Analysis of the boundary value problem associated with

the nonrelativistic ThomasFermi equation

for heavy atoms in intense magnetic elds
Christopher C Tisdell and Mark Holzer
October 8, 2010
Department of Applied Mathematics
School of Mathematics & Statistics
The University of New South Wales
UNSW Sydney 2052
Australia.
Email: cct@unsw.edu.au
mholzer@unsw.edu.au
Abstract
This article presents a rm mathematical foundation for the boundary value prob-
lem (BVP) associated with the nonrelativistic ThomasFermi equation for heavy
atoms in intense magnetic elds. It is shown that the BVP admits a unique solution
that lies between two simple bounding functions. The approach uses: an application
of dierential inequalities, known as the technique of lower and upper solutions; and
ideas from nonlinear analysis, namely, xedpoint theory. The results prove that the
ThomasFermi model leads to a robust theory of heavy atoms in intense magnetic
elds in spite of the severe approximations that it employs.
Short title: Analysis of the ThomasFermi BVP
Keywords: ThomasFermi equation; boundary value problem; heavy atoms in intense
magnetic elds; qualitative analysis of solutions; existence and uniqueness of solutions.
MSC (2010): 34B60; 34B15
Corresponding Author: Christopher C Tisdell
1 Introduction
A mathematically rigorous theory of the electronic structure of atoms in intense magnetic
elds is of fundamental physical interest. In astrophysics, for example, one would like to
understand the behaviour of atoms in the magnetic elds of neutron stars, which ranges
from 10
12
G for pulsars to 10
15
G for magnetars, for example, see [26]. Atoms with high
nuclear binding energies and hence stability, such as iron (atomic number Z = 26), are of
particular interest, for example, see [2].
The semiclassical ThomasFermi theory [18, 42] of atoms has appealing physical sim-
plicity. Its venerable history spans some eighty years with a vast literature, for example,
see [1, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 19, 21, 23, 28, 31, 32, 33, 34, 36, 38] and references
therein. In spite of its failure to predict an atoms shell structure and chemical bonding
1
[41], the theory has nevertheless been shown to be an asymptotically exact description of
heavy atoms in the Z limit, when interpreted as describing the core and mantle
of the electron distribution [33].
The standard case with no magnetic eld leads to the singular dierential equation
x
1/2
y

= y
1/2
. Coupled with various boundary conditions, this problem has been exten-
sively studied from theoretical, physical and computational points of view, for example,
see [18, 20, 25, 42] and references therein.
The case of heavy atoms in intense magnetic elds has been studied within the frame-
work of the ThomasFermi theory and much is known, both quantitatively and quali-
tatively, for example, see [19, 24, 27, 29, 30, 35] and references therein. In particular,
variational methods have been used to prove the existence and uniqueness of the electron
density that minimizes the ThomasFermi energy functional [30].
Herein a dierent approach is taken by analyzing the associated boundary value prob-
lem (BVP) for the electron density in the case of atoms in strong magnetic elds which has
been subjected to only limited mathematical analysis. The associated ordinary dierential
equation is
y

