \left\langle x,e\right\rangle
4271: \right\vert ^{2} \label{ch2x.2.33} \\
4272: =\left( \func{Re}\Gamma -\func{Re}\left\langle x,e\right\rangle \right)
4273: \left( \func{Re}\left\langle x,e\right\rangle -\func{Re}\gamma \right) \\
4274: +\left( \func{Im}\Gamma -\func{Im}\left\langle x,e\right\rangle \right)
4275: \left( \func{Im}\left\langle x,e\right\rangle -\func{Im}\gamma \right) \\
4276: +\left\Vert x-\frac{\gamma +\Gamma }{2}e\right\Vert ^{2}-\frac{1}{4}%
4277: \left\vert \Gamma -\gamma \right\vert ^{2}.
4278: \end{multline}
4279: \end{lemma}
4280:
4281: \begin{proof}
4282: We start with the following known equality (see for instance \cite[eq. (2.6)]%
4283: {SSD1a})%
4284: \begin{multline}
4285: \left\Vert x\right\Vert ^{2}-\left\vert \left\langle x,e\right\rangle
4286: \right\vert ^{2} \label{ch2x.2.34} \\
4287: =\func{Re}\left[ \left( \Gamma -\left\langle x,e\right\rangle \right) \left(
4288: \overline{\left\langle x,e\right\rangle }-\bar{\gamma}\right) \right] -\func{%
4289: Re}\left\langle \Gamma e-x,x-\gamma e\right\rangle
4290: \end{multline}%
4291: holding for $x\in H,$ $e\in H,$ $\left\Vert e\right\Vert =1$ and $\gamma
4292: ,\Gamma \in \mathbb{K}$.
4293:
4294: We also know that (see for instance \cite{SSD1b})%
4295: \begin{equation}
4296: -\func{Re}\left\langle \Gamma e-x,x-\gamma e\right\rangle =\left\Vert x-%
4297: \frac{\gamma +\Gamma }{2}e\right\Vert ^{2}-\frac{1}{4}\left\vert \Gamma
4298: -\gamma \right\vert ^{2}. \label{ch2x.2.35}
4299: \end{equation}%
4300: Since%
4301: \begin{multline}
4302: \func{Re}\left[ \left( \Gamma -\left\langle x,e\right\rangle \right) \left(
4303: \overline{\left\langle x,e\right\rangle }-\bar{\gamma}\right) \right]
4304: \label{ch2x.2.36} \\
4305: =\left( \func{Re}\Gamma -\func{Re}\left\langle x,e\right\rangle \right)
4306: \left( \func{Re}\left\langle x,e\right\rangle -\func{Re}\gamma \right) \\
4307: +\left( \func{Im}\Gamma -\func{Im}\left\langle x,e\right\rangle \right)
4308: \left( \func{Im}\left\langle x,e\right\rangle -\func{Im}\gamma \right) ,
4309: \end{multline}%
4310: hence, by (\ref{ch2x.2.34}) -- (\ref{ch2x.2.36}), we deduce the desired
4311: identity (\ref{ch2x.2.33}).
4312: \end{proof}
4313:
4314: The following general result providing a reverse of the Schwarz inequality
4315: may be stated \cite{DRAG02}.
4316:
4317: \begin{proposition}
4318: \label{ch2x.t2.15}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
4319: \right) $ be an inner product space over $\mathbb{K},$ $e\in H,$ $\left\Vert
4320: e\right\Vert =1,$ $x\in H$ and $\gamma ,\Gamma \in \mathbb{K}$. Then we have
4321: the inequality:%
4322: \begin{equation}
4323: \left( 0\leq \right) \left\Vert x\right\Vert ^{2}-\left\vert \left\langle
4324: x,e\right\rangle \right\vert ^{2}\leq \left\Vert x-\frac{\gamma +\Gamma }{2}%
4325: \cdot e\right\Vert ^{2}. \label{ch2x.2.37}
4326: \end{equation}%
4327: The constant $\frac{1}{2}$ is best possible in (\ref{ch2x.2.37}). The case
4328: of equality holds in (\ref{ch2x.2.37}) if and only if%
4329: \begin{equation}
4330: \func{Re}\left\langle x,e\right\rangle =\func{Re}\left( \frac{\gamma +\Gamma
4331: }{2}\right) ,\qquad \func{Im}\left\langle x,e\right\rangle =\func{Im}\left(
4332: \frac{\gamma +\Gamma }{2}\right) . \label{ch2x.2.38}
4333: \end{equation}
4334: \end{proposition}
4335:
4336: \begin{proof}
4337: Utilising the elementary inequality for real numbers%
4338: \begin{equation*}
4339: \alpha \beta \leq \frac{1}{4}\left( \alpha +\beta \right) ^{2},\qquad \alpha
4340: ,\beta \in \mathbb{R};
4341: \end{equation*}%
4342: with equality iff $\alpha =\beta ,$ we have%
4343: \begin{equation}
4344: \left( \func{Re}\Gamma -\func{Re}\left\langle x,e\right\rangle \right)
4345: \left( \func{Re}\left\langle x,e\right\rangle -\func{Re}\gamma \right) \leq
4346: \frac{1}{4}\left( \func{Re}\Gamma -\func{Re}\gamma \right) ^{2}
4347: \label{ch2x.2.39}
4348: \end{equation}%
4349: and%
4350: \begin{equation}
4351: \left( \func{Im}\Gamma -\func{Im}\left\langle x,e\right\rangle \right)
4352: \left( \func{Im}\left\langle x,e\right\rangle -\func{Im}\gamma \right) \leq
4353: \frac{1}{4}\left( \func{Im}\Gamma -\func{Im}\gamma \right) ^{2}
4354: \label{ch2x.2.40}
4355: \end{equation}%
4356: with equality if and only if%
4357: \begin{equation*}
4358: \func{Re}\left\langle x,e\right\rangle =\frac{\func{Re}\Gamma +\func{Re}%
4359: \gamma }{2}\qquad \text{and\qquad }\func{Im}\left\langle x,e\right\rangle =%
4360: \frac{\func{Im}\Gamma +\func{Im}\gamma }{2}.
