Professional Documents
Culture Documents
Art:10.1007/s13324 016 0125 9 PDF
Art:10.1007/s13324 016 0125 9 PDF
(2016) 6:403424
DOI 10.1007/s13324-016-0125-9
Mykhailo Bilogliadov1
Abstract We consider the minimal energy problem on the unit sphere S2 in the Euclid-
ean space R3 immersed in an external field Q, where the charges are assumed to interact
via Newtonian potential 1/r, r being the Euclidean distance. The problem is solved by
finding the support of the extremal measure, and obtaining an explicit expression for
the equilibrium density. We then apply our results to an external field generated by a
point charge, and to a quadratic external field.
1 Introduction
1 1
U (x) := d(y), I () := d(x)d(y).
|x y| |x y|
B Mykhailo Bilogliadov
mykhail@okstate.edu
Let
Define the Newtonian capacity of E as cap(E) := 1/W (E). We say that a property
holds quasi-everywhere (q.e.), if the exceptional set has a Newtonian capacity zero.
When cap(E) > 0, there is a unique E such that I ( E ) = W (E). Such E is called
the Newtonian equilibrium measure for E.
An external field is defined as a non-negative lower-semicontinuous function Q :
E [0, ], such that Q(x) < on a set of positive Lebesgue surface measure.
The weighted energy associated with Q(x) is then defined by
I Q () := I () + 2 Q(x)d(x).
The energy problem on S2 in the presence of the external field Q(x) refers to the
minimal quantity
VQ := inf{I Q () : M(E)}.
that the charges interact via Newtonian potential 1/r, with r denoting the Euclidean
distance between the point charges. We obtain an equation that describes the support
of the extremal measure, while also giving an explicit expression for the equilibrium
density. We then apply our results to an external field produced by a positive point
charge placed on the positive polar semi-axis, thus, in particular, recovering the cor-
responding results of [2,4] for the case of Newtonian potential in R3 . An application
to an external field, given by a quadratic polynomial, is also considered.
We remark that Theorems 4 and 5, as well as Proposition 4 below recover the
corresponding statements first obtained in [2,4], for the case of Newtonian potential
in R3 . The equilibrium density appearing in Theorem 6 below, was first found in [8]
for the case of general Riesz potentials.
Let (r, , ) be spherical coordinates in R3 , with 0 r , 0 , and
0 2 . A spherical cap on the sphere S2 , centered at the North Pole, is defined
via an angle , 0 < , as
C N , := {(r, , ) : r = 1, 0 , 0 2 }.
Similarly, a spherical cap centered at the South Pole, is defined in terms of an angle
, 0 < , as
C S, := {(r, , ) : r = 1, , 0 2 }.
We solve this equation using the results of [6], and thus obtain the following two
theorems, that describe explicitly the equilibrium density when support S Q is either
a spherical cap C N , , centered at the North Pole, or C S, , a spherical cap centered at
the South Pole.
where
1 d t Q( ) sin d
g(t) = . (5)
4 dt 0 cos cos t
FQ = 1 2 F() sin d . (7)
sin + 0
An analogous statement for the support being a spherical cap C S, centered at the
South Pole, is of the following nature.
where
1 d Q( ) sin d
g(t) = . (9)
4 dt t cos t cos
FQ = 1 2 F() sin d . (11)
sin +
2 Applications
We begin by considering the case of no external field, when the support is a spherical
cap centered at the South Pole, that is S Q = C S, , Q = 0. The equilibrium density for
the spherical cap C N , , centered at the North Pole, for the case of no external field,
was first found by Collins [6]. Note that the equilibrium measure obtained in [6], does
not have mass 1. The equilibrium measure for the spherical cap centered at the South
Pole C S, , was also found in [8], and in [2].
1 2 1 cos 1 1 cos
f () = 1+ tan ,
4( + sin ) cos cos cos cos
. (12)
The capacity of C S, is
1
cap(C S, ) = ( + sin ). (13)
Suppose now that the sphere S2 is immersed in an external field Q, that satisfies
the conditions of Proposition 1. Then the support S Q of the weighted equilibrium
measure Q will be a spherical cap C S, , centered at the South Pole. One of the
ways to determine the angle , which defines the extremal support C S, , is via the
Newtonian analog of the MhaskarSaff F-functional. Originally introduced for the
case of logarithmic potential [15], it was extended by Brauchart et al. [2] to the case
of the Riesz potentials on the sphere Sd1 , for d 3. Here we adapt the approach
from [2]. We begin with a definition.
where W (E) is the Newtonian energy of the compact E and E is the equilibrium
measure (with no external field) on E.