xy, x [0, x
0
]
subject to appropriate (Dirichlet) boundary conditions.
In Section 2 a brief review and derivation of the ThomasFermi BVP for heavy atoms
in intense magnetic elds is presented.
In Section 3 contains a proof that the BVP for heavy atoms in an intense magnetic
eld is well posed with a unique solution that can rigorously be bounded to lie between
simple functions on a nite interval [0, x
0
] of the nondimensional radial distance,.
The methods chosen herein involve dierential inequalities and are ideas from nonlinear
analysis. This includes such approaches as: an application of the technique of lower
and upper solutions [16]; and a particular application of xedpoint theory known as
Schauders xedpoint theorem [43, Theorem 2.A, p.57].
In Section 4 contains some known results that are used in the proof of the main result,
Theorem 3.2. In doing this, the aim is to keep the paper somewhat selfcontained but not
to distract the reader during the proof of Theorem 3.2 with long excursions away from its
core elements.
2 The ThomasFermi model for heavy atoms in an intense
magnetic eld
To place the boundary value problem (BVP) considered here in its physical context, a
derivation of the ThomasFermi model is now sketched as applied to heavy atoms in
intense magnetic elds following [27, 35]. Gaussian units are adopted and m will denote
the mass of the electron.
The basic idea of the ThomasFermi approach is to model the atomic electrons as a
degenerate Fermi gas, which allows both kinetic and potential energies of the electrons
to be expressed entirely in terms of the number density n(r) of the electrons. Exchange
energies are neglected but can be included in renements of the theory [17, 15, 40].
In the presence of an intense magnetic eld (precisely what is meant by intense will
be quantied below), the electrons are modelled as a onedimensional Fermi gas moving
along the eld lines but localized to the atom by the Coulomb potential of the nucleus.
The kinetic energy of the electrons associated with their cyclotron orbits is assumed to
be unperturbed by the Coulomb potential so that it does not contribute to the energy
2
relative to the free state in the magnetic eld. The kinetic energy associated with the
orbital motion in the Coulomb potential is assumed to be negligible compared to the
kinetic energy of the motion along the eld. The range of validity of these assumptions
can be justied a posteriori for large Z, for example, see, [27, 35, 30] and also below.
The number of states per unit volume is then set by the density of Landau levels per
unit transverse area, (eB)/(hc), and the density of momentum states per unit length
and per unit momentum for the motion parallel to the magnetic eld, 2/h. Assuming
temperatures suciently low so that all electron spins remain antialigned with the eld
and so that the Fermi momentum cannot be exceeded, there is exactly one electron per
state up to the Fermi momentum p
F
so that n(r) = 2eBp
F
/(h
2
c). Thus, the kinetic
energy density per unit volume is given by
eB
hc
2
h
_
p
F
0
p
2
2m
dp = e
2
L
5
n
3
, where the length
L = (c
2
h
4
)
1/5
/(24mB
2
e
4
)
1/5
.
The energy functional of the electron gas is given by the sum of its kinetic energy, its
electric potential energy in the eld of the nucleus, and the electric potential energy due
to Coulomb interactions of the electrons with each other:
E
TF
[n(r)] = e
2
L
5
_
n
3
(r)d
3
r e
2
Z
_
n(r)
|r|
d
3
r +
e
2
2
_
n(r)n(r

)
|r r

|
d
3
rd
3
r

.
The corresponding groundstate energy scales with eld strength B, number of electrons
N, and atomic number Z like (B/B
0
)
2/5
N
3/5
Z
6/5
for N Z. Thus, for neutral atoms with
N = Z, the groundstate energy scales like (B/B
0
)
2/5
Z
9/5
, compared with the neutral
zeroeld case whose ground-state energy scales like Z
7/3
. Here the natural scale for the
magnetic eld, B
0
, is the eld for which the energy of the lowest Landau level equals the
ionization energy of a hydrogen atom in its ground state. Thus, B
0
cm
2
e
3
/
3
210
9
G.
Interestingly, the presence of the magnetic eld does not spoil the spherical symmetry of
the electron density for the magnetic eld range for which this model is valid, for example,
see [27, 35, 29, 30] and see below. Thus, n(r) = n(r), where r = |r|.
The electron density that minimizes E
TF
subject to the constraint that
_
n(r)d
3
r = N,
the number of electrons N Z, is obtained by variation with respect to n. Setting
(E
TF
[n] eN
0
)/n = 0, where
0
is the Lagrange multiplier associated with the con-
straint, one immediately obtains
3e
2
L
5
n
2
(r) = e[(r)
0
], (2.1)
where (r) is the total Coulomb potential. For neutral atoms, the Lagrange multiplier

0
= 0. This potential is forced to be selfconsistent with the electron density by demand-
ing that n(r) be its source in Gauss law

2
(r) = 4en(r), (2.2)
subject to the boundary condition that the nuclear potential dominates at the origin:
lim
r0
r(r) = Ze. (2.3)
Before dening the associated BVP, it is appropriate to review the limits of validity
of the radially symmetric ThomasFermi electron density, see also [35]. For there to be
well dened momentum states parallel to the magnetic eld, the majority of electrons
cannot be in deep bound states with nodeless wave functions. The condition for this is
that the cyclotron radius of the m
th
Landau level with m N Z be much larger than
the Bohr radius for nuclear charge Ze, which gives the condition B/B
0
Z
3
. At the
3
lower range of B, the requirement that the binding energy in the presence of a strong
magnetic eld be signicantly greater than that with no eld, gives Z
4/3
B/B
0
, which
follows immediately from the dierent scaling of the energy with B and Z in the two cases.
With respect to the classication scheme of Lieb et al. [29, 30] for the dierent asymptotic
regimes that are realized depending on the powerlaw with which B as Z , we
are thus concerned with region 3 for which Z
4/3
B/B
0
Z
3
.
2.1 The boundary value problem
To combine (2.1) and (2.2) into a single nonlinear ordinary dierential equation one denes
the nondimensional function y(r) :=
r
Ze
[(r)
0
] and nondimensional radial distance
x = r/R, with R = (4)
2/5
(3Z)
1/5
L.
Changing the variable from r to x, (2.1) and (2.2) combine to give
y