4361: \end{equation*}%
4362: Finally, on making use of (\ref{ch2x.2.39}), (\ref{ch2x.2.40}) and the
4363: identity (\ref{ch2x.2.33}), we deduce the desired result (\ref{ch2x.2.37}).
4364: \end{proof}
4365:
4366: The following result may be stated as well \cite{DRAG02}.
4367:
4368: \begin{proposition}
4369: \label{ch2x.t2.16}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
4370: \right) $ be an inner product space over $\mathbb{K},$ $e\in H,$ $\left\Vert
4371: e\right\Vert =1,$ $x\in H$ and $\gamma ,\Gamma \in \mathbb{K}$. If $x\in H$
4372: is such that%
4373: \begin{equation}
4374: \func{Re}\gamma \leq \func{Re}\left\langle x,e\right\rangle \leq \func{Re}%
4375: \Gamma \qquad \text{and\qquad }\func{Im}\gamma \leq \func{Im}\left\langle
4376: x,e\right\rangle \leq \func{Im}\Gamma , \label{ch2x.2.41}
4377: \end{equation}%
4378: then we have the inequality%
4379: \begin{equation}
4380: \left\Vert x\right\Vert ^{2}-\left\vert \left\langle x,e\right\rangle
4381: \right\vert ^{2}\geq \left\Vert x-\frac{\gamma +\Gamma }{2}e\right\Vert ^{2}-%
4382: \frac{1}{4}\left\vert \Gamma -\gamma \right\vert ^{2}. \label{ch2x.2.42}
4383: \end{equation}%
4384: The constant $\frac{1}{4}$ is best possible in (\ref{ch2x.2.42}). The case
4385: of equality holds in (\ref{ch2x.2.42}) if and only if%
4386: \begin{equation*}
4387: \func{Re}\left\langle x,e\right\rangle =\func{Re}\Gamma \text{ or }\func{Re}%
4388: \left\langle x,e\right\rangle =\func{Re}\gamma
4389: \end{equation*}%
4390: and%
4391: \begin{equation*}
4392: \func{Im}\left\langle x,e\right\rangle =\func{Im}\Gamma \text{ or }\func{Im}%
4393: \left\langle x,e\right\rangle =\func{Im}\gamma .
4394: \end{equation*}
4395: \end{proposition}
4396:
4397: \begin{proof}
4398: From the hypothesis we obviously have%
4399: \begin{equation*}
4400: \left( \func{Re}\Gamma -\func{Re}\left\langle x,e\right\rangle \right)
4401: \left( \func{Re}\left\langle x,e\right\rangle -\func{Re}\gamma \right) \geq 0
4402: \end{equation*}%
4403: and%
4404: \begin{equation*}
4405: \left( \func{Im}\Gamma -\func{Im}\left\langle x,e\right\rangle \right)
4406: \left( \func{Im}\left\langle x,e\right\rangle -\func{Im}\gamma \right) \geq
4407: 0.
4408: \end{equation*}%
4409: Utilising the identity (\ref{ch2x.2.33}) we deduce the desired result (\ref%
4410: {ch2x.2.42}). The case of equality is obvious.
4411: \end{proof}
4412:
4413: Further on, we can state the following reverse of the quadratic Schwarz
4414: inequality \cite{DRAG02}:
4415:
4416: \begin{proposition}
4417: \label{ch2x.t2.17}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
4418: \right) $ be an inner product space over $\mathbb{K},$ $e\in H,$ $\left\Vert
4419: e\right\Vert =1.$ If $\gamma ,\Gamma \in \mathbb{K}$ and $x\in H$ are such
4420: that either%
4421: \begin{equation}
4422: \func{Re}\left\langle \Gamma e-x,x-\gamma e\right\rangle \geq 0
4423: \label{ch2x.2.43}
4424: \end{equation}%
4425: or, equivalently,%
4426: \begin{equation}
4427: \left\Vert x-\frac{\gamma +\Gamma }{2}e\right\Vert \leq \frac{1}{2}%
4428: \left\vert \Gamma -\gamma \right\vert , \label{ch2x.2.44}
4429: \end{equation}%
4430: then%
4431: \begin{align}
4432: (0& \leq )\left\Vert x\right\Vert ^{2}-\left\vert \left\langle
4433: x,e\right\rangle \right\vert ^{2} \label{ch2x.2.45} \\
4434: & \leq \left( \func{Re}\Gamma -\func{Re}\left\langle x,e\right\rangle
4435: \right) \left( \func{Re}\left\langle x,e\right\rangle -\func{Re}\gamma
4436: \right) \notag \\
4437: & \qquad \qquad \qquad +\left( \func{Im}\Gamma -\func{Im}\left\langle
4438: x,e\right\rangle \right) \left( \func{Im}\left\langle x,e\right\rangle -%
4439: \func{Im}\gamma \right) \notag \\
4440: & \leq \frac{1}{4}\left\vert \Gamma -\gamma \right\vert ^{2}. \notag
4441: \end{align}%
4442: The case of equality holds in (\ref{ch2x.2.45}) if it holds either in (\ref%
4443: {ch2x.2.43}) or (\ref{ch2x.2.44}).
