1: \documentclass[12pt]{article}

2: \usepackage{amssymb,amsmath}

3: \begin{document}

4: \newtheorem{theorem}{Theorem}

5: \newtheorem{lemma}{Lemma}

6: \newtheorem{proposition}{Proposition}

7: \newtheorem{definition}{Definition}

8: \begin{center}{\Large Introductory Clifford Analysis}

9: \end{center}

10: \begin{center}{by John Ryan}\end{center}

11: \begin{center}{Department of Mathematics, University of Arkansas}

12: \end{center}

13: \begin{center}{Fayetteville, AR 72701, U. S. A.}\end{center}

14: \begin{center}{\Large Introduction}\end{center}

15: 

16: \ We want in this chapter to regard Clifford algebras as natural generalizations of the complex number system. First let us note that if $z$ is a complex number then

17: $\overline{z}z=\|z\|^{2}$. Now for a quaternion $q$ we also have

18: $\overline{q}q=\|q\|^{2}$. In this way quaternions may be regarded

19: as a generalization of the complex number system. It seems natural

20: to ask if one can extend basic results of one complex variable

21: analysis on holomorphic function theory to four dimensions using

22: quaternions. The answer is yes. This was developed by the Swiss

23: mathematician Rudolph Fueter in the 1930's and 1940's and also my Moisil and Teodorecu \cite{mt}. See for

24: instance \cite{f}. An excellent review of this work is given in the

25: survey article "Quaternionic analysis" by A. Sudbery, see \cite{su}. There is

26: also earlier work of Dixon \cite{di}.

27: However in previous lectures we have seen that for a vector $x\in

28: R^{n}$ when we consider $R^{n}$ as embedded in the Clifford

29: algebra $Cl_{n}$ then $x^{2}=-\|x\|^{2}$. So it seems reasonable

30: to ask if all that is known in the quaternionic setting extends to

31: the Clifford algebra setting. Again the answer is yes. The earlier aspects

32: of this study was developed by amongst others Richard

33: Delanghe \cite{d}, Viorel Iftimie \cite{i} and David Hestenes \cite{h}. The subject that

34: has grown from these works is now called Clifford analysis.

35: 

36: \ In more recent times Clifford analysis has found a wealth of unexpected

37: applications in a number of branches of mathematical analysis particularly

38: classical harmonic analysis, see for instance the work of Alan McIntosh

39: and his collaborators, for instance \cite{lms,lmq}, Marius Mitrea \cite{m1,m2}

40: and papers in \cite{r4}. Links to representation theory and several complex variables may be

41: found in \cite{gm,r1,r2,r3} and elsewhere.

42: 

43: \ The purpose of this paper is to present a review of many of the basic

44: aspects of Clifford analysis.

45: 

46: \ Alternative accounts of much of this work together with other related

47: results can be found in \cite{bds,dss,gs,gm,ks,o,r4,r5}.

48: 

49: 

50: \begin{center}{\Large Foundations of Clifford Analysis}\end{center}

51: 

52: \ We start by replacing the vector $x=x_{1}e_{1}+\ldots+x_{n}e_{n}$ by the

53: differential operator $D=\Sigma_{j=1}^{n}e_{j}\frac{\partial}{\partial x_{j}}$.

54: One basic but interesting property of $D$ is that $D^{2}=-\triangle_{n}$,

55: the Laplacian $\Sigma_{j=1}^{n}\frac{\partial^{2}}{\partial x_{j}^{2}}$ in

56: ${\bf{R}}^{n}$. The differential operator $D$ will be called a Dirac operator.

57: This is because the classical Dirac operator constructed over four dimensional

58: Minkowski space squares to give the wave operator.

59: 

60: \begin{definition}

61: Suppose that $U$ is a domain in ${\bf{R}}^{n}$ and $f$ and $g$ are $C^{1}$

62: functions defined on $U$ and taking values in $Cl_{n}$. Then $f$ is called a

63: left monogenic function if $Df=0$ on $U$ while $g$ is called a right

64: monogenic function on $U$ if $gD=0$ where $gD=\Sigma_{j=1}^{n}

65: \frac{\partial g}{\partial x_{j}}e_{j}$.

66: \end{definition}

67: 

68: \ Alternatively left monogenic functions are also called left regular

69: functions and perhaps most appropriately left Clifford holomorphic

70: functions. The term Clifford holomorphic functions or Clifford analytic functions appears to be due to Stephen Semmes, see \cite{se} and elsewhere. We shall most often use the term Clifford holomorphic functions here.

71: 

72: \ Examples of such functions include the gradients of real valued harmonic

73: functions on $U$. So if $h$ is harmonic on $U$ and it is also real valued

74: then $Dh$ is a vector valued left monogenic function. It is also a right

75: monogenic function. Such a function is more commonly referred to as a

76: conjugate harmonic function or a harmonic $1$-form. See for instance

77: \cite{sw}. An example of such a function is $G(x)=\frac{x}{\|x\|^{n}}$.

78: 

79: \ It should be noted that if $f$ and $g$ are left monogenic functions then

80: due to the lack of commutativity of the Clifford algebra, it is not in

81: general true that their product $f(x)g(x)$ is left monogenic.

82: 

83: \ To introduce other possible examples of left monogenic functions suppose

84: that $\mu$ is a $Cl_{n}$ valued measure with compact support $[\mu]$ in

85: ${\bf{R}}^{n}$. Then the convolution $\int_{[\mu]}G(x-y)d\mu(y)$ defines a

86: left monogenic function on the maximal domain lying in

87: ${\bf{R}}^{n}\backslash[\mu]$. The previously defined integral is the

88: Cauchy transform of the measure $[\mu]$.

89: 

90: \ Another way to construct examples of left monogenic functions was introduced

91: by Littlewood and Gay in \cite{lg} for the case $n=3$ and independently

92: re-introduced for all $n$ by Sommen \cite{s2}. Suppose $U'$ is a domain in

93: ${\bf{R}}^{n-1}$, the span of $e_{2},\ldots,e_{n}$. Suppose also that

94: $f'(x')$ is a $Cl_{n}$ valued function such that at each point $x'\in U'$

95: there is a multiple series expansion in $x_{2},\ldots,x_{n}$ that converges

96: uniformly on some neighbourhood of $x'$ in $U'$ to $f'$. Such a function is

97: called a real analytic function. The series

98: \[\Sigma_{k=0}^{\infty}\frac{1}{k!}x_{1}^{k}(-e_{1}D'f'(x')=\exp(-x_{1}e_{1}D')

99: f'(x')\]

100: where $D'=\Sigma_{j=2}^{n}e_{j}\frac{\partial}{\partial x_{j}}$, defines a

101: left monogenic function $f$ in some neighborhood $U(f')$ in ${\bf{R}}^{n}$ of

102: $U'$. The left monogenic function $f$ is the Cauchy-Kowalewska extension of

103: $f'$.

104: 

105: \ It should be noted that if $f$ is a left monogenic function then

106: $\overline{f}$ and $\tilde{f}$ are both right monogenic functions.

107: 

108: \ We now turn to analogues of Cauchy's Theorem and Cauchy's integral formula.

109: \begin{theorem}{\bf{(The Clifford-Cauchy Theorem):}} Suppose that $f$ is

110: a

111: left Clifford holomorphic function on $U$ and $g$ is a right Clifford holomorphic function on $U$.

112: Suppose also that $V$ is a bounded subdomain of $U$ with piecewise

113: differentiable boundary $S$ lying in $U$. Then

114: \begin{equation}

115: \int_{S}g(x)n(x)f(x)d\sigma(x)=0

116: \end{equation}

117: where $n(x)$ is the outward pointing normal vector to $S$ at $x$ and $\sigma$

118: is the Lebesgue measure on $S$.

119: \end{theorem}

120: 

121: \ The proof follows directly from Stokes' Theorem. One important point to

122: keep

123: in mind though is that as $Cl_{n}$ is not a commutative algebra then it is

124: important to place the vector $n(x)$ between $f$ and $g$. One then has that

125: \[\int_{S}g(x)n(x)f(x)d\sigma(x)=\int_{V}((g(x)D)f(x)+g(x)(Df(x)))dx^{n}=0.\]

126: 

127: \ Suppose that $g$ is the gradient of a real valued harmonic function and

128: $f=1$. Then the real part of Equation 1 gives the following well known integral

129: formula.

130: \[\int<grad g(x),n(x)>d\sigma(x)=0.\]

131: 

132: \ We now turn to the analogue of a Cauchy integral formula.

133: \begin{theorem}{\bf{(Clifford-Cauchy Integral Formula):}}

134: Suppose that $U$, $V$, $S$, $f$ and $g$ are all as in Theorem 1 and that

135: $y\in V$. Then

136: \[f(y)=\frac{1}{\omega_{n}}\int_{S}G(x-y)n(x)f(x)d\sigma(x)\]

137: and

138: \[g(y)=\frac{1}{\omega_{n}}\int_{S}g(x)n(x)G(x-y)d\sigma(x)\]

139: where $\omega_{n}$ is the surface area of the unit sphere in ${\bf{R}}^{n}$.

140: \end{theorem}

141: {\bf{Proof:}} The proof follows very similar lines to the argument in one

142: variable complex analysis. We shall establish the formula for $f(y)$ the proof

143: being similar for $g(y)$. First let us take a sphere $S^{n-1}(y,r)$ centered

144: at $y$ and of radius $r$. The radius $r$ is chosen sufficently small so that

145: the closed disc with boundary $S^{n-1}(y,r)$ lies in $V$. Then by the

146: Clifford-Cauchy theorem

147: \[\int_{S}G(x-y)n(x)f(x)d\sigma(x)=

148: \int_{S^{n-1}(y,r)}G(x-y)n(x)f(x)d\sigma(x).\]

149: However on $S^{n-1}(y,r)$ the vector $n(x)=\frac{y-x}{\|x-y\|}$. So

150: $G(x-y)n(x)=\frac{1}{r^{n-1}}$. So

151: \[\int_{S^{n-1}(y,r)}G(x-y)n(x)f(x)d\sigma(x)=\int_{S^{n-1}(y,r)}

152: \frac{1}{r^{n-1}}(f(x)-f(y))d\sigma(x)\]

153: \[+\int_{S^{n-1}(y,r)}\frac{f(y)}{r^{n-1}}

154: d\sigma(x).\]

155: The right side of this previous expression reduces to

156: \[\int_{S^{n-1}(y,r)}\frac{(f(x)-f(y))}{r^{n-1}}d\sigma(x)+f(y)\int_{S^{n-1}}

157: d\sigma(x).\]

158: Now $\int_{S^{n-1}}d\sigma(x)=\omega_{n}$ and by continuity

159: $\lim_{r\rightarrow 0}\int_{S^{n-1}(y,r)}

160: \frac{(f(x)-f(y)}{r^{n-1}}d\sigma(x)=0$. The result follows. $\Box$

161: 

162: \ One important feature is to note that Kelvin inversion, ie $x^{-1}=\frac{-x}{\|x\|^{2}}$ whenever $x$ is non-zero, plays a

163: fundamental

164: role in this proof. Moreover the proof is almost exactly the same as the

165: proof of Cauchy's Integral Formula for piecewise $C^{1}$ curves in one

166: variable complex analysis.

167: 

168: \ Having obtained a Cauchy Integral Formula in ${\bf{R}}^{n}$ a number of

169: basic results that one might see in a first course in one variable complex

170: analysis carry over more or less automatically to the context described here.

171: This includes a Liouville Theorem and Weierstrass Convergence Theorem. We

172: leave it as an exercise to the interested reader to set up and establish the

173: Clifford analysis analogues of these results. Their statements and proofs can

174: be found in \cite{bds}.

175: 

176: \ Theorems 1 and 2 show us that the individual components of the equations

177: $Df=0$ and $gD=0$ comprise generalized Cauchy-Riemann equations. In the

178: particular case where $f$ is just vector valued so

179: $f=\Sigma_{j=1}^{n}f_{j}e_{j}$ then the generalized Cauchy-Riemann equations

180: become

181: $\frac{\partial f_{j}}{\partial x_{i}}=\frac{\partial f_{i}}{\partial x_{j}}$

182: whenever $i\neq j$ and $\Sigma_{j=1}^{n}\frac{\partial f_{j}}{\partial x_{j}}

183: =0$. This system of equations is often referred to as the Riesz system.

184: 

185: \ Having obtained an analogue of Cauchy's integral formula in euclidean

186: space we shall now exploit this result to show how many

187: consequences of the classical Cauchy integral carry over to the context

188: described here. We begin with the Mean Value Theorem.

189: 

190: \begin{theorem}

191: Suppose that $D(y, R)$ is a closed disc centered at $y$, of radius $R$ and

192: lying in $U$. Then for each left Clifford holomorphic function $f$ on $U$

193: \[f(y)=\frac{1}{R\omega_{n}}\int_{D(y, R)}\frac{f(x)}{\|x-y\|^{n-1}}dx^{n}.\]

194: \end{theorem}

195: {\bf{Proof:}}

196: We have already seen that for each $r\in(0, R)$

197: \[f(y)=\frac{1}{\omega_{n}}\int_{S^{n-1}(y, r)}\frac{f(x)}{\|x-y\|^{n-1}}

198: d\sigma(x),\]

199: where $S^{n-1}(y, r)$ is the $(n-1)$-dimensional sphere centered at $y$ and

200: of radius $r$. We obtain the result by integrating both sides of this

201: expression with respect to the variable $r$ and dividing throughout by $R$.