xy. (2.4)
Equation (2.4) must be solved under suitable boundary conditions, namely (2.3), which
becomes
y(0) = 1, (2.5)
and to ensure a nitesize atom one must also have
y(x
0
) = 0, (2.6)
for suciently large x
0
.
1
2.2 Necessary notation and denitions
A few mathematics denitions are presented that are necessary for the sections that follow.
A solution to (2.4) (2.6) is dened to be a twice continuously dierentiable function
y = y(x), denoted by y C
2
([0, x
0
]), such that y satises (2.4) on [0, x
0
] and satises
(2.5) (2.6).
The analysis will be set in the Banach space (Y,
0
) where Y is the set of continuous
functions on [0, x
0
], denoted by C([0, x
0
]); coupled with the maximum norm, that is,
y
0
:= max
x[0,x
0
]
|y(x)|, for all y C([0, x
0
]).
Let T be a map between Banach spaces, i.e., T : Y Y . The map T is dened to be
compact if: T is continuous; and T maps bounded sets into relatively compact sets.
The following result is known as Schauders xedpoint theorem [43, Theorem 2.A,
p.57] and will be required in Section 3.
Theorem 2.1 (Schauder) Let be a nonempty, closed, bounded and convex subset of
the Banach space Y . If T : is a compact map then there is at least one y such
that Ty = y.
1
I am grateful to Dr Mark Holzer, School of Mathematics and Statistics, UNSW for providing insight
into the history and formulation of the ThomasFermi theory.
4
3 Main results
The aim of this section is to consider the BVP (2.4)(2.6) and to illustrate that a unique
solution exists, with the unique solution lying between two simple functions, generally
known as lower and upper solutions.
The following result ensures that the BVP (2.4)(2.6) has, at most, one solution. This
result will be built upon shortly.
Theorem 3.1 The BVP (2.4)(2.6) has, at most, one solution on [0, x
0
].
Proof. The proof involves exploiting the monotonicity of the righthandside of (2.4) in
y. Let u and v be two solutions to (2.4)(2.6). We show that u v on [0, x
0
] so that there
is, at most, one solution to (2.4)(2.6).
Let
r(x) := u(x) v(x), for all x [0, x
0
]. (3.1)
Since r is continuous on [0, x
0
], it must achieve its maximum (and minimum) values on
[0, x
0
]. Let x
1
[0, x
0
] be such that
r(x
1
) = max
x[0,x
0
]
r(x) > 0 (3.2)
that is, u(x
1
) > v(x
1
). We show that (3.2) cannot hold in the sense that r(x
1
) cannot be
positive.
If x
1
= 0 then (2.5) ensures r(0) = 0 and so (3.2) cannot hold for x
1
= 0. Similarly,
(2.6) ensures that (3.2) cannot hold for x
1
= x
0
.
If x
1
(0, x
0
) then the maximum principle [39, p.1] gives r

(x
1
) = 0 and r

(x
1
) 0.
From (3.1) we also have
r

(x
1
) = u

(x
1
) v

(x
1
)
= (x
1
u(x
1
))
1/2
(x
1
v(x
1
))
1/2
> 0
as u(x
1
) > v(x
1
). Thus, we reach a contradiction and so (3.2) cannot hold for any
x
1
(0, x
0
).
Combining the above cases we see that r 0 on [0, x
0
].
In a similar fashion to the above argument, it can be shown that r 0 on [0, x
0
] by
applying the maximum principle to r on [0, x
0
].
Thus, r 0 on [0, x
0
] and so u v on [0, x
0
], that is, there is a most one solution to
(2.4)(2.6).
The next result guarantees that the BVP (2.4)(2.6) has a unique solution and also
gives a location on the solution, that is, two functions are furnished that bound the solution
from below and from above, thus sandwiching the solution between two functions.
Theorem 3.2 The BVP (2.4)(2.6) has a unique solution y C
2
([0, x
0
]) such that
0 y(x) 1
x
x
0
, for all x [0, x
0
]. (3.3)
Proof. The basic idea is to modify (2.4) to form a related dierential equation whose
righthandside is continuous and uniformly bounded on [0, x
0
] R. At least one solution
to the modied BVP is then guaranteed to exist by Schauders xedpoint theorem. These
solutions are shown to be solutions to (2.4)(2.6) by a forming a particular pair of lower
5
and upper solutions. Finally, an application of Theorem 3.1 shows that this solution is
unique.
Consider the modied dierential equation
y