4444: \end{proposition}
4445:
4446: The proof is obvious by Lemma \ref{ch2x.l2.14} and we omit the details.
4447:
4448: \begin{remark}
4449: \label{ch2x.r2.18}We remark that the inequality (\ref{ch2x.2.45}) may also
4450: be used to get, for instance, the following result%
4451: \begin{multline}
4452: \left\Vert x\right\Vert ^{2}-\left\vert \left\langle x,e\right\rangle
4453: \right\vert ^{2} \label{ch2x.2.46} \\
4454: \leq \left[ \left( \func{Re}\Gamma -\func{Re}\left\langle x,e\right\rangle
4455: \right) ^{2}+\left( \func{Im}\Gamma -\func{Im}\left\langle x,e\right\rangle
4456: \right) ^{2}\right] ^{\frac{1}{2}} \\
4457: \times \left[ \left( \func{Re}\left\langle x,e\right\rangle -\func{Re}\gamma
4458: \right) ^{2}+\left( \func{Im}\left\langle x,e\right\rangle -\func{Im}\gamma
4459: \right) ^{2}\right] ^{\frac{1}{2}},
4460: \end{multline}%
4461: that provides a different bound than $\frac{1}{4}\left\vert \Gamma -\gamma
4462: \right\vert ^{2}$ for the quantity $\left\Vert x\right\Vert ^{2}-\left\vert
4463: \left\langle x,e\right\rangle \right\vert ^{2}.$
4464: \end{remark}
4465:
4466: The following result may be stated as well \cite{DRAG02}.
4467:
4468: \begin{theorem}[Dragomir, 2004]
4469: \label{ch2x.t2.19}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
4470: \right) $ be an inner product space over $\mathbb{K}$ and $\alpha ,\gamma
4471: >0, $ $\beta \in \mathbb{K}$ with $\left\vert \beta \right\vert ^{2}\geq
4472: \alpha \gamma .$ If $x,a\in H$ are such that $a\neq 0$ and%
4473: \begin{equation}
4474: \left\Vert x-\frac{\beta }{\alpha }a\right\Vert \leq \frac{\left( \left\vert
4475: \beta \right\vert ^{2}-\alpha \gamma \right) ^{\frac{1}{2}}}{\alpha }%
4476: \left\Vert a\right\Vert , \label{ch2x.2.47}
4477: \end{equation}%
4478: then we have the following reverses of Schwarz's inequality%
4479: \begin{align}
4480: \left\Vert x\right\Vert \left\Vert a\right\Vert & \leq \frac{\func{Re}\beta
4481: \cdot \func{Re}\left\langle x,a\right\rangle +\func{Im}\beta \cdot \func{Im}%
4482: \left\langle x,a\right\rangle }{\sqrt{\alpha \gamma }} \label{ch2x.2.48} \\
4483: & \leq \frac{\left\vert \beta \right\vert \left\vert \left\langle
4484: x,a\right\rangle \right\vert }{\sqrt{\alpha \gamma }} \notag
4485: \end{align}%
4486: and%
4487: \begin{equation}
4488: \left( 0\leq \right) \left\Vert x\right\Vert ^{2}\left\Vert a\right\Vert
4489: ^{2}-\left\vert \left\langle x,a\right\rangle \right\vert ^{2}\leq \frac{%
4490: \left\vert \beta \right\vert ^{2}-\alpha \gamma }{\alpha \gamma }\left\vert
4491: \left\langle x,a\right\rangle \right\vert ^{2}. \label{ch2x.2.49}
4492: \end{equation}
4493: \end{theorem}
4494:
4495: \begin{proof}
4496: Taking the square in (\ref{ch2x.2.47}), it becomes equivalent to%
4497: \begin{equation*}
4498: \left\Vert x\right\Vert ^{2}-\frac{2}{\alpha }\func{Re}\left[ \bar{\beta}%
4499: \left\langle x,a\right\rangle \right] +\frac{\left\vert \beta \right\vert
4500: ^{2}}{\alpha ^{2}}\left\Vert a\right\Vert ^{2}\leq \frac{\left\vert \beta
4501: \right\vert ^{2}-\alpha \gamma }{\alpha ^{2}}\left\Vert a\right\Vert ^{2},
4502: \end{equation*}%
4503: which is clearly equivalent to%
4504: \begin{align}
4505: \alpha \left\Vert x\right\Vert ^{2}+\gamma \left\Vert a\right\Vert ^{2}&
4506: \leq 2\func{Re}\left[ \bar{\beta}\left\langle x,a\right\rangle \right]
4507: \label{ch2x.2.50} \\
4508: & =2\left[ \func{Re}\beta \cdot \func{Re}\left\langle x,a\right\rangle +%
4509: \func{Im}\beta \cdot \func{Im}\left\langle x,a\right\rangle \right] . \notag
4510: \end{align}%
4511: On the other hand, since%
4512: \begin{equation}
4513: 2\sqrt{\alpha \gamma }\left\Vert x\right\Vert \left\Vert a\right\Vert \leq
4514: \alpha \left\Vert x\right\Vert ^{2}+\gamma \left\Vert a\right\Vert ^{2},
4515: \label{ch2x.2.51}
4516: \end{equation}%
4517: hence by (\ref{ch2x.2.50}) and (\ref{ch2x.2.51}) we deduce the first
4518: inequality in (\ref{ch2x.2.48}).