The main objective of introducing the F-functional is its following extremal property,
proved in [2] for the general Riesz potentials.
408 M. Bilogliadov
2 1 1 cos
Q() tan sin d . (15)
cos cos
It is clear that Q(x) = Q(x3 ) = q(1 + h 2 2hx3 )1/2 is a convex increasing function
on [1, 1], and is symmetric with respect to rotations about the polar axis x3 . From
Proposition 1 it then follows that the support of the equilibrium measure Q will be
a spherical cap, centered at the South Pole, that is S Q = C S, , for some [0, ).
The first step is to compute the F-functional for such external field Q.
Proposition 4 The F-functional for the spherical cap C S, in the case of Q as in
(16), has the form
q(h + 1)
F(C S, ) = 1+ 1
+ sin 2h
q(h 1) h 1
tan1 cot . (17)
h 2 h+1
For the external field Q generated by a point charge outside of the sphere S2 , we
can, in fact, specify when the extremal support S Q is a proper subset of the sphere S2 .
Indeed, the answer is given in terms of solution to Gonchars problem, as explained in
[4,5]. It states that if a point charge, located on the positive polar semi-axis, is too far
from the surface of the sphere S2 , or the magnitude q of the point charge is too small,
the electrostatic field, created by this point charge, is too weak to force the equilibrium
distribution from occupying the whole surface of the sphere S2 . More specifically, in
order for the extremal support S Q to be a spherical cap C S, , the height of the point
charge of magnitude q must satisfy h < h + , where h + is defined as the unique (real)
zero in (1, ) of the cubic equation
1 h+1 1
= . (18)
q (h 1)2 h
Weighted energy problem on the unit sphere 409
For instance, in the case of the positive unit point chargeq = 1, the critical height h +
is precisely 1 + + , where + is the golden ratio (1 + 5)/2 1.6180339887.
If a positive point charge of magnitude q is placed inside the sphere S2 , then in
order that the extremal support S Q to be a spherical cap C S, , the height of the point
charge of magnitude q must satisfy h > h , where h is defined as the unique (real)
zero in (0, 1) of the cubic equation (see [4])
1 h+1
= 1. (19)
q (1 h)2
We next write down a transcendental equation, whose solution gives the angle ,
defining the support C S, of the extremal measure Q .
Theorem 4 For the external field Q of (16), with h chosen such that h < h + or
h > h , the support S Q is a spherical cap C S, , centered at the South Pole, with
= 0 (0, ). The angle 0 is defined as a unique solution of the transcendental
equation
q(h + 1)
1+ 1
+ sin 2h
q(h 1) h 1 h+1
tan1 cot =q 2 . (20)
h 2 h+1 h + 1 2h cos
Now, as the support C S, is found via Eq. (20), we can recover the equilibrium measure
Q from integral Eq. (3). Using Theorem 3, by a direct calculation we obtain the
following result.
Theorem 5 Let 0 be a unique solution to (20). Set
q(h + 1) 1 1 cos 0
F() =
2 2 1 + h 2 2h cos cos 0 cos
h1
+
(1 + h 2 2h cos )3/2
1 h1 cos 0 cos
tan , (21)
1 + h 2 2h cos 1 cos 0 )
q(h + 1) 0 q(h 1)
FQ = 1+ 1
0 + sin 0 2h h
h 1
tan1
0
cot . (22)
h+1 2
From Theorem 4 it follows that in the case when the point charge is placed at the North
Pole, Eq. (20) takes especially simple form. Furthermore, expression (21) suggests that
in this case the equilibrium density will have a simple structure as well.
Theorem 6 For the external field Q of (16), with the point charge placed at the
North Pole, the support S Q is a spherical cap C S, , centered at the South Pole, with
= 0 (0, ). The angle 0 is defined as a unique solution of the transcendental
equation
+ q( 0 ) 2 1 cos 0
f () = 1+
4(sin 0 + 0 ) cos 0 cos
1 1 cos 0 q 1 1 cos 0
tan , (25)
cos 0 cos 2 1 cos cos 0 cos
2
where 0 .