= g(x, y), x [0, x

0
] (3.4)
subject to (2.5) and (2.6), where
g(x, y) :=
_

_
_
x
_
1
x
x
0
__
1/2
+
y
_
1
x
x
0
_
1 +
_
y
_
1
x
x
0
__
2
, for y 1
x
x
0
;
(xy)
1/2
, for 0 y 1
x
x
0
;
y
1 +y
2
, for y 0.
Note that g is continuous and uniformly bounded on [0, x
0
] R and we let this bound be
denoted by M > 0. The BVP (3.4), (2.5), (2.6) is equivalent to the integral equation
y(x) :=
_
1
x
x
0
_
+
_
x
0
0
G(x, s)g(s, y(s))ds, x [0, x
0
] (3.5)
where
G(x, s) :=
_

(x
0
x)s
x
0
, for 0 s x x
0
;

x(x
0
s)
x
0
, for 0 x s x
0
.
The proof of this equivalence is found in Theorem 5.1 of the appendix.
It is straightforward to show that G satises
_
x
0
0
|G(x, s)| ds
x
2
0
8
, for all x [0, x
0
] (3.6)
see, for example, [22, Chap.XII, Sec.4].
Now, form a Banach space by choosing Y = C([0, x
0
]) and coupling it with the maxi-
mum norm. Consider the map T dened by
[Ty](x) :=
_
1
x
x
0
_
+
_
x
0
0
G(x, s)g(s, y(s))ds, x [0, x
0
] (3.7)
The continuity of G, g and the properties of the integral ensure that for all y C([0, x
0
])
we have Ty C([0, x
0
]). In addition, comparing (3.5) with (3.4) we see that y is a solution
to (3.4), subject to (2.5) and (2.6), if and only if Ty = y.
If we dene the nonempty, closed, bounded and convex set
:=
_
y C([0, x
0
]) : y
0
= max
x[0,x
0
]
|y(x)| 1 +
Mx
2
0
8
_
C([0, x
0
]) (3.8)
then the continuity of g and the bounds (3.6) and M ensure that for all y C([0, x
0
]) we
have
Ty
0
= max
x[0,x
0
]
|[Ty](x)| 1 +
Mx
2
0
8
6
with the calculations contained in the proof of Theorem 5.2 in the appendix. Thus,
T : C([0, x
0
]) and so T : .
Furthermore, T is a compact map, as shown in Theorem 5.5 in the appendix (or see
[43, p.5556]).
Hence, all of the conditions of Schauders theorem are satised so there is at least one
y such that Ty = y. These y are not only in C([0, x
0
]), but are actually in C
2
([0, x
0
])
due to the continuity of f (and g).
Thus, the modied BVP (3.4), (2.5), (2.6) has at least one solution y C
2
([0, x
0
]).
We now show that solutions y to (3.4), (2.5), (2.6) must satisfy (3.3) and so they must
be solutions to the unmodied BVP (2.4)(2.6).
We rst show that
y(x) 1
x
x
0
, for all x [0, x
0
]. (3.9)
The case showing y 0 on [0, x
0
] is similar and so is omitted for brevity.
Let
w(x) := y(x)
_
1
x
x
0
_
, for all x [0, x
0
]. (3.10)
Since w is continuous on [0, x
0
], it must achieve its maximum value on [0, x
0
]. Let x
2

[0, x
0
] be such that
w(x
2
) = max
x[0,x
0
]
w(x) > 0 (3.11)
that is, y(x
2
) > 1 x
2
/x
0
. We show that (3.11) cannot hold in the sense that w(x
2
)
cannot be positive.
The boundary conditions (2.5) and (2.6) ensure that (3.11) cannot hold, respectively,
for x
2
= 0 or x
2
= x
0
.
If x
2
(0, x
0
) then the maximum principle [39, p.1] gives w

(x
2
) = 0 and w

(x
2
) 0.
From (3.10) and (3.4) we also have
w

(x
2
) = y

(x
2
)
= g(x
2
, y(x
2
))
=
_
x
2
_
1
x
2
x
0
__
1/2
+
y(x
2
)
_
1
x
2
x
0
_
1 +
_
y(x
2
)
_
1
x
2
x
0
__
2
> 0.
Thus, we reach a contradiction and so (3.11) cannot hold for any x
2
(0, x
0
).
Combining the above cases we conclude that (3.9) must hold.
In a similar fashion to the above, we also have y 0 on [0, x
0
]. Thus, (3.3) holds and
the solutions to (3.4), (2.5), (2.6) must also be solutions to (2.4)(2.6). By Theorem 3.1,
there is, at most, one solution and so solutions to (2.4)(2.6) are unique.
The following result provides estimates on the slope of the the solution to (2.4)(2.6)
at x
0
.
Theorem 3.3 The solution y to (2.4)(2.6) furnished by Theorem 3.2 satises