4519:
4520: The other inequalities are obvious.
4521: \end{proof}
4522:
4523: \begin{remark}
4524: \label{ch2x.r2.20}The above inequality (\ref{ch2x.2.48}) contains in
4525: particular the reverse (\ref{ch2x.1.11}) of the Schwarz inequality. Indeed,
4526: if we assume that $\alpha =1,$ $\beta =\frac{\delta +\Delta }{2},$ $\delta
4527: ,\Delta \in \mathbb{K}$, with $\gamma =\func{Re}\left( \Delta \bar{\gamma}%
4528: \right) >0,$ then the condition $\left\vert \beta \right\vert ^{2}\geq
4529: \alpha \gamma $ is equivalent to $\left\vert \delta +\Delta \right\vert
4530: ^{2}\geq 4\func{Re}\left( \Delta \bar{\gamma}\right) $ which is actually $%
4531: \left\vert \Delta -\delta \right\vert ^{2}\geq 0.$ With this assumption, (%
4532: \ref{ch2x.2.47}) becomes%
4533: \begin{equation*}
4534: \left\Vert x-\frac{\delta +\Delta }{2}\cdot a\right\Vert \leq \frac{1}{2}%
4535: \left\vert \Delta -\delta \right\vert \left\Vert a\right\Vert ,
4536: \end{equation*}%
4537: which implies the reverse of the Schwarz inequality%
4538: \begin{align*}
4539: \left\Vert x\right\Vert \left\Vert a\right\Vert & \leq \frac{\func{Re}\left[
4540: \left( \bar{\Delta}+\bar{\delta}\right) \left\langle x,a\right\rangle \right]
4541: }{2\sqrt{\func{Re}\left( \Delta \bar{\delta}\right) }} \\
4542: & \leq \frac{\left\vert \Delta +\delta \right\vert }{2\sqrt{\func{Re}\left(
4543: \Delta \bar{\delta}\right) }}\left\vert \left\langle x,a\right\rangle
4544: \right\vert ,
4545: \end{align*}%
4546: which is (\ref{ch2x.1.11}).
4547: \end{remark}
4548:
4549: The following particular case of Theorem \ref{ch2x.t2.19} may be stated \cite%
4550: {DRAG02}:
4551:
4552: \begin{corollary}
4553: \label{ch2x.c2.21}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
4554: \right) $ be an inner product space over $\mathbb{K},$ $\varphi \in \lbrack
4555: 0,2\pi ),$ $\theta \in \left( 0,\frac{\pi }{2}\right) .$ If $x,a\in H$ are
4556: such that $a\neq 0$ and%
4557: \begin{equation}
4558: \left\Vert x-\left( \cos \varphi +i\sin \varphi \right) a\right\Vert \leq
4559: \cos \theta \left\Vert a\right\Vert , \label{ch2x.2.52}
4560: \end{equation}%
4561: then we have the reverses of the Schwarz inequality%
4562: \begin{equation}
4563: \left\Vert x\right\Vert \left\Vert a\right\Vert \leq \frac{\cos \varphi
4564: \func{Re}\left\langle x,a\right\rangle +\sin \varphi \func{Im}\left\langle
4565: x,a\right\rangle }{\sin \theta }. \label{ch2x.2.52a}
4566: \end{equation}%
4567: In particular, if%
4568: \begin{equation*}
4569: \left\Vert x-a\right\Vert \leq \cos \theta \left\Vert a\right\Vert ,
4570: \end{equation*}%
4571: then%
4572: \begin{equation*}
4573: \left\Vert x\right\Vert \left\Vert a\right\Vert \leq \frac{1}{\cos \theta }%
4574: \func{Re}\left\langle x,a\right\rangle ;
4575: \end{equation*}%
4576: and if%
4577: \begin{equation*}
4578: \left\Vert x-ia\right\Vert \leq \cos \theta \left\Vert a\right\Vert ,
4579: \end{equation*}%
4580: then%
4581: \begin{equation*}
4582: \left\Vert x\right\Vert \left\Vert a\right\Vert \leq \frac{1}{\cos \theta }%
4583: \func{Im}\left\langle x,a\right\rangle .