202: $\Box$

203: 

204: \ Let us now turn to explore the real analyticity properties of Clifford holomorphic

205: functions. First it may be noted that when $n$ is even

206: $G(x-y)=(-1)^{\frac{n-2}{2}}(x-y)^{-n+1}$. Also $(x-y)^{-1}=

207: x^{-1}(1-yx^{-1})^{-1}=(1-x^{-1}y)^{-1}x^{-1}$, and $\|x^{-1}y\|=\|yx^{-1}\|

208: =\frac{\|y\|}{\|x\|}$. So for $\|y\|<\|x\|$

209: \[(x-y)^{-1}=x^{-1}(1+yx^{-1}+\ldots+yx^{-1}\ldots yx^{-1}+\ldots)\]

210: \[=(1+x^{-1}y+\ldots+x^{-1}y\ldots x^{-1}y+\ldots)x^{-1}.\]

211: Hence these two sequences converge uniformly to $(x-y)^{-1}$ provided

212: $\|y\|\leq r<\|x\|$ and they converge pointwise to $(x-y)^{-1}$ provided

213: $\|y\|<\|x\|$. One can now take $(-1)^{\frac{n-2}{2}}$ times the $(n-1)$-fold

214: product of the series expansions of $(x-y)^{-1}$ with itself to obtain a

215: series expansion for $G(x-y)$. In this process of multiplying series

216: together in order to maintain the same radius of convergence one needs to

217: group together all linear combinations of monomials in $y_{1},\ldots, y_{n}$

218: that are of the same order. Thus we have deduced that when $n$ is even the

219: multiple Taylor series expansion

220: \[\Sigma_{j=0}^{\infty}(\Sigma_{\stackrel{j_{1},\ldots,j_{n}}{j_{1}+

221: \ldots+j_{n}=j}}

222: \frac{y_{1}^{j_{1}}\ldots y_{n}^{j_{n}}}{j_{1}!\ldots j_{n}!}

223: \frac{\partial^{j}G(x)}{\partial x_{1}^{j_{1}}\ldots \partial

224: x_{n}^{j_{n}}})\]

225: converges uniformly to $G(x-y)$ provided $\|y\|<r<\|x\|$ and converges

226: pointwise to $G(x-y)$ provided $\|y\|<\|x\|$.

227: 

228: \ A similar argument may be developed when $n$ is odd.

229: 

230: \ Returning to Cauchy's integral formula let us suppose that $f$ is a left

231: Clifford holomorphic function defined in a neighbourhood of the closure of some ball

232: $B(0, R)$. Then

233: \[f(y)=\frac{1}{\omega_{n}}\int_{\partial B(0,R)}G(x-y)n(x)f(x)d\sigma(x)=\]

234: \[\frac{1}{\omega_{n}}\int_{\partial B(0,R)}\Sigma_{j=0}^{\infty}

235: (\Sigma_{\stackrel{j_{1}\ldots j_{n}}{j_{1}+\ldots+j_{n}=j}}

236: \frac{y_{1}^{j_{1}}\ldots y_{n}^{j_{n}}}{j_{1}!\ldots j_{n}!}

237: \frac{\partial^{j}G(x)}{\partial x_{1}^{j_{1}}\ldots\partial x_{n}^{j_{n}}})

238: n(x)f(x)d\sigma(x)\]

239: provided $\|y\|<\|x\|$. As this series converges uniformly on each ball

240: $B(0, r)$ for each $r<R$ then this last integral can be re-written as

241: \[\frac{1}{\omega_{n}}\Sigma_{j=0}^{\infty}\int_{\partial B(0, R)}

242: (\Sigma_{\stackrel{j_{1}\ldots j_{n}}{j_{1}+\ldots+j_{n}=j}}

243: \frac{y_{1}^{j_{1}}\ldots y_{n}^{j_{n}}}{j_{1}!\ldots j_{n}!}

244: \frac{\partial^{j}G(x)}{\partial x_{1}^{j_{1}}\ldots \partial x_{n}^{j_{n}}}

245: n(x)f(x))d\sigma(x).\]

246: As the summation within the parentheses is a finite summation this last

247: expression easily reduces to

248: \[\frac{1}{\omega_{n}}\Sigma_{j=0}^{\infty}

249: (\Sigma_{\stackrel{j_{1}\ldots j_{n}}{j_{1}+\ldots+j_{n}=j}}

250: \frac{y_{1}^{j_{1}}\ldots y_{n}^{j_{n}}}{j_{1}!\ldots j_{n}!}

251: \int_{\partial B(0, R)}\frac{\partial^{j}G(x)}{\partial x_{1}^{j_{1}}\ldots

252: \partial x_{n}^{j_{n}}}n(x)f(x))d\sigma(x).\]

253: On placing

254: \[\frac{1}{\omega_{n}}\int_{\partial B(0, R)}\frac{\partial^{j}G(x)}

255: {\partial x_{1}^{j_{1}}\ldots\partial x_{n}^{j_{n}}}n(x)f(x)d\sigma(x)

256: =a_{j_{1}\ldots j_{n}}\]

257: it may be seen that on $B(0, R)$ the series

258: \[\Sigma_{j=0}^{\infty}

259: (\Sigma_{\stackrel{j_{1}\ldots j_{n}}{j_{1}+\ldots+j_{n}=j}}

260: \frac{x_{1}^{j_{1}}\ldots x_{n}^{j_{n}}}{j_{1}!\ldots j_{n}!}

261: a_{j_{1}\ldots j_{n}})\]

262: converges pointwise to $f(y)$. Convergence is uniform on each ball

263: $B(0, r)$ provided $r<R$.

264: 

265: \ Similarly if $g$ is a right Clifford holomorphic function defined in a neighbourhood

266: of the closure of $B(0, R)$ then the series

267: \[\Sigma_{j=0}^{\infty}

268: (\Sigma_{\stackrel{j_{1}\ldots j_{n}}{j_{1}+\ldots+j_{n}=j}}

269: b_{j_{1}\ldots j_{n}}\frac{y_{1}^{j_{1}}\ldots y_{n}^{j_{n}}}

270: {j_{1}!\ldots j_{n}!})\]

271: converges pointwise on $B(0, R)$ to $g(y)$ and converges uniformly on

272: $B(0, r)$ for $r<R$, where

273: \[b_{j_{1}\ldots j_{n}}=\frac{1}{\omega_{n}}\int_{\partial B(0, R)}

274: g(x)n(x)\frac{\partial^{j}G(x)}

275: {\partial x_{1}^{j_{1}}\ldots \partial x_{n}^{j_{n}}}d\sigma(x).\]

276: 

277: \ By translating the ball $B(0, R)$ to the ball $B(w, R)$ where

278: $w=w_{1}e_{1}+\ldots+w_{n}e_{n}$ one may readily observe that for any left

279: Clifford holomorphic function $f$ defined in a neighbourhood of the closure of $B(w, R)$

280: the series

281: \[\Sigma_{j=0}^{\infty}

282: (\Sigma_{\stackrel{j_{1}\ldots j_{n}}{j_{1}+\ldots j_{n}=j}}

283: \frac{(y_{1}-w_{1})^{j_{1}}\ldots (y_{n}-w_{n})^{j_{n}}}

284: {j_{1}!\ldots j_{n}!}a'_{j_{1}\ldots j_{n}})\]

285: converges pointwise on $B(w, R)$ to $f(y)$, where

286: \[a'_{j_{1}\ldots j_{n}}=\frac{1}{\omega_{n}}\int_{\partial B(w, R)}

287: \frac{\partial^{j}G(x-w)}{\partial x_{1}^{j_{1}}\ldots x_{n}^{j_{n}}}

288: n(x)f(x)d\sigma(x).\]

289: Again the series converges uniformly on $B(w, r)$ for each $r<R$.

290: A similar series may be readily obtained for any right Clifford holomorphic function

291: defined in a neighbourhood of the closure of $B(w, R)$.

292: 

293: \ The types of power series that we have developed for left Clifford holomorphic

294: functions are not entirely satisfactory. In particular, unlike their complex

295: analogues the homogeneous polynomials

296: \[\Sigma_{\stackrel{j_{1}\ldots j_{n}}{j_{1}+\ldots+j_{n}=j}}

297: \frac{x_{1}^{j_{1}}\ldots x_{n}^{j_{n}}}{j_{1}!\ldots j_{n}!}

298: a_{j_{1}\ldots j_{n}}\]

299: are not expressed as a linear combination of left Clifford holomorphic polynomials. To

300: rectify this situation let us first take a closer look at the Taylor

301: expansion for the Cauchy kernel $G(x-y)$ where all the Taylor coefficients are

302: real. Let us first look at the first order terms in the Taylor expansion.

303: This is the expression

304: \[y_{1}\frac{\partial G(x)}{\partial x_{1}}+\ldots+y_{n}

305: \frac{\partial G(x)}{\partial x_{n}}.\]

306: As $G$ is a Clifford holomorphic function then $\frac{\partial G(x)}{\partial x_{1}}=

307: -\Sigma_{j=2}^{n}e_{1}^{-1}e_{j}\frac{\partial G(x)}{\partial x_{j}}$.

308: Therefore the first order terms of the Taylor expansion for $G(x-y)$ can be

309: re-expressed as

310: \[\Sigma_{j=2}^{n}(y_{j}-e_{1}^{-1}e_{j}y_{1})\frac{\partial G(x)}

311: {\partial x_{j}}.\]

312: Moreover, for $2\leq j\leq n$ the first order polynomial

313: $y_{j}-e_{1}^{-1}e_{j}y_{1}$ is a left Clifford holomorphic polynomial.

314: Let us now go to second order terms. Again we will replace the operator

315: $\frac{\partial}{\partial x_{1}}$ by the operator $-\Sigma_{j=2}^{n}

316: e_{1}^{-1}e_{j}\frac{\partial}{\partial x_{j}}$ whenever it arises. Let us

317: consider the term $\frac{\partial^{2}G(x)}{\partial x_{i}\partial x_{j}}$

318: where $i\neq j\neq 1$. We end up with the polynomial

319: $y_{i}y_{j}-y_{i}y_{1}e_{1}^{-1}e_{j}-y_{j}y_{1}e_{1}^{-1}e_{i}

320: =\frac{1}{2}((y_{i}-y_{1}e_{1}^{-1}e_{i})(y_{j}-y_{1}e_{1}^{-1}e_{j})

321: +(y_{j}-y_{1}e_{1}^{-1}e_{j})(y_{i}-y_{1}e_{1}^{-1}e_{i})$.

322: Similarly the polynomial attached to the term $\frac{\partial^{2}G(x)}

323: {\partial x_{i}^{2}}$ is $(y_{i}-y_{1}e_{1}^{-1}e_{i})^{2}$. Using the

324: Clifford algebra anti-commutation relationship $e_{i}e_{j}+e_{j}e_{i}=-2

325: \delta_{ij}$ and on replacing the differential operator $\frac{\partial}

326: {\partial x_{1}}$ by the operator $-\Sigma_{j=2}^{n}e_{1}^{-1}e_{j}

327: \frac{\partial}{\partial x_{j}}$ it may be determined that the power

328: series we previously obtained for $G(x-y)$ can be replaced by the

329: series $\Sigma_{j=0}^{\infty}(\Sigma_{\stackrel{j_{2}\ldots j_{n}}

330: {j_{2}+\ldots+j_{n}=j}}P_{j_{2}\ldots j_{n}}(y)\frac{\partial^{j}G(x)}

331: {\partial x_{2}^{j_{2}}\ldots \partial x_{n}^{j_{n}}})$, where $\|y\|<

332: \|x\|$ and

333: \[P_{j_{2}\ldots j_{n}}(y)=\frac{1}{j!}\Sigma(y_{\sigma(1)}

334: -y_{1}e_{1}^{-1}e_{\sigma(1)})\ldots

335: (y_{\sigma(j)}-y_{1}e_{1}^{-1}e_{\sigma(j)}).\]

336: Here $\sigma(i)\in\{2,\ldots, n\}$ and the previous summation is taken over

337: all permutations of the

338: monomials $(y_{\sigma(i)}-y_{1}e_{1}^{-1}e_{\sigma(i)})$ without repetition.

339: The quaternionic monogenic analogues for these polynomials were introduced

340: by Fueter \cite{f} while the Clifford analogues, $P_{j_{2}\ldots j_{n}}$,

341: described here were introduced by Delanghe in \cite{d}. It should be noted that

342: each polynomial $P_{j_{2}\ldots j_{n}}(y)$ takes its values in the space

343: spanned by $1$, $e_{1}e_{2},\dots, e_{1}e_{n}$. Also each such polynomial is

344: homogeneous of degree $j$. Similar arguments to those

345: just outlined give that $G(x-y)=\Sigma_{j=0}^{\infty}

346: (\Sigma_{\stackrel{j_{2}\ldots j_{n}}{j_{2}+\ldots+j_{n}=j}}

347: \frac{\partial^{j}G(x)}{\partial x_{2}^{j_{2}}\ldots x_{n}^{j_{n}}}

348: \widetilde{P_{j_{2}\ldots j_{n}}}(y)$ provided $\|y\|<\|x\|$.

349: \begin{proposition}

350: Each of the polynomials $P_{j_{2}\ldots j_{n}}(y)$ is a left Clifford holomorphic

351: polynomial.

