Upload
aud-in
View
0
Download
0
Embed Size (px)
Citation preview
Mathematical and Computer Modelling 54 (2011) 3189–3196
Contents lists available at SciVerse ScienceDirect
Mathematical and Computer Modelling
journal homepage: www.elsevier.com/locate/mcm
Subclasses of spirallike multivalent functionsNantu Sarkar a, Pranay Goswami b,∗, Teodor Bulboacă c
a Department of Mathematics, Jadavpur University, Kolkata-700032, Indiab Department of Mathematics, AMITY University Rajasthan, Jaipur-302002, Indiac Faculty of Mathematics and Computer Science, Babeş-Bolyai University, 400084 Cluj-Napoca, Romania
a r t i c l e i n f o
Article history:Received 24 February 2011Received in revised form 5 August 2011Accepted 8 August 2011
Keywords:Multivalent functionsSubordinationHadamard productDziok–Srivastava operator
a b s t r a c t
We studied convolution properties of spirallike, starlike and convex functions, and somespecial cases of the main results are also pointed out. Further, we also obtained inclusionand convolution properties for some new subclasses on p-valent functions defined by usingthe Dziok–Srivastava operator.
Crown Copyright© 2011 Published by Elsevier Ltd. All rights reserved.
1. Introduction and preliminaries
Let Ap denote the class of functions of the form
f (z) = zp +
∞−k=p+1
akzk, (p ∈ N = {1, 2, . . .}) , (1.1)
which are analytic and p-valent in the open unit disc U = {z ∈ C : |z| < 1}.If f and g are analytic functions in U, we say that f is subordinate to g , written f (z) ≺ g(z), if there exists a Schwarz
functionw, which (by definition) is analytic in U withw(0) = 0, and |w(z)| < 1 for all z ∈ U, such that f (z) = g(w(z)), forall z ∈ U. Furthermore, if the function g is univalent in U, then we have the following equivalence:
f (z) ≺ g(z) ⇔ f (0) = g(0) and f (U) ⊂ g(U).
We will define the following two subclasses of a multivalent function:
Definition 1.1. Let λ ∈ R with |λ| < π2 , let p ∈ N, and let φ be a univalent function in the unit disc U with φ(0) = 1, such
that
Reφ(z) > 1 −1p, z ∈ U. (1.2)
We define the classes Sλp (φ) and Cλp (φ) by
Sλp (φ) =
f ∈ Ap : eiλ
zf ′(z)f (z)
≺ p (φ(z) cos λ+ i sin λ)
∗ Corresponding author.E-mail addresses: [email protected] (N. Sarkar), [email protected] (P. Goswami), [email protected] (T. Bulboacă).
0895-7177/$ – see front matter Crown Copyright© 2011 Published by Elsevier Ltd. All rights reserved.doi:10.1016/j.mcm.2011.08.015
AUTHOR'S PERSONAL COPY
3190 N. Sarkar et al. / Mathematical and Computer Modelling 54 (2011) 3189–3196
and
Cλp (φ) =
f ∈ Ap : eiλ
[1 +
zf ′′(z)f ′(z)
]≺ p (φ(z) cos λ+ i sin λ)
.
Remarks 1.1. 1. It is easy to see that
f ∈ Cλp (φ) ⇔zf ′(z)
p∈ Sλp (φ). (1.3)
2. For φ(z) =1+Az1+Bz (−1 ≤ B < A ≤ 1), we set Sλp [A, B] := Sλp (φ) and Cλp [A, B] := Cλp (φ).
3. For λ = 0 we write
S0p (φ) =: S∗
p (φ) and C0p (φ) =: Cp(φ),
S0p [A, B] =: S∗
p [A, B] and C0p [A, B] =: Cp[A, B].
Moreover, for the special case p = 1, A = 1 and B = −1, the class C1[1,−1] represents the class of convex (univalent)and normalized functions (with the conditions f (0) = f ′(0)− 1 = 0) in U.
4. For 0 ≤ α < 1, the classes Sλp [1 − 2α,−1] and Cλp [1 − 2α,−1] reduce to the classes Sλp (α) and Cλp (α) of multivalentλ-spirallike functions of order α, and multivalent λ-spirallike convex functions of order α in U respectively.
Lemma 1.1. 1. If f ∈ Sλp (φ), then the function F(z) =f (z)zp−1 is a λ-spirallike (univalent) function in U.