4584: \end{equation*}
4585: \end{corollary}
4586:
4587: \subsection{Reverses of the Generalised Triangle Inequality}
4588:
4589: In \cite{SSD4a}, the author obtained the following reverse result for the
4590: generalised triangle inequality%
4591: \begin{equation}
4592: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \geq \left\Vert
4593: \sum_{i=1}^{n}x_{i}\right\Vert , \label{ch2x.3.1}
4594: \end{equation}%
4595: provided $x_{i}\in H,$ $i\in \left\{ 1,\dots ,n\right\} $ are vectors in a
4596: real or complex inner product $\left( H;\left\langle \cdot ,\cdot
4597: \right\rangle \right) :$
4598:
4599: \begin{theorem}[Dragomir, 2004]
4600: \label{ch2x.t3.1}Let $e,x_{i}\in H,$ $i\in \left\{ 1,\dots ,n\right\} $ with
4601: $\left\Vert e\right\Vert =1.$ If $k_{i}\geq 0,$ $i\in \left\{ 1,\dots
4602: ,n\right\} $ are such that%
4603: \begin{equation}
4604: \left( 0\leq \right) \left\Vert x_{i}\right\Vert -\func{Re}\left\langle
4605: e,x_{i}\right\rangle \leq k_{i}\qquad \text{for each \qquad }i\in \left\{
4606: 1,\dots ,n\right\} , \label{ch2x.3.2}
4607: \end{equation}%
4608: then we have the inequality%
4609: \begin{equation}
4610: \left( 0\leq \right) \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\left\Vert
4611: \sum_{i=1}^{n}x_{i}\right\Vert \leq \sum_{i=1}^{n}k_{i}. \label{ch2x.3.3}
4612: \end{equation}%
4613: The equality holds in (\ref{ch2x.3.3}) if and only if%
4614: \begin{equation}
4615: \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \geq \sum_{i=1}^{n}k_{i}
4616: \label{ch2x.3.4}
4617: \end{equation}%
4618: and%
4619: \begin{equation}
4620: \sum_{i=1}^{n}x_{i}=\left( \sum_{i=1}^{n}\left\Vert x_{i}\right\Vert
4621: -\sum_{i=1}^{n}k_{i}\right) e. \label{ch2x.3.5}
4622: \end{equation}
4623: \end{theorem}
4624:
4625: By utilising some of the results obtained in Section \ref{ch2x.s2}, we point
4626: out several reverses of the generalised triangle inequality (\ref{ch2x.3.1})
4627: that are corollaries of the above Theorem \ref{ch2x.t3.1} \cite{DRAG02}.
4628:
4629: \begin{corollary}
4630: \label{ch2x.c3.2}Let $e,$ $x_{i}\in H\backslash \left\{ 0\right\} ,$ $i\in
4631: \left\{ 1,\dots ,n\right\} $ with $\left\Vert e\right\Vert =1.$ If%
4632: \begin{equation}
4633: \left\Vert \frac{x_{i}}{\left\Vert x_{i}\right\Vert }-e\right\Vert \leq
4634: r_{i}\qquad \text{for each \qquad }i\in \left\{ 1,\dots ,n\right\} ,
4635: \label{ch2x.3.6}
4636: \end{equation}%
4637: then%
4638: \begin{align}
4639: (0& \leq )\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\left\Vert
4640: \sum_{i=1}^{n}x_{i}\right\Vert \label{ch2x.3.7} \\
4641: & \leq \frac{1}{2}\sum_{i=1}^{n}r_{i}^{2}\left\Vert x_{i}\right\Vert \notag
4642: \\
4643: & \leq \frac{1}{2}\times \left\{
4644: \begin{array}{ll}
4645: \left( \max\limits_{1\leq i\leq n}r_{i}\right)
4646: ^{2}\sum\limits_{i=1}^{n}\left\Vert x_{i}\right\Vert ; & \\
4647: & \\
4648: \left( \sum\limits_{i=1}^{n}r_{i}^{2p}\right) ^{\frac{1}{p}}\left(
4649: \sum\limits_{i=1}^{n}\left\Vert x_{i}\right\Vert ^{q}\right) ^{\frac{1}{q}},
4650: & p>1,\ \frac{1}{p}+\frac{1}{q}=1; \\
4651: & \\
4652: \max\limits_{1\leq i\leq n}\left\Vert x_{i}\right\Vert
4653: \sum\limits_{i=1}^{n}r_{i}^{2}. &
4654: \end{array}%
4655: \right. \notag
4656: \end{align}
4657: \end{corollary}
4658:
4659: \begin{proof}
4660: The first part follows from Proposition \ref{ch2x.p2.1} on choosing $%
4661: x=x_{i}, $ $y=e$ and applying Theorem \ref{ch2x.t3.1}. The last part is
4662: obvious by H\"{o}lder's inequality.