352: \end{proposition}

353: {\bf{Proof:}} As $DP_{j_{2}\ldots j_{n}}(y)=e_{1}(\frac{\partial}

354: {\partial y_{1}}+e_{1}^{-1}\Sigma_{j=2}^{n}e_{j}\frac{\partial}

355: {\partial y_{j}}P_{j_{2}\ldots j_{n}}(y))$ then we shall consider the

356: expression $(\frac{\partial}{\partial y_{1}}+\Sigma_{j=2}^{n}e_{1}^{-1}e_{j}

357: \frac{\partial}{\partial y_{j}})P_{j_{2}\ldots j_{n}}(y)$. This term is equal

358: to

359: \[(\frac{\partial}{\partial y_{j}}+\Sigma_{j=2}^{n}e_{1}^{-1}e_{j}

360: \frac{\partial}{\partial y_{j}})

361: \Sigma(y_{\sigma(1)}-e_{1}^{-1}e_{\sigma(1)}y_{1})\ldots\]

362: \[\ldots(y_{\sigma(i-1)}-e_{1}^{-1}e_{\sigma(i-1)}y_{1})

363: (y_{\sigma(i)}-e_{1}^{-1}

364: e_{\sigma(i)}y_{1})(y_{\sigma(i+1)}-e_{1}^{-1}e_{\sigma(i+1)}y_{1})\ldots\]

365: \[\ldots(y_{\sigma(j)}-e_{1}^{-1}e_{\sigma(j)}y_{1}).\]

366: This is equal to

367: \[\Sigma(y_{\sigma(1)}-e_{1}^{-1}e_{\sigma(1)}y_{1})\ldots

368: (y_{\sigma(i-1)}-e_{1}^{-1}e_{\sigma(i-1)}y_{1})(-e_{1}^{-1}e_{\sigma(i)})

369: (y_{\sigma(i+1)}-e_{1}^{-1}e_{\sigma(i+1)}y_{1})\]

370: \[\ldots (y_{\sigma(j)}-e_{1}^{-1}e_{\sigma(j)}y_{1})+

371: \Sigma e_{1}^{-1}e_{\sigma(i)}

372: (y_{\sigma(1)}-e_{1}^{-1}e_{\sigma(1)}y_{1})\ldots

373: (y_{\sigma(i-1)}-e_{1}^{-1}e_{\sigma(i-1)}y_{1})\]

374: \[(y_{\sigma(i+1)}-e_{1}^{-1}e_{\sigma(i+1)}y_{1})\ldots

375: (y_{\sigma(j)}-e_{1}^{-1}e_{\sigma(j)}y_{1}).\]

376: 

377: \ If we multiply the previous term by $y_{1}$ and add to it the following

378: term, which is equal to zero,

379: \[\Sigma(y_{\sigma(1)}-e_{1}^{-1}e_{\sigma(1)}y_{1})\ldots

380: (y_{\sigma(i-1)}-e_{1}^{-1}e_{\sigma(i-1)}y_{1})

381: (y_{\sigma(i)}-y_{\sigma(i)})\]

382: \[(y_{\sigma(i+1)}-e_{1}^{-1}e_{\sigma(i+1)}y_{1})\ldots

383: (y_{\sigma(j)}-e_{1}^{-1}e_{\sigma(i)}y_{1})\]

384: we get, after regrouping terms,

385: \[\Sigma(y_{\sigma(1)}-e_{1}^{-1}e_{\sigma(1)}y_{1})\ldots

386: (y_{\sigma(i-1)}-e_{1}^{-1}e_{\sigma(i-1)}y_{1})

387: (y_{\sigma(i)}-e_{1}^{-1}e_{\sigma(i)}y_{1})\]

388: \[(y_{\sigma(i+1)}-e_{1}^{-1}e_{\sigma(i+1)}y_{1})\ldots

389: (y_{\sigma(j)}-e_{1}^{-1}e_{\sigma(j)}y_{1})\]

390: \[-\Sigma(y_{\sigma(i)}-e_{1}^{-1}e_{\sigma(i)}y_{1})

391: (y_{\sigma(1)}-e_{1}^{-1}e_{\sigma(1)}y_{1})\ldots

392: (y_{\sigma(i-1)}-e_{1}^{-1}e_{\sigma(i-1)}y_{1})\]

393: \[(y_{\sigma(i+1)}-e_{1}^{-1}e_{\sigma(i+1)}y_{1})\ldots

394: (y_{\sigma(j)}-e_{1}^{-1}e_{\sigma(j)}y_{j}).\]

395: As summation is taken over all possible permutations without

396: repetition this last term vanishes. $\Box$

397: 

398: \ Using Proposition 1 and the results we previously obtained on series

399: expansions we can obtain the following generalization of Taylor expansions

400: from one variable complex analysis.

401: \begin{theorem}{\bf{(Taylor Series)}}

402: Suppose that $f$ is a left Clifford holomorphic function defined in an open

403: neighbourhood of the closure of the ball $B(w, R)$. Then

404: \[f(y)=\Sigma_{j=0}^{\infty}(\Sigma_{\stackrel{j_{2}\ldots j_{n}}

405: {j_{2}+\ldots+j_{n}=j}}P_{j_{2}\ldots j_{n}}(y-w)a_{j_{2}\ldots j_{n}}),\]

406: where $a_{j_{2}\ldots j_{n}}=\frac{1}{\omega_{n}}\int_{\partial B(w, R)}

407: \frac{\partial^{j}G(x-w)}{\partial x_{2}^{j_{2}}\ldots\partial x_{n}^{j_{n}}}

408: n(x)f(x)d\sigma(x)$ and $\|y-w\|<R$. Convergence is uniform

409: provided $\|x-w\|<r<R$.

410: \end{theorem}

411: 

412: \ A simple application of Cauchy's theorem now tells us that the Taylor

413: series that we obtained for $f$ in the previous theorem remains

414: valid on the largest open ball on which $f$ is defined and the largest open

415: ball on which $g$ is defined. Also the previous identities immediately yield

416: the mutual linear independence of the collection of the left Clifford holomorphic

417: polynomials $\{P_{j_{2}\ldots j_{n}}:j_{2}+\ldots j_{n}=j$ and $0\leq j

418: <\infty\}$.

419: 

420: \begin{center}{\Large Other Types of Clifford Holomorphic Functions}\end{center}

421: 

422: \ Unlike the the classical Cauchy-Riemann operator $\frac{\partial}

423: {\partial\overline{z}}=\frac{\partial}{\partial x}+i\frac{\partial}

424: {\partial y}$ the generalized Cauchy-Riemann operator $D$ that we have

425: introduced here does not have an identity component. Instead we could have

426: considered the differential operator

427: $D'=\frac{\partial}{\partial x_{0}}+\Sigma_{j+1}^{n-1}e_{j}\frac{\partial}

428: {\partial x_{j}}$. Also for $U'$ a domain in $R\oplus R^{n-1}$, the span of

429: $1, e_{1},\ldots, e_{n-1}$, one can consider $Cl_{n-1}$ valued differentiable

430: functions $f'$ and $g'$ defined on $U'$ such that $D'f'=0$ and $g'D'=0$,

431: where $g'D'=\frac{\partial g'}{\partial x_{0}}+\Sigma_{j=1}^{n-1}

432: \frac{\partial g'}{\partial x_{j}}e_{j}$. Traditionally such functions are

433: also called left monogenic and right monogenic functions. To avoid confusion

434: we shall call such functions unital left monogenic and unital right

435: monogenic respectively. In the case where $n=2$ the operator $D'$ corresponds

436: to the usual Cauchy-Riemann operator and unital monogenic functions are the

437: usual holomorphic functions studied in one variable complex analysis.

438: The function $G'(\underline{x})=\frac{\overline{\underline{x}}}

439: {\|\underline{x}\|^{n}}=\underline{x}^{-1}\|\underline{x}\|^{-n+2}$ is an

440: example of a function which is both unital

441: left monogenic and unital right monogenic. It is a simple matter to observe

442: that $f'$ is unital left monogenic if and only if $\tilde{f}'$ is unital

443: right monogenic. However $\overline{f}'$ is not unital right monogenic

444: whenever $f'$ is unital left monogenic. Instead $\overline{f}'$ satisfies

445: the equation $\overline{f}'D'=0$.

446: 

447: \ The function theory for unital left monogenic functions is much the same

448: as for left monogenic functions. For instance if $f'$ is unital left

449: monogenic on $U'$ and $g'$ is unital right monogenic on the same domain and

450: $S'$ is a piecewise smooth, compact surface lying in $U'$ and bounding a

451: subdomain $V'$ then $\int_{S'}g'(\underline{x})n(\underline{x})

452: f'(\underline{x})d\sigma(\underline{x})=0$ where $n(\underline{x})$ is the

453: outward pointing normal vector to $S'$ at $\underline{x}$. Also for each

454: $\underline{y}\in V'$ there is the following version of Cauchy's integral

455: formula

456: \[f'(\underline{y})=\frac{1}{\omega_{n}}\int_{S'}G'(\underline{x}-

457: \underline{y})n(\underline{x})f'(\underline{x})d\sigma(\underline{x}).\]

458: 

459: \ To get from the operator $D$ to the operator $D'$ one first rewrites $D$

460: as $e_{n}(\frac{\partial}{\partial x_{n}}+\Sigma_{j=1}^{n-1}e_{n}^{-1}e_{j}

461: \frac{\partial}{\partial x_{j}})$. On multiplying on the left by $e_{n}$ and

462: changing the variable $x_{n}$ to $x_{0}$ we get the operator

463: $D"=\frac{\partial}{\partial x_{0}}+\Sigma_{j=1}^{n-1}e_{n}^{-1}e_{j}

464: \frac{\partial}{\partial x_{j}}$. This operator takes its values in the

465: even subalgebra $Cl_{n}^{+}$ of $Cl_{n}$. On applying the isomorphism

466: \[\theta: Cl_{n-1}\rightarrow Cl_{n}^{+}:\theta(e_{j_{1}}\ldots e_{j_{r}})=e_{n}^{-1}e_{j_{1}}\ldots e_{n}^{-1}e_{j_{r}}\]

467:  it immediately follows that

468: $\theta(D')=D"$. So if $f'$ is unital left monogenic then $D"\theta(f)=0$.

469: If we change the variable $x_{0}$ of the function $\theta(f(\underline{x}))$

470: to $x_{n}$ we now get a left monogenic function, which we denote by

471: $\theta'(f)(x)$, where $x=x_{1}e_{1}+\ldots+x_{n}e_{n}\in U\subset R^{n}$

472: if and only if $\underline{x}=x_{n}+x_{1}e_{1}+\ldots+x_{n-1}e_{n-1}\in U'

473: \subset R\oplus R^{n-1}$.

474: 

475: \ It should be noted that $D'\overline{D'}=\overline{D'}D'=\triangle_{n}$.

476: 

477: \ When $n=3$ the algebra$Cl_{3}$ is split by the two projection

478: operators $E_{\pm}=\frac{1}{2}(1\pm e_{1}e_{2}e_{3})$ into a direct sum $E_{+}Cl_{3}\oplus E_{-}Cl_{3}$ and that

479: each of these subalgebras are isomorphic to the quaternion algebra ${\bf{H}}$.

480: In this setting the differential operator $E_{\pm}D'$ can best be written as

481: $\frac{\partial}{\partial t}+i\frac{\partial}{\partial x}+j\frac{\partial}

482: {\partial y}+k\frac{\partial}{\partial z}$ and the operator $E_{\pm}D$ can

483: best be written as $i\frac{\partial}{\partial x}+j\frac{\partial}

484: {\partial y}+k\frac{\partial}{\partial z}$. We shall denote the first of

485: these two operators by $D'_{{\bf{H}}}$ and the second by $D_{{\bf{H}}}$. The

486: operator $D'_{{\bf{H}}}$ is sometimes referred to as the

487: Cauchy-Riemann-Fueter operator. The function theory associated to the

488: differential operators $D_{{\bf{H}}}$ and $D'_{{\bf{H}}}$ is much the same as

489: that associated to the operators $D$ and $D'$. In fact historically the

490: starting point for Clifford analysis was to study the function theoretic

491: aspects of the operators $D'_{{\bf{H}}}$ and $D_{{\bf{H}}}$, see for instance

492: \cite{d,f} and the excellent review article of Sudbery \cite{su}.

493: 

494: \ Over all these function theories have proved itself to be much the same as

495: that for the operators $D$ and $D'$. It is a simple enough matter to set up

496: analogues of Cauchy's theorem and Cauchy's integral formula for the

497: quaternionic valued differentiable functions that are either annihilated by

498: $D'_{{\bf{H}}}$ or $D_{{\bf{H}}}$, either acting on the left or on the right.

499: When dealing with the operator $D'_{{\bf{H}}}$ such functions are called

500: quaternionic monogenic. The quaternionic monogenic Cauchy kernel is the

501: function $q^{-1}\|q\|^{-2}$. Consequently for each quaternionic left

502: monogenic function $f(q)$ defined on a domain $U"\subset{\bf{H}}$ and each

503: $q_{0}$ lying in a bounded subdomain with piecewise $C^{1}$ boundary  $S"$

504: \[f(q_{0})=\frac{1}{\omega_{3}}\int_{S"}(q-q_{0})^{-1}\|q-q_{0}\|^{-2}n(q)

505: f(q)d\sigma(q).\]

506: Similarly if $g$ is right quaternionic monogenic on $U"$ then

507: \begin{equation}

508: g(q_{0})=\frac{1}{\omega_{3}}\int_{S"}g(q)n(q)(q-q_{0})^{-1}\|q-q_{0}\|^{-2}

509: d\sigma(q).

510: \end{equation}

511: 

512: \begin{center}{\Large The Equation $D^{k}f=0$}\end{center}

513: 

514: \ It is reasonably well known that if $h$ is a real valued harmonic function

515: defined on a domain $U\subset R^{n}$ then for each $y\in U$ and each

516: compact, piecewise $C^{1}$ surface $S$ lying in $U$ such that $S$ bounds a

517: subdomain $V$ of $S$ and $y\in V$, then

518: \[h(y)=\frac{1}{\omega_{n}}\int_{S}(H(x-y)<n(x),grad h(x)>-<G(x-y),n(x)>h(x))

519: d\sigma(x),\]

520: where $H(x-y)=\frac{1}{(n-2)\|x-y\|^{n-2}}$. This formula is Green's formula

521: for a harmonic function, and it heavily relies on the standard inner product

522: on $R^{n}$. Introducing the Clifford algebra $Cl_{n}$ the right side  of

523: Green's formula is the real part of

524: \[\frac{1}{\omega_{n}}\int_{S}(G(x-y)n(x)h(x)-H(x-y)n(x)Dh(x))d\sigma(x).\]

525: 

526: \ Assuming that the function $h$ is $C^{2}$ then on applying Stokes' theorem

527: the previous integral becomes

528: \[\frac{1}{\omega_{n}}\int_{S^{n-1}(y, r(y))}

529: (G(x-y)n(x)h(x)-H(x-y)n(x)Dh(x))d\sigma(x),\]

530: where $S^{n-1}(y, r(y))$ is a sphere centered at $y$, of radius $r(y)$ and

531: lying in $V$. On letting the radius $r(y)$ tend to zero the first term of the

532: integral tends to $h(y)$ while the second term tends to zero. Consequently

533: the Clifford analysis version of Green's formula is

534: \[h(y)=\frac{1}{\omega_{n}}\int_{S}(G(x-y)n(x)h(x)-H(x-y)n(x)Dh(x))

535: d\sigma(x).\]

536: 

537: \ This formula was obtained under the assumption that $h$ is real valued and

538: $C^{2}$. The fact that we have assumed $h$ to be real valued can easily be

539: observed to be irrelevant, and so we can assume that $h$ is $Cl_{n}$ valued.

540: From now on we shall assume that all harmonic functions take their values in

541: $Cl_{n}$. If $h$ is also a left monogenic function then the Clifford analysis

542: version of Green's formula becomes Cauchy's integral formula.

543: 

544: 

545: \begin{proposition}

546: Suppose that $f$ is a Clifford holomorphic function on some domain $U$. Then $xf(x)$ is

547: harmonic.

548: \end{proposition}

549: {\bf{Proof:}}

550: $Dxf(x)=-nf(x)-\Sigma_{j=1}^{n}x_{j}\frac{\partial f(x)}{\partial x_{j}}

551: -\Sigma_{\stackrel{j, k}{j\neq k}}x_{k}e_{k}e_{j}

552: \frac{\partial f(x)}{\partial x_{j}}$.