2. Consequently, for all γ ∈ C∗= C \ {0}, the multivalued function
f (z)zp
γhas an analytic branch in U, with
f (z)zp
γ z=0
= 1.
Proof. Since f ∈ Ap we have F ∈ A1. Differentiating the definition formula of F we get zF ′(z)F(z) =
zf ′(z)f (z) − p+ 1, and using the
fact that f ∈ Sλp (φ) it follows
eiλzF ′(z)F(z)
≺ [pφ(z)+ 1 − p] cos λ+ i sin λ =: H(z). (1.4)
Using the assumption (1.2) we deduce
ReH(z) = cos λ · Re [pφ(z)+ 1 − p] > 0, z ∈ U,
whenever λ ∈ R with |λ| < π2 . Since the function H is univalent, the subordination (1.4) yields that Re
eiλ zF ′(z)
F(z)
> 0,
z ∈ U, and this last inequality shows that F is a λ-spirallike function, hence it is univalent in U. It follows F(z)z = 0 for all
z ∈ U, i.e. f (z)zp = 0, z ∈ U, that implies the second part of our lemma. �
Using the above lemma we will introduce the following two subclasses of Ap:
Definition 1.2. Let γ ∈ C∗ be an arbitrary number.
1. A function g ∈ Ap is said to be in the class Sλ,γp (φ), if there exists a function f ∈ Sλp (φ) such that g(z) = zp
f (z)zp
γ,
wheref (z)zp
γ z=0
= 1.
2. A function g ∈ Ap is said to be in the class Cλ,γp (φ), if G ∈ S
λ,γp (φ), where G(z) =
zg ′(z)p .
Remarks 1.2. 1. Note that for the special case γ = 1 the classes Sλ,γp (φ) and C
λ,γp (φ) coincide with Sλp (φ) and Cλp (φ)
respectively.2. Also, for p = 1 and φ(z) =
1+Az1+Bz (−1 ≤ B < A ≤ 1)we have S
λ,γ
1 [A, B] =: Sλ,γ [A, B], which was studied by Ahuja [1].
For the functions f ∈ Ap, given by (1.1) and g ∈ Ap of the form
g(z) = zp +
∞−k=p+1
bkzk,
the Hadamard (or convolution) product of f and g is defined by
(f ∗ g)(z) = zp +
∞−k=p+1
akbkzk.
For ai ∈ C (i = 1, 2, . . . , q) and bj ∈ C\Z0−(j = 1, 2, . . . , s), whereZ0
−= {0,−1,−2,−3, . . .}, we define the generalized
hypergeometric function qFs(a1, . . . , aq; b1, . . . , bs; z) (see, for example, [2]) by the following infinite series
AUTHOR'S PERSONAL COPY
N. Sarkar et al. / Mathematical and Computer Modelling 54 (2011) 3189–3196 3191
qFs(a1, . . . , aq; b1, . . . , bs; z) =
∞−k=0
(a1)k · · · (aq)k(b1)k · · · (bs)k
zk
k!, z ∈ U, (1.5)
where q, s ∈ N0 = N ∪ {0} with q ≤ s + 1, and (λ)r is the Pochhammer symbol defined in terms of the Gamma function by
(λ)r =Γ (λ+ r)Γ (λ)
=
1, for r = 0, λ ∈ C \ {0},λ(λ+ 1) · · · (λ+ r − 1), for r ∈ N, λ ∈ C.
Corresponding to a function hp(a1, . . . , aq; b1, . . . , bs; z) given by
hp(a1, . . . , aq; b1, . . . , bs; z) = zp ·q Fs(a1, . . . , ai, . . . , aq; b1, . . . , bs; z).
Dziok and Srivastava (see [3,4]) considered the linear operator
Hp(a1, . . . , aq; b1, . . . , bs) : Ap → Ap
defined by the following Hadamard product:
Hp(a1, . . . , aq; b1, . . . , bs)f (z) = hp(a1, . . . , aq; b1, . . . , bs; z) ∗ f (z).
Thus, for a function f of the form (1.1) we have
Hp(a1, . . . , aq; b1, . . . , bs)f (z) = zp +
∞−k=p+1
Γk−p[a1; b1]akzk,
where
Γk−p[a1; b1] :=1
(k − p)!(a1)k−p · · · (aq)k−p
(b1)k−p · · · (bs)k−p,
and for convenience, we write
Hp;q,s(a1) := Hp(a1, . . . , aq; b1, . . . , bs).