The second application of our results is concerned with the quadratic external fields.
In particular, we consider a quadratic external field
where coefficients a, b, c chosen such that 4a 2 < b2 4ac, with a, b > 0. The choice
of the coefficients implies that the external field Q is a convex increasing function on
[1, 1], while its invariance with respect to rotations about the polar axis x3 is obvious.
It then follows from Proposition 1 that the support of the extremal measure is a cap
C S, , centered at the South Pole. Calculating F-functional for such Q, we obtain the
following statement.
Proposition 5 The F-functional for the spherical cap C S, , in the case of the
quadratic external field Q, as in (26), is given by
1
F(C S, ) = tan 32 a cos3
36( + sin ) 2
+ 4(2a + 9b) cos2 + 4(9c 5a) cos + 4a 36b + 36c
Theorem 7 For the quadratic external field Q of (26), the support S Q is a spherical
cap C S, , centered at the South Pole, with = 0 (0, ). The angle 0 is defined
as a unique solution of the transcendental equation
Having the support C S, now determined, we compute the density of the equilibrium
measure Q by invoking Theorem 3.
1 cos 0 cos
6 tan (15a cos + 9b cos 4a + 3c) ,
2
(29)
1 cos 0
1 0
FQ = tan 32 a cos3 0
36( 0 + sin 0 ) 2
+ 4(2a + 9b) cos2 0 + 4(9c 5a) cos 0 + 4a 36b + 36c
3 Proofs
Proof of Theorem 1 The strategy of the proof is to show that the equilibrium potential
U Q is Lipschitz continuous on C S, . If that is established, it will imply that the normal
derivatives of U Q exist a.e. on C S, . Hence the measure Q can be recovered from
its potential by the formula
1 U Q U Q
d Q = + d S := f () sin d d,
4 n + n
412 M. Bilogliadov
U Q (x) + Q(x) = FQ , x S Q ,
Q
U (x) + Q(x) FQ , x S2 ,
and also Q C 2 (S2 ). To show that it is indeed possible, we first remark that the
external field Q is rotationally symmetric about the polar axis. Therefore, Q is a
function of the polar angle only, that is Q = Q(). This symmetry is also inherited
by potential, hence on the sphere S2 we have U Q (x) = U Q (), x = (r, , ) S2 .
Next, note that N = {(r, , ) : r = 1, < , 0 2 }, with some
(0, ). Pick a number
such that <
< . We define a new external field Q
as follows: set Q() = Q(), for
< , while on [0,
] we tweak Q to Q in
such a way that
U Q () + Q() FQ ,
and Q C 2 (S2 ). Applying Theorem 4.2.14 from [1], we infer that = Q and
Q
FQ = FQ .
Let u and v denote the equilibrium potentials for the minimum energy problem on
C S, and S2 , respectively. Since the equilibrium measure is the same for those two
sets, it immediately follows that u = v. Now observe that the spherical cap C S, is a
part of the sphere S2 , which is a closed smooth surface. Thus we can invoke the result
of Gtz [9] to conclude that v is C 2 in an open neighborhood U of S2 . Now recalling
that u = U Q , we deduce that U Q is Lipschitz continuous on C S, .
be such two points, written in spherical coordinates. Then, for the distance |x y|,
one has
Q
2 f ( ) sin d d
U (r, , ) = , (32)
0 0 1 + r 2 2 r
2 f ( ) sin d d
U Q (, ) = .
0 0 2 2
2 f ( ) sin d d
= FQ Q(), 0 , (33)
0 0 2 2
where
a := 2 sin cos ,
2 2
b := 2 sin cos .
2 2
Note that since 0 < , it is clear that a > 0 and b > 0 for all and 0 and
0 . We need a special case of Copsons lemma [7, p. 16].
In view of Lemma 1, the inner integral in left hand side of Eq. (33) is transformed as
2 2
d d
=
0 2 2 0 a 2 + b2 2ab cos( )
min (a,b) dt
=4 .
0 a t 2 b2 t 2
2
One can easily check that a < b for < , while for > , we have a > b. Splitting
the interval of integration of the outer integral in the left hand side of Eq. (33), we
rewrite Eq. (33) as follows,
b a
dt dt
f ( ) sin d + f ( ) sin d
0 0 a 2 t 2 b2 t 2 0 a t 2 b2 t 2
2
1
= (FQ Q()), 0 .