1
x
0
y

(x
0
) 0 (3.12)
so that, for large (but nite) x
0
, the solution will satisfy y

(x
0
) 0.
7
Proof. From a geometric point of view, inequality (3.12) is immediately apparent as
Theorem 3.2 ensures the solution y to (2.4)(2.6) is squeezed between the two functions
0 and (x) := 1 x/x
0
at x = x
0
. Thus, y

(x
0
) must lie between

(x
0
) and

(x
0
).
More formally, for all x [0, x
0
] with x < x
0
, inequality (3.3) and (2.6) ensure
_
1
x
x
0
_
y(x
0
)
x x
0

y(x) y(x
0
)
x x
0
0 (3.13)
and taking limits as x x

0
in (3.13) we obtain (3.12).
In Theorem 3.2, the solution to (2.4)(2.6) is bounded between two simple functions.
In fact, the solution can be bounded by a range of functions, with the general technique
known as the method of upper and lower solutions [16].
Theorem 3.4 If there exist functions l, u C
2
([0, x
0
]) such that 0 u l on [0, x
0
] and
l

(x)
_
xl(x), for all x [0, x
0
] (3.14)
u

(x)
_
xu(x), for all x [0, x
0
] (3.15)
l(0) 0, u(0) 1
l(x
0
) = 0, u(x
0
) = 0
then the BVP (2.4)(2.6) has a unique solution y C
2
([0, x
0
]) such that l(x) y(x)
u(x) for all x [0, x
0
]. In addition, u

(x
0
) y

(x
0
) l

(x
0
).
Proof. The proof is very similar to that of Theorem 3.2 and so is omitted.
Remark 3.5 In Theorem 3.2 we constructed the following lower and upper solutions,
respectively, to (2.4)(2.6)
l 0, u(x) = 1
x
x
0
. (3.16)
The advantage in this choice is in its simplicity.
4 Discussion and Conclusions
This article considered the ThomasFermi model for heavy atoms in intense magnetic
elds in the regime Z
4/3
B/B
0
Z
3
, where spherical symmetry is preserved. In
this regime, the model reduces to the Thomas-Fermi equation y

=

xy, where x is the
non-dimensional distance from the nucleus, and the non-dimensionalized electron number
density n n
_
3L
5
R/Z is given by n(x) =
_
y(x)/x. This ThomasFermi equation is
highly nonlinear and in the absence of exact solutions requires approximations, such as
numerical solutions. It is therefore natural to focus on the boundary value problem over
the nite interval x [0, x
0
], with the boundary conditions y(0) = 1 and y(x
0
) = 0. Given
the severe approximations made in the ThomasFermi approach and the nonlinearity of
the resulting equation, neither the wellposedness of the boundary value problem, nor the
existence of unique solutions can be taken for granted. Indeed, the Thomas-Fermi model
is known to give an exact description of atoms only in the asymptotic limit of innite
atomic number and innitely strong magnetic elds [33]. This paper provides proof that
the boundary value problem is wellposed with a unique solution for any nite x
0
. In this
sense the ThomasFermi model constitutes a robust theory.
The analysis also brackets the solution between upper and lower bounds. The lower
bound of zero guarantees a positive denite electron density as physically required. It is
8
noted in passing that the nonnegativity of solutions to the boundary value problem of a
wide class of dierential equations that includes the ThomasFermi equation considered
here has been proved also for the semiinnite domain x [0, ) by ORegan [37, p,117],
although uniqueness was not obtained. The upper bound constrains possible solutions
y(x) to lie below the straight line 1x/x
0
. Combined with the condition that the number
density must integrate to the total number of electrons, i.e.,
_

0
n(r)4r
2
dr = Z for
the neutral atom, which in terms of y(x) and the boundary condition y(x
0
) = 0 maps
to the constraint
_
x
0
0
x
3/2
y
1/2
dx = 1, the solution to the ThomasFermi equation for
heavy atoms in intense magnetic elds is strongly constrained. These constraints provide
powerful checks on any numerical or other approximate solutions.
There are a number of possible directions for further research. For example, the choice
of upper and lower solutions in Section 3 involve very simple functions. In view of the more
general result, Theorem 3.4, it may be interesting to construct upper and lower solutions
with a more detailed structure to gain further qualitative information about the electron
density. In particular, it may be possible to obtain sharper bounds on the solution to the
BVP (2.4)(2.6) at the end points of the interval [0, x
0
].
5 Appendix
To keep the paper reasonably selfcontained, some known results are now presented in
this section that are used in the proof of the main result, Theorem 3.2.
Theorem 5.1 The BVP (3.4), (2.5), (2.6) is equivalent to the integral equation (3.5).
Proof. Let y be a solution to the integral equation (3.5) on [0, x
0
]. Direct dierentiation
of both sides of (3.5) will yield (3.4). Direct substitution of x = 0 and x = x
0
into (3.5)
will yield (2.5) and (2.6), respectively. Thus, every solution to (3.5) must also be a solution
to (3.4), (2.5), (2.6).
Now let y be a solution to the BVP (3.4), (2.5), (2.6) on [0, x
0
]. Integration of (3.4)
from 0 to z yields
y