553: Now

554: \[\Sigma_{\stackrel{j, k}{j\neq k}}x_{k}e_{k}e_{j}

555: \frac{\partial f(x)}{\partial x_{j}}=

556: \Sigma_{k=1}^{n}\Sigma_{j\neq k}x_{k}e_{k}e_{j}

557: \frac{\partial f(x)}{\partial x_{j}}.\]

558: As $f$ is left monogenic this last expression simplifies to

559: $\Sigma_{k=1}^{n}x_{k}\frac{\partial f(x)}{\partial x_{k}}$. Moreover

560: $D\Sigma_{j=1}^{n}x_{j}\frac{\partial f(x)}{\partial x_{j}}=0$. Consequently

561: $D^{2}xf(x)=0$. $\Box$

562: 

563: \ The previous proof is a generalization of the statement- "if $h(x)$ is a

564: real valued harmonic function then so is $<x,grad h(x)>$".

565: 

566: \ In fact in the previous proof we determine that

567: $Dxf(x)=-nf(x)-2\Sigma_{j=1}^{n}x_{j}\frac{\partial f(x)}{\partial x_{j}}$.

568: In the special case where $f(x)=P_{k}(x)$, a left Clifford holomorphic polynomial of

569: order $k$, this equation simplifies to $DxP_{k}(x)=-(n+2k)P_{k}(x)$. Suppose

570: now that $h(x)$ is a harmonic function defined in a neighbourhood of the

571: ball $B(0,R)$. Now $Dh$ is a left Clifford holomorphic function so we know that there

572: is a series $\Sigma_{l=0}^{\infty}P_{l}(x)$ of left Clifford holomorphic polynomials

573: with each $P_{l}$ homogeneous of degree $l$ and such that the series

574: converges locally uniformly on $B(0, R)$ to $Dh(x)$. Now consider

575: the series $\Sigma_{l=0}^{\infty}\frac{-1}{n+2l}P_{l}(x)$. As

576: $\frac{1}{n+2l}\|P_{l}(x)\|<\|P_{l}(x)\|$ then this new series converges

577: locally uniformly on $B(0, R)$ to a left Clifford holomorphic function $f_{1}(x)$.

578: Moreover, $Dxf_{1}(x)=Dh(x)$ on $B(0, R)$. Consequently $h(x)-xf_{1}(x)$ is

579: equal to a left Clifford holomorphic function $f_{2}(x)$ on $B(0, R)$. Thus we have

580: established:

581: \begin{proposition}

582: Suppose that $h$ is a harmonic function defined in a neighbourhood of

583: $B(0, R)$ then there are left Clifford holomorphic functions $f_{1}$ and $f_{2}$ defined

584: on $B(0, R)$ such that $h(x)=xf_{1}(x)+f_{2}(x)$ for each $x\in B(0, R)$.

585: \end{proposition}

586: 

587: 

588: \ This result remains invariant under translation. As a consequence

589: it shows us that all harmonic functions are real analytic functions. So

590: there is no need to specify whether or not a harmonic function is $C^{2}$.

591: The result also provides an Almansi type decomposition of harmonic functions

592: in terms of Clifford holomorphic functions over any ball in $R^{n}$.

593: 

594: \ It should be noted that Proposition 3 remains true if $h$ is only real

595: valued.

596: 

597: \ Proposition 3 gives rise to an alternative proof of the Mean Value

598: Theorem

599: for harmonic functions.

600: 

601: \begin{theorem}

602: For any harmonic function $h$ defined in a neighbourhood of a ball $B(a, R)$

603: \[h(a)=\frac{1}{\omega_{n}}\int_{\partial B(a, r)}h(x)d\sigma(x)\]

604: for any $r<R$.

605: \end{theorem}

606: {\bf{Proof:}}

607:  Proposition 3 tells us that there is a pair of left Clifford holomorphic functions

608: $f_{1}$ and $f_{2}$ such that $h(x)=(x-a)f_{1}(x)+f_{2}(x)$ on $B(a, R)$.

609: So $h(a)=f_{2}(a)$, and we have previously shown that $\frac{1}{\omega_{n}}

610: \int_{\partial B(a, r)}f_{2}(x)d\sigma(x)=f_{2}(a)$. Now

611: $\int_{\partial B(a,r)}(x-a)f_{1}(x)d\sigma(x)=

612: r\int_{\partial B(a, r)}n(x)f_{1}(x)d\sigma(x)=0$. $\Box$

613: 

614: \ The following is an immediate consequence of Proposition 3.

615: \begin{proposition}

616: If $h_{l}(x)$ is a harmonic polynomial homogeneous of degree $l$ then

617: \[h_{l}(x)=p_{l}(x)+xp_{l-1}(x)\]

618: where $p_{l}$ is a left Clifford holomorphic polynomial homogeneous of degree $l$ while

619: $p_{l-1}$ is a left monogenic polynomial which is homogeneous of degree $l-1$.

620: \end{proposition}

621: 

622: \ It is well known that pairs of homogeneous harmonic polynomials of

623: differing degrees of homogeneity are orthogonal with respect to the usual

624: inner product over the unit sphere. Proposition 4 offers a further

625: refinement

626: to this. Suppose that $f$ and $g$ are $Cl_{n}$ valued functions defined on

627: $S^{n-1}$ and each component of $f$ and $g$ is square integrable. If we define

628: the $Cl_{n}$ inner product of $f$ and $g$ to be

629: \[<f, g>=\frac{1}{\omega_{n}}\int_{S^{n-1}}\overline{f(x)}g(x)d\sigma(x)\]

630: then if $f$ and $g$ are both real valued this inner product is equal to

631: \[\frac{1}{\omega_{n}}\int_{S^{n-1}}f(x)g(x)d\sigma(x)\]

632: which is the usual inner product for real valued square integrable functions

633: defined on $S^{n-1}$. Now

634: \[<xp_{l-1}(x), p_{l}(x)>=

635: -\frac{1}{\omega_{n}}\int_{S^{n-1}}\overline{p}_{l-1}(x)xp_{l}(x)d\sigma(x)\]

636: \[=-\frac{1}{\omega_{n}}\int_{S^{n-1}}\overline{p}_{l-1}(x)n(x)p_{l}(x)

637: d\sigma(x)=0.\]

638: The evaluation of the last integral is an application of Cauchy's theorem.

639: 

640: \ Let us denote the space of $Cl_{n}$ valued functions defined on $S^{n-1}$

641: and such that each component is square integrable by $L^{2}(S^{n-1},Cl_{n})$.

642: Clearly the space of real valued square integrable functions defined on

643: $S^{n-1}$ is a subset of $L^{2}(S^{n-1},Cl_{n-1})$. The space

644: $L^{2}(S^{n-1},Cl_{n})$ is a $Cl_{n}$ module.

645: 

646: \ We have shown that by introducing the module $L^{2}(S^{n-1}, Cl_{n})$

647: Proposition 4 provides a further orthogonal decomposition of harmonic

648: polynomials using left Clifford holomorphic polynomials. We shall return to this theme

649: later. This decomposition was introduced for the case $n=4$ by Sudbery

650: \cite{su} and independently extended for all $n$ by Sommen \cite{s2}.

651: 

652: 

653: \ Let us now consider higher order iterates of the Dirac operator $D$. In the

654: same way as we have that $DH(x)=G(x)$ there is a function $G_{3}(x)$ defined

655: on $R^{n}\backslash\{0\}$ such that $DG_{3}(x)=H(x)$. Specifically

656: $G_{3}(x)=C(n, 3)\frac{x}{\|x\|^{n-2}}$ for some dimensional constant

657: $C(n,3)$. Continuing inductively we may find a function $G_{k}(x)$ on $R^{n}

658: \backslash\{0\}$ such that $DG_{k}(x)=G_{k-1}(x)$. Specifically

659: \[G_{k}(x)=C(n,k)\frac{x}{\|x\|^{n-k+1}}\]

660: when $n$ is odd and so is $k$.

661: \[G_{k}(x)=C(n,k)\frac{1}{\|x\|^{n-k}}\]

662: when $n$ is odd and $k$ is even

663: \[G_{k}(x)=C(n,k)\frac{x}{\|x\|^{n-k+1}}\]

664: when $n$ is even, $k$ is odd and $k<n$

665: \[G_{k}(x)=C(n,k)\frac{1}{\|x\|^{n-k}}\]

666: when $n$ is even, $k$ is even and $k<n$

667: \[G_{k}(x)=C(n,k)(x^{k-n}\log\|x\|+A(n,k)x^{k-n})\]

668: when $n$ is even and $k\geq n$. In the last expression $A(n,k)$ is a real

669: constant dependent on $n$ and $k$. $C(n,k)$ is a constant dependent on $n$

670: and $k$ throughout.

671: 

672: \ It should be noted that $G_{1}(x)=G(x)$ and $G_{2}(x)=H(x)$.  It should

673: also be noted that $D^{k}G_{k}(x)=0$.

674: 

675: \ Here is a simple technique for constructing solutions to the equation

676: $D^{k}g=0$ from left Clifford holomorphic functions. The special case $k=2$ was

677: illustrated in Proposition 2.

678: \begin{proposition}

679: Suppose that $f$ is a left Clifford holomorphic function on $U$ then

680: $D^{k}x^{k-1}f(x)=0$.

681: \end{proposition}

682: {\bf{Proof}} The proof is by induction. We have already seen the result to be

683: true in the case $k=2$ in Proposition 2. If $k$ is odd then $Dx^{k-1}f(x)

684: =(k-1)x^{k-2}f(x)$. If $k$ is even then

685: \[Dx^{k-1}f(x)=-n(k-1)x^{k-2}f(x)

686: +x^{k-2}\Sigma_{j=1}^{n}e_{j}x\frac{\partial f(x)}{\partial x_{j}}.\]

687: By arguments presented in Proposition 5 this expression is equal to

688: \[-n(k-1)x^{k-2}f(x)+x^{k-2}\Sigma_{j=1}^{n}x_{j}\frac{\partial f(x)}

689: {\partial x_{j}}.\]

690: The induction hypothesis tells us that the only term we need consider is

691: $x^{k-2}\Sigma_{j=1}^{n}x_{j}\frac{\partial f(x)}{\partial x_{j}}$.

692: However $\Sigma_{j=1}^{n}x_{j}\frac{\partial f(x)}{\partial x_{j}}$ is a left

693: Clifford holomorphic function. So proof by induction is now complete. $\Box$

694: 

695: \ In future we shall refer to a function $g:U\rightarrow Cl_{n}$ which

696: satisfies the equation $D^{k}g=0$ as a left $k$-monogenic function. Similarly

697: if $h:U\rightarrow Cl_{n}$ satisfies the equation $hD^{k}=0$ then $h$ is a

698: right $k$-monogenic function. In the case where $k=1$ we return to the

699: setting of left, or right, Clifford holomorphic functions and when $k=2$ we return to the

700: setting of harmonic functions. When $k=4$ the equations $D^{4}g=0$ and

701: $gD^{4}=0$ correspond to the equations $\triangle_{n}^{2}g=0$ and

702: $\triangle_{n}^{2}h=0$. So left or right $4$-monogenic functions are in fact

703: biharmonic functions. In greater generality if $k$ is even then a left or

704: right $k$-monogenic function $f$ automatically satisfies the equation

705: $\triangle_{n}^{\frac{k}{2}}f=0$.

706: 

707: \begin{proposition}

708: Suppose that $p$ is a left $k$-monogenic polynomial homogeneous of degree $q$

709: then there are left Clifford holomorphic polynomials $f_{0},\ldots, f_{k-1}$ such that

710: \[p(x)=f_{0}(x)+\ldots+x^{k-1}f_{k-1}(x)\]

711: and each polynomial $f_{j}$ is homogeneous of degree $q-j$ whenever $q-j\geq

712: 0$ and is identically zero otherwise.

713: \end{proposition}

714: {\bf{Proof:}} The proof is via induction on $k$. The case $k=2$ is

715: established

716: immediately after the proof of Proposition 2. Let us now consider $Dp(x)$.

717: This is a left $k-1$-monogenic polynomial homogeneous of degree $q-1$. So by

718: the induction hypothesis $Dp(x)=g_{1}(x)+\ldots+x^{k-2}g_{k-1}(x)$ where

719: each $g_{j}$ is a left Clifford holomorphic polynomial homogeneous of degree $q-j$

720: whenever $q-j\geq 0$ and is equal to zero otherwise. Using Euler's lemma and

721: the observations made after the proof of Proposition 5 one may now find

722: left Clifford holomorphic polynomials $f_{1}(x),\ldots,f_{k-1}(x)$ such that

723: $D(xf_{1}(x)+\ldots+x^{k-1}f_{k-1}(x))=Dp(x)$ and $f_{j}(x)=c_{j}g_{j}(x)$

724: for some $c_{j}\in R$ and for $1\leq j\leq k-1$. It follows that $p(x)-

725: \Sigma_{j=1}^{k-1}x^{j}f_{j}(x)$ is a left Clifford holomorphic polynomial $f_{0}$

726: homogeneous of degree $q$. $\Box$

727: 

728: \ One may now use Proposition 6 and the arguments used to establish

729: Proposition 3 to deduce:

730: \begin{theorem}

731: Suppose that $f$ is a left $k$-monogenic function defined in a neighbourhood

732: of the ball $B(0, R)$ then there are left monogenic functions $f_{0},\ldots,

733: f_{k-1}$ defined on $B(0,R)$ such that

734: $f(x)=f_{0}(x)+\ldots+x^{k-1}f_{k-1}(x)$ on $B(0,R)$.

735: \end{theorem}

736: 

737: \ Theorem 6 establishes an Almansi decomposition for left $k$-monogenic

738: functions in terms of left Clifford holomorphic functions over any open ball. It also

739: follows from this theorem that each left $k$-monogenic function is a real

740: analytic function. It is also reasonably well known that if $h$ is a

741: biharmonic function defined in a neighbourhood of $B(0,R)$ then there are

742: harmonic functions $h_{1}$ and $h_{2}$ defined on $B(0, R)$ and such that

743: $h(x)=h_{1}(x)+\|x\|^{2}h_{2}(x)$. In the special case where $k=4$ Theorem

744: 6 both establishes this result and refines it.

745: 

746: \ As each left $k$-monogenic function is a real analytic function then we

747: can immediately use Stokes' theorem to deduce the following Cauchy-Green

748: type formula.