4663: \end{proof}
4664:
4665: \begin{remark}
4666: \label{ch2x.r3.3}One would obtain the same reverse inequality (\ref{ch2x.3.7}%
4667: ) if one were to use Theorem \ref{ch2x.t2.4}. In this case, the assumption (%
4668: \ref{ch2x.3.6}) should be replaced by%
4669: \begin{equation}
4670: \left\Vert \left\Vert x_{i}\right\Vert x_{i}-e\right\Vert \leq
4671: r_{i}\left\Vert x_{i}\right\Vert \qquad \text{for each \qquad }i\in \left\{
4672: 1,\dots ,n\right\} . \label{ch2x.3.8}
4673: \end{equation}
4674: \end{remark}
4675:
4676: On utilising the inequalities (\ref{ch2x.2.5}) and (\ref{ch2x.2.13.a}) one
4677: may state the following corollary of Theorem \ref{ch2x.t3.1} \cite{DRAG02}.
4678:
4679: \begin{corollary}
4680: \label{ch2x.c3.4}Let $e,$ $x_{i}\in H\backslash \left\{ 0\right\} ,$ $i\in
4681: \left\{ 1,\dots ,n\right\} $ with $\left\Vert e\right\Vert =1.$ Then we have
4682: the inequality%
4683: \begin{equation}
4684: (0\leq )\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\left\Vert
4685: \sum_{i=1}^{n}x_{i}\right\Vert \leq \min \left\{ A,B\right\} ,
4686: \label{ch2x.3.9}
4687: \end{equation}%
4688: where%
4689: \begin{equation*}
4690: A:=2\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert \left( \frac{\left\Vert
4691: x_{i}-e\right\Vert }{\left\Vert x_{i}\right\Vert +1}\right) ^{2},
4692: \end{equation*}%
4693: and%
4694: \begin{equation*}
4695: B:=\frac{1}{2}\sum_{i=1}^{n}\frac{\left( \left\Vert x_{i}\right\Vert
4696: +1\right) ^{2}\left\Vert x_{i}-e\right\Vert ^{2}}{\left\Vert
4697: x_{i}\right\Vert }.
4698: \end{equation*}
4699: \end{corollary}
4700:
4701: For vectors located outside the closed unit ball $\bar{B}\left( 0,1\right)
4702: :=\left\{ z\in H|\left\Vert z\right\Vert \leq 1\right\} ,$ we may state the
4703: following result \cite{DRAG02}.
4704:
4705: \begin{corollary}
4706: \label{ch2x.c3.5}Assume that $x_{i}\notin \bar{B}\left( 0,1\right) ,$ $i\in
4707: \left\{ 1,\dots ,n\right\} $ and $e\in H,$ $\left\Vert e\right\Vert =1.$
4708: Then we have the inequality:%
4709: \begin{align}
4710: (0& \leq )\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\left\Vert
4711: \sum_{i=1}^{n}x_{i}\right\Vert \label{ch2x.3.10} \\
4712: & \leq \left\{
4713: \begin{array}{ll}
4714: \dfrac{1}{2}p^{2}\sum\limits_{i=1}^{n}\left\Vert x_{i}\right\Vert
4715: ^{p-1}\left\Vert x_{i}-e\right\Vert ^{2}, & \text{if \ }p\geq 1 \\
4716: & \\
4717: \dfrac{1}{2}\sum\limits_{i=1}^{n}\left\Vert x_{i}\right\Vert
4718: ^{1-p}\left\Vert x_{i}-e\right\Vert ^{2}, & \text{if \ }p<1.%
4719: \end{array}%
4720: \right. \notag
4721: \end{align}
4722: \end{corollary}
4723:
4724: The proof follows by Proposition \ref{ch2x.p2.5.a} and Theorem \ref%
4725: {ch2x.t3.1}.
4726:
4727: For complex spaces one may state the following result as well \cite{DRAG02}.
4728:
4729: \begin{corollary}
4730: \label{ch2x.c3.6}Let $\left( H;\left\langle \cdot ,\cdot \right\rangle
4731: \right) $ be a complex inner product space and $\alpha _{i}\in \mathbb{C}$
4732: with $\func{Re}\alpha _{i},$ $\func{Im}\alpha _{i}>0,$ $i\in \left\{ 1,\dots
4733: ,n\right\} . $ If $x_{i},e\in H,$ $i\in \left\{ 1,\dots ,n\right\} $ with $%
4734: \left\Vert e\right\Vert =1$ and%
4735: \begin{equation}
4736: \left\Vert x_{i}-\frac{\func{Im}\alpha _{i}}{\func{Re}\alpha _{i}}\cdot
4737: e\right\Vert \leq d_{i},\qquad i\in \left\{ 1,\dots ,n\right\} ,
4738: \label{ch2x.3.11}
4739: \end{equation}%
4740: then%
4741: \begin{equation}
4742: (0\leq )\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\left\Vert
4743: \sum_{i=1}^{n}x_{i}\right\Vert \leq \frac{1}{2}\sum_{i=1}^{n}\frac{\func{Re}%
4744: \alpha _{i}}{\func{Im}\alpha _{i}}\cdot d_{i}^{2}. \label{ch2x.3.12}
4745: \end{equation}
4746: \end{corollary}
4747:
4748: The proof follows by Theorems \ref{ch2x.t2.8} and \ref{ch2x.t3.1} and the
4749: details are omitted.