For the various properties and special cases of this operator we may refer to [2,5,3,4,6,7], and many others.
Definition 1.3. Let λ ∈ R with |λ| < π2 , let p ∈ N, and let φ be an univalent function in the unit disc U with φ(0) = 1, that
satisfies (1.2). For q, s ∈ N0, with q ≤ s + 1, we define the following two subclasses of analytic functions:
Sλ,γ
p;q,s[a1;φ] := Sλ,γp [a1, . . . , aq; b1, . . . , bs;φ]
=
f ∈ Ap : Hp;q,s(a1)f ∈ Sλ,γp (φ)
and
Cλ,γ
p;q,s[a1;φ] := Cλ,γp [a1, . . . , aq; b1, . . . , bs;φ]
=
f ∈ Ap : Hp;q,s(a1)f ∈ Cλ,γp (φ)
.
In particular, for γ = 1 we denote Sλp;q,s[a1;φ] := Sλ,1p;q,s[a1;φ] and Cλp;q,s[a1;φ] := Cλ,1p;q,s[a1;φ].
In the present paper we studied several convolution and inclusion properties of the subclasses introduced in this section,that extend some known results obtained by Ahuja [1], and Aouf and Soudy [5].
2. Convolution properties
Theorem 2.1. Let f ∈ Ap, let γ ∈ C∗= C \ {0} be an arbitrary number, and let g(z) = zp
f (z)zp
γ, where the power function
is considered to the main branch, i.e.
f (z)zp
γ z=0
= 1. Then, g ∈ Sλp (φ) if and only if
1zp
[f (z) ∗
zp − Czp+1
(1 − z)2
]= 0, z ∈ U, (2.1)
and for all
C = Cγ (x) =[1 − φ(x)] p cos λ− γ eiλ
[1 − φ(x)] p cos λ, |x| = 1, (2.2)
and also for C = 1, if γ ∈ C∗\ {1}.
AUTHOR'S PERSONAL COPY
3192 N. Sarkar et al. / Mathematical and Computer Modelling 54 (2011) 3189–3196
Proof. For any f ∈ Ap, we may easily deduce that
f (z) = f (z) ∗zp
1 − zand zf ′(z) = f (z) ∗
[zp+1
(1 − z)2+
pzp
1 − z
]. (2.3)
1. Suppose that g ∈ Sλp (φ), where g(z) = zpf (z)zp
γ. Since
γ eiλzf ′(z)f (z)
+ eiλp(1 − γ ) = eiλzg ′(z)g(z)
,
it follows that g ∈ Sλp (φ) if and only if
γ eiλzf ′(z)f (z)
+ eiλp(1 − γ ) ≺ p (φ(z) cos λ+ i sin λ) . (2.4)
Using the fact that the function from the right-hand side of the above subordination is univalent, from (2.4) it followsthat
γ eiλzf ′(z)f (z)
+ eiλp(1 − γ )− pφ(x) cos λ− ip sin λ = 0, ∀|x| = 1, ∀ z ∈ U. (2.5)
From the convolution relations (2.3), a simple computation shows that (2.5) is equivalent to (2.1), where C is givenby (2.2).