749: \begin{theorem}

750: Suppose that $f$ is a left $k$-monogenic function defined on some domain

751: $U$ and suppose that $S$ is a piecewise $C^{1}$ compact surface lying in $U$

752: and bounding a bounded subdomain $V$ of $U$. Then for each $y\in V$

753: \[f(y)=\frac{1}{\omega_{n}}\int_{S}(\Sigma_{j=1}^{k}(-1)^{j-1}G_{j}(x-y)n(x)

754: D^{j-1}f(x))d\sigma(x).\]

755: \end{theorem}

756: 

757: \begin{center}{\Large Conformal Groups and Clifford Analysis}\end{center}

758: 

759: \ Here we will examine the role played

760: by the conformal group within parts of Clifford analysis. Our starting point

761: is to ask what type of diffeomorphisms acting on subdomains of $R^{n}$

762: preserve Clifford holomorphic functions. If a diffeomorphism $\phi$ can transform the

763: class of left Clifford holomorphic functions on one domain $U$ to a class of left

764: Clifford holomorphic functions on the domain $\phi(U)$ and do the same for the class

765: of right Clifford holomorphic functions on $U$ then it must preserve Cauchy's theorem.

766: So if $f$ is left Clifford holomorphic on $U$ and $g$ is right Clifford holomorphic on $U$ and

767: these functions are transformed to $f'$ and $g'$ respectively left and right

768: Clifford holomorphic functions on $\phi(U)$ then

769: \[\int_{S}g(x)n(x)f(x)d\sigma(x)=0=\int_{\phi(S)}g'(y)n(y)f'(y)d\sigma(y)\]

770: where $S$ is a piecewise $C^{1}$ compact surface lying in $U$ and $y=\phi(x)$.

771: An important point to note here is that we need to assume that $\phi$

772: preserves vectors orthogonal to the tangent spaces at $x$ and $\phi(x)$. As

773: the choice of $x$ and $S$ is arbitrary it follows that the diffeomorphism

774: $\phi$ is angle preserving. In other words $\phi$ is a conformal

775: transformation. A theorem of Liouville \cite{l} tells us that for dimensions

776: $3$ and greater the only conformal transformations on domains are

777: M\"{o}bius transformations.

778: 

779: \ In order to deal with M\"{o}bius transformations using Clifford algebras

780: we have seen in a previous chapter that one can use Vahlen matrices. We

781: now proceed to show that each M\"{o}bius transformation

782: preserves monogenicity. Sudbery \cite{su} and also Bojarski \cite{b} have

783: established this fact. We will need the following two lemmata.

784: 

785: \begin{lemma}

786: Suppose that $\phi(x)=(ax+b)(cx+d)^{-1}$ is a M\"{o}bius transformation

787: then

788: \[G(u-v)=J(\phi, x)^{-1}G(x-y)\tilde{J}(\phi, y)^{-1}\]

789: where $u=\phi(x)$, $v=\phi(y)$ and $J(\phi, x)=\frac{(\widetilde{cx+d})}

790: {\|cx+d\|^{n}}$.

791: \end{lemma}

792: {\bf{Proof}} The proof essentially follows from the fact that

793: \[(x^{-1}-y^{-1})=x^{-1}(y-x)y^{-1}.\]

794: Consequently $\|x^{-1}-y^{-1}\|=\|x\|^{-1}\|x-y\|\|y\|^{-1}$.

795: Also $ax\tilde{a}-ay\tilde{a}=a(x-y)\tilde{a}$.

796: 

797: \ If one breaks the transformation down into terms arising from the generators

798: of the M\"{o}bius group and use the previous set of equations then one will

799: readily arrive at the result. $\Box$

800: \begin{lemma}

801: Suppose that $y=\phi(x)=(ax+b)(cx+d)^{-1}$ is a M\"{o}bius transformation

802: and for domains $U$ and $V$ we have $\phi(U)=V$ then

803: \[\int_{S}f(u)n(u)g(u)d\sigma(u)=\int_{\psi^{-1}(S)}f(\psi(x))\tilde{J}(\psi,x)

804: n(x)J(\psi,x)g(\psi(x))d\sigma(x)\]

805: where $u=\psi(x)$, $S$ is a orientable hypersurface lying in $U$ and

806: $J(\psi,x)=\frac{\widetilde{cx+d}}{\|cx+d\|^{n}}$.

807: \end{lemma}

808: {\bf{Outline Proof}} On breaking $\psi$ up into the generators of the

809: M\"{o}bius group the result follows from noting that

810: \[\frac{\partial x^{-1}}{\partial x_{j}}=-x^{-1}e_{j}x^{-1}.\]

811: $\Box$

812: 

813: \ It follows from Cauchy's Theorem that if $g(u)$ is a left Clifford holomorphic

814: function in the variable $u$ then $J(\psi,x)f(\psi(x))$ is left Clifford holomorphic

815: in the variable $x$.

816: 

817: 

818: \ When $\phi(x)$ is the Cayley transformation $y=(e_{n}x+1)(x+e_{n})^{-1}$

819: we can use this transformation to establish a Cauchy-Kowalewska extension

820: in a neighbourhood of the sphere. If $f(x)$ is a real analytic function

821: defined on an open subset $U$ of $S^{n-1}\backslash\{e_{n}\}$ then $l(y)=

822: J(\phi^{-1},y)^{-1}f(\phi(y))$ is a real analytic function on the open set

823: $V=\phi^{-1}(U)$. This function has a Cauchy-Kowalewska extension to a left

824: Clifford holomorphic function $L(y)$ defined on an open neighbourhood $V(g)\subset R^{n}$

825: of $V$. Consequently $F(x)=J(\phi^{-1}, x)L(\phi^{-1}(x))$ is a left

826: Clifford hholomorphic defined on an open neighbourhood $U(f)=\phi^{-1}(V(g))$ of $U$.

827: Moreover $F_{|U}=f$. Combing with similar arguments for the other Cayley

828: transformation $y=(-e_{n}x+1)(x-e_{n})^{-1}$ one can deduce:

829: \begin{theorem}

830: {\bf{(Cauchy-Kowalewska Theorem)}}

831: Suppose that $f$ is a $Cl_{n}$ valued real analytic function defined on

832: $S^{n-1}$. Then there is a unique left Clifford holomorphic function $F$ defined on an

833: open neighbourhood $U(f)$ of $S^{n-1}$ such that $F_{|S^{n-1}}=f$.

834: \end{theorem}

835: 

836: \ In fact if $f(u)$ is defined on some domain and satisfies the equation

837: $D^{k}f=0$ then the function $J_{k}(\psi,x)f(\psi(x))$ satisfies the same

838: equation, where $J_{k}(\psi,x)=\frac{\widetilde{cx+d}}{\|cx+d\|^{n-k+1}}$.

839: \begin{theorem}{\bf{(Fueter-Sce Theorem)}}

840: Suppose that $f=u+iv$ is a holomorphic function on a domain $\Omega\subset

841: {\bf{C}}$ and that $\Omega=\overline{\Omega}$ and $f(\overline{z})=

842: \overline{f(z)}$. Then the function $F(\underline{x})=u(x_{1},\|x'\|)

843: +e_{1}^{-1}\frac{x'}{\|x'\|}v(x_{1},\|x'\|)$ is a unital left

844: $n-1$-monogenic function

845: on the domain $\{\underline{x}:x_{1}+i\|x'\|\in\Omega\}$ whenever $n$ is

846: even. Here $x'=x_{2}e_{2}+\ldots+x_{n}e_{n}$.

847: \end{theorem}

848: {\bf{Proof:}}

849: First let us note that $x^{-1}e_{1}$ is left $n-1$ monogenic whenever

850: $n$ is even. It follows that $\frac{\partial^{k}}{\partial

851: x_{1}^{k}}x^{-1}e_{1}=c_{k}x^{-k-1}e_{1}$ is $n-1$ left monogenic for each

852: positive integer $k$. Here $c_{k}$ is some non-zero real number. Using

853: Kelvin inversion it follows that $x^{k}e_{1}$ is left $n-1$ monogenic for

854: each positive integer $k$. By taking translations and Taylor series

855: expansions for the function $f$ the result follows. $\Box$

856: 

857: \ This result was first established for the case $n=4$ by Fueter, \cite{f}, see

858: also Sudbery \cite{su}. It was extended to all even dimensions by Sce \cite{sc},

859: though the methods used do not make use of the conformal group. This result has been applied in \cite{q1,q2} to study various types of singular integral operators acting on $L^{p}$ spaces of Lipschitz perturbations of the sphere.

860: 

861: \begin{center}{\Large Conformally Flat Spin Manifolds}\end{center}

862: 

863: \ The invariance of monogenic functions under M\"{o}bius transformations described in the previous section makes use of a conformal weight factor $J(\psi,x)$. This invariance can be seen as an automorphic form invariance. This leads to a natural generalization of the concept of a Riemann surface to the euclidean setting. A manifold $M$ is said to be conformally flat if there is an atlas ${\cal A}$ of $M$ whose transition functions are M\"{0}bius transformations. For instance via the Cayley transformations $(e_{n+1}x+1)(x+e_{n+1})^{-1}$ and $(-e_{n+1}x+1)(x-e_{n+1})^{-1}$ one can see that the sphere $S^{n}\subset R^{n+1}$ is an example of a conformally flat manifold. Another way of constructing conformally flat manifolds is to take a simply connected domain $U$ of $R^{n}$ and consider a Kleinian subgroup $\Gamma$ of the M\"{o}bius group that acts discontinuously on $U$. Then the factorization $U\backslash \Gamma$ is a conformally flat manifold. For instance let $U=R^{n}$ and let $\Gamma$ be the integer lattice $Z^{k}=Ze_{1}+\ldots+Ze_{k}$ for some positive integer $k\leq n$. In this case $R^{n}\backslash Z^{k}$ gives the cylinder $C_{k}$ and when $k=n$ we get the $n$-torus. Also if we let $U=R^{n}\backslash\{0\}$ and $\Gamma=\{2^{k}:k\in Z\}$ the resulting manifod is $S^{1}\times S^{n-1}$.

864: 

865: \ We locally construct a spinor bundle over $M$ by making the identification $(u,X)$ with either $(x,\pm J(\psi,x)X)$ where $u=\psi(x)=(ax+b)(cx+d)^{-1}=(-ax-b)(-cx-d)^{-1}$. If we can compatibly choose the signs then we have created a spinor bundle over the conformally flat manifold. Note, it might be possible to create more than one spinor bundle over $M$. For instance consider the cylinder $C_{k}$ if we make the identification $(x,X)$ with $x+\underline{m},(-1)^{m_{1}+\ldots+m_{l}}X)$ where $l$ is a fixed integer with $l\leq k$ and $\underline{m}=m_{1}e_{1}+\ldots+m_{l}e_{l}+\ldots+m_{k}e_{k}$ then we have created $k$ different spinor bundles $E^{1},\ldots E^{k}$ over $C_{k}$.

866: 

867: \ As we have used the conformal weight function function $J(\psi,x)$ to construct the spinor bundle $E$ then it is easy to see that a section $f:M\rightarrow E$ could be called a left monogenic section if it is locally a left monogenic function. It is now natural to ask if one can construct Cauchy integral formulas for such sections. To do this we essentially need to construct a kernel over the euclidean domain $U$ that is periodic with respect to $\Gamma$ and then use the projection map $p:U\rightarrow M$ to construct from this kernel a Cauchy kernel for $U$. In \cite{kr2} we show that the Cauchy kernel for

868: $C_{k}$ with spinor bundle $E^{l}$ is constructed from the kernel

869: \[\cot_{k,l}(x,y)=\Sigma_{\underline{m}\in Z^{l},\underline{n}\in Z^{k-l}}(-1)^{m_{1}+\ldots+m_{l}}G(x-y+\underline{m}+\underline{n}),\]

870: where $\underline{n}=n_{l+1}e_{l+1}+\ldots+n_{n}e_{n}$.

871: While for the conformally flat spin manifold $S^{1}\times S^{n-1}$ with trivial bundle $Cl_{n}$ the Cauchy kernel is constructed from the kernel

872: \[\Sigma_{k=0}^{\infty}G(2^{k}x-2^{k}y)+2^{2-2n}G(x)(\Sigma_{k=-1}^{-\infty}G(2^{-k}x^{-1}-2^{-k}y^{-1}))G(y).\]

873: 

874: \ See \cite{k,kr1,kr2} for more details and related results.

875: 

876: \ It should be noted that one may set up a Dirac operator over arbitrary Riemannian manifolds ,see for instance \cite{bbw}, and one may set up Cauchy integral formulas for functions annihilated by these Dirac operators, see for instance \cite{c,m2} for details.

877: 

878: \begin{center}{\Large Boundary Behaviour and Hardy Spaces}\end{center}

879: 

880: \ Possibly the main topic that unites all that has been previously

881: discussed here on Clifford analysis is its applications to boundary value

882: problems. This in turn leads to a study of boundary behaviour of classes

883: of Clifford holomorphic functions and Hardy spaces. Let us look first at one of the

884: simplest cases. Previously we noted that if $\theta$ is a square integrable

885: function defined on the sphere $S^{n-1}$ then there is a harmonic function

886: $h$ defined on the unit ball in $R^{n}$ with boundary value $\theta$

887: almost everywhere. Previously we have seen that $h(x)=f_{1}(x)+xf_{2}(x)$

888: where $f_{1}$ and $f_{2}$ are left Clifford holomorphic. However on $S^{n-1}$ the

889: function $G(x)$ equals $x$. So one can see that on $S^{n-1}$ we have

890: $\theta(x)=f_{1}(x)+g(x)$ almost everywhere. Here $f_{1}$ is left

891: monogenic on the unit ball $B(0,1)$ and $g$ is left Clifford holomorphic on

892: $R^{n}\backslash clB(0,1)$, where $clB(0,1)$ is the closure of the open

893: unit ball. Let $H^{2}(B(0,1))$ denote the space of Clifford holomorphic functions

894: defined on $B(0,1)$ with extension to a square integrable function on

895: $S^{n-1}$ and let $H^{2}(R^{n}\backslash clB(0,1)$ denote the class of

896: left Clifford holomorphic functions defined on $R^{n}\backslash clB(0,1)$ with square

897: integrable extension to the $S^{n-1}$. What we have so far outlined is

898: that $L^{2}(S^{n-1})=H^{2}(B(0,1))\oplus H^{2}(R^{n}\backslash clB(0,1))$,

899: where $L^{2}(S^{n-1})$ is the space of $Cl_{n}$ valued Lebesgue square

900: integrable functions defined on $S^{n-1}$. This is the Hardy $2$-space

901: decomposition of $L^{2}(S^{n-1})$. It is also true if we replace $2$ by

902: $p$ where $1<p<\infty$. We will not go into this detail here as it is

903: beyond the scope of the material covered here.