4750:
4751: Finally, by the use of Theorem \ref{ch2x.t2.13}, we can state \cite{DRAG02}:
4752:
4753: \begin{corollary}
4754: \label{ch2x.c3.7}If $x_{i},e\in H,$ $i\in \left\{ 1,\dots ,n\right\} $ with $%
4755: \left\Vert e\right\Vert =1$ and $p\geq 1,$ then we have the inequalities:%
4756: \begin{align}
4757: (0& \leq )\sum_{i=1}^{n}\left\Vert x_{i}\right\Vert -\left\Vert
4758: \sum_{i=1}^{n}x_{i}\right\Vert \label{ch2x.3.13} \\
4759: & \leq \frac{1}{2}\times \left\{
4760: \begin{array}{l}
4761: \sum\limits_{i=1}^{n}\left[ \left( \left\Vert x_{i}\right\Vert +1\right)
4762: ^{2p}-\left\Vert x_{i}+e\right\Vert ^{2p}\right] ^{\frac{1}{p}}, \\
4763: \\
4764: \sum\limits_{i=1}^{n}\left[ \left\Vert x_{i}-e\right\Vert ^{2p}-\left\vert
4765: \left\Vert x_{i}\right\Vert -1\right\vert ^{2p}\right] ^{\frac{1}{p}}.%
4766: \end{array}%
4767: \right. \notag
4768: \end{align}
4769: \end{corollary}
4770:
4771: \begin{thebibliography}{99}
4772: \bibitem{BL} C. BLATTER, Zur Riemannschen Geometrie im Grossen auf dem M\"{o}%
4773: biusband. (German) \textit{Compositio Math.} \textbf{15} (1961), 88--107.
4774:
4775: \bibitem{BR} N.G. de BRUIJN, Problem 12, \textit{Wisk. Opgaven}, \textbf{21}
4776: (1960), 12-14.
4777:
4778: \bibitem{B} M.L. BUZANO, Generalizzazione della diseguaglianza di
4779: Cauchy-Schwarz. (Italian), \textit{Rend. Sem. Mat. Univ. e Politech. Torino,}
4780: \textbf{31} (1971/73), 405--409 (1974).
4781:
4782: \bibitem{SSD3a} S.S. DRAGOMIR, \textit{Advances in Inequalities of the
4783: Schwarz, Gr\"{u}ss and Bessel Type in Inner Product Spaces}, RGMIA
4784: Monographs, Victoria University, 2004. [ONLINE: \texttt{%
4785: http://rgmia.vu.edu.au/monographs/}].
4786:
4787: \bibitem{SSD1a} S.S. DRAGOMIR, A generalisation of Gr\"{u}ss' inequality in
4788: inner product spaces and applications, \textit{J. Mathematical Analysis and
4789: Applications}, \textbf{237} (1999), 74-82.
4790:
4791: \bibitem{SSD2} S.S. DRAGOMIR, A generalisation of Kurepa's inequality,
4792: \textit{RGMIA Res. Rep. Coll., }\textbf{7}(E) (2004), Art. 23. [ONLINE
4793: \texttt{http://rgmia.vu.edu.au/v7(E).html}]
4794:
4795: \bibitem{DRAG02} S.S. DRAGOMIR, A potpourri of Schwarz related inequalities
4796: in inner product spaces, ArXiv:math. MG/0501129v1 [ONLINE ]
4797:
4798: \bibitem{SSD4} S.S. DRAGOMIR, A refinement of Cauchy-Schwarz's inequality,
4799: \textit{Gazeta Mat. Metod}. (Bucharest, Romania),\textbf{\ 8} (1987), 94-95.
4800:
4801: \bibitem{SSD1} S.S. DRAGOMIR, \textit{Discrete Inequalities of the
4802: Cauchy-Bunyakovsky-Schwarz Type, }Nova Science Publishers, NY, 2004.
4803:
4804: \bibitem{DRAG3} S.S. DRAGOMIR, Generalizations of Precupanu's inequality for
4805: orthornormal families of vectors in inner product spaces, \textit{RGMIA Res.
4806: Rep. Coll., }\textbf{7}(E) (2004), Art. 26. [ONLINE \texttt{%
4807: http://rgmia.vu.edu.au/v7(E).html}]
4808:
4809: \bibitem{DRAG2} S.S. DRAGOMIR, Inequalities for orthornormal families of
4810: vectors in inner product spaces related to Buzano's, Richard's and Kurepa's
4811: results, \textit{RGMIA Res. Rep. Coll., }\textbf{7}(E) (2004), Art. 25.
4812: [ONLINE \texttt{http://rgmia.vu.edu.au/v7(E).html}]
4813:
4814: \bibitem{DRAG01} S.S. DRAGOMIR, Refinements of the Schwarz and Heisenberg
4815: inequalities in Hilbert spaces, \textit{J. Inequal. Pure \& Appl. Math.},
4816: \textbf{5}(3) (2004), Art. 60. [ONLINE: \texttt{%
4817: http://jipam.vu.edu.au/article.php?sid=446}]
4818:
4819: \bibitem{SSD4a} S.S. DRAGOMIR, Reverses of the triangle inequality in inner
4820: product spaces, \textit{RGMIA Res. Rep. Coll}., \textbf{7}(E) (2004),
4821: Article 7. [ONLINE: \texttt{http://rgmia.vu.edu.au/v7(E).html}].