If γ ∈ C∗\ {1}, since g(z) = zp
f (z)zp
γis analytic in U it follows that f (z)
zp = 0 for all z ∈ U, and according to (2.3) thisis equivalent to
1zp
[f (z) ∗
zp
1 − z
]= 0, z ∈ U, (2.6)
that represents (2.1) for C = 1.2. Reversely, if γ ∈ C∗
\ {1}, since the assumption (2.1) holds for C = 1, it follows that (2.6) holds, and according to (2.3)
this shows that f (z)zp = 0 for all z ∈ U. Thus, the power function
f (z)zp
γhas uniform branches in the unit disc U, hence
the function g is well defined.Since it was shown in the first part of the proof that the assumption (2.1) is equivalent to (2.5), we obtain that
γ eiλzf ′(z)f (z)
+ eiλp(1 − γ ) = p (φ(x) cos λ+ i sin λ) , ∀|x| = 1, ∀ z ∈ U. (2.7)
Denoting
ϕ(z) = γ eiλzf ′(z)f (z)
+ eiλp(1 − γ ), ψ(z) = p (φ(z) cos λ+ i sin λ) ,
the relation (2.7) shows that ϕ(U)∩ψ(∂U) = ∅, and thus the simply-connected domain ϕ(U) is included in a connectedcomponent of C \ ψ(∂U). From here, and using the fact that ϕ(0) = ψ(0) together with the univalence of the functionψ , it follows that ϕ(z) ≺ ψ(z), which represents in fact the subordination (2.4), i.e. g ∈ Sλp (φ). �
Theorem 2.2. If g ∈ Ap, then g ∈ Sλ,γp (φ) if and only if
1zp
[g(z) ∗
zp − Dzp+1
(1 − z)2
]= 0, z ∈ U, (2.8)
for all
D = Dγ (x) =[1 − φ(x)] pγ cos λ− eiλ
[1 − φ(x)] pγ cos λ, |x| = 1. (2.9)
Proof. From Definition 1.2, g ∈ Sλ,γp (φ) if there exists a function f ∈ Sλp (φ) such that g(z) = zp
f (z)zp
γ, where
f (z)zp
γ z=0
= 1.Since
γzf ′(z)f (z)
+ p(1 − γ ) =zg ′(z)g(z)
,
AUTHOR'S PERSONAL COPY
N. Sarkar et al. / Mathematical and Computer Modelling 54 (2011) 3189–3196 3193
and using the fact that g ∈ Sλ,γp (φ) if and only if f ∈ Sλp (φ), it follows that g ∈ S
λ,γp (φ) is equivalent to
eiλzg ′(z)g(z)
− eiλp(1 − γ ) = γ eiλzf ′(z)f (z)
≺ pγ (φ(z) cos λ+ i sin λ) .
Using the same arguments as in the proof of Theorem 2.1, we find that the above subordination holds if and only if therelation (2.8) is satisfied whenever D has the form (2.9). �
Remark 2.1. Settingφ(z) =1+Az1+Bz (−1 ≤ B < A ≤ 1) andwriting−x as x in Theorems 2.1 and 2.2,we obtain the convolution
results obtained by Ahuja [1].
For γ = 1 we have g(z) = f (z), where g is defined like in Theorem 2.1, and C1(x) ≡ D1(x), hence both of the results ofTheorems 2.1 and 2.2 coincide with the following one:
Corollary 2.1. If f ∈ Ap, then f ∈ Sλp (φ) if and only if (2.1) holds for all
C = C1(x) =[1 − φ(x)] p cos λ− eiλ
[1 − φ(x)] p cos λ, |x| = 1. (2.10)
Theorem 2.3. If f ∈ Ap, then f ∈ Cλp (φ) if and only if
1zp
[f (z) ∗
pzp − [p + C(p + 1)− 2] zp+1+ C(p − 1)zp+2
(1 − z)3
]= 0, z ∈ U, (2.11)
for all C , where C = C1(x) is given by (2.10).
Proof. If we let G(z) =zp−Czp+1
(1−z)2, then
zG′(z) =
[pzp − [p + C(p + 1)− 2] zp+1
+ C(p − 1)zp+2
(1 − z)3
].
Using the duality relation (1.3), we have
f ∈ Cλp (φ) ⇔zf ′(z)
p∈ Sλp (φ),
and according to Corollary 2.1 we deduce that f ∈ Cλp (φ) if and only if
1zp
[zf ′(z)
p∗ G(z)
]= 0, z ∈ U, (2.12)
for all C given by (2.2). Since the identity zf ′(z) ∗ g(z) = f (z) ∗ zg ′(z) holds for all f , g ∈ Ap, in particular we havezf ′(z) ∗ G(z) = f (z) ∗ zG′(z), hence the condition (2.12) reduces to
1zp
f (z) ∗ zG′(z)
= 0, z ∈ U.
This last relation is equivalent to (2.11), which gives the required result. �
Theorem 2.4. If f ∈ Ap, then f ∈ Cλ,γp (φ) if and only if
1zp
[f (z) ∗
pzp − [p + D(p + 1)− 2] zp+1+ D(p − 1)zp+2
(1 − z)3
]= 0, z ∈ U, (2.13)
for all D, where D = Dγ (x) is given by (2.9).