904: 

905: \ Let us now take an alternative look at a way of obtaining this

906: decomposition. This method will generalize to all reasonable surfaces. We

907: will clarify what we mean by a reasonable surface later. Instead of

908: considering an arbitrary square integrable function on $S^{n-1}$ let us

909: instead assume that $\theta$ is a continuously differentiable function. Let

910: us now consider the integral

911: \[\frac{1}{\omega_{n}}\int_{S^{n-1}}G(x-y)n(x)\theta(x)d\sigma(x)\]

912: where $y\in B(0,1)$. This defines a left Clifford holomorphic function on $B(0,1)$.

913: Now let us allow the point $y$ to approach a point $z\in S^{n-1}$ along a

914: differentiable path $y(t)$. Let us also assume that $\frac{dy(t)}{dt}$

915: evaluated at $t=1$, so $y(t)=z$, is not tangential to $S^{n-1}$ at $z$. We

916: can essentially ignore this last point at a first read. We want to

917: evaluate

918: \[\lim_{t\rightarrow

919: 1}\frac{1}{\omega_{n}}\int_{S^{n-1}}G(x-y(t))n(x)\theta(x)d\sigma(x).\]

920: We do this by removing a small ball on $B(0,1)$ from $S^{n-1}$. The ball

921: is centered at $z$ and is of radius $\epsilon$. We denote this ball by

922: $b(z,\epsilon)$. The previous integral now splits into an integral over

923: $b(z,\epsilon)$ and an integral over $S^{n-1}\backslash b(z,\epsilon)$. On

924: $b(z, \epsilon)$ we can express $\theta(x)$ as

925: $(\theta(x)-\theta(z))+\theta(z)$. As $\theta$ is continuously

926: differentiable then $\|\theta(x)-\theta(z)\|<C\|x-z\|$ for some $C\in

927: R^{+}$. It follows that

928: \[\lim_{\epsilon\rightarrow 0}\lim_{t\rightarrow

929: 1}\int_{b(z,\epsilon)}\|G(x-y(t))n(x)(\theta(x)-\theta(z))\|d\sigma(x)=0.\]

930: Moreover, the term

931: \[\lim_{\epsilon\rightarrow 0}\lim_{t\rightarrow

932: 1}\frac{1}{\omega_{n}}\int_{b(z,\epsilon}G(x-y(t))n(x)\theta(z)d\sigma(x)\]

933: can, as $\theta(z)$ is a Clifford holomorphic function, be replaced by the term

934: \[\lim_{\epsilon\rightarrow 0}\lim_{t\rightarrow

935: 1}\int_{B(0,1)\cap\partial

936: B(z,\epsilon)}G(x-y(t))n(x)\theta(z)d\sigma(x).\]

937: By the residue theorem the limit of this integral evaluates to

938: $\frac{1}{2}\theta(z)$.

939: 

940: \ We leave it to the interested reader to note that the singular integral

941: or principal valued integral

942: \[\lim_{\epsilon\rightarrow 0}\lim_{t\rightarrow

943: 1}\frac{1}{\omega_{n}}\int_{S^{n-1}\backslash

944: b(z,\epsilon)}G(x-y(t))n(x)\theta(x)d\sigma(x)=\]

945: \[P.V.\frac{1}{\omega_{n}}\int_{S^{n-1}}G(x-z)n(x)\theta(x)d\sigma(x)\]

946: is bounded.

947: 

948: \ We have established that

949: \[\lim_{t\rightarrow 1}\int_{S^{n-1}}G(x-y(t))n(x)\theta(x)d\sigma(x)=\]

950: \[\frac{1}{2}\theta(z)+

951: P.V.\frac{1}{\omega_{n}}\int_{S^{n-1}}G(x-z)n(x)\theta(x)d\sigma(x).\]

952: 

953: \ If we now assumed that $y(t)$ is a path tending to $z$ on the complement

954: of $clB(0,1)$, then similar arguments give

955: \[\lim_{t\rightarrow 1}\int_{S^{n-1}}G(x-y(t))n(x)\theta(x)d\sigma(x)=\]

956: \[-\frac{1}{2}\theta(z)+

957: P.V.\frac{1}{\omega_{n}}\int_{S^{n-1}}G(x-z)n(x)\theta(x)d\sigma(x).\]

958: We will write these expressions as

959: \[(\pm\frac{1}{2}I+C_{S^{n-1}})\theta.\]

960: 

961: \ If we now consider the limit

962: \[\lim_{t\rightarrow 1}\frac{1}{\omega_{n}}\int_{S^{n-1}}G(x-y(t))n(x)

963: (\frac{1}{2}I+C_{S^{n-1}})\theta(x)d\sigma(x)\]

964: we may determine that

965: \[(\frac{1}{2}I+C_{S^{n-1}})^{2}=\frac{1}{2}I+C_{S^{n-1}}.\]

966: Furthermore

967: \[(\frac{1}{2}I+C_{S^{n-1}})(-\frac{1}{2}I+C_{S^{n-1}})=0\]

968: and

969: \[(-\frac{1}{2}I+C_{S^{n-1}})^{2}=-\frac{1}{2}I+C_{S^{n-1}}.\]

970: 

971: \ It is known that each function $\psi\in L^{2}(S^{n-1})$ can be

972: approximated by a sequence of functions each with the same properties as

973: $\theta$. This tells us that the previous formulas can be repeated but

974: this time simply for $\theta\in L^{2}(S^{n-1})$. It follows that for such

975: a $\theta$ we have

976: \[\theta=(\frac{1}{2}I+C_{S^{n-1}})\theta+(\frac{1}{2}I-C_{S^{n-1}})\theta.\]

977: This formula gives the Hardy space decomposition of $L^{2}(S^{n-1})$. In

978: fact if one looks more carefully at the previous calculations to obtain

979: these formulas we see that it is not so significant that the surface used

980: is a sphere and we can re-do the calculations for any "reasonable" hypersurface

981: $S$. In this case we get

982: \[\theta=(\frac{1}{2}I+C_{S})\theta+(\frac{1}{2}I-C_{S})\theta\]

983: where $\theta$ now belongs to $L^{2}(S)$ and

984: \[C_{S}\theta=P.V.\frac{1}{\omega_{n}}\int_{S}G(x-y)n(x)\theta(x)d\sigma(x).\]

985: This gives rise to the Hardy space decomposition

986: \[L^{2}(S)=H^{2}(S^{+})\oplus H^{2}(S^{-})\]

987: where $S^{\pm}$ are the two domains that complement the surface $S$ (we

988: are assuming that $S$ divides $R^{n}$ into two complementary domains.

989: 

990: \ Last one should address the smoothness of $S$. In some parts of the

991: literature one simply assumes that $S$ is compact and $C^{2}$, while in

992: more advanced and recent aspects of the literature one assumes that $S$

993: has rougher conditions, usually one assumes that the surface is simply

994: Lipschitz continuous, see for instance \cite{lmq,lms,m1}. The formulas given above involving the singular

995: integral operator $C_{S}$ are called Plemelj formulas. It is a simple

996: exercise to see that these formulas are conformally invariant. So using

997: Kelvin inversion or even a Cayley transformation one can see that these

998: formulas and the Hardy space decompositions are also valid on unbounded

999: surfaces and domains. A great deal of modern Clifford analysis has been

1000: devoted to the study of such Hardy spaces and singular integral operators. This is in fact due to an idea of R. Coifman that various hard problems in classical harmonic analysis studied in euclidean space might be more readily handled using tools from Clifford analysis, particularly the singular Cauchy transform and associated Hardy spaces. In particular Coifman speculated that a more direct proof of the celebrated Coifman-McIntosh-Meyer Theorem \cite{cmm}could be derived using Clifford analysis. The Coifman-McIntosh-Meyer Theorem establishes the $L^{2}$ boundedness of the double layer potential operator for Lipschitz graphs in $R^{n}$. Coifman's observation was that the double layer potential operator is the real or scalar part of the singular Cauchy transform arising in Clifford analysis and discussed earlier. If one can establish the $L^{2}$ boundedness of the singular Cauchy transform for a Lipschitz graph in $R^{n}$ then one automatically has the $L^{2}$ boundedness for the double layer potential operator for the same graph. The $L^{2}$ boundedness of the singular Cauchy transform was first established for Lipschitz graphs with small constant by M. Murray \cite{m} and extended to the general case by A. McIntosh, see for instance \cite{mc}, see also \cite{m1}. One very importand reason for needing to know that the double layer potential operator is $L^{2}$ bounded for Lipschitz graphs is to be able to solve bounndary value problems for domains with Lipschitz graphs as boundaries. such boundary value problems would include the Dirichlet problem and Neuaman problem for the Laplacian. See \cite{mc,m1} for more details. Moreover in \cite{w} Clifford analysis and more precisely the Hardy space decomposition mentioned here is specifically used to solve the water wave problem in three dimensions.

1001: 

1002: \begin{center}{\Large More on Clifford Analysis on the Sphere}\end{center}

1003: 

1004: \ In the previous section We saw that $L^{2}(S^{n-1})$ splits into a direct sum of Hardy spaces for the corresponding complemetary domains $B(0,1)$ and $R^{n}\backslash cl(B(0,1))$. In an earlier section we saw that any left Clifford holomorphic function $f(x)$ can be expressed as a locally uniformly convergent series $\Sigma_{j=0}^{\infty}f_{j}(x)$ where each $f_{j}(x)$ is left Clifford holomorphic and homogeneous of degree $j$. Now following \cite{su} consider the operator

1005: \[D=x^{-1}xD=x^{-1}(\Sigma_{i<k}e_{i}e_{k}(x_{i}\frac{\partial}{\partial x_{k}}-x_{k}\frac{\partial}{\partial x_{i}}-\Sigma_{j=1}^{n}x_{j}\frac{\partial}{\partial x_{j}}).\]

1006:  By letting the last term in this expression act on homogeneous polynomials one may determine from Euler's lemma that in fact $\Sigma_{j=1}^{n}x_{j}\frac{\partial}{\partial x_{j}}$ is the radial operator $r\frac{\partial}{\partial r}$. So $r\frac{\partial}{\partial r}f_{j}(x)=jf_{j}(x)$. As each polynomial $f_{k}$ is Clifford holomorphic it follows that each $f_{j}$ is an eigenvector of the spherical Dirac operator $x\Lambda_{n-1}=x\Sigma_{i<k}e_{i}e_{k}(x_{i}\frac{\partial}{\partial x_{k}}-x_{k}\frac{\partial}{\partial x_{i}})$ with eigenvalue $k$. Now using Kelvin inversion we know that $f_{k}$ is  homogeneous of degree $k$ and left Clifford holomorphic if and only if $G(x)f_{k}(x^{-1})$ is homogeneous of degree $-n+1-k$ and is left Clifford holomorphic. On restricting $G(x)f_{k}(x^{-1})$ to the unit sphere this function becomes $xf_{k}(x^{-1})$ and this function is an eigenvector for the spherical Dirac operator $x\Lambda_{n-1}$. As each $f\in H^{2}(R^{n}\backslash cl(B(0,1)))$ can be written as $\Sigma_{k=0}^{\infty}G(x)f_{k}(x^{-1})$ where each $f_{k}$ is homogeneous of degree $k$ and is left Clifford holomorphic it follows that if $h\in L^{2}(S^{n-1})$ then $\Lambda_{n-1} xh(x)=(1-n)xh(x)-x\Lambda_{n-1} h(x)$. Similarly if we replace $S^{n-1}$ by the $n$-sphere $S^{n}$ embedded in $R^{n+1}$ then we have the identity $\Lambda_{n} xh(x)=-nxh(x)-x\Lambda_{n}h(x)$ for each $h\in L^{2}(S^{n})$. As all $C^{\infty}$ functions defined on $S^{n}$ belong to $L^{2}(S^{n})$ this identity holds for all such functions too.

1007: 

1008: \ It should be noted that for each $x\in S^{n}$ if we restrict the operator $x\Lambda_{n}$ to the tangent bundle $TS^{n}_{x}$ then we obtain the Euclidean Dirac operator acting on this tangent space.

1009: 

1010: \ By using the Cayley transformation $x=\psi(y)(e_{n+1}y+1)(y+e_{n+1})^{-1}$ from $R^{n}$ to $S^{n}\backslash\{e_{n+1}\}$ one can transform left Clifford holomorphic functions from domains in $R^{n}$ to functions defined on domains lying on the sphere. Namely if $f(y)$ is left Clifford holomorphic on the domain $U$ lying in $R^{n}$ then we obtain a function $f'(x)=J(\psi^{-1},x)f(\psi^{-1}(x)$ defined on the domain $U'=\psi(U)$ lying on $S^{n}$. Here $J(\psi^{-1},x)=\frac{x+1}{\|x+1\|^{n}}$. Similarly if $g(y)$ is right Clifford holomorphic on $U$ then $g'(x)=g(\psi^{-1}(x)J(\psi^{-1},x)$ is a well defined function on $U'$. Moreover for any smooth, compact hypersurface $S$ bounding a subdomain $V$ of $U$ we have from the conformal invariance of Cauchy's Theorem $\int_{S'}g'(x)n(x)f'(x)d\sigma'(x)=0$ where $S'=\psi(S)$, and $n(x)$ is the unit vector lying in the tangent space $TS^{n}_{x}$ of $S^{n}$ at $x$ and outer normal to

1011: $S'$ at $x$. Furthermore $sigma'$ is the Lebesgue measure on $S'$.

1012: 

1013: \ From Lemma 1 it now follows that for each point $y'\in V'=\psi(V)$ we have the following version of Cauchy's Integral Formula:

1014: \[f'(y')=\frac{1}{\omega_{n}}\int_{S'}G(x-y')n(x)f'(x)d\sigma(x)\]

1015: where as before $G(x-y')=\frac{x-y'}{\|x-y'\|^{n}}$, but now $x$ and $y'\in S^{n}$.