4
Inverting the order of integration in the first integral in the left hand side of (35), as
d
f ( ) sin d
0 0 cos cos cos cos
d f ( ) sin d
= ,
0 cos cos cos cos
Letting
1 f ( ) sin d
S( ) = 0 , (37)
2 cos cos
Equation (38) is an Abel type integral equation. Since Q C 2 , the solution of (38) is
[13, p. 50, # 23]
1 d (FQ Q()) sin d
S( ) = , 0 . (39)
4 d 0 cos cos
Observing that Eq. (37) is again an Abel type integral equation with respect to
f ( ) sin , we solve it and obtain
2 1 d S( ) sin d
f ( ) = , 0 . (40)
sin d cos cos
416 M. Bilogliadov
Denote
1 d sin Q( ) d
g( ) = , (41)
4 d 0 cos cos
and let
2 1 d g( ) sin d
F( ) = , 0 . (42)
sin d cos cos
FQ 1 d sin d sin d
f ( ) = d
2 2 sin d cos cos d 0 cos cos
+F( ), 0 . (43)
Executing the sequence of integrations and differentiations on the right hand side of
(43), we arrive to
FQ 2 1 + cos 1 1 + cos
f ( ) = 1+ tan + F( ), (44)
4 cos cos cos cos
FQ 2 1 + cos 1 1 + cos
+ 1+ tan sin d.
2 0 cos cos cos cos
(46)
1 + cos
sin d = 2 sin .
0 cos cos
Weighted energy problem on the unit sphere 417
Making the change of variables t = (cos cos )/(1 + cos ), and then integrating
by parts, one can show that
1 + cos (1 cos )
tan1 sin d = + sin .
0 cos cos 2
2 1 + cos 1 1 + cos
1+ tan sin d
0 cos cos cos cos
4 2 2
= 1 cos + sin (1 cos ) sin +
2
= (sin + ).
Therefore,
FQ
1 = 2 F() sin d + (sin + ),
0
and thus
FQ = 1 2 F() sin d .
+ sin 0
where = cos cos +sin sin cos( ). Via the change of variables
= ,
we transform (47) into
2 f 0 ( ) sin d d
= FQ Q 0 (), 0
, (48)
0 0 2 2
with = , f 0 (
) = f (
) and Q 0 (
) = Q(
), and
= cos
cos +
sin
sin cos( ). The integral Eq. (48) is of the form (33). Hence Theorem 2
applies, and we obtain
2 1 d sin t g0 (t) dt
F0 (
) = ,
sin
d
cos cos t
418 M. Bilogliadov
where
1 d t sin Q 0 ( ) d
g0 (t) = .
4 dt 0 cos cos t
FQ 2 1 + cos 1 + cos
f 0 (
) = 1+ tan1 + F0 (
),
4 cos
cos cos
cos
where 0
. Going back to the variable via =
, after some algebra,
we find
1 d sin Q( ) d 2 1 d sin t g(t) dt
g(t) = , F() = ,
4 dt t cos t cos sin d cos t cos
For the Robin constant FQ , in this case we similarly obtain the expression
FQ = 1 2 F() sin d .
sin +
q
Q() = . (49)
1 + h 2h cos
2
We now insert expression (49) into (15), and evaluate the integrals on the right hand
side of (15). The first integral on the right hand side of (15) evaluates as follows,
sin d
Q() sin d = q
1 + h 2 2h cos
q
= (1 + h) 1 + h 2 2h cos . (50)
h
Weighted energy problem on the unit sphere 419
Next, we deal with the second integral on the right hand side of (15). We have
1 cos
Q() sin d
cos cos
sin d
= q 1 cos
1 + h 2h cos cos cos
2
2q 1 + h + 2 h cos(/2)
= sin log . (51)
h 2 1 + h 2 2h cos
We now turn to the last remaining integral on the right hand side of (15). Namely,
1 cos
1
Q() tan sin d
cos cos
1 cos sin d
=q tan1
cos cos 1 + h 2 2h cos
a
q 1 1 1a
= tan du,
2h 1 c u au
a 1 1 1a 1 1a
tan du = c a + 2 c + 1 tan
1 cu a u 1+a
c1 (1 + a)(c 1)
+2 1a tan1
1a (1 a)(c + 1)
1
+ log( c + 1 + a + 1) log(c a) .
2
1 cos
1
Q() tan sin d
cos cos
q 1 h1
= (h + 1) + 2(h 1) tan cot
2h 2 h+1
2q h + 1 + 2 h cos(/2)
1 + h 2h cos + sin
2 log .
h 2 1 + h 2 2h cos
420 M. Bilogliadov
Inserting all of the above auxiliary calculations into expression (15), after simplifica-
tions, we obtain the following expression for F(C S, ),
q(h + 1)
F(C S, ) = 1+ 1
+ sin 2h
q(h 1) h 1
tan1 cot .
h 2 h+1
Proof of Proposition 5 Inserting expression (26) into the formula for the Mhaskar
Saff functional (15), we see that
F(C S, ) = 2+ (a cos2 + b sin + c) sin d
2( + sin )
2 1 cos
+ (a cos2 + b sin + c) sin d
cos cos
2 1 1 cos
(a cos + b sin + c) tan
2
sin d .
cos cos
Each of the integrals on the right hand side of the last expression, can be easily evaluated
via the substitution u = cos . Evaluating those integrals, and inserting them into the
expression for F(C S, ), after some algebra, we arrive at the desired expression (27).
(52)
1
w () = (), (53)
3( + sin )
where
() = 6 sin2 w() + 8a cos3 + 2(3b 2a) cos2
2
+(3c 5a 3b) cos + a 3b 3c. (54)
Weighted energy problem on the unit sphere 421
Therefore, the critical points of w() are given by solutions of the transcendental
equation
6 sin2 w() + 8a cos3 + 2(3b 2a) cos2
2
+ (3c 5a 3b) cos + a 3b 3c = 0,
or, equivalently,
+ sin() 3w() 24a cos2 4(3b 2a) cos (3c 5a 3b) .
(55)
Our goal is to show that w ( ) > 0. Indeed, from (55) it is clear that
sin( )
w ( ) =
3( + sin )
3w( ) 24a cos2 4(3b 2a) cos (3c 5a 3b) ,
422 M. Bilogliadov
Taking into account that the sign of w ( ) is the same as the sign of w ( )(1 cos ),
it is a straightforward calculation to see that the sign of w ( ) is, in fact, determined
by the sign of
m( ) = 16 cos3 + 2(3
b 14) cos2 + 4(2 3
b) cos + 4 + 6
b,
where we denoted b := b/a > 2. Setting t = cos (1, 1), abusing notation, we
consider the function
First assume that 0 t < 1. Then m (t) < 0 on [0, 1), so that m(t) is a strictly
decreasing function on [0, 1). This obviously entails m(t) > m(1), for all t [0, 1).
A direct calculation shows that m(1) = 0, and thus m(t) > 0 for 0 t < 1. If
1 < t 0, it follows that we need to demonstrate the positivity of
Using trivial estimates, we see that p(t) > 0, for 0 t < 1. Combining the previous
two estimates, we deduce that m( ) > 0. Thus w ( ) > 0, as desired.
We are now able to conclude that any critical point of w() is a minimum of w().
Since w() is smooth on (0, ), we see that 0 is a unique local minimum for w()
on (0, ). Thus we are only left to show that 0 is a global minimum as well. Indeed,
we observe that w () > 0 on (0 , ), and w () < 0 on (0, 0 ). Hence () > 0
on (0 , ), and () < 0 on (0, 0 ). This means that w() has exactly one global
minimum on [0, ), which is either a unique solution 0 (0, ) of the transcendental
Eq. (28), if it exists, or 0 = 0, if such a solution fails to exist.
Proof of Theorem 8 Inserting Q() = a cos2 + b cos + c into (9), we easily find
sin t
g(t) = (8a cos2 t + 2(2a + 3b) cos t a + 3b + 3c). (56)
12 1 + cos t
sin2 t dt
+(3b + 3c a) .
0 1 + cos t cos t cos
All of the integrals on the right hand side of the last expression are readily evaluated
via the elementary substitution u = cos t. Evaluating the integrals and simplifying,
we find that
1 1
I () = 8a 1 cos 0 cos 0 cos (16 cos2 0
12 48
+4(5 cos 0 2) cos ) 4 cos 0 + 15 cos(2) + 9)
1 1 cos 0 cos
+ (1 cos )(4 cos + 5 cos(2) + 7) tan
16 1 cos 0
1
+2(2a + 3b) 1 cos 0 cos 0 cos (2 cos 0 + 3 cos 1)
4
1 1 cos 0 cos
+ (1 cos )(3 cos + 1) tan
4 1 cos 0
cos 0 cos
+(3b + 3c a) (1 cos ) tan1
1 cos 0
+ 1 cos 0 cos 0 cos . (57)
Differentiating the last expression, and upon inserting the result into (8), after some
algebra, we deduce that
1
F() = 1 cos 0 cos 0 cos (20a cos 0 + 60a cos + 10a + 27b)
36 2
1 cos 0
(8a cos2 0 + 10a cos 0 cos + (4a + 9b) cos 0
cos 0 cos
+(20a + 27b) cos + 15a cos(2) + 9a + 18b + 18c)
cos 0 cos
6 tan1 (15a cos2 + 9b cos 4a + 3c) .
1 cos 0
At last, we compute the Robin constant FQ . By (11) and (8), taking into account that
2 1 d g(t) sin t dt
F() =
sin d 0 cos t cos
424 M. Bilogliadov
2 1
= I (),
sin
FQ = 1 + 4(I (0 ) I ( )) .
sin 0 + 0
Inserting (57) into the latter expression, after simplifications, we infer that
1 0
FQ = 32 a cos3 0 + 4(2a + 9b) cos2 0
tan
36( 0 + sin 0 ) 2
Acknowledgments The author would like to thank his doctoral advisor Prof. Igor E. Pritsker for suggesting
the problem and stimulating discussions. The author expresses gratitude to Prof. Edward B. Saff for his
kind permission to use a reference from the forthcoming book [1], and the referee for helpful comments
and suggestions.
References
1. Brauchart, J., Dragnev, P., Saff, E.: An electrostatic problem on the sphere arising from a nearby point
charge. In: Ivanov K., Nikolov G., Uluchev R. (eds) Constructive Theory of Functions, Sozopol 2013,
pp. 1155. Drinov Academic Publishing House, Sofia (2014)
2. Brauchart, J., Dragnev, P., Saff, E.: Riesz Extremal measures on the sphere for axis-supported external
fields. J. Math. Anal. Appl. 356, 769792 (2009)
3. Brauchart, J., Dragnev, P., Saff, E.: Riesz External field problems on the hypersphere and optimal point
separation. Potential Anal. 41, 647678 (2014)
4. Brauchart, J., Dragnev, P., Saff, E.: An electrostatic problem on the sphere arising from a nearby point
charge. In: Ivanov, K., Nikolov, G., Uluchev, R. (eds.) Constructive Theory of Functions, Sozopol
2013, pp. 1155. Drinov Academic Publishing House, Sofia (2014)
5. Brauchart, J., Dragnev, P., Saff, E., Van de Woestijne, C.: A fascinating polynomial sequence arising
from an electrostatics problem on the sphere. Acta Math. Hung. 137, 1026 (2012)
6. Collins, W.D.: Note on the electrified spherical cap. Proc. Camb. Philos. Soc. 55, 377379 (1959)
7. Copson, E.T.: On the problem of the electrified disc. Proc. Edinb. Math. Soc. 8, 1419 (1947)
8. Dragnev, P., Saff, E.: Riesz spherical potentials with external fields and minimal energy points separa-
tion. Potential Anal. 26, 139162 (2007)
9. Gtz, M.: On the distribution of weighted extremal points on a surface in Rd , d 3. Potential Anal.
13, 345359 (2000)
10. Kellogg, O.: Foundations of potential theory. Dover, New York (1954)
11. Landkof, N.: Foundations of modern potential theory. Springer, Heidelberg (1972)
12. Mizuta, Y.: Potential theory in Euclidean spaces. Gakkotosho Co. Ltd, Tokyo (1996)
13. Polyanin, A., Manzhirov, A.: Handbook of Integral Equations, 2nd edn. CRC Press, Boca
Raton (2008)
14. Pritsker, I.: Weighted energy problem on the unit circle. Constr. Approx. 23, 103120 (2006)
15. Saff, E., Totik, V.: Logarithmic potentials with external fields. Springer, Berlin Heidelberg (1997)