Proof. Letting φ(z) =zp−Dzp+1
(1−z)2, we have
zH ′(z) =
[pzp − [p + D(p + 1)− 2] zp+1
+ D(p − 1)zp+2
(1 − z)3
].
From Definition 1.2 we have that f ∈ Cλ,γp (φ) if and only if zf ′(z)
p ∈ Sλ,γp (φ), and according to Theorem 2.2 we deduce
that f ∈ Cλ,γp (φ) if and only if
1zp
[zf ′(z)
p∗ φ(z)
]= 0, z ∈ U,
for all D given by (2.9). Now, applying exactly the same method used in the proof of Theorem 2.3 we easily obtain thedesired result. �
AUTHOR'S PERSONAL COPY
3194 N. Sarkar et al. / Mathematical and Computer Modelling 54 (2011) 3189–3196
Remark 2.2. For γ = 1, since C1(x) ≡ D1(x), the results of Theorems 2.3 and 2.4 coincide.
Theorem 2.5. If f ∈ Ap is of the form (1.1), then f ∈ Sλ,γ
p;q,s[a1;φ] if and only if
1 +
∞−k=p+1
[[1 − φ(x)] pγ cos λ+ (k − p)eiλ
[1 − φ(x)] pγ cos λ
]Γk−p[a1; b1]akzk−p
= 0, z ∈ U, (2.14)
for all |x| = 1.
Proof. From Theorem 2.2 we find that f ∈ Sλ,γ
p;q,s[a1;φ] if and only if
1zp
[Hp;q,s(a1)f (z) ∗
zp − Dzp+1
(1 − z)2
]= 0, z ∈ U, (2.15)
for all |x| = 1, where D is given by (2.9). Since
zp
(1 − z)2= zp +
∞−k=p+1
(k − p + 1)zk,zp+1
(1 − z)2=
∞−k=p+1
(k − p)zk, (2.16)
after some simple computations in Eq. (2.15), with the help of (2.16) we get
1zp
zp +
∞−k=p+1
[1 + (1 − D)(k − p)]Γk−p[a1; b1]akzk
= 0, z ∈ U,
for all |x| = 1, which simplifies to (2.14), and the proof of the theorem is complete. �
Theorem 2.6. If f ∈ Ap is of the form (1.1), then f ∈ Cλ,γ
p;q,s[a1;φ] if and only if
1 +
∞−k=p+1
kp
[[1 − φ(x)] pγ cos λ+ (k − p)eiλ
[1 − φ(x)] pγ cos λ
]Γk−p[a1; b1]akzk−p
= 0, z ∈ U, (2.17)
for all |x| = 1.
Proof. From Theorem 2.4 we have that f ∈ Cλ,γ
p;q,s[a1;φ] if and only if
1zp
[Hp;q,s(a1)f (z) ∗
pzp − [p + D(p + 1)− 2] zp+1+ D(p − 1)zp+2
(1 − z)3
]= 0, z ∈ U, (2.18)
for all |x| = 1, where D is given by (2.9). Now, it can be easily shown that
zp
(1 − z)3= zp +
∞−k=p+1
(k − p + 1)(k − p + 2)2
zk, z ∈ U,
zp+1
(1 − z)3=
∞−k=p+1
(k − p)(k − p + 1)2
zk, z ∈ U,
zp+2
(1 − z)3=
∞−k=p+1
(k − p − 1)(k − p)2
zk z ∈ U.
Using these identities in (2.18), after some simple computations we get
1zp
pzp +
∞−k=p+1
k[1 + (1 − D)(k − p)]Γk−p[a1; b1]akzk
= 0, z ∈ U,
for all |x| = 1, which simplifies to (2.17). �
3. Inclusion properties
In this sectionwewill discuss some inclusion relations for the classes Sλp;q,s[a1;φ] andCλp;q,s[a1;φ]. To prove these resultswe shall require the following lemma:
AUTHOR'S PERSONAL COPY
N. Sarkar et al. / Mathematical and Computer Modelling 54 (2011) 3189–3196 3195
Lemma 3.1 ([8]). Let φ be convex (univalent) inU, with Re[βφ(z)+γ ] > 0 for all z ∈ U. If q is analytic inU, with q(0) = φ(0),then
q(z)+zq′(z)
βq(z)+ γ≺ φ(z) ⇒ q(z) ≺ φ(z).