1016: It would now appear that the functions $f'$ and $g'$ are solutions to some spherical Dirac equations. We need to isolate this Dirac operator. We shall achieve this by applying the operator $x\Lambda_{n}$ to the kernel $G_{s}(x,y')=G(x-y')$. As $x$ and $y'\in S^{n}$ then $\|x-y'\|^{2}=2-2<x,y'>$, where $<x,y'>$ is the inner product of $x$ and $y'$. So $G_{s}(x,y')=2^{\frac{-n}{2}}\frac{x-y'}{(1-<x,y'>)^{\frac{n}{2}}}$. So in calculating $x\Lambda_{n}G_{s}(x,y')$ we will need to know what $\Lambda_{n}<x,y'>$ evaluates to. It is a simple exercise to determine that $\Lambda_{n}<x,y'>=xy'+<x,y'>$ which is simply the wedge product, $x\wedge y'$, of $x$ with $y'$.

1017: Now let us calculate $x\Lambda_{n}G_{s}(x,y')$.

1018: Now

1019: \[x\Lambda_{n}G_{s}(x,y')=2^{\frac{-n}{2}}(x\Lambda_{n}\frac{x}{(1-<x,y'>)^{\frac{n}{2}}}-x\Lambda_{n}\frac{y'}{(1-<x,y'>)^{\frac{n}{2}}})\]

1020: \[=2^{\frac{n}{2}}(-x\frac{nx}{(1-<x,y'>)^{\frac{n}{2}}}+\Lambda_{n}\frac{1}{(1-<x,y'>)^{\frac{-n}{2}}}-x\Lambda_{n}\frac{1}{(1-<x,y'>)^{\frac{n}{2}}}y').\]

1021: First

1022: \[\Lambda_{n}\frac{1}{(1-<x,y'>)^{\frac{n}{2}}}=\frac{n}{2}\frac{x\wedge y'}{(1-<x,y'>)^{\frac{n+2}{2}}}.\]

1023: So

1024: \[x\Lambda_{n}G_{s}(x,y')=2^{\frac{-n}{2}}\frac{n}{2(1-<x,y>)^{\frac{n+2}{2}}}(2(1-<x,y'>)+x\wedge y'-x(x\wedge y')y').\]

1025: The expression

1026: \[2(1-<x,y'>)+x\wedge y'-x(x\wedge y')y'\]

1027: is equal to

1028: \[2-2<x,y'>+xy'+<x,y'>-x(xy+<x,y'>)y'.\]

1029: This expression simplifies to

1030: \[1-<x,y'>+xy'-xy'<x,y'>\]

1031: which in turn simplifies to

1032: \[(1-<x,y'>)(1+xy')=-x(1-<x,y'>)(x-y').\]

1033: So

1034: \[x\Lambda_{n}G_{s}(x,y')=\frac{-n}{2}xG_{s}(x,y').\]

1035: Hence $x(\Lambda_{n}+\frac{n}{2})G_{s}(x,y')=0$. So the Dirac operator, $D_{s}$, over the sphere is $x(\Lambda_{n}+\frac{n}{2})$. It follows from our Cauchy integral formula for the sphere that $D_{s}f'(x)=0$. For more details on this operator, related operators and their properties see \cite{cnm,lr,r5,r6,v}

1036: 

1037: \ Besides the operator $D_{s}$ we also need a Laplacian $\triangle_{s}$ acting on functions defined on domains on $S^{n}$. To do this we will work backwards, and look first for a fundamental solution to $\triangle_{n}$. A strong candidate for such a fundamental solution is the kernel $H_{s}(x,y')=\frac{1}{n-2}\frac{1}{\|x-y'\|^{n-2}}$. By similar considerations to those made in the previous calculation we find that $D_{s}H_{s}(x,y')=-xH_{s}(x,y')+G_{s}(x,y')$. So $(D_{s}+x)H_{s}(x,y')=G_{s}(x,y')$. Therefore we may define our Laplacian $\triangle_{s}$ to be $D_{s}(D_{s}+x)$.

1038: 

1039: \begin{definition}

1040: Suppose $h$ is a $Cl_{n}$ valued function defined on a domain $U'$ of $S^{n}$. Then $h$ is called a harmonic function on $U'$ if $\triangle_{s}h=0$.

1041: \end{definition}

1042: 

1043: \ In much the same way as one would derive Green's Theorem in $R^{n}$ one now has:

1044: \begin{theorem}

1045: Suppose $U'$ is a domain on $S^{n}$ and $h:U'\rightarrow Cl_{n}$ is a harmonic function on $U'$. Suppose also that $S'$ is a smooth hypersurface lying in $U'$ and that $S'$ bounds a subdomain $V'$ of $U'$ and that $y'\in V'$. Then

1046: \[h(y')=\frac{1}{\omega_{n}}\int_{S'}(G_{s}(x,y')n(x)h(x)+H_{s}(x,y')n(x)D_{s}h(x))d\sigma'(x).\]

1047: \end{theorem}

1048: 

1049: \ See \cite{lr} for more details.

1050: 

1051: \begin{center}{\Large The Fourier Transform and Clifford Analysis}\end{center}

1052: 

1053: \ Closely related to Hardy spaces is the Fourier transform. Here we will consider $R^{n}$ as divided into upper and lower half space, $R^{n+}$ and $R^{n-}$, respectively. Where $R^{n+}=\{x=x_{1}e_{1}+\ldots+x_{n}e_{n}:x_{n}>o\}$ and $R^{n-}=\{X=x_{1}e_{1}+\ldots+x_{n}e_{n}:x_{n}<0\}$. These two domains have $R^{n-1}=$ span$<e_{1},\ldots,e_{n-1}>$ as common boundary. As before we have that $L^{2}(R^{n-1})=H^{2}(R^{n+})\otimes H^{2}(R^{n-})$. Let us now consider a function $\psi\in L^{2}(R^{n-1})$. Then \[\psi(y)=(\frac{1}{2}\psi(y)+\frac{1}{\omega_{n}}PV\int_{R^{n-1}}G(x'-y)e_{n}\psi(x')dx^{n-1})\]

1054: \[+(\frac{1}{2}\psi(y)-\frac{1}{\omega_{n}}PV\int_{R^{n-1}}G(x'-y)e_{n}\psi(x')dx^{n-1})\]

1055: almost everywhere. Here $\frac{1}{2}\psi(y)+\frac{1}{\omega_{n}}PV\int_{R^{n-1}}G(x'-y)e_{n}\psi(x')dx^{n-1}$ is the nontangential limit of $\frac{1}{\omega_{n}}\int_{R^{n-1}}G(x'-y(t))e_{n}\psi(x')dx^{n-1}$ as $y(t)$ tends to $y$ nontangentially through a smooth path in upper half space, and $\frac{1}{2}\psi(y)-\frac{1}{\omega_{n}}PV\int_{R^{n-1}}G(x'-y)e_{n}\psi(x')dx^{n-1}$ is the nontangential limit of $\frac{-1}{\omega_{n}}\int_{R^{n-1}}G(x'-y(t))e_{n}\psi(x')dx^{n-1}$ as $y(t)$ tends nontangentially to $y$ through a smooth path in lower half space.

1056: 

1057: \ Consider now the Fourier transform, ${\cal F}(\psi)=\hat{\psi}$, of $\psi$. In order to proceed we need to calculate

1058: \[{\cal F}(\frac{1}{2}\psi\pm\frac{1}{\omega_{n}}PV\int_{R^{n-1}}G(x'-y)e_{n}\psi(x')dx^{n-1}).\]

1059: In particular we need to determine ${\cal F}(\frac{1}{\omega_{n}}PV\int_{R^{n-1}}G(x'-y)e_{n}\psi(x')dx^{n-1})$. Following \cite{bla} it may be determined that this is $\frac{1}{2}i\frac{\zeta}{\|\zeta\|} e_{n}\hat{\psi}(\zeta)$ where $\zeta=\zeta_{1}e_{1}+\ldots+\zeta_{n-1}e_{n-1}$. So

1060: \[{\cal F}(\frac{1}{2}\psi\pm\frac{1}{\omega_{n}}\int_{R^{n-1}}G(x'-y)e_{n}\psi(x')d\sigma(x')=\frac{1}{2}(1\pm i\frac{\zeta}{\|\zeta\|} e_{n}).\]

1061: Now as observed in \cite{blabla}

1062: \[(\frac{1}{2}(1\pm i\frac{\zeta}{\|\zeta\|} e_{n})^{2}=\frac{1}{2}(1\pm i\frac{\zeta}{\|\zeta\|} e_{n})\]

1063: \[\frac{1}{2}(1\pm i\frac{\zeta}{\|\zeta\|} e_{n})\frac{1}{2}(1\mp i\frac{\zeta}{\|\zeta\|} e_{n})=0\]

1064: and

1065: \[i\zeta e_{n}\frac{1}{2}(1\pm i\frac{\zeta}{\|\zeta\|} e_{n})=\|\zeta\|\frac{1}{2}(1\pm i\frac{\zeta}{\|\zeta\|} e_{n}).\]

1066: Now take the Cauchy Kowalewska extension of $e^{i<x',\zeta>}$. We get

1067: \[\exp(i<x',\zeta>-ix_{n}e_{n}\zeta)\]

1068: defined on some neighbourhood in $R^{n}$ of $R^{n-1}$. Now consider

1069: \[\exp(i<x',\zeta>-ix_{n}e_{n}\zeta)\frac{1}{2}(1\pm i\frac{\zeta}{\|\zeta\|} e_{n}).\]

1070: This simplifies to

1071: \[e^{i<x',\zeta>+x_{n}\|\zeta\|}\frac{1}{2}(1+i\frac{\zeta}{\|\zeta\|} e_{n})\]

1072: if $x_{n}>0$ and to

1073: \[e^{i<x',\zeta>-x_{n}\|\zeta\|}\frac{1}{2}(1-i\frac{\zeta}{\|\zeta\|} e_{n})\]

1074: if $x_{n}<0$. The first of these series converges locally uniformly on upper half space while the second series converges locally uniformly on lower half space. We denote these two functions by $e_{\pm}(x,\zeta)$ respectively. So the integrals $\frac{1}{\omega_{n}}\int_{R^{n-1}}e_{\pm}(x,\zeta)\hat{\psi}(\zeta)d\sigma(\zeta)$ define left monogenic functions $\Psi_{\pm}(x)$. These left monogenic functions are defined on upper and lower half space respectively. Moreover $\Psi_{\pm}\in H^{2}(R^{n\pm})$ respectively, and $\Psi_{\pm}$ explicitly give the Hardy space decomposition of $\psi\in L^{2}(R^{n-1})$. This approach to the links between the Fourier transform and Clifford analysis were first carried out in \cite{s3} and later independently rediscovered and applied in \cite{lmq}.

1075: 

1076: \ The integrals $\frac{1}{\omega_{n}}\int_{R^{n-1}}e_{\pm}(x,\zeta)d\zeta^{n-1}$ can be expressed in polar co-ordinates as an integral $\frac{1}{\omega_{n}}\int_{S^{n-2}}\int_{0}^{\infty}e^{ir<x',\zeta'>\pm x_{n}r}r^{n-2}\frac{1}{2}(1\pm\frac{\zeta}{\|\zeta\|}e_{n})drdS^{n-2}$, where $\zeta'=\frac{\zeta}{\|\zeta\|}$. In \cite{lmq} Chun Li observed that the integrals $\int_{0}^{\infty}e^{ir<x',\zeta'>\pm x_{n}r}\frac{1}{2}(1\pm \zeta' e_{n})r^{n-2}dr$ are Laplace transforms of the function $f(R)=R^{n-2}$. So the integral

1077: \[\int_{0}^{\infty}e^{ir<x',\zeta'>-x_{n}r}\frac{1}{2}(1-i\zeta' e_{n})dr\]

1078: evalutes to $(-i)^{n}(n-2)!(<x',\zeta'>+ix_{n})^{-n+1}$. Hence the integral

1079: \[\frac{1}{\omega_{n}}\int_{R^{n-1}}e_{+}(x,\zeta)d\zeta^{n-1}\]

1080: becomes

1081: \[\frac{1}{\omega_{n}}\int_{S^{n-2}}(-i)^{n}(n-2)!(<x',\zeta'>+ix_{n})^{-n+1}\frac{1}{2}(1+ie_{n}\zeta')dS^{n-2}.\]

1082: 

1083: \ For any complex number $a+ib$ the product $(a+ib)\frac{1}{2}(1+ie_{n}\zeta')$ is equal to $(a-be_{n})\frac{1}{2}(1+ie_{n}\zeta')$. Thus the previous integral becomes

1084: \[\frac{1}{\omega_{n}}\int_{S^{n-2}}(-i)^{n}(<x',\zeta'>-x_{n}e_{n}\zeta')^{-n+1}\frac{1}{2}(1+ie_{n}\zeta')dS^{n-2}.\]

1085: The imaginary, or $iCl_{n}$ part of this integral is the integral of an odd function so when $n$ is even the integral becomes

1086: \[\frac{1}{\omega_{n}}\int_{S^{n-2}}\frac{(n-2)!e_{n}\zeta'}{2(<x',\zeta'>-x_{n}e_{n}\zeta')^{n-1}}dS^{n-2}\]

1087: and when $n$ is odd the integral becomes

1088: \[\frac{-1}{\omega_{n}}\int_{S^{n-2}}\frac{(n-2)!}{(<x',\zeta'>-x_{n}e_{n}\zeta')^{n-1}}dS^{n-2}.\]

1089: These integrals are the plane wave decompositions of the Cauchy kernel for upper half space described by Sommen in \cite{s4}.

