ISSN 1068-3623, Journal of Contemporary Mathematical Analysis, 2015, Vol. 50, No. 2, pp. 5362.


c Allerton Press, Inc., 2015.
Original Russian Text 
c A. R. Meszaros, K. Shamseddine, 2015, published in Izvestiya NAN Armenii. Matematika, 2015, No. 2, pp. 53-68.

DIFFERENTIAL EQUATIONS

On the Solutions of Linear Ordinary Dierential Equations

and Bessel-type Special Functions on the Levi-Civita Field
A. R. Meszaros1* and K. Shamseddine2**
1
Universite Paris-Sud, France
2
University of Manitoba, Winnipeg, Canada
Received January 31, 2014

AbstractBecause of the disconnectedness of a non-Archimedean ordered eld in the topology

induced by the order, it is possible to have non-constant functions with zero derivatives everywhere.
In fact the solution space of the dierential equation y  = 0 is innite dimensional. In this paper,
we give sucient conditions for a function on an open subset of the Levi-Civita eld to have zero
derivative everywhere and we use the nonconstant zero-derivative functions to obtain non-analytic
solutions of systems of linear ordinary dierential equations with analytic coecients. Then we use
the results to introduce Bessel-type special functions on the Levi-Civita eld and to study some of
their properties.
MSC2010 numbers : 26E30, 12J25, 33C10
DOI: 10.3103/S1068362315020016
Keywords: Levi-Civita eld; ordinary dierential equations; Bessel-type special functions;
generalized Bessel functions of the rst kind.

1. INTRODUCTION
Solutions of linear ordinary dierential equations and some Bessel-type special functions on the Levi-
Civita eld R [5, 6] are presented in this paper. 1 We recall that the elements of R are functions from Q
to R with left-nite support (denoted by supp). That is, below every rational number q, there are only
nitely many points where the given function does not vanish. For the further discussion, it is convenient
to introduce the following terminology.
Denition 1.1 (, , , =r ). For x = 0 in R, we let (x) = min(supp(x)), which exists because of
the left-niteness of supp(x), and we let (0) = +.
Given x, y = 0 in R, we say that x y if (x) = (y), and x y if (x) = (y) and x[(x)] =
y[(y)].
Given x, y R and r R, we say that x =r y if x[q] = y[q] for all q r.

At this point, these denitions may look somewhat arbitrary, but after having introduced an order
on R, we will see that describes orders of magnitude, the relation corresponds to agreement up
to innitely small relative error, while corresponds to agreement of order of magnitude. The set R is
endowed with formal power series multiplication and componentwise addition, which make it into a eld
(see [3]) in which we can isomorphically embed R as a subeld via the map : R R dened by

x if q = 0
(x)[q] = (1.1)
0 otherwise.

*
E-mail: alpar.meszaros@math.u-psud.fr
**
E-mail: khodr.shamseddine@umanitoba.ca
1
The research of the rst author was conducted in the frames of TAMOP 4.2.4. A/2-11-1-2012-0001 National Excellence
Program Elaborating and operating an inland student and researcher personal support system. The project was
subsidized by the European Union and co-nanced by the European Social Fund.

53
54 MESZAROS, SHAMSEDDINE

Denition 1.2 (Order in R). Let x, y R be given. Then we say x y if x = y or [x = y and

(x y)[(x y)] > 0].
It is easy to check that the relation "" is a total order and (R, +, , ) is an ordered eld (which
we denote simply by R). Moreover, the embedding in equation (1.1) of R into R is compatible with
the order. The order induces an absolute value on R in the natural way: |x| = x if x 0 and |x| = x if
x < 0. We also note that , as dened above, is a valuation. Moreover, the relation is an equivalence
relation, and the set of equivalence classes (the value group) is (isomorphic to) Q. Besides the usual
order relations, some other notations are also convenient.
Denition 1.3 (
, ). Let x, y R be non-negative. We say that x is innitely smaller than y
(and write x
y) if nx < y for all n N; we say that x is innitely larger than y (and write
x  y) if y
x. If x
1, then we say that x is innitely small; if x  1, then we say that x is
innitely large. Innitely small numbers are also called innitesimals or dierentials. Innitely
large numbers are also called innite. Non-negative numbers that are neither innitely small nor
innitely large are called nite numbers.
Denition 1.4 (The Number d). Let d be the element of R given by d[1] = 1 and d[q] = 0 for q = 1.
It is easy to check that dq
1 if q > 0 and dq  1 if q < 0. Moreover, for all x R, the elements of
supp(x) can be arrangedin ascending order, say supp(x) = {q1 , q2 , . . .} with qj < qj+1 for all j, and x
q
can be written as x = j=1 x[qj ]d , where the series converges in the topology induced by the absolute
j

value (see [3]).

Altogether, it follows that R is a non-Archimedean eld extension of R. For a detailed study of this
eld, we refer the reader to [10, 20], and references therein. In particular, it is shown that R is complete
with respect to the topology induced by the absolute value, that is, every Cauchy sequence of elements
of R converges to an element of R. In the wider context of valuation theory, it is interesting to note that
the topology induced by the absolute value is the same as that introduced via the valuation , as it was
shown in [19].
It follows therefore that the eld R is just a special case of the class of elds discussed in [9]. For a
general overview of the algebraic properties of formal power series elds in general, we refer the reader to
the comprehensive overview by Ribenboim [8], and for an overview of the related valuation theory to the
books by Krull [4], Schikhof [9] and Alling [1]. A thorough and complete treatment of ordered structures
can also be found in [7].
Besides being the smallest ordered non-Archimedean eld extension of the real numbers that is
both complete in the order topology and real closed, the Levi-Civita eld R is of particular interest
because of its practical usefulness. Since the supports of the elements of R are left-nite, it is possible to
represent these numbers on a computer (see [3]). Having innitely small numbers, the errors in classical
numerical methods can be made innitely small, and hence irrelevant in all practical applications. One
such application is the computation of derivatives of real functions representable on a computer, where
both the accuracy of formula manipulators and the speed of classical numerical methods are achieved
(see [16]).
In this paper we present some tools to construct a large class of solutions for equation y  = 0 on R.
Then as an application of that, we dene and study the properties of Bessel-type special functions on
(open subsets of) R.

2. PRELIMINARY RESULTS
For an easier study of systems of linear ordinary dierential equations on R, it is benecial to introduce
matrix exponentials on R. We dene matrices on R and matrix operations: addition, multiplication,
determinant, just as we do in the real case.
Denition 2.1. Let Mn (R) denote the set of all n n matrices with entries in R. For A Mn (R),
we dene | | : Mn (R) R by
|A| = max {|aij |},
1i,jn

where |aij | = max {aij , aij }.

JOURNAL OF CONTEMPORARY MATHEMATICAL ANALYSIS Vol. 50 No. 2 2015

 ON SOLUTIONS OF LINEAR ORDINARY DIFFERENTIAL EQUATIONS 55

In what follows we deal only with square matrices whose entries are at most nite in absolute value,
and we denote this class of matrices by Mfn (R).
Denition 2.2. Let A Mfn (R), and for each k N {0} let ck R be given. We say that the
 
series k f
k=0 ck A is convergent in Mn (R) if the series k=0 ck |A | is convergent in R with respect
k

to the weak topology discussed in [3, 12, 17].

Denition 2.3. Let A Mfn (R) be given. We dene the exponential of A by the series eA =
 1 k
k=0 k! A , where A = In is the n n identity matrix.
0

In the next theorem we show that the series in Denition 2.3 converges in the sense of Denition 2.2.
Theorem 2.1. For any A Mfn (R), the series


A 1 k
e = A
k!
k=0

is always convergent, and hence eA is well-dened.

Proof. Let A = (aij ), and let A2 = (bij ). Then, by the way we perform matrix multiplication, for all
i, j {1, . . . , n} we have
  
|bij | n max {|aij |} max {|aij |} = n|A|2 .
1i,jn 1i,jn

It follows that
 2
A  = max {|bij |} n|A|2 .
1i,jn

Using induction on k, it is then easy to show that

 
 k
A  nk1 |A|k , k N. (2.1)
Since


 nk1 1  1
|A|k = (n|A|)k
k! n k!
k=1 k=1

n|A| 
converges in the weak topology of R to e 1 /n (because |A| is at most nite), it follows from
equation (2.1) and the properties of weak convergence of innite series (see [12, 17]) that
1  k  1  k 

 
A  = 1 + A 
k! k!
k=0 k=1


converges weakly in R. Hence the series 1 k
k! A converges in the sense of Denition 2.2. Theorem 2.1
k=0
is proved.
Taking into account that power series on R can be dierentiated term by term within their domain of
convergence (see [19]), we obtain the following result.
Theorem 2.2. Let A Mfn (R) and let F : R Mn (R) be given by F (t) = etA . Then F is dier-
entiable at each t R, with derivative F  (t) = AetA .

Proof. Since F (t) = tk k 
k=0 k! A , we can obtain F (t) by dierentiating the series term by term as a
function of t:

  k
tk1 t k
F  (t) = Ak = A A = AetA = AF (t).
(k 1)! k!
k=1 k=0
Note that all the series in the last equation are well-dened.

JOURNAL OF CONTEMPORARY MATHEMATICAL ANALYSIS Vol. 50 No. 2 2015

56 MESZAROS, SHAMSEDDINE

3. THE MAIN RESULTS

3.1. Linear ordinary dierential equations on R
One of the main goals of this paper is to obtain solutions of linear ordinary dierential equations in
the case where the coecients are analytic functions of the independent variable, including non-analytic
solutions in addition to the analytic ones. The basic idea for the construction of non-analytic solutions
of linear ordinary dierential equations on R is based on the following theorem (proved in [11]), which
shows that even the simplest dierential equation y  = 0 over R has innitely many linearly independent
solutions on [1, 1] R.
Theorem 3.1. The solution space of the dierential equation y  = 0 on [1, 1] is innite dimen-
sional.
Proof. For each n N, let gn : [1, 1] R be given by gn (x)[q] = x[q/(n + 1)]. We show that, for all
n N, gn is dierentiable on [1, 1] with gn (x) = 0 for all x [1, 1]. So let n N be given. We rst
observe that gn (x + y) = gn (x) + gn (y) for all x, y [1, 1]. Now let x [1, 1] and  > 0 in R be
given. Let = min{2 , d}, and let y [1, 1] be such that 0 < |y x| < . Then, taking into account
that g(y x) (y x)n+1 , we have
   
 gn (y) gn (x)   gn (y x) 
 = 
  y x  |y x| .
n
 yx
Next, since |y x| < min{2 , d}, we obtain that |y x|n
. Hence
 
 gn (y) gn (x) 
  <  for all y [1, 1] satisfying 0 < |y x| < .
 yx 

It follows that gn is dierentiable at x, with gn (x) = 0. This is true for all x [1, 1] and for all n N.
Hence gn is a solution of the dierential equation y  = 0 on [1, 1] for all n N.
Next, we show that the set S = {gn : n N} is linearly independent on [1, 1]. So let j N and let
n1 < n2 < < nj in N be given. It is enough to show that gn1 , gn2 , . . . , gnj are linearly independent
on [1, 1]. To this end, we suppose that c1 gn1 + c2 gn2 + + cj gnj = 0 for some c1 , c2 , . . . , cj in R,
and show that c1 = c2 = = cj = 0. Indeed, since c1 gn1 + c2 gn2 + + cj gnj = 0, we obtain that
c1 gn1 (d) + c2 gn2 (d) + + cj gnj (d) = 0. Hence c1 dn1 + c2 dn2 + + cj dnj = 0, from which we infer
that c1 = c2 = = cj = 0.
Remark 3.1. For each n N, it is easy to check that the mapping gn in the proof of Theorem
3.1 is an order preserving eld automorphism of R; this is a special property of non-Archimedean
structures since it is well-known that the only eld automorphism of R is the identity map ([13]).
In Propositions 3.1 - 3.3 that follow, we give sucient conditions for a nonconstant function to be a
solution of the dierential equation y  = 0 on an open subset of R.
Proposition 3.1. Let M R be open and let f : M R be such that, for some xed p > 1 in Q
and for some positive
1 in R, we have
x, y M, (x y) () |f (x) f (y)| |x y|p .
Then f is dierentiable on M with derivative f  (x) = 0 for all x M , that is, f is a solution for
the dierential equation y  = 0 on M .
Proof. Let x M and  > 0 in R be given. Since M is open, there exists 0 > 0 in R such that
(x 0 , x + 0 ) M . Let
2
= min{0 ,  p1 , }.
Then for all y R satisfying 0 < |y x| < , we have that y M and (y x) () (). Hence
  
 f (x) f (y)   
  |x y|p1 < p1 = min p1 , 2 , p1 min 2 , p1
.
 xy  0

JOURNAL OF CONTEMPORARY MATHEMATICAL ANALYSIS Vol. 50 No. 2 2015

 ON SOLUTIONS OF LINEAR ORDINARY DIFFERENTIAL EQUATIONS 57

The last step is justied by the fact that if 

1, then 2
 and if  is nite or innitely large,
then p1
 since
1 and p 1 > 0. So in both cases, we have min{2 , p1 }
. Thus, f is
dierentiable at x for all x M with f  (x) = 0.

Proposition 3.2. Let M R be open and let f : M R be such that, for some xed p > 1 in Q
and for some positive
1 and positive in R, we have
x, y M, (x y) () |f (x) f (y)| |x y|p .
Then f is a solution for the dierential equation y  = 0 on M .

Proof. Let x M and  > 0 in R be given. Then there exists 0 > 0 in R such that (x 0 , x + 0 ) M .
Let

2 1/(p1)
1/(p1)
= min 0 , ,  / , (/) .

Then > 0 and for y R satisfying 0 < |y x| < , we have that y M and (y x) () ().
Hence
  
 f (x) f (y)   
  |x y|p1 < p1 = min p1 , p1 , 2 , min 2 ,
.
 xy  0

This shows that f is dierentiable at x for all x M with f  (x) = 0.

Remark 3.2. We note that Proposition 3.1 follows from Proposition 3.2 if we take to be any
innitely large positive number.
Denition 3.1. Let M R and h : M R. We say that h is level preserving on M and write
h P (M ) if x, y M satisfying (x) = (y) and x =r y it follows that (h(x)) = (h(y)) and
h(x) =q h(y), where q (h(x)) + r (x).

Example 3.1. Let f : R R be given by f (x)[q] = x[q 1]. Then it is easy to check that f P (R).

Proposition 3.3. Let M R be open and such that (x) 0 for all x M . Let h : M R be a
level preserving function on M , and let Q, > 1 be given. Then the function f : M R, given
by
  
h(x) q(h(x))
(x) , if (x) > 0,
f (x)[q] = (3.1)
h(x) [q + (h(x))] , if (x) = 0
is dierentiable on M with derivative f  (x) = 0 for all x M .

Proof. Let x R and  > 0 in R be given. Since M is open, there exists > 0 in R such that
1 and
2
(x , x + ) M . Let = min{ 1 , }. Then 0 <
1 and (x , x + ) M . Now assuming
y M such that 0 < |y x| < , we show that
 
 f (x) f (y) 
  < . (3.2)
 xy 

We note that, since |y x|

1, then either (x) = 0 = (y) or [(x) > 0 and (y) > 0].
First assume that (x) > 0 (and hence (y) > 0). We distinguish three cases.
Case 1: (x) = (y). In this case, we have (f (x)) = (f (y)), and it follows that
 
 f (x) f (y) 
  (x y)1 .
 xy 

Hence
 
f (x)f (y)
xy = ( 1)(y x) ( 1)()

JOURNAL OF CONTEMPORARY MATHEMATICAL ANALYSIS Vol. 50 No. 2 2015

58 MESZAROS, SHAMSEDDINE

= ( 1) max 2
1 (), () = max {2(), ( 1)()} > (),

implying that
 
 f (x) f (y) 
 
. (3.3)
 xy 

Case 2: x y and x[(x)] = y[(y)]. In this case the arguments are similar to those of Case 1.
Case 3: x =r y for some r Q with r (x). Then obviously (|x y|) = r+ , where r+ is a rational
number such that r+ > r. It follows that (|f (x) f (y)|) = r+ . Thus, we have
 
f (x) f (y)
= (f (x) f (y)) (x y) = ( 1)r+ > ( 1)(),
xy
from which, as in the Case 1, we obtain (3.3). Finally, if (x) = 0 = (y) then the proof of the inequality
(3.2) follows by the same arguments as above (when (x) > 0), except that we have to use the
appropriate expression for f from equation (3.1).
In the following denition we introduce the class of all functions that are dierentiable with derivative
equal to 0 everywhere on an open subset of R.

Denition 3.2. Let M R be open and let f : M R. We dene the class of functions D01 (M ) as
follows:
 
D01 (M ) = f : M R|f is dierentiable on M, f  (x) = 0 x M .

3.2. Systems of linear ordinary dierential equations on R

In this subsection we investigate the solutions of systems of linear ordinary dierential equations on
R, using the functions of class D01 . The main goal of this subsection is to obtain solutions of systems
of linear ODEs of the form: Y  (t) = A(t)Y (t) + B(t), where Y (t) is a vector of dimension n > 0,
n N, which contains the unknown functions, Y  (t) contains the derivatives of the functions from Y (t),
A(t) Mfn (R), for all t R, which contains the coecient functions of the system, and B(t) is a vector
of dimension n, which contains functions that ensure the inhomogeneity of the system. In order to realize
this we study a few cases, going from the most special to the most general ones.
Theorem 3.2. Consider the linear homogeneous system of ordinary dierential equations with
constant coecients which are at most nite in absolute value Y  (t) = AY (t). Then the solution
is given by
Y (t) = eAt C + eAt Una (t),
where C Mn,1 (R) is a vector containing constants, and Una (t) Mn,1 (D01 ) is a vector which
contains functions of class D01 .

Proof. We rewrite the system in the form: Y  (t) AY (t) = On,1 , which is equivalent to

At 
eAt Y  (t) eAt AY (t) = On,1 or e Y (t) = On,1 .
It follows that eAt Y (t) = C + Una (t), where C Mn,1 (R) and the elements of Una (t) are functions of
class D01 , and hence Y (t) = eAt C + eAt Una (t).

Remark 3.3. Theorem 3.2 shows that the solutions of linear homogeneous systems with constant
coecients over R are very similar to those of the real case, except that the solutions in the non-
Archimedean case may also involve non-analytic functions with zero-derivatives.

Since we know how to integrate R-analytic functions [14] in the Lebesgue-like theory developed
in [15, 18], we can study next those inhomogeneous systems, where the functions ensuring the
inhomogeneity are R-analytic.

JOURNAL OF CONTEMPORARY MATHEMATICAL ANALYSIS Vol. 50 No. 2 2015

 ON SOLUTIONS OF LINEAR ORDINARY DIFFERENTIAL EQUATIONS 59

Theorem 3.3. Consider the inhomogeneous system of linear ordinary dierential equations with
constant coecients: Y  (t) = AY (t) + B(t) on the interval [a, b] R, where |A|, |a| and |b| are at
most nite in absolute value, and B(t) is a vector, which contains functions that are R-analytic
on [a, b]. Then the solution is given by the equation

At At
Y (t) = e C + e Una (t) + eAt
eAs B(s),
[a,t]

where C Mn,1 (R) and the elements of Una (t) are functions of class D01 .


Proof. We rewrite the system in the form: Y  (t) AY (t) = B(t), which is equivalent to eAt Y (t) =
eAt B(t). It follows that

At
e Y (t) = C + Una (t) + eAs B(s),
[a,t]

where C Mn,1 (R), and Una (t) Mn,1 (D01 ), and hence

At At
Y (t) = e C + e Una (t) + e At
eAs B(s).
[a,t]

Note that we have used the fact that eAt B(t) is a vector whose components are products of functions
that are R-analytic on [a, b], and hence the components themselves are R-analytic on [a, b].
The proofs of the next two theorems (Theorems 3.4 and 3.5) are similar to those of Theorems 3.2 and
3.3 above, and therefore they are stated without proofs.

Theorem 3.4. Consider the homogeneous system of linear ordinary dierential equations with
non-constant coecients Y  (t) = A(t)Y (t) on [a, b], where |a| and |b| are at most nite, and where
A(t) is an n n matrix whose elements are R-analytic on [a, b] and such that |A(t)| is at most
nite for all t [a, b]. Then the solution is given by
 
A(s) A(s)
Y (t) = e [a,t] C +e [a,t] Una (t),
where C Mn,1 (R) and Una (t) Mn,1 (D01 ).

Theorem 3.5. Consider the inhomogeneous system of linear ordinary dierential equations with
non-constant coecients: Y  (t) = A(t)Y (t) + B(t) on [a, b], where |a| and |b| are at most nite, A(t)
is an n n matrix whose elements are R-analytic on [a, b] and such that |A(t)| is at most nite
for all t [a, b], and B(t) is a vector whose components are functions that are R-analytic on [a, b].
Then the solution is given by
    
A(s) A(s) A(s) A(r)
Y (t) = e [a,t] C+e [a,t] Una (t) + e [a,t] e [a,s] B(s),
[a,t]

where C Mn,1 (R) and Una (t) Mn,1 (D01 ).

3.3. Bessel-type special functions on R

In this subsection we study Bessel-type special functions on R, with the help of the solutions for
systems of linear ordinary dierential equations that we developed in Subsection 3.2. We introduce such
functions with the following problem.

Problem 1. Lets study the solutions of the dierential equation

t2 y  + ty  + (t2 2 )y = 0. (3.4)
on [a, 1] R, where 0 < a < 1, a is nite, and Q \ Z.

JOURNAL OF CONTEMPORARY MATHEMATICAL ANALYSIS Vol. 50 No. 2 2015

60 MESZAROS, SHAMSEDDINE

We call equation (3.4) the Bessel equation of order , and we study its solutions below. It is easy to
check, similar to the real case, that the functions
    2n
t (1)n t
J (t) = , (3.5)
2 n=0 n!(n + + 1) 2
  
 2n
t (1)n t
J (t) = (3.6)
2 n!(n + 1) 2
n=0
are two linearly independent analytic solutions of equation (3.4), for all t [a, 1], where is the Eulers
Gamma function. Moreover, it follows from Corollary 3.10 in [17] that the power series in equations (3.5)
and (3.6) converge weakly in [a, 1].
The main objective of Problem 1 is to construct solutions of equation (3.4) that involve functions of
the class D01 , and to study their properties. We set w1 = y, w2 = y  , and consider the following system
of two linear ordinary dierential equations:

w = w
2
1
 
w = 1 2 w + 1 w ,
2 t2 1 t 2

which we can write in matrix form as

W  (t) = A(t)W (t), (3.7)
where

w1 (t) w1 (t) 0 1
W (t) = ; W  (t) = ; A(t) = .
w2 (t) w2 (t) 1 2
t2
1
t
Because the elements of A(t) are R-analytic functions on [a, 1] and they are at most nite in absolute
value for all t [a, 1], we can use Theorem 3.4 to write the solutions of equation (3.7), which are also the
solutions of equation (3.4). Let


0 t a
D(t) = A(s) = ,
[a,t] t a + 2 ( 1t a1 ) ln at
where the function ln is R-analytic on [a, 1].
Thus, the solution of equation (3.7), and hence of equation (3.4), on [a, 1] has the form
W (t) = eD(t) C + eD(t) Una (t),
where C M2,1 (R) and Una (t) M2,1 (D01 ).
Taking into account that the analytic part of the solution has the form c1 J (t) + c2 J (t), we
conclude that in the rst
row D(t) we can take the entries to be D
c1
 of the matrix e 11 = J (t) and
D12 = J (t). With C = c2 , it follows that the rst component of W (t), which is y = w1 (t), has the
form:
y = c1 [J (t) + J (t)g (t)] + c2 [J (t) + J (t)g (t)] ,
where g , g D01 ([a, 1]).
Now we are in a position to dene Bessel functions of the rst kind on R.
Denition 3.3. For 0 < a < 1, a R nite, we dene the functions J , J : [a, 1] R R by
J (t) = J (t) + J (t)g (t), and
J (t) = J (t) + J (t)g (t),
where Q \ Z and g , g We call the functions J , J Bessel functions of the rst
D01 ([a, 1]).
kind and of order , and , respectively.

JOURNAL OF CONTEMPORARY MATHEMATICAL ANALYSIS Vol. 50 No. 2 2015

 ON SOLUTIONS OF LINEAR ORDINARY DIFFERENTIAL EQUATIONS 61

Next, we study some properties of the Bessel functions J and J .

Theorem 3.6. Under the notation of Problem 1 and Denition 3.3 above, the following two
statements are true for all t [a, 1]:
2   
(a) J (t) = J1 (t) + J+1 (t) 1 + g (t) ,
t   
(b) 2J (t) = J1 (t) J+1 (t) 1 + g (t) .

Proof. Using equation (3.5), we can write

 
J (t)  1  (1)n t2n1 J+1 (t)
= = .
t 2 (n 1)!(n + + 1)22n1 t
n=1

Then, using the fact that J (t) = J (t) + J (t)g (t) and g (t) = 0, t [a, 1], we obtain
 
J (t)  J+1 (t) J+1 (t)
= g (t),
t t t
implying that

J (t) = J (t) J+1 (t) J+1 (t)g (t) (3.8)
t
Similarly, we can show that
1 
t J (t) = t1 J1 (t) + t1 J1 (t)g (t),
t
implying that

J (t) = J1 (t) J (t) + J1 (t)g (t). (3.9)
t
If we subtract equation (3.9) from equation (3.8), we get statement (a) of the theorem, and if we add the
two equations, we get (b).
Remark 3.4. Just as we did in Theorem 3.6, we can obtain other recursive relations for J and
J that would extend the classical recursive relations for J and J from Real Calculus to the
non-Archimedean calculus on R.

3.4. Generalized Bessel functions of the rst kind on R

We give some basic denitions based on those given by A. Baricz (see [2]) in the classical case (real
and complex). As before, let 0 < a < 1, a R nite, p Q \ Z, and let b, c R in the rest of this paper.
Denition 3.4. The dierential equation
t2 w (t) + btw(t) + [ct2 p2 + (1 b)p]w(t) = 0 (3.10)
will be referred to as the generalized Bessel equation of order p (see [2]), and any solution of it
will be called a generalized Bessel function of order p.
The generalized Bessel functions permit the study of Bessel functions, spherical Bessel functions and
modied Bessel functions together. That is why it is very important to extend this kind of functions to
the eld R.
Remark 3.5. As in the classical case (see [2]), it can easily be veried that the function

  2n+p
(c)n t
wp (t) = b+1
n!(p + n + 2 ) 2
n=0
is a solution of equation (3.10). Moreover, when c = b = 1 we get the Bessel function of the rst
kind of order p discussed in the previous subsection.

JOURNAL OF CONTEMPORARY MATHEMATICAL ANALYSIS Vol. 50 No. 2 2015

62 MESZAROS, SHAMSEDDINE

Similar to Denition 3.3, we introduce generalized Bessel functions of the rst kind on R as follows.

Denition 3.5. Let 0 < a < 1, a R nite, p Q \ Z, and let b, c R. Then the function Wp (t) =
wp (t) + wp (t)gp (t), where gp D01 ([a, 1]), and
  2n+p
(c)n t
wp (t) = b+1
n!(p + n + 2 ) 2
n=0

is a solution of equation (3.10). We call Wp (t) a generalized Bessel function of order p on [a, 1].

The proof of the next result is similar to that of Theorem 3.6, as well as to the proof of the
corresponding result in the classical case (see [2], Lemma 1.1), and therefore we state it without proof.

Theorem 3.7. Under the notation in Denition 3.5, the following statements are true for all
t [a, 1]
  
2p+b1
(a) t W p (t) = wp1 (t) + cwp+1 (t) 1 + gp (t) ,
  
(b) (2p + b 1)Wp (t) = pwp1 (t) (p + b 1)cwp+1 (t) 1 + gp (t) .

REFERENCES
1. N. L. Alling, Foundations of Analysis over Surreal Number Fields (North Holland, Amsterdam, 1987).
2. A. Baricz, Generalized Bessel Functions of the First Kind (Springer, Berlin, 2010).
3. M. Berz, Calculus and numerics on Levi-Civita elds, In M. Berz, C. Bischof, G. Corliss, and A. Griewank,
editors, SIAM, Computational Dierentiation: Techniques, Applications, and Tools, 1935, 1996.
4. W. Krull, Allgemeine Bewertungstheorie, J. Reine Angew. Math., 167, 160196, 1932.
5. T. Levi-Civita, Sugli inniti ed innitesimi attuali quali elementi analitici, Atti Ist. Veneto di Sc., Lett. ed
Art., 7a, 4:1765, 1892.
6. T. Levi-Civita, Sui numeri transniti, Rend. Acc. Lincei, 5a, 7:91,113, 1898.
7. S. Priess-Crampe, Angeordnete Strukturen: Gruppen, Korper, projektive Ebenen (Springer, Berlin,
1983).
8. P. Ribenboim, Fields: Algebraically Closed and Others, Manuscripta Mathematica, 75, 115150, 1992.
9. W. H. Schikhof, Ultrametric Calculus: An Introduction to p-Adic Analysis (Cambridge University Press,
Camridge, 1985).
10. K. Shamseddine, New Elements of Analysis on the Levi-Civita Field, PhD thesis, Michigan State
University, report MSUCL-1147, 1999.
11. K. Shamseddine, On the existence and uniqueness of solutions of ordinary dierential equations on the Levi-
Civita eld, Int. J. Dier. Equ. Appl., 4, 375386, 2002.
12. K. Shamseddine, On the topological structure of the Levi-Civita eld, J. Math. Anal. Appl., 368, 281292,
2010.
13. K. Shamseddine, Nontrivial order preserving automorphisms of non-Archimedean elds, Contemp. Math.,
547, 217225, 2011.
14. K. Shamseddine, A brief survey of the study of power series and analytic functions on the Levi-Civita elds,
Contemp. Math., 596, 269280, 2013.
15. K. Shamseddine, New results on integration on the Levi-Civita eld, Indag. Math. (N.S.), 24 (1), 199211,
2013.
16. K. Shamseddine and M. Berz, Exception handling in derivative computation with non-Archimedean cal-
culus, In M. Berz, C. Bischof, G. Corliss, and A. Griewank, editors, SIAM, Computational Dierentiation:
Techniques, Applications, and Tools, 3751, 1996. .
17. K. Shamseddine and M. Berz, Convergence on the Levi-Civita eld and study of power series, In Proc.
Sixth International Conference on p-adic Functional Analysis, New York, Marcel Dekker, 283299, 2000.
18. K. Shamseddine and M. Berz, Measure theory and integration on the Levi-Civita eld, Contemp. Math.,
319, 369387, 2003.
19. K. Shamseddine and M. Berz, Analytical properties of power series on Levi-Civita elds, Ann. Math. Blaise
Pascal, 12 (2), 309329, 2005.
20. K. Shamseddine and M. Berz, Analysis on the Levi-Civita eld, a brief overview, Contemp. Math., 508,
215237, 2010.

JOURNAL OF CONTEMPORARY MATHEMATICAL ANALYSIS Vol. 50 No. 2 2015