Theorem 3.1. Suppose that the function φ is convex (univalent) in U and satisfies the inequality
p Reφ(z) > (p − Re a1) sec2 λ− p [Imφ(z)+ tan λ] tan λ, z ∈ U. (3.1)
If f ∈ Sλp;q,s[a1 + 1;φ] such that Hp;q,s(a1)f (z) = 0 for all z ∈ U \ {0}, then f ∈ Sλp;q,s[a1;φ].
Proof. Suppose that f ∈ Sλp;q,s[a1 + 1;φ], and let us define
q(z) =1
cos λ
eiλ
zHp;q,s(a1)f (z)
′
pHp;q,s(a1)f (z)− i sin λ
. (3.2)
Then q is analytic in U, and using the well-known relation
zHp;q,s(a1)f (z)
′= a1Hp;q,s(a1 + 1)f (z)− (a1 − p)Hp;q,s(a1)f (z),
from (3.2) we obtain
q(z) cos λ+a1 − p
peiλ + i sin λ = eiλ
a1Hp;q,s(a1 + 1)f (z)pHp;q,s(a1)f (z)
. (3.3)
Differentiating logarithmically (3.3) and then using (3.2), we deduce that
q(z)+zq′(z)
p [q(z) cos λ+ i sin λ] e−iλ + a1 − p≺ φ(z). (3.4)
Since the inequality Rep [φ(z) cos λ+ i sin λ] e−iλ
+ a1 − p> 0, z ∈ U, is equivalent to (3.1), according to Lemma 3.1
the subordination (3.4) implies q(z) ≺ φ(z), which proves that f ∈ Sλp;q,s[a1;φ]. �
From the duality formula (1.3) we have
Hp;q,s(a1)f ∈ Cλp (φ) ⇔zHp;q,s(a1)f (z)
′
p∈ Sλp (φ),
and using the fact that
Hp;q,s(a1)zf ′(z)
p
=
zHp;q,s(a1)f (z)
′
p,
in the same way we may prove the following inclusion:
Theorem 3.2. Suppose that the function φ satisfies the inequality (3.1). If f ∈ Cλp;q,s[a1 +1;φ] such that Hp;q,s(a1)
zf ′(z)p
= 0
for all z ∈ U \ {0}, then f ∈ Cλp;q,s[a1;φ].
Acknowledgments
The second author (Dr. Pranay Goswami) is grateful to Prof. S.P. Goyal, Emeritus Scientist, Department of Mathematics,University of Rajasthan, Jaipur, India, for his guidance and constant encouragement. The authors are also thankful to refereesfor their helpful comments.
References
[1] O.P. Ahuja, Hadamard products and neighbourhoods of spirallike functions, Yokohama Math. J. 40 (1993) 95–103.[2] M.K. Aouf, Certain subclasses of meromorphically multivalent functions associated with generalized hypergeometric function, Comput. Math. Appl. 55
(2008) 494–509.[3] J. Dziok, H.M. Srivastava, Classes of analytic functions associated with the generalized hypergeometric function, Appl. Math. Comput. 103 (1) (1999)
1–13.[4] J. Dziok, H.M. Srivastava, Certain subclasses of analytic functions associated with the generalized hypergeometric function, Integral Transforms Spec.
Funct. 14 (2003) 7–18.
AUTHOR'S PERSONAL COPY
3196 N. Sarkar et al. / Mathematical and Computer Modelling 54 (2011) 3189–3196
[5] M.K. Aouf, T.M. Seoudy, Classes of analytic functions related to the Dziok–Srivastava operator, Integral Transforms Spec. Func. 22 (6) (2011) 423–430.doi:10.1080/10652469.2010.518567.
[6] J. Sokól, On some applications of the Dziok–Srivastava operator, Appl. Math. Comput. 201 (1–2) (2008) 774–780.[7] J.-L. Liu, H.M. Srivastava, Subclasses of meromorphically multivalent functions associated with a certain linear operator, Math. Comput. Modelling 39
(1) (2004) 35–44.[8] P.J. Eenigenburg, S.S.Miller, P.T.Mocanu, O.M. Reade, Second order differential inequalities in the complex plane, J.Math. Anal. Appl. 65 (1978) 289–305.
AUTHOR'S PERSONAL COPY