(z) = y

(0) +
_
z
0
g(q, y(q)) dq, z [0, x
0
]
and yet another integration from 0 to x gives
y(x) = y(0) +xy

(0) +
_
x
0
_
z
0
g(s, y(s)) ds dz (5.1)
= 1 +xy

(0) +
_
x
0
(x s)g(s, y(s)) ds, x [0, x
0
]. (5.2)
Now, (5.2) evaluated at x = x
0
and coupled with (2.6) yields
0 = y(x
0
) = 1 +x
0
y

(0) +
_
x
0
0
(x
0
s)g(s, y(s)) ds. (5.3)
A rearrangement in (5.3) gives
y

(0) =
1
x
0
_
1 +
_
x
0
0
(x
0
s)g(s, y(s)) ds
_
which may be substituted into (5.2) and rearranged to form (3.5).
9
Theorem 5.2 If T is dened in (3.7) and dened in (3.8) then T() is uniformly
bounded.
Proof. Let y and consider
Ty
0
:= max
x[0,x
0
]
|[Ty](x)|
= max
x[0,x
0
]

_
1
x
x
0
_
+
_
x
0
0
G(x, s)g(s, y(s)) ds

max
x[0,x
0
]
_

1
x
x
0

+M
_
x
0
0
|G(x, s)| ds
_
= 1 +
Mx
2
0
8
.
Thus T() is uniformly bounded by 1 +Mx
2
0
/8.
Theorem 5.3 If T is dened in (3.7) and dened in (3.8) then T() is equicontinuous.
Proof. Let y and x
1
, x
2
[0, x
0
]. To show T() is equicontinuous, we illustrate that
for any given > 0 there exists a = () such that
|[Ty](x
1
) [Ty](x
2
)| < whenever |x
1
x
2
| <
Since G has continuous partial derivatives we have
|G(x
1
, s) G(x
2
, s)| =

_
x
1
x
2
G
x
(x, s) ds

|x
1
x
2
|.
Thus, for any given > 0 we have
|[Ty](x
1
) [Ty](x
2
)|
|x
1
x
2
|
x
0
+
_
x
0
0
|G(x
1
, s) G(x
2
, s)||g(s, y(s))| ds

|x
1
x
2
|
x
0
+M
_
x
0
0
|x
1
x
2
| ds
= |x
1
x
2
|
_
1
x
0
+Mx
0
_
<
whenever |x
1
x
2
| < for the choice
() =

1
x
0
+Mx
0
.
Theorem 5.4 If T is dened in (3.7) and dened in (3.8) then T continuous on .
Proof. Let y
n
be a sequence of functions in that converges uniformly on [0, x
0
] to some
y , that is
y
n
y, as n .
We show that
Ty
n
Ty, as n .
10
Consider the rectangular region in R
2
dened by

1
:=
_
(x, p) [0, x
0
] R : |p| 1 +
Mx
2
0
8
_
.
Now, as g is uniformly continuous on
1
we have
g(x, y
n
(x)) g(x, y(x)) as n
with the convergence being uniform on [0, x
0
]. Thus, on [0, x
0
] we have
Ty
n
Ty
0
= max
x[0,x
0
]

_
x
0
0
G(x, s)[g(s, y
n
(s)) g(s, y(s))] ds

max
x[0,x
0
]
_
x
0
0
|G(x, s)||g(s, y
n
(s)) g(s, y(s))| ds
0, as n .
Theorem 5.5 If T is dened in (3.7) and dened in (3.8) then T is a compact map.
Proof. In view of the denition of compactness given in Section 2.2 we need to show that
T is continuous and T maps bounded sets into relatively compact sets.
Firstly, continuity of T on is ensured from Theorem 5.4.
Secondly, the uniform boundedness and equicontinuity of T() illustrated in Theorems
5.2 and 5.3 may be applied to show that T() is relatively compact, that is, T() is a
compact set. The properties of uniform boundedness and equicontinuity of T() ensure
that the Arzel`aAscoli theorem [43, (24g), p.772] may be applied to T() to guarantee it
is relatively compact.
References
[1] Akkermans, Eric; Ghosh, Sankalpa; Musslimani, Ziad H. Numerical study of one
dimensional and interacting BoseEinstein condensates in a random potential. J. Phys. B:
At. Mol. Opt. Phys. 41 (2008), no. 4, Article ID 045302 (12pp).
[2] Ballantyne, D. R. Reection spectra from an accretion disc illuminated by a neutron star
X-ray burst. Mon. Not. R. Astron. Soc. 351 (2004), 5762.
[3] Bautista, Manuel A. Electronic correlations and polarizability of the ThomasFermiDirac
Amaldi potential: applications to the singly ionized ironpeak species. J. Phys. B: At. Mol.
Opt. Phys. 41 (2008), no. 6, Article ID 065701 (8pp).
[4] Benilan, Philippe; Brezis, Ham. Nonlinear problems related to the ThomasFermi equation.
Dedicated to Philippe Benilan. J. Evol. Equ. 3 (2003), no. 4, 673770.
[5] Benguria, Rafael; Brezis, Ham; Lieb, Elliott H. The ThomasFermivon Weizsacker theory
of atoms and molecules. Comm. Math. Phys. 79 (1981), no. 2, 167180.
[6] Benguria, R.; Lieb, E.H. Manybody atomic potentials in ThomasFermi theory. Ann. Phys.
NY 110, (1978) 3445.
[7] Benguria, R.; Lieb, E.H. The positivity of the pressure in ThomasFermi theory. Comm.
Math. Phys. 63 (1978), 193218.
[8] Bhaduri, R. K.; Murthy, V. M. N.; Brack, M. Fermionic ground state at unitarity and
Haldane exclusion statistics. J. Phys. B: At. Mol. Opt. Phys. 41 (2008), no. 11, Article ID
115301 (4pp).
11
[9] Brezis, Ham. Nonlinear problems related to the ThomasFermi equation. Contemporary
developments in continuum mechanics and partial dierential equations (Proc. Internat.
Sympos., Inst. Mat., Univ. Fed. Rio de Janeiro, Rio de Janeiro, 1977), pp. 8189, North-
Holland Math. Stud., 30, NorthHolland, AmsterdamNew York, 1978
[10] Brezis, Ham. Nonlinear equations of Thomas Fermi type. Proceedings of the First Franco
Southeast Asian Mathematical Conference (Nanyang Univ., Singapore, 1979). Southeast
Asian Bull. Math. 1979, Special Issue, 167171.
[11] Brezis, H. Some variational problems of the ThomasFermi type. Variational inequalities and
complementarity problems (Proc. Internat. School, Erice, 1978), pp. 5373, Wiley, Chich-
ester, 1980.
[12] Brezis, Ham. A free boundary problem in quantum mechanics: ThomasFermi equation.
Free boundary problems, Vol. II (Pavia, 1979), pp. 8591, Ist. Naz. Alta Mat. Francesco
Severi, Rome, 1980.
[13] Brezis, Ham; Lieb, Elliott H. Long range atomic potentials in ThomasFermi theory. Comm.
Math. Phys. 65 (1979), no. 3, 231246.
[14] Caracanhas, M. A.; Seman, J. A.; Ramos, E. R. F.; Henn, E. A. L.; Magalhaes, K. M.
F.; Helmerson K.; Bagnato, V. S. Finite temperature correction to the ThomasFermi ap-
proximation for a BoseEinstein condensate: comparison between theory and experiment.
J. Phys. B: At. Mol. Opt. Phys. 42 (2009), no. 14, Article ID 145304, (8pp).
[15] Danz, R. W.; Glasser, M. L. Exchange energy of an electron gas in a magnetic eld. Phys.
Rev. B 4, (1971) 9499.
[16] De Coster, C.; Habets, P. The lower and upper solutions method for boundary value prob-
lems. Handbook of dierential equations, 69160, Elsevier NorthHolland, Amsterdam, 2004.
[17] Dirac, P. A. M. Note on the exchange phenomena in the Thomas atom. Proc. Cambridge
Philos. Soc. 26 (1930), 376.
[18] Fermi, E. Un metodo statistico per la determinazione di alcune prioriet`a dellatome. Rend.
Addad. Naz. Lincei 6 (1927), 602.
[19] Glossman, M. D.; Castro, E. A. Approximate analytical solution of the ThomasFermi
equation for positive ions in a strong magnetic eld., Z. Phys. D - Atoms, Molecules and
Clusters 6 (1987), 8182.
[20] Granas, A.; Guenther, R. B.; Lee, J. W. A note on the ThomasFermi equation. Z. Angew.
Math. Mech. 61 (1981), no. 3, 204205.
[21] Gupta, Moumita,; Krishna, Krishna Rai. An analytical approach for the nonlinear modied
ThomasFermi equation to derive the ground-state and dynamic properties of a spherically
and cylindrically trapped BoseEinstein condensate. J. Phys. B: At. Mol. Opt. Phys. 41
(2008), no. 19, Article ID 195302, (8pp).
[22] Hartman, Philip. Ordinary dierential equations. Corrected reprint of the second (1982)
edition. With a foreword by Peter Bates. Classics in Applied Mathematics, 38. Society for
Industrial and Applied Mathematics (SIAM), Philadelphia, PA, 2002.
[23] Hill, S. H.; Grout, P. J.; March, N. H. Relativistic ThomasFermi equation in an extremely
high magnetic eld, and its solution for atomic ions. J. Phys. B: At. Mol. Phys. 18 (1985),
46654673.
[24] Hill, S. H.; Grout, P. J.; March, N. H. Chemical potential and total energy of heavy positive
ions in extremely strong magnetic elds, near the weak ionisation limit. J. Phys. B: At. Mol.
Phys. 16 (1983), 23012307.
12
[25] Hille, Einar. On the ThomasFermi equation. Proc. Nat. Acad. Sci. U.S.A. 62 (1969), 710.
[26] Ibrahim A. I; Sa-Harb S.; Swank J. H.; Parke W.; Zane S; Turolla, R. Discovery of cyclotron
resonance features in the soft gamma repeater SGR 180620. The Astrophysical Journal,
574:L51L55, (2002).
[27] Kadomstev, B. B. Heavy atom in an ultrastrong magnetic eld. Soviet Physics JETP 31
(1970), no. 5, 945947.
[28] Kornev, A. S.; Zon, B. A. Rydberg spectra of atoms and positive ions in the ThomasFermi
model. J. Phys. B: At. Mol. Opt. Phys. 36 (2003), no. 19, 40274034.
[29] Lieb E. H.; Solovej J. P.; Yngvason J. Asymptotics of heavy atoms in high magnetic Fields:
I. Lowest Landau band regions. Commun. Pure. Appl. Math. 47 (1994), 513591.
[30] Lieb E. H.; Solovej J. P.; Yngvason J. Asymptotics of heavy atoms in high magnetic elds:
II. Semiclassical regions. Commun. Math. Phys. 161 (1994), 77-124.
[31] Lieb, Elliott H. ThomasFermi and HartreeFock theory. Proceedings of the international
congress of mathematicians (Vancouver, B. C., 1974), Vol. 2, pp. 383386. Canad. Math.
Congress, Montreal, Que., 1975.
[32] Lieb, E. H. The stability of matter. Rev. Mod. Phys. 48 (1976), 553569.
[33] Lieb, E. H.; Simon, B. ThomasFermi theory revisited. Phys. Rev. Lett. 31 (1973), 681683.
[34] Lieb, E. H.; Simon, B. The ThomasFermi theory of atoms, molecules and solids. Adv.
Math. 23, (1977), 22116.
[35] Mueller. R. O.; Rau, A. R. P.; Spruch L. Statistical model of atoms in intense magnetic
elds. Phys. Rev. Lett. 26 (1971), 11361139.
[36] Msezane, A. Z.; Feli, Z.; Sokolovski, D. Nearthreshold resonances in electron elastic scat-
tering cross sections for Au and Pt atoms: identication of electron anities. J. Phys. B:
At. Mol. Opt. Phys. 41 (2008), no. 10, Article ID 105201, (6pp).
[37] ORegan, Donal. Theory of singular boundary value problems. World Scientic Publishing
Co., Inc., River Edge, NJ, 1994.
[38] Penna, Andre L. A.; Diniz, Joao B.; Oliveria, Fernando A. Charge degeneracy removal in
the screened hydrogen atom. J. Phys. B: At. Mol. Opt. Phys. 41 (2008), no. 19, Article ID
195302, (8pp).
[39] Protter, Murray H.; Weinberger, Hans F. Maximum principles in dierential equations.
Corrected reprint of the 1967 original. SpringerVerlag, New York, 1984.
[40] Rau, A. R. P.; Mueller, R. O.; Spruch, L. Simple model and wave function for atoms in
intense magnetic elds. Phys. Rev. A 11, (1975) 18651879.
[41] Teller, E. On the stability of molecules in the ThomasFermi theory. Rev. Mod. Phys. 34
(1962), 627631.
[42] Thomas, L. H. The calculation of atomic elds. Proc. Cambridge Phil. Soc. 23 (1927), 542.
[43] Zeidler, Eberhard. Nonlinear functional analysis and its applications. I. Fixedpoint theo-
rems. Translated from the German by Peter R. Wadsack. SpringerVerlag, New York, 1986.
13