1090: \ It should be noted that while introducing the Fourier transform and exploring some of its links with Clifford analysis we have also been forced to complexify the Clifford algebra $Cl_{n}$ so that we now work with the complex Clifford algebra $Cl_{n}(C)$. Further the functions $\frac{1}{2}(1\pm i\frac{\zeta}{\|\zeta\|e_{n}})$ can be seen as defined on spheres lying in the null cone $\{x_{n}e_{n}+iw': w'\in R^{n-1}$ and $x_{n}^{2}-\|w'\|^{2}=0\}$. This leads naturally to the question: what domains in $C^{n}$ do the functions $e_{\pm}(x,\zeta)$ extend to? Here we are replacing the real vector variable $x\in R^{n}$ by a complex vector variable $\underline{z}=z_{1}e_{1}+\ldots+z_{n}e_{n}\in C^{n}$, where $z_{1},\ldots,z_{n}\in C$. On placing $\underline{z}=x+iy$ where $x$ and $y$ are real vector variables, it may be noted that the term is $e^{-<\zeta,y'>-x_{n}\|\zeta\|}$ is well defined for $x_{n}\|\zeta\|>|<\zeta,y'>|$. So in this case $iy'+x_{n}e_{n}$ varies over the interior of the forward null cone $\{iy'+x_{n}e_{n}:x_{n}>0$ and $x_{n}>\|y'\|\}$. So $e_{+}(\underline{z},\zeta)$ is well defined for each $\underline{z}=x+iy=x'+iy'+(x_{n}+iy_{n})e_{n}\in C^{n}$ where $x'\in R^{n-1}$, $y_{n}\in R$, $x_{n}>0$ and $\|y'\|<x_{n}$. Similarly $e_{-}(x,\zeta)$ holomorphically extends to $\{\underline{z}=x'+iy'+(x_{n}+iy_{n})e_{n}:x'\in R^{n-1}$, $x_{n}<0$, $y_{n}\in R$ and $\|y'\|<|x_{n}|\}$. We denote these domains by $C^{\pm}$ respectively. It should be noted that $\Psi^{\pm}$ holomorphically continue to $C^{\pm}$ respectively. We denote these holomorphic continuations of $\Psi^{\pm}$ by $\Psi'^{\pm}$. The domains $C^{\pm}$ are examples of tube domains.

1091: 

1092: \begin{center}{\Large Complex Clifford Analysis}\end{center}

1093: 

1094: \ In the previous section we ended by showing that for any left Clifford holomorphic function $f\in H^{2}(R^{n,+})$ may be holomorphically continued to a function $f^{\dagger}$ defied on a tube domain $C^{+}$ in $C^{n}$. In this section we will briefly show how this type of holomorphic continuation happens for all Clifford holomorphic functions defined on a domain $U\subset R^{n}$. First let $S$ be a compact smooth hypersurface lying in $U$ and suppose that $S$ bounds a subdomain $V$ of $U$. Cauchy's integral formula gives us

1095: \[f(y)=\frac{1}{\omega_{n}}\int_{S}G(x-y)n(x)f(x)d\sigma(x)\]

1096: for each $y\in V$. Let us now consider complexifying the Cauchy kernel. The function $G(x)$ holomorphically continues to $\frac{\underline{z}}{(\underline{z}^{2})^{\frac{n}{2}}}$. In even dimensions this is a well defined function on $C^{n}\backslash N(0)$ where $N(0)=\{\underline{z}\in C^{n}:\underline{z}^{2}=0\}$. In odd dimensions this lifts to a well defined function on complex $n$-dimensional Riemann surface double covering $C^{n}\backslash N(0)$. Though things work out well in odd dimensions we will for ease just work with the cases where $n$ is even. In holomorphically extending $G(x-y)$ in the variable $y$ we obtain a function $G^{\dagger}(x-\underline{z})=\frac{x-\underline{z}}{((x-\underline{z})^{2})^{\frac{n}{2}}}$. This function is well defined on $C^{n}\backslash N(x)$ where $N(x)=\{\underline{z}\in C^{n}:(x-\underline{z})^{2}=0\}$. So the integral $\frac{1}{\omega_{n}}\int_{S}G^{\dagger}(x-\underline{z})n(x)f(x)d\sigma(x)$ is well defined provided $\underline{z}$ is not in $N(x)$ for any $x\in S$. It may be determined that the set $C^{n}\backslash\cup_{x\in S}N(x)$ is an open set in $C^{n}$. We shall take the component of this open set which contains $V$. We shall denote this connected open subset of $C^{n}$ by $V^{\dagger}$. It now follows that the left Clifford holomorphic function $f(y)$ now has a holomorphic extension $f^{\dagger}(\underline{z})$ to $V^{\dagger}$. Furthermore this function is given by the integral formula

1097: \[f^{\dagger}(\underline{z})=\frac{1}{\omega_{n}}\int_{S}G^{\dagger}(x-\underline{z})n(x)f(x)d\sigma(x).\]

1098: 

1099: \ Furthermore the holomorphic function $f^{\dagger}$ is now a solution to the complex Dirac equation $D^{\dagger}f^{\dagger}=0$, where $D^{\dagger}=\Sigma_{j=1}^{n}e_{j}\frac{\partial}{\partial z_{j}}$. By letting allowing the hypersurface to deform and move out to include more of $U$ in its interior we see that $f^{\dagger}$ is a well defined holomorphic function on $U^{\dagger}$ where $U^{\dagger}$ is the component of $C^{n}\backslash\cup_{x\in cl(U)\backslash U}N(x)$ which contains $U$. In the special cases where $U$ is either of $R^{n,\pm}$ then $U^{\dagger}=C^{\pm}$. See \cite{r1,r2,r3} for more details.

1100: 

1101: \begin{thebibliography}{999}

1102: \bibitem{a} L. V. Ahlfors, {\em M\"{o}bius transformations in $R^{n}$ expressed

1103: through

1104: $2\times 2$ matrices of Clifford numbers}, Complex Variables, 5, 1986,

1105: 215-224.

1106: \bibitem{abs} M. F. Atiyah, R. Bott and A. Shapiro {\em Clifford modules},

1107: Topology, 3, 1965, 3-38.

1108: \bibitem{b} B. Bojarski {\em Conformally covariant differential operators},

1109: Proceedings, XX th Iranian Math. Congress, Tehran, 1989.

1110: \bibitem{bbw} B. Booss-Bavnbek and K. Wojciechowski, {\em Elliptic Boundary

1111: Problems for Dirac Operators}, Birkhauser, Basel, 1993.

1112: \bibitem{bds} F. Brackx, R. Delanghe and F. Sommen {\em Clifford Analysis},

1113: Pitman, London, 1982.

1114: \bibitem{c} D. Calderbank, {\em Clifford analysis for Dirac operators on manifolds

1115: with boundary}, Max-Plank-Institut f\"{u}r Mathematik preprint, Bonn, 1996.

1116: \bibitem{cmm} R. Coifman, A. McIntosh and Y. Meyer, {\em L'int\'{e}grale de Cauchy d\'{e}finit un op\'{e}rator born\'{e} sur $L^{2}$ pour les courbes lipschitziennes}, Annals of Mathematics, 116, 1982, 361-387.

1117: \bibitem{cnm} J. Cnops and H. Malonek {\em An introduction to Clifford analysis},

1118: Univ. Coimbra, Coimbra, 1995.

1119: \bibitem{d} R. Delanghe, {\em On regular-analytic functions with values in a

1120: Clifford algebra}, Math. Ann., 185, 1970, 91-111.

1121: \bibitem{dss} R. Delanghe, F. Sommen and V. Soucek, {\em Clifford Algebra and

1122: Spinor-Valued Functions}, Kluwer, Dordrecht, 1992.

1123: \bibitem{di} A. C. Dixon, {\em On the Newtonian potential}, Quarterly Journal of

1124: Mathematics, 35, 1904, 283-296.

1125: \bibitem{f} R. Fueter, {\em Die Funktionentheorie der Differentialgleichungen

1126: $\triangle u=0$ und $\triangle \triangle u=0$ mit

1127: Vier Reallen Variablen}, Commentarii Mathematici Helvetici 7, 1934-1935,

1128: 307-330.

1129: \bibitem{gs} K. Guerlebeck and W. Sproessig, {\em Quaternionic and Clifford Calculus

1130: for Physicists and Engineers}, Wiley, Chichester, 1998.

1131: \bibitem{gm} J. Gilbert and M. A. M. Murray, {\em Clifford Algebras and Dirac

1132: Operators in Harmonic Analysis}, CUP, Cambridge, 1991.

1133: \bibitem{h} D. Hestenes, {\em Multivector functions}, J. Math. Anal. Appl., 24,

1134: 1968, 467-473.

1135: \bibitem{i} V. Iftimie, {\em Fonctions hypercomplexes}, Bull. Math. de la Soc. Sci.

1136: Math. de la R. S. de Roumanie 9, 1965, 279-332.

1137: \bibitem{k}R. S. Krausshar, {\em Automorphic forms in Clifford analysis}, Complex Variables, 47, 2002, 417-440.

1138: \bibitem{kr1} R. S. Krausshar and J. Ryan, {\em Clifford and harmonic analysis on cylinders and tori}, to appear.

1139: \bibitem{kr2} R. S. Krausshar and J. Ryan, {\em Some conformally flat spin manifolds, Dirac operators and automorphic forms}, to appear.

1140: \bibitem{ks} V. Kravchenko and M. Shapiro, {\em Integral Representations for Spatial

1141: Models of Mathematical Physics}, Pitman Research Notes in Mathematics, London,

1142: no 351, 1996.

1143: \bibitem{lms} C. Li, A. McIntosh and S. Semmes, {\em Convolution singular integrals

1144: on Lipschitz surfaces}, J. of the AMS, 5, 1992, 455-481.

1145: \bibitem{lmq} C. Li, A. McIntosh and T. Qian, {\em Clifford algebras, Fourier

1146: transforms, and singular convolution operators on Lipschitz surfaces},

1147: Revista Mathematica Iberoamericana, 10, 1994, 665-721.

1148: \bibitem{lg} T. E. Littlewood and C. D. Gay, {\em Analytic spinor fields}, Proc.

1149: Roy. Soc., A313, 1969, 491-507.

1150: \bibitem{l} J. Liouville {\em Extension au cas des trois dimensions de la

1151: question du trace geographique. Applications de l'analyse a geometrie}, G.

1152: monge, Paris, 1850, 609-616.

1153: \bibitem{lr} H. Liu and J. Ryan {\em Clifford analysis techniques for spherical pde}, Journal of Fourier Analysis and its Applications, 8, 2002, 535-564.

1154: \bibitem{mc} A. McIntosh, {\em Clifford algebras, Fourier theory, singular integrals, and harmonic functions on Lipschitz domains}, Clifford Algebras in Analysis and Related Topics, ed J. Ryan, 33-88, CRC Press, Boca Raton, 1996.

1155: \bibitem{m1} M. Mitrea, {\em Clifford Wavelets, Singular Integrals and Hardy Spaces},

1156: Lecture Notes in Mathematics, Springer-Verlag, Heidelberg, No 1575, 1994.

1157: \bibitem{m2} M. Mitrea, {\em Generalized Dirac operators on non-smooth manifolds and

1158: Maxwell's equations}, Journal of Fourier Analysis and its Applications, 7, 2001, 207-256.

1159: \bibitem{mt} Gr. C. Moisil and N. Theodorescu, {\em Fonctions holomorphes dans

1160: l'espace}, Mathematica (Cluj), 5, 1931, 142-159.

1161: \bibitem{m} M. Murray, {\em The cauchy integral, Calderon commutation, and conjugation of singular integrals in $R^{n}$}, Trans. of the AMS, 298, 1985, 497-518.

1162: \bibitem{o} E. Obolashvili, {\em Partial Differential Equations in Clifford Analysis},

1163: Pitman Monographs and Surveys in Pure and Applied Mathematics no 96, Harlow,

1164: 1998.

1165: \bibitem{q1} T. Qian, {\em Singular integrals on star shaped Lipschitz surfaces in the quaternionic space}, Math Ann, 310, 1998, 601-630.

1166: \bibitem{q2} T. Qian, {\em Fourier analysis on starlike Lipschitz surfaces}, Journal of Functional Analysis, 183, 2001, 370-412.

1167: \bibitem{r1} J. Ryan, {\em Cells of Harmonicity and generalized Cauchy integral

1168: formulae}, Proc. London Math. Soc, 60, 1990, 295-318.

1169: \bibitem{r2} J. Ryan, {\em Intrinsic Dirac operators in $C^{n}$}, Advances in

1170: Mathematics, 118, 1996, 99-133.

1171: \bibitem{r3} J. Ryan, {\em Complex Clifford analysis and domains of holomorphy},

1172: Journal of the Australian Mathematical Society, Series A, 48, 1990, 413-433.

1173: \bibitem{r4} J. Ryan, Ed {\em Clifford Algebras in Analysis and Related Topics},

1174: Studies in Advanced Mathematics, CRC Press, Boca Raton, 1995.

1175: \bibitem{r5} J. Ryan, {\em Basic Clifford analysis}, Cubo Matem\/{a}tica

1176: Educacional, 2, 2000, 226-256.

1177: \bibitem{r6} J. Ryan, {\em Dirac operators on spheres and hyperbolae}, Bolletin de la Sociedad Matematica a Mexicana, 3, 1996, 255-270.

1178: \bibitem{sc} M. Sce {\em Osservaziono sulle serie di potenzi nei moduli

1179: quadratici}, Atti. Acad. Nax. Lincei Rend Sci Fis Mat. Nat, 23, 1957,

1180: 220-225.

1181: \bibitem{se} S. Semmes, {\em Some remarks concerning integrals of curvature on curves and surfaces}, to appear.

1182: \bibitem{s1} F. Sommen {\em Spherical monogenics and analytic functionals on the

1183: unit sphere}, Tokyo J. Math., 4, 1981, 427-456.

1184: \bibitem{s2} F. Sommen {\em A product and an exponential function in hypercomplex

1185: function theory}, Appl. Anal., 12, 1981, 13-26.

1186: \bibitem{s3} F. Sommen, {\em Microfunctions with values in a Clifford algebra II}, Scientific Papers of the College of Arts and Sciences, The University of Tokyo, 36, 1986, 15-37.

1187: \bibitem{s4} F. Sommen, {\em Plane wave decomposition of monogenic functions}, Ann. Polon. Math., 49,  1988, 101-114.

1188: \bibitem{sw} E. M. Stein and G. Weiss {\em Introduction to Fourier Analysis on

1189: Euclidean Space}, Princeton University Press, Princeton, 1971.

1190: \bibitem{su} A. Sudbery {\em Quaternionic analysis}, Math. Proc. of the Cambridge

1191: Philosophical Soc., 85, 1979, 199-225.

1192: \bibitem{v} P. Van Lancker, {\em Clifford analysis on the sphere}, Clifford Algebras and their Applications in Mathematical Physics, V. Diedrich et al, editors, Kluwer, Dordrecht, 1998, 201-215.

1193: \bibitem{w} S. Wu, {\em Well-posedness in Sobolev spaces of the full water wave problem in $3$-D}, Journal of the AMS, 12, 1999, 445-495.

1194: \end{thebibliography}

1195: \end{document}

1196: