JOURNAL OF MATHEMATICAL PHYSICS 56, 012102 (2015)

On spectral deformations and singular Weyl functions

for one-dimensional Dirac operators
Alexander Beigl,1,a) Jonathan Eckhardt,2,b) Aleksey Kostenko,1,c)
and Gerald Teschl1,3,d)
1
Faculty of Mathematics, University of Vienna, Oskar-Morgenstern-Platz 1,
1090 Wien, Austria
2
School of Computer Science and Informatics, Cardiff University, Queens Buildings,
5 The Parade, Roath, Cardiff CF24 3AA, Wales, United Kingdom
3
International Erwin Schrdinger Institute for Mathematical Physics, Boltzmanngasse 9,
1090 Wien, Austria
(Received 5 October 2014; accepted 11 December 2014; published online 8 January 2015)

We investigate the connection between singular WeylTitchmarshKodaira the-

ory and the double commutation method for one-dimensional Dirac operators. In
particular, we compute the singular Weyl function of the commuted operator in
terms of the data from the original operator. These results are then applied to
radial Dirac operators in order to show that the singular Weyl function of such
an operator belongs to a generalized Nevanlinna class N0 with 0 = || + 21 ,
where R is the corresponding angular momentum. C 2015 AIP Publishing LLC.
[http://dx.doi.org/10.1063/1.4905166]

I. INTRODUCTION
WeylTitchmarshKodaira theory is the fundament of direct and inverse spectral theory for
ordinary differential operators. The classical theory usually assumes that one endpoint is regu-
lar. However, it has been shown by Kodaira,15 Kac,14 and more recently by Fulton,11 Gesztesy
and Zinchenko,13 Fulton and Langer,12 Kurasov and Luger,22 and Kostenko, Sakhnovich, and
Teschl1619 that many aspects of this classical theory still can be established at a singular endpoint. It
has recently proven to be a powerful tool for inverse spectral theory for these operators and further
refinements were given in Refs. 37, 10, and 20. The analogous theory for one-dimensional Dirac
operators was developed by Brunnhuber and three of us in Ref. 2 (for further extensions see also
Refs. 8 and 9). Nevertheless, such operators are still difficult to understand.
One approach, originating from ideas of Krein,21 is to use spectral deformation methods to
reduce a given spectral problem to a simpler one. In the case of one-dimensional Schrdinger oper-
ators, there is by now an enormous literature on this subject and we refer to Kostenko, Sakhnovich,
and Teschl18 for further references. Moreover in Ref. 18, the connection between these methods and
singular WeylTitchmarshKodaira theory was established and it is our present aim to do the same
for Dirac operators. However, here the situation is slightly different. In fact, since Dirac operators
are not bounded from below, a factorization of the type A A is not possible, and hence there is
no analog of the classical CrumDarboux method for Dirac operators. However, an analog of the
double commutation method was established by Teschl.25 Moreover, this method has been used by
Albeverio, Hryniv, and Mykytyuk1 to reduce the inverse spectral problem of radial Dirac operators
on a compact interval to the case of regular operators following the general idea of Krein.21 Here,
we will further extend these results in a more general setting. As an application, we will find a

a) E-mail address: alex_beigl@gmx.at

b) E-mail address: EckhardtJ@cardiff.ac.uk
c) E-mail addresses: duzer80@gmail.com and Oleksiy.Kostenko@univie.ac.at
d) E-mail address: Gerald.Teschl@univie.ac.at. URL: http://www.mat.univie.ac.at/gerald/.

0022-2488/2015/56(1)/012102/11/$30.00 56, 012102-1 2015 AIP Publishing LLC

012102-2 Beigl et al. J. Math. Phys. 56, 012102 (2015)

representation formula for the singular Weyl function of radial Dirac operators and show that it is in
a generalized Nevanlinna class N0 with 0 = || + 21 , extending the results from Ref. 2.

II. WEYL-TITCHMARSH-KODAIRA
Let I = (a, b) R (with a < b ) be an arbitrary interval. We will be concerned with
Dirac operators in the Hilbert space L 2(I, C2) equipped with the inner product
b
f , g = f ( y)g( y) dy, f 2 = f , f . (2.1)
a
To this end, we consider the differential expression
1 d
= 2 + Q(x). (2.2)
i dx
Here, the potential matrix Q(x) is given by
Q(x) = qel(x)1 + qam(x)1 + (m + qsc(x))3, (2.3)
where 1, 2, 3 denote the Pauli matrices

0 1+ 0 i + 1 0+
1 = * , 2 = * , 3 = * , (2.4)
, 1 0 - , i 0 - , 0 1 -
and m, qsc, qel, and qam are interpreted as mass, scalar potential, electrostatic potential, and anoma-
lous magnetic moment, respectively (see Ref. 26, Chap. 4). As usual, we require that m [0, ) and
qsc, qel, qam L 1loc(I) are real-valued.
We do not include a magnetic moment = + 2qmg (x) as it can be easily eliminated by a
x
simple gauge transformation = 1, where = exp(i qmg(r)dr).
If is in the limit point case at both a and b, then gives rise to a unique self-adjoint operator
H when defined maximally (cf., e.g., Refs. 23, 27, and 28). Otherwise, we fix a boundary condition
at each endpoint where is in the limit circle case. Explicitly, such an operator H is given by

H : D(H) L 2(I, C2),

(2.5)
f f,
where
D(H) = { f L 2(I, C2) | f ACloc(I, C2), f L 2(I, C2),
(2.6)
Wa (u, f ) = Wb (u+, f ) = 0},
with
W x ( f , g) = i f (x), 2g(x) = f 1(x)g2(x) f 2(x)g1(x) (2.7)
the usual Wronskian (we remark that the limit Wa,b (., ..) = lim x a,b W x (., ..) exists for functions
as in (2.6)). Here, the function u (respectively, u+) used to generate the boundary condition at a
(respectively, b) can be chosen to be a non-trivial solution of u = 0 if is in the limit circle case at
a (resp. b) and zero else.
D D
For a given point c I, consider the operators H(a,c) and H(c,b) which are obtained by re-
stricting H to (a, c) and (c, b) with a Dirichlet boundary condition f 1(c) = 0 at c, respectively. The
corresponding operators with a Neumann boundary condition f 2(c) = 0 will be denoted by H(a,c) N

N
and H(c,b) .
Our main object will be singular WeylTitchmarshKodaira theory which, for Dirac operators,
was developed only recently in Ref. 2. Hence, we will start by reviewing some relevant facts from
Ref. 2. We will need a system of real entire solutions (z, x), (z, x) of the underlying homoge-
neous equation u = zu such that (z, x) lies in the domain of H near a and W ((z), (z)) = 1. To
this end we introduce the following hypothesis:
012102-3 Beigl et al. J. Math. Phys. 56, 012102 (2015)

D
Hypothesis 2.1. Suppose that the spectrum of H(a,c) is purely discrete for one (and hence for
all) c (a, b).

Lemma 2.2 (Ref. 2, Lemma 2.2). The following are equivalent:

D
(i) The spectrum of H(a,c) is purely discrete for some c I.
(ii) There is a real entire solution (z, x) that is non-trivial and lies in the domain of H near a
for each z C.
(iii) There are real entire solutions (z, x), (z, x) with W ((z), (z)) = 1, such that (z, x) is
non-trivial and lies in the domain of H near a for each z C.

Assuming Hypothesis 2.1, we can introduce the singular Weyl function

W ((z),u+(z))
M(z) = (2.8)
W ((z),u+(z))
such that the solution which is in the domain of H near b is given by
(z, x) = (z, x) + M(z)(z, x). (2.9)
Here, u+(z) is the (up to scalar multiples unique) non-trivial solution of u = zu which lies in the
domain of H near b. We stress the fact that there is no natural choice of a fundamental system and
. The regular solution can be multiplied by a (zero free) real entire function (z, x) = eg (z)(z, x)
for some real entire function g. Due to the requirement that the Wronskian has to be normalized,
the singular solution needs to be of the form (z, x) = eg (z)(z, x) f (z)(z, x), where f is again
some real entire function. This will change the Weyl function according to
M(z) = e2g (z) M(z) + eg (z) f (z). (2.10)
Associated with M(z) is a corresponding spectral measure given by the StieltjesLivic inversion
formula

1 1 1
(0, 1) + [0, 1] = lim Im M( + i) d.





(2.11)
2 0 0
Then there exists a spectral transformation which maps the Dirac operator H in L 2(I, C2) to the
multiplication operator with the independent variable in L 2(R, d). Conversely, M(z) can be recon-
structed from up to an entire function.

Theorem 2.3 (Ref. 17, Theorem 4.1). Let M(z) be a singular Weyl function and its associ-
ated spectral measure. Then there exists an entire function g(z) such that g() 0 for R and
eg () L 2(R, d).
Moreover for any entire function g(z) such that g() > 0 for all R and (1 + 2)1g()1
L (R, d) (e.g., g(z) = e2g (z)), we have the integral representation
1

( )
1 d()
M(z) = E(z) + g(z) , z C \ (H), (2.12)
R z 1+ 2 g()
where E(z) is a real entire function.
If the endpoint a is regular, or at least limit circle, then, with the usual choice for and , the
Weyl function is a HerglotzNevanlinna function and we can choose g(z) 1 and E(z) = Re(M(i))
in the previous theorem. However, this will not be true in general. The following theorem gives
a criterion when the singular Weyl function belongs to the class N of generalized Nevanlinna
functions with no non-real poles and the only generalized pole of nonpositive type at (for further
information on generalized Nevanlinna functions we refer to Ref. 24, see also Ref. 19, Appendix B).

Theorem 2.4 (Ref. 17, Theorem 4.3). Fix the solution (z, x) and k N {0}. Then there is
a corresponding solution (z, x) such that M(z) N for some k if and only if (1 + 2)k1
L 1(R, d). Moreover, = k if k = 0 or (1 + 2)k < L 1(R, d).
012102-4 Beigl et al. J. Math. Phys. 56, 012102 (2015)

III. THE DOUBLE COMMUTATION METHOD

As explained in the Introduction, we want to investigate the effect of the double commutation
method introduced for Dirac operators in Ref. 25 on the singular Weyl function. Hence, we will
review some prerequisites on this method first.
Given a one-dimensional Dirac operator H associated with in H = L 2(I, C2), we denote
by u(z, x), u+(z, x) its corresponding Weyl solutions, that is, solutions of u = zu which are in
the domain of H near a, b, respectively. In general, u(z, x) (u+(z, x)) might not exist unless
D D
z C\ess (H(a,c) ) (z C\ess (H(c,b) )) and by using them we will always implicitly suppose their
existence in such a situation. Without loss of generality, we will also assume u(z , x) = u(z, x)
such that u(, x) is real whenever R.
Given u(, x), R, and [u() 2, ], let us set (here and henceforth, we employ the
convention 1 = 0; the case = 0 has to be read as H0 = H and this case is of course trivial)
x
u(, x) 1
u(, x) = , c(, x) = + u(, y)u(, y)d y (3.1)
c(, x) a
and define
H f = f B ( + Q ) f , D(H) = { f H | f ACloc(I, C2), f H,
(3.2)
Wa (u(), f ) = Wb (u(), f ) = 0},

2 (1 )
Q (x) = Re 2u(, x)u(, x)
c(, x) i
(3.3)
u,1(, x)2 u,2(, x)2 u,1(, x)u,2(, x)
= 1 2 3 .
c(, x) c(, x)
Then, the main result from Ref. 25 states that H and H are unitarily equivalent up to possibly some
one-dimensional subspaces. More precisely, denote by P and P the orthogonal projections onto the
one-dimensional subspaces of H spanned, respectively, by u and u (set P, P = 0 if u, u < H).
Then, we have
Theorem 3.1 (Ref. 25). Let u(, x), R, and [u() 2, ] be given and define H as
in (3.2). If u() H, then we also require p (H).
Suppose first that u() < H.
(i) If > 0, then H and (1 P())H are unitarily equivalent. Moreover, H has the additional
eigenvalue with eigenfunction u,().
(ii) If = , then H and H are unitarily equivalent.
Suppose that u() H and p (H) (i.e., is an eigenvalue of H).
(i) If (u() 2, ), then H and H are unitarily equivalent.
(ii) If = u() 2 or , then (1 P())H and H are unitarily equivalent, that is, the eigen-
value is removed.

Furthermore, the solutions of the new operator H can be expressed in terms of the solutions
of H.
Lemma 3.2 (Ref. 25, Lemma 3.4). Let u ACloc(I, C2) fulfill u = zu, z C\{}, and set
u(, x)
v(z, x) = u(z, x) + W x (u(),u(z)). (3.4)
z
Then, v ACloc(I, C2) and v fulfills v = zv. We also note that if u, v are constructed analogously,
then
1
W x (v(z), v( z)) = W x (u(z), u( z))
c(, x)
(3.5)
z z
W x (u(),u(z))W x (u(), u( z)).
(z )( z )
012102-5 Beigl et al. J. Math. Phys. 56, 012102 (2015)

In addition, the solutions

u(, x)
u,(z, x) = u(z, x) + W x (u(),u(z)) (3.6)
z
are square integrable near a, b and satisfy the boundary condition of H at a, b, respectively.

Remark 3.3. All the previous considerations still hold if one starts from the right endpoint b
instead of the left endpoint a. One just needs to interchange the following roles:
b
1
u(, x) u+(, x), c(x) u+(, y)u+(, y)dy. (3.7)
x

Now, we are ready to relate this method to singular WeylTitchmarshKodaira theory.

Theorem 3.4. Let H be constructed from u(, x) = (, x), R, with [() 2, )

(, x) = (, x)/c(, x). Moreover, if () H, we require p (H).
and set
The operator H has a system of real entire solutions
x
(z, x) = (z, x) (, x)
(, y)(z, y)dy
a (3.8)
1
= (z, x) + (, x)W x ((), (z)),
z
1 ( )
(z, x) = (z, x) + (, x)W x ((), (z)) + (z, x) , (3.9)
z
with W ((z), (z)) = 1. In fact, for z = , we have
(, x) = 1
(, x), (3.10)
(, x) = (, x) + (, x)W x ((), ()) + (, x), (3.11)
where the dot denotes the derivative with respect to the spectral parameter z. In particular, H
satisfies again Hypothesis 2.1.
The Weyl solutions of H are given by
1
(z, x), (z, x) = (z, x) + (, x)W x ((), (z))
z (3.12)
= (z, x) + M(z)(z, x),
where

M(z) = M(z) (3.13)
z
is the singular Weyl function of H.

Proof. Using Lemma 3.2, it is straightforward to check that (z, x), (z, x) is a real entire
system of solutions whose Wronskian equals one. The extra multiple of (z, x) has been added to
(z, x) to remove the pole at z = . The rest is a straightforward calculation. 
Note that in the previous theorem, the singularity at the left endpoint is not changed, which is
reflected by the fact that also the asymptotic behavior of the Weyl function is almost unchanged.
In the limiting case = , we obtain

Theorem 3.5. Let H be constructed from u(, x) = (, x), R, with = and set
(, x) = (, x)/c(, x). Moreover, if () H, we require p (H).

The operator H has a system of real entire solutions

( x )
1
(z, x) = (, x)
(z, x) (, y)(z, y)d y , (3.14)
z a

(z, x) = (z )(z, x) +
(, x)W x ((), (z)), (3.15)
012102-6 Beigl et al. J. Math. Phys. 56, 012102 (2015)

with W ((z), (z)) = 1. In fact, for z = , we have

x
(, x) = (, x)
(, x) (, y)(, y)d y, (3.16)
a
(, x) =
(, x). (3.17)
In particular, H satisfies again Hypothesis 2.1.
The Weyl solutions of H are given by

(z, x), (z, x) = (z )(z, x) +

(, x)W x ((), (z))
(3.18)
= (z, x) + M(z)(z, x),
where
M(z) = (z )2 M(z) (3.19)
is the singular Weyl function of H.

Proof. In the limiting case , the definition from the previous theorem would give
(, x) = 0 (see (3.10)) and we simply need to remove this zero. The rest follows as before. 

Note that in this case, the singularity at the left endpoint is changed, however, the growth of
M(z) is increased, whereas it would be desirable to have a transformation which decreases the
growth. Hence, we need to invert the above procedure. To this end note that the new operator H
has (, x) = u,+(, x). So this shows that we should look at the case where H is computed from
u+(, x) = (, x) (cf. Remark 3.3). Let us also stress that the following result is essential for the
application to the perturbed radial Dirac operator in Sec. IV.

Theorem 3.6. Let H be constructed from u+(, x) = (, x) < H, R, with (0, ] and set
b
(, x) 1
(, x) =
, c(, x) = (, y)(, y)d y. (3.20)
c(, x) x

The operator H has a system of real entire solutions

( 1 )
(z, x) = (z ) (z, x) + (, x)W x ((), (z)) , (3.21)
z
1 1
(z, x) = (z, x) + (, x)W x ((), (z))
z z
(1 )
+ Wb ((), ()) (z, x) , (3.22)

with W ((z), (z)) = 1. Moreover,

(, x) =
(, x). (3.23)
In particular, H satisfies again Hypothesis 2.1.
The Weyl solutions of H are given by
1 1
(z, x), (z, x) = (z, x) + (, x)W x ((), (z))
z z (3.24)
= (z, x) + M(z)(z, x),
where
M(z) + Wb ((), ())(z ) 1
M(z) = (3.25)
(z )2 z
is the singular Weyl function of H.
012102-7 Beigl et al. J. Math. Phys. 56, 012102 (2015)

Proof. That is entire is obvious. For , use lHpitals rule,

( 1 )
lim (z, x) + (, x)W x ((), (z))
z z
= (, x) + (, x)W x ((), ())
(1 )
= Wb ((), ()) (, x)

since
b
W x ((), ()) = Wb ((), ()) + (, y)(, y)dy,
x

which is obtained by differentiating the Lagrange identity

b
( z) (, y)(z, y) d y = Wb ((), (z)) W x ((), (z))
x

with respect to z and evaluating at z = . Hence, the pole of (z) at z = is removed and
the solution is entire. The claim about the Wronskian follows from (3.5). The rest follows by a
straightforward calculation as before. 

Note that by (2.8), we have M() = 0 in the above situation. Moreover, the first summand
in (3.25) has no residue at z = since the residue of M(z) must be given by () 2 =
() 2 = 1. Furthermore, if H is limit circle at b and < then H will be again limit

circle at b by Ref. 25, Theorem 3.7 (clearly H is always limit point at b). In the limit circle case,
the boundary condition of H will be generated by (, x) = (, x) H and hence we can repeat
this procedure at every zero of

(), (z)) = 1 1
z Wb ( Wb ((), (z))
c(, b) z
(1 ) (3.26)
Wb ((), ()) Wb ((), (z)) .

Since we have u,+(z, b) = C(z) (, b) with a nonzero entire function C(z), this implies W (u,+(z),
(z)) = C(z)Wb ((), (z)). Now Eq. (2.8) implies that the zeros of this Wronskian coincide

with the zeros of M(z). But the residues of M(z) are always negative and hence there must be
an odd number of zeros between two consecutive poles of M(z). In particular, we see that the
above Wronskian has an infinite number of zeros and we can iterate this procedure which will be
important later on. We also mention that if the function E(z) in the representation (2.12) is zero,
then the derivative at every zero of M(z) is positive and there will be precisely one zero between
each pole.
Finally, one could also consider the case u+(, x) = (, x). In this case, is an eigenvalue
and the procedure coincides with the one from Theorem 3.4 if one makes the replacement 1
1 + () 2.

IV. APPLICATIONS TO RADIAL DIRAC OPERATORS

In this section, we are going to apply the double commutation method to perturbed radial Dirac
operators

1 d Q(x) L 1loc[0, b), || , 12 ,

H= 2 + 1 + Q(x),


(1 + | log(x)|)Q(x) L 1 [0, b), || = 1 . (4.1)
i dx x loc 2

In the case Q 0, the underlying differential equation can be solved in terms of Bessel functions
and in the general case, standard perturbation arguments can be used to show the following:
012102-8 Beigl et al. J. Math. Phys. 56, 012102 (2015)

Lemma 4.1 (Ref. 2, Sec. 8). If 0, then the operator (4.1) has a unique real entire solution
satisfying

0

(z, x) = x ..

/ + o(x )
* +/
(4.2)
, 2 ( + 1/2) -
as x 0.
In addition, this solution satisfies the growth restriction

|(z, x)| = O |z| e|Im(z)|x

(4.3)
as |z| for all x and has the asymptotics
x

qel( y)dy +


sin zx
2 0 x
*.
(z, x) (z) ..
/

// (4.4)
cos zx qel( y)dy
, 2 0 -
as |z| in any sector | arg(z)| < .

Lemma 4.2 (Ref. 1, Appendix A). If > 21 , then (4.1) has a second real entire solution (z, x)
with W ((z), (z)) = 1 satisfying

x
1(z, x)
(z, x) = * +1 +, (4.5)
,x 2(z, x) -
1(z, .) C[0, b) and
with 2(z, .) L 1 [0, b).
loc
Moreover,
c
|(z, y)|2 d y = x 2+1w(z, x), w
(z, .) W 1,1(0, c), w
(z, 0) > 0, (4.6)
x

for every c (0, b).

Now our strategy is the usual one (cf. also Ref. 1): We iteratively apply the double commutation
method to lower until we end up in the limit circle case || [0, 1/2). To be able to satisfy
the requirement u+(, x) = (, x) from Theorem 3.6, we will assume that the right endpoint b is
regular.

Lemma 4.3. Let H be given by (4.1) with > 12 . Moreover, let H be constructed from u+(, x)
= (, x), R, with (0, ] as in Theorem 3.6. Then
1 d 1
2
H = + 1 + Q(x),
L 1 [0, b).
Q loc (4.7)
i dx x
Moreover, if 1 and (z, x) is normalized according to Lemma 4.1, then so is i2(z, x).

Proof. By (3.3) and Remark 3.3, the commuted operator is of the form H = H + Q , where
1(, x)2 2(, x)2 1(, x)2(, x)
Q (x) = 1 2 3
c(, x) c(, x)

c(, x) 22(, x)2 1(, x)2(, x)
= 1 1 2 3 .
c(, x) c(, x) c(, x)
By Lemma 4.2, the denominator is of the form
b
1 ( 1 )
c(, x) = |(, y)|2 d y = x 2+1 x 21 + w(x) ,
x
012102-9 Beigl et al. J. Math. Phys. 56, 012102 (2015)

with w W 1,1(0, b) and w > 0 on [0, b]. Note that since > 21 , the mapping x x 21 lies in
W 1,1(0, b) too and therefore c(, x) = x 2+1w
(x), where w
shares the same properties as w. Hence
c (, x) d 2 + 1 w(x)
= log(c(, x)) = + ,
c(, x) dx x w
(x)
with w /

w L (0, b). Using the properties of the singular solution from Lemma 4.2, one infers that
1

2(, .)2/c(, .) and (1(, .)2(, .))/c(, .) lie in L 1(0, b) too and the first part follows.
To see the last part, recall that every entire solution of H which lies in the domain of H near
a = 0 is of the form eg (z)(z, x). Since (z, x) z(z, x) as |z| , Eqs. (4.3) and (4.4) show
that g(z) 0. 

Remark 4.4. The new operator H has a negative angular momentum if > 1. In this case, we
employ the gauge transform 2 H2, resulting in a positive angular momentum. The correspond-
ing system of fundamental solutions is given by i2(z, x) and i2(z, x). Note that we again
have W (i2(z), i2(z)) = W ((z), (z)) = 1 and the formula (3.25) remains unchanged. If
(1/2, 1], then 1 [0, 1/2) but (z, x) will not be normalized according to Lemma 4.1 but
will correspond to a different boundary condition (e.g., for = 1 it corresponds to the boundary
condition (z, 0) = (1, 0)).
In order to iterate this procedure, we will assume that our operator is regular at b. Then H will
again be regular at b as long as (0, ). Moreover, by the discussion after Theorem 3.6, there will
be another choice such that u,+() = () (i.e., such that () satisfies the boundary condition
of H at b).
Since the singular Weyl function in the limit circle case will be a Herglotz function, combining
these results with Theorem 3.6, one obtains by induction:

Theorem 4.5. Let H be given by (4.1) with > 12 , + 12 < N, and let b be regular. Assume also
that (z, x) is normalized according to Lemma 4.1. Then there is a singular Weyl function of the
form
12

M(z) = P+ 1 (z) M0(z)
2
cn Pn (z)2(n z), (4.8)
2
n=0
where M0(z) is a HerglotzNevanlinna function and
cn = n1 Wb (n (n ), n (n )), (4.9)
n1

Pn (z) = (z j ), P0(z) = 1, (4.10)
j=0

depend on the choice of n and n in every step of Lemma 4.3. The corresponding spectral measure
is given by
d(t) = P+ 1 (t)2d0(t), (4.11)
2
d0(t)
where the measure 0 satisfies R d0(t) = and R 1+t 2 < .
Proof. As pointed out before, we can reduce by 1 using the above method until we reach the
case (1/2, 3/2). If (1, 3/2), the above procedure will lead us to [0, 1/2) with a properly
normalized and the theorem is proven. In the case (1/2, 1), Lemma 4.3 will give us an
operator H of type (4.1) with 1 . Moreover, as in the proof of the previous lemma, we see
C
(, x) = x 1 * + + o(x 1),
C , 0,
,0-
and by inspection of (3.21) (see also (3.23)) we see
C
(z, x) = x 1 * + + o(x 1).
,0-
012102-10 Beigl et al. J. Math. Phys. 56, 012102 (2015)

Moreover from W ((z), (z)) = 1, we conclude

0
(z, x) = x 1 * 1 + + o(x 1 ).
,C -
The last two equations imply
W0((z), ( z)) = 0, W0((z), ( z)) = 1,
whereas (3.5) implies
W0((z), ( z)) = 0.
Hence, a direct computation (cf. the proof of Theorem A.7 in Ref. 17) shows
b
Im(M(z)) = Im(z) |(z, x)|2dx,
0
which implies that the corresponding Weyl function M(z) is a HerglotzNevanlinna function as
required. 
As another consequence, we obtain

Corollary 4.6. Let H be given by (4.1) with > 12 , + 12 < N, and let b be regular. Then, there
is a corresponding system of entire solutions (z, x), (z, x) with as in Lemma 4.1 such that
M(z) N0 with 0 = + 12 .

Proof. Combining (4.11) with (1 + t 2)1d0(t) < and d0(t) = , the claim follows by
applying Theorem 2.4 with 0 = + 12 . 

Remark 4.7. It is easy to see that the assumption that b is regular is superfluous. Indeed,
observe that Lemma 7.1 from Ref. 2 shows that the asymptotics of M(z) as Im(z) depends
only on the behavior of the potential near a = 0. Furthermore, Ref. 17, Lemma C.2 shows that the
required integrability properties of the spectral measure d depend only on the asymptotics of M(z)
and hence also depend only on the behavior of the potential near a = 0.
This generalizes Theorem 8.4 from Ref. 2 where the bound 0 was given.

Remark 4.8. There is a straightforward connection with the standard theory for radial Schr-
dinger operators if our Dirac operator is supersymmetric, that is, qel = qsc = 0 (see Ref. 2, Sec. 3).
Using the results of this section, we can extend Theorem 4.5 from Ref. 19 to Schrdinger operators
defined in L 2(0, b) by differential expressions

( )( )
d d
= aq aq B

+ + qam(x) + + qam(x) . (4.12)
dx x dx x
Note that in the case qam L 2loc[0, b), this differential expression can be written in the potential form

d2 ( + 1) 2
= + + q+, q+(x) = qam (x) + qam (x)2 qam

(x), (4.13)
dx 2 x2 x
1,2
where q+ is a Wloc [0, b) distribution. For further details, we refer to Ref. 9.

ACKNOWLEDGMENTS
We are very grateful to Annemarie Luger for helpful discussions. A.B., J.E., and G.T. gratefully
acknowledge the stimulating atmosphere at the Institut Mittag-Leffler during summer 2014 where
parts of this paper were written during the workshop on Modern aspects of the TitchmarshWeyl
m-function and its multidimensional analogues. Research supported by the Austrian Science Fund
(FWF) under Grant Nos. J3455, P26060, and Y330.
012102-11 Beigl et al. J. Math. Phys. 56, 012102 (2015)

1 S. Albeverio, R. Hryniv, and Ya. Mykytyuk, Reconstruction of radial Dirac operators, J. Math. Phys. 48, 043501 (2007).
2 R. Brunnhuber, J. Eckhardt, A. Kostenko, and G. Teschl, Singular Weyl-Titchmarsh-Kodaira theory for one-dimensional
Dirac operators, Monatsh. Math. 174, 515547 (2014).
3 J. Eckhardt, Inverse uniqueness results for Schrdinger operators using de Branges theory, Complex Anal. Oper. Theory

8(1), 3750 (2014).

4 J. Eckhardt, Direct and inverse spectral theory of singular left-definite SturmLiouville operators, J. Differ. Equations

253(2), 604634 (2012).

5 J. Eckhardt, F. Gesztesy, R. Nichols, and G. Teschl, WeylTitchmarsh theory for SturmLiouville operators with distribu-

tional potentials, Opusc. Math. 33(3), 467563 (2013).

6 J. Eckhardt, F. Gesztesy, R. Nichols, and G. Teschl, Inverse spectral theory for SturmLiouville operators with distributional

potentials, J. London Math. Soc. 88(2), 801828 (2013).

7 J. Eckhardt and A. Kostenko, Quadratic operator pencils associated with the conservative CamassaHolm flow, e-print

arXiv:1406.3703.
8 J. Eckhardt, A. Kostenko, and G. Teschl, Inverse uniqueness results for one-dimensional weighted Dirac operators, Spec-

tral Theory and Differential Equations: V. A. Marchenko 90th Anniversary Collection, Advances in Mathematical Sciences
edited by E. Khruslov, L. Pastur, and D. Shepelsky, (American Mathematical Society, Providence, 2014), 233, pp. 117133.
9 J. Eckhardt, A. Kostenko, and G. Teschl, Spectral asymptotics for canonical systems (unpublished); e-print arXiv:1412.

0277.
10 J. Eckhardt and G. Teschl, Uniqueness results for one-dimensional Schrdinger operators with purely discrete spectra,

Trans. Am. Math. Soc. 365, 39233942 (2013).

11 C. Fulton, TitchmarshWeyl m-functions for second order SturmLiouville problems, Math. Nachr. 281, 14181475

(2008).
12 C. Fulton and H. Langer, SturmLiouville operators with singularities and generalized Nevanlinna functions, Complex

Anal. Oper. Theory 4, 179243 (2010).

13 F. Gesztesy and M. Zinchenko, On spectral theory for Schrdinger operators with strongly singular potentials, Math.

Nachr. 279, 10411082 (2006).

14 I. S. Kac, The existence of spectral functions of generalized second order differential systems with boundary conditions

at the singular end, Am. Math. Soc. Transl. 62(2), 204262 (1967).
15 K. Kodaira, The eigenvalue problem for ordinary differential equations of the second order and Heisenbergs theory of

S-matrices, Am. J. Math. 71, 921945 (1949).

16 A. Kostenko, A. Sakhnovich, and G. Teschl, Inverse eigenvalue problems for perturbed spherical Schrdinger operators,

Inverse Probl. 26, 105013 (2010).

17 A. Kostenko, A. Sakhnovich, and G. Teschl, WeylTitchmarsh theory for Schrdinger operators with strongly singular

potentials, Int. Math. Res. Not. 2012, 16991747 (2012).

18 A. Kostenko, A. Sakhnovich, and G. Teschl, Commutation methods for Schrdinger operators with strongly singular

potentials, Math. Nachr. 285, 392410 (2012).

19 A. Kostenko and G. Teschl, On the singular WeylTitchmarsh function of perturbed spherical Schrdinger operators, J.

Differ. Equations 250, 37013739 (2011).

20 A. Kostenko and G. Teschl, Spectral asymptotics for perturbed spherical Schrdinger operators and applications to quantum

scattering, Commun. Math. Phys. 322, 255275 (2013).

21 M. G. Krein, On a continual analogue of a Christoffel formula from the theory of orthogonal polynomials, Dokl. Akad.

Nauk SSSR (N. S.) 113, 970973 (1957) (in Russian).

22 P. Kurasov and A. Luger, An operator theoretic interpretation of the generalized TitchmarshWeyl coefficient for a singular

SturmLiouville problem, Math. Phys. Anal. Geom. 14, 115151 (2011).

23 B. M. Levitan and I. S. Sargsjan, SturmLiouville and Dirac Operators (Kluwer Academic Publishers, Dordrecht, 1991).
24 A. Luger, Generalized Nevanlinna functions. Operator representations, asymptotic behavior, in Springer Reference on

operator theory (in press).

25 G. Teschl, Deforming the point spectra of one-dimensional Dirac operators, Proc. Am. Math. Soc. 126, 28732881 (1998).
26 B. Thaller, The Dirac Equation (Springer, Berlin, 1991).
27 J. Weidmann, Spectral Theory of Ordinary Differential Operators, Lecture Notes in Mathematics (Springer, Berlin, 1987),

Vol. 1258.
28 J. Weidmann, Lineare Operatoren in Hilbertrumen, Teil 2: Anwendungen (B. G. Teubner, Stuttgart, 2003), http://link.

springer.com/referenceworkentry/10.1007%2F978-3-0348-0692-3_35-1.
Journal of Mathematical Physics is copyrighted by AIP Publishing LLC (AIP). Reuse of AIP
content is subject to the terms at: http://scitation.aip.org/termsconditions. For more
information, see http://publishing.aip.org/authors/rights-and-permissions.