4822:
4823: \bibitem{SSD1b} S.S. DRAGOMIR, Some Gr\"{u}ss type inequalities in inner
4824: product spaces, \textit{J. Inequal. Pure \& Appl. Math.}, \textbf{4}(2)
4825: (2003), Article 42. [Online: http://jipam.vu.edu.au/article.php?sid=280].
4826:
4827: \bibitem{SSD3} S.S. DRAGOMIR, Some refinements of Schwartz inequality,
4828: Simpozionul de Matematici \c{s}i Aplica\c{t}ii, Timi\c{s}oara, Romania, 1-2
4829: Noiembrie 1985, 13--16.
4830:
4831: \bibitem{DRAG1} S.S. DRAGOMIR, Refinements of Buzano's and Kurepa's
4832: inequalities in inner product spaces, \textit{RGMIA Res. Rep. Coll., }%
4833: \textbf{7}(E) (2004), Art. 24. [ONLINE \texttt{%
4834: http://rgmia.vu.edu.au/v7(E).html}]
4835:
4836: \bibitem{DM} S.S. DRAGOMIR and B. MOND, On the superadditivity and
4837: monotonicity of Schwarz's inequality in inner product spaces, \textit{%
4838: Contributios, Macedonian Acad. Sci. Arts., }\textbf{15}(2) (1994), 5-22.
4839:
4840: \bibitem{DS} S.S. DRAGOMIR and J. S\'{A}NDOR, Some inequalities in
4841: prehilbertian spaces, \textit{Studia Univ., Babe\c{s}-Bolyai, Mathematica, }%
4842: \textbf{32}(1)(1987), 71-78 MR 89h: 46034.
4843:
4844: \bibitem{DS1} S.S. DRAGOMIR and J. S\'{A}NDOR, Some inequalities in
4845: prehilbertian spaces, \textit{Conferin\c{t}a Na\c{t}ional\u{a} de Geometrie
4846: \c{s}i Topologie}, Targovi\c{s}te, Romania, 12-14 Aprilie, 1986, 73-76.
4847:
4848: \bibitem{DW} C.F. DUNKL and K.S. WILLIAMS, A simple norm inequality, \textit{%
4849: The Amer. Math. Monthly, }\textbf{71}(1) (1964), 43-54.
4850:
4851: \bibitem{FK} M. FUJII and F. KUBO, Buzano's inequality and bounds for roots
4852: of algebraic equations, \textit{Proc. Amer. Math. Soc., }\textbf{117}(2)
4853: (1993), 359-361.
4854:
4855: \bibitem{GRC} A.A. GOLDSTEIN, J.V. RYFF and L.E. CLARKE, Problem 5473,
4856: \textit{The Amer. Math. Monthly, }\textbf{75}(3) (1968), 309.
4857:
4858: \bibitem{HLP} G.H. HARDY, J.E. LITTLEWOOD and G. POLYA, \textit{Inequalities}%
4859: , Cambridge University Press, Cambridge, United Kingdom, 1952.
4860:
4861: \bibitem{H} G.N. HILE, Entire solution of linear elliptic equations with
4862: Laplacian principal part, \textit{Pacific J. Math., }\textbf{62} (1976),
4863: 127-148.
4864:
4865: \bibitem{KU} S. KUREPA, On the Buniakowsky-Cauchy-Schwarz inequality,
4866: \textit{Glasnick Mathemati\v{c}ki, }\textbf{1}(\textbf{21})(2) (1966),
4867: 147-158.
4868:
4869: \bibitem{MIT} D.S. MITRINOVI\'{C}, \textit{Analytic Inequalities}, Springer
4870: Verlag, 1970.
4871:
4872: \bibitem{M} M.H. MOORE, An inner product inequality, \textit{SIAM J. Math.
4873: Anal.}, \textbf{4}(1973), No. 3, 514-518.
4874:
4875: \bibitem{PE} J.E. PE\v{C}ARI\'{C}, On some classical inequalities in unitary
4876: spaces, \textit{Mat. Bilten, (Macedonia) }\textbf{16} (1992), 63-72.
4877:
4878: \bibitem{P} T. PRECUPANU, On a generalisation of Cauchy-Buniakowski-Schwarz
4879: inequality, \textit{Anal. St. Univ. \textquotedblleft Al. I.
4880: Cuza\textquotedblright\ Ia\c{s}i}, \textbf{22}(2) (1976), 173-175.
4881:
4882: \bibitem{R} U. RICHARD, Sur des in\'{e}galit\'{e}s du type Wirtinger et
4883: leurs application aux \'{e}quations diff\'{e}rentielles ordinaires,
4884: Collquium of Analysis held in Rio de Janeiro, August, 1972, pp. 233-244.
4885: \end{thebibliography}
4886: