Toward the theory of semi-linear equations

arXiv:1711.00041v1 [math.CV] 31 Oct 2017

Vladimir Gutlyanski, Olga Nesmelova, Vladimir Ryazanov

Abstract
In this paper we study the semilinear partial dierential equations in
the plane, the linear part of which is written in a divergence form. The
main result is given as a factorization theorem. This theorem states that
every weak solution of such an equation can be represented as a composi-
tion of a weak solution of the corresponding isotropic equation in a canon-
ical domain and a quasiconformal mapping agreed with a matrix-valued
measurable coecient appearing in the divergence part of the equation.
The latter makes it possible, in particular, to remove the regularity re-
strictions on the boundary in the study of boundary value problems for
such semilinear equations.

2010 Mathematics Subject Classification. AMS: Primary 30C62,

31A05, 31A20, 31A25, 31B25, 35J61, 35Q15 Secondary 30E25, 31C05, 34M50,
35F45

Keywords : semi-linear elliptic, parabolic and hyperbolic equations, quasi-

linear Poisson equation, inhomogeneous and anisotropic media, quasiconformal
mappings

1 Introduction
It is well known that the first order linear partial differential equation

z = (z) z , z C, (1.1)

1 1
where z = 2 (x iy ), z = 2 (x + iy ), z = x + i y, which has come
to be known as the Beltrami equation, turned out to be instrumental in the
study of Riemann surfaces, Teichmuller spaces, Kleinian groups, meromorphic
functions, low dimensional topology, holomorphic motion complex dynamics,
Clifford analysis and control theory. As known, a Kquasiconformal mapping
1,2
: C, K 1, is just an orientation-preserving homeomorphic Wloc ()

1
 2

solution to the Beltrami equation when the measurable coefficient satisfies the
strong ellipticity condition |(z)| (K 1)/(K + 1) almost everywhere in .
Note in particular, that if = 0 in a domain C, then the Beltrami equation
reduces to the CauchyRiemann equation and a solution is analytic in , see
e.g. [1], [25], see also [19], and the references therein.
We also would like to pay attention to a strong interaction between linear and
non-linear elliptic systems in the plane and quasiconformal mappings. The most
general first order linear homogeneous elliptic system with real coefficients can
be written in the form z + (z) z + (z) z = 0, with measurable coefficients
and such that || + || (K 1)/(K + 1) < 1. This equation is a particular
case of a non-linear first order system z = H(z, z ) where H : G C C is
Lipschitz in the second variable,

K 1
|H(z, w1 ) H(z, w2 )| |w1 w2 | , H(z, 0) 0 .
K +1

The principal feature of the above equation is that the difference of two solu-
tions need not solve the same equation but each solution can be represented as
a composition of a quasiconformal homeomorphism and an analytic function.
Thus quasiconformal mappings become the central tool for the study of these
non-linear systems. A rather comprehensive treatment of the present state of
the theory is given in the excellent book of Astala, Iwaniec and Martin [2].
This book contains also an exhaustive bibliography on the topic. In particular,
the following fundamental Harmonic Factorization Theorem for the uniformly
elliptic divergence equations

div A(z, u) = 0, z , (1.2)

1,2
holds, see [2], Theorem 16.2.1: Every solution u Wloc () of the equation (1.2)
can be represented as u(z) = h((z)), where : G is Kquasiconformal
and h is harmonic on G.
The main goal of this paper is to point out another application of quasicon-
formal mappings to the study of some nonlinear partial differential equations
 3

in the plane. Namely, we will deal with co-called semi-linear partial differential
equations, linear part of which contains the elliptic operator in the divergence
form div [A(z)u(z)], where the matrix function A(z) is assumed to be sym-
metric, measurable and satisfy the uniform ellipticity condition

1
||2 hA(z) , i K ||2 a.e. in (1.3)
K

for every C, where 1 K < . Some examples of such semi-linear equations

include in anisotropic case the nonlinear heat equation like

ut div [A(z)u(z)] = f (u), (1.4)

(the same equation describes the brownian motion, diffusion models of popula-
tion dynamics, and many other phenomena), nonlinear wave equation

utt div [A(z)u(z)] = f (u), (1.5)

with continuous functions f or nonlinear Schrodinger equation

iut + div [A(z)u(z)] = k|u|p u, (1.6)

as well as their stationary counterparts. We show that every weak solution of

semi-linear equations of such type can be factorized as the composition of a
solution to the corresponding isotropic equation and a quasiconformal mapping
agreed with the matrix function A(z). If we deal, for example, with the semi-
linear equation
div [A(z)u(z)] = f (u), f 0, (1.7)

then we prove the following generalization of the mentioned above Factorization

1,2
theorem: Every solution u Wloc () of the semi-linear equation (1.7) can be
represented as u(z) = h((z)), where : G is a quasiconformal mapping
agreed with the matrix function A and h is subharmonic on G.
The paper is organized as follows. In the next section we recall basic results
from the theory of quasiconformal mappings and their connection with linear
elliptic differential equations in the divergence form. In Section 3 we prove
 4

the basic identity concerning the linear elliptic operator in the divergence form.
The main result of this section is Proposition 3.1. In Section 4 we prove the
principal Factorization Theorem 4.1 and their corollaries for some elliptic semi-
linear model equations in the plane. Some applications of Factorization Theorem
to existence theorems for elliptic semi-linear equations are given in Section 5.
Its applications to the free boundary problems ("dead zones") can be found in
Section 6. A similar discussion of equations of the heat type is in Section 7.
Finally, Section 8 contains a discussion of boundary value problems.

2 QK-maps and divergent equations

Let be a domain in the complex plane C. Denote by M 22 () the class of all
2 2 symmetric matrix function A(z) = {ajk (z)} with measurable entries and
det A(z) = 1, satisfying the uniform ellipticity condition

1
||2 hA(z) , i K ||2 a.e. in (2.1)
K

for every C where 1 K < .

Given A M 22 (), let us first consider the following second order elliptic
homogeneous equation in the divergent form

div (A(z) u) = 0 a.e. in . (2.2)

Equation (2.2) is interpreted in the distributional sense. That is, a function u

is a weak solution to the equation if it has locally integrable gradient u with
Z
hA(z)u, i = 0 C0 () . (2.3)

1,1 1,p
This is meaningful at least for u Wloc (), where Wloc () stands for the well-
known Sobolev space. Here we will assume a little more regularity, namely that
1,2
u C Wloc (), because in this paper we will be dealing with quasiconformal
mappings generated by A with the uniform ellipticity condition (2.1). We plan
to study also semi-linear degenerate partial differential equations in subsequent
1,1
works. In that case we will consider the weaker solutions u C Wloc ().
 5

1,2
Let A M 22 () and u C Wloc () be a weak solution to (2.2). Then
1,2
there exists v C Wloc (), called the stream function of u, such that
 
0 1
v = H A u a.e in , where H = . (2.4)
1 0

Note by the way that

 
1 0
H 2 = I , where I = ,
0 1

i.e. H takes a part of the imaginary unit in the space of 2 2 matrices.

Setting (z) = u(z) + i v(z), one writes in complex notations that satisfies
the Beltrami equation

z (z) = (z) z (z) a.e. in , (2.5)

where the complex dilatation (z) is given by

a22 (z) a11 (z) 2ia12 (z)

(z) = . (2.6)
det (I + A(z))

The condition of ellipticity (2.1) now is written as

K 1
|(z)| a.e. in . (2.7)
K +1

And vice versa, given measurable complex valued function , satisfying (2.7),
one can invert the algebraic system (2.6) to obtain that, see e.g. Theorem 16.1.6
in [2],
|1|2 2Im
!
1||2 1||2
A(z) = 2Im |1+|2
. (2.8)
1||2 1||2

Thus, given any A M 22 (), one produces by (2.6) the complex dilatation
(z) for which in turn, by the Measurable Riemann mapping theorem, see e.g.
Theorem V.B.3 in [1] and Theorem V.1.3 in [25], the Beltrami equation (2.5)
generates as its solution a quasiconfomal homeomorphism : G. As the
domain G one can take any plane domain which is conformally equivalent to .
1,p 2K
It is well known that Wloc (), 2 p < K1 , Cloc (), = 1/K,
and we have seen that if (z) = u(z) + i v(z), then div (A(z) u(z)) = 0,
 6

1,2
div (A(z) v(z)) = 0 a.e. in and u and v C Wloc (). In what follows we
will say that the matrix function A generates the corresponding quasiconformal
mapping , or that A and are agreed.
Note also in passing the very often applied fact that quasiconformal map-
pings admit the change of variables in integrals because homeomorphisms of
1,2
the class Wloc are absolute continuous with respect to the area measure, see
e.g. Theorem III.6.1, Lemmas III.2.1 and III.3.3 in [25].
We complete this section with the following very important result on the
composition operators in the Sobolev spaces.
Let be a domain in the ndimensional Euclidean space Rn , n 2. For the
sake of completeness in the exposition we recall that the Sobolev space L1p (),
p 1, consists of locally integrable functions : R with distributional
partial derivatives of the first order with the seminorm
1/p
Z
||||L1p () = ||||Lp () = ||p dm < ,

where m is the Lebesgue measure in Rn , is the distributional gradient of

the function , = (/x1 , ..., /xn ) , x = (x1 , ..., xn ), defined by the
conditions

Z Z
dm = dm C0 (), i = 1, 2, ..., n.
xi xi

Here C0 () stands for the space of all infinitely smooth functions with a
compact support in . Similarly, a vector-function is said to belong to the
Sobolev class L1p () if its coordinate functions belong to L1p (). The classes
W 1,p () = L1p () Lp (), which will be used later on, differ from the classes
L1p () only by the norm

||||W 1,p () = ||||Lp () + ||||Lp () .

The following statement will play an important role in our further conside-
rations, see [11], [34] and [39].
 7

Lemma 2.1. Let : G be a homeomorphism. Then the following

conditions are equivalent:
1) the composition = generates the bounded operator

: L1p (G) L1q () , 1 q p < ,

1,1
2) the mapping belongs to the class Wloc () and the function
n o
Kp (x; ) := inf k(x) : ||D||(x) k(x)|J (x)|1/p

belongs to Lr (), where r is determined from the equality

1 1 1
= .
r q p

Here ||D||(x) stands for the operator norm of the Jacobian matrix D of
the mapping at the point x, i.e., ||D||(x) := sup D h.
hC,|h|=1
Assuming that is a quasiconformal homeomorphism and, consequently,
1,2
Wloc () and K2 (x, ) L (), we arrive at the the following conclusion,
when n = 2, p = 2, q = 2 and r = .

Proposition 2.1. Let : G be a quasiconformal homeomorphism

1,2
and let : G C belong to the class Wloc (G). Then the composition function
1,2
Wloc (G).

The study of superposition operators on Sobolev spaces stems from the clas-
sical article [33], see also e.g. [10], [36] and [37] for the detailed history and
bibliography. The Reshetnyak problem on the description of all isomorphisms
of the Sobolev space L1n generated by a quasiconformal mapping in Rn , n 2,
was posed in 1968 at the first Donetsk colloquium in the theory of quasiconfor-
mal mappings. The problem was solved in [38], see also [32]. As a consequence,
the composition = generates the bounded operator

: L12 (G) L12 ()

if and only if the homeomorphism is quasiconformal, see also [31].

 8

3 A basic identity
It is well-known that every positive defined quadratic form

ds2 = a(x, y)dx2 + 2b(x, y)dxdy + c(x, y)dy 2 , (3.1)

defined in a plane domain , can be reduced, by means of a suitable quasicon-

formal change of variables, to the canonical form

ds2 = (du2 + dv 2 ), 6= 0 a.e. in , (3.2)

provided that ac b2 0 > 0, a > 0, a.e. in , see, e.g., [7], pp. 10

12. This key result can be extended to every linear divergent operator of the
form div [A(z)u(z)], z = x + iy, for matrix function A M 22 (). Namely,
given such a matrix function A and a quasiconformal mapping : G,
1,2
Wloc () agreed with A, we have already seen by direct computation, that
if the function T and the entries of A are sufficiently smooth, then

div [A(z)(T ((z)))] = J (z)T ((z)), z , (3.3)

see, e.g., [15], [16]. Here J (z) stands for the Jacobian of the mapping (z), e.g.,
J (z) = det D (z), where D (z) is the Jacobian matrix of the mapping at the
point z . Making use of the standard procedure, the regularity requirements
in the equality (3.3) may be substantially weakened. The equality (3.4) below,
that will be applied to the study of weak solutions to some semi-linear partial
differential equations, can be viewed as a weak counterpart to equality (3.3).

Proposition 3.1. Let be a domain in C, A M 22 () and : G

be a quasiconformal mapping agreed with A. Then the equality
Z Z
hA(z)(T ((z))), (z)i dmz = hD1 (z)T ((z)), (z)iJ (z) dmz

(3.4)
1,2
holds for every T Wloc (G) and for all W01,2 ().
 9

1,2
Proof. Assuming that T Wloc (G) and that : G is a quasiconformal
mapping agreed with A(z), we see, by Proposition 2.1, that u : = T
1,2
Wloc (). Since
u(z) = Dt (z)T ((z)), (3.5)

where Dt (z) stands for the transpose matrix to D (z), and satisfies the
Beltrami equation (2.5), that can be written in the matrix form as

A(z)Dt (z) = D1 (z) J (z), (3.6)

we arrive successively to the required equality (3.4):

Z
hA(z)(T ((z))), (z)i dmz

Z
= hA(z)Dt (z)T ((z)), (z)i dmz

Z
= hD1 (z)T ((z)), (z)iJ (z) dmz . (3.7)

Remark 3.1. It is easy to see that if T and A are smooth and (3.4) holds,
then (3.3) also holds. One can use the Greens formula as above, but in the
opposite direction, to show that
Z
{div [A(z)(T ((z)))] J (z)T ((z))} (z) dmz = 0, C01 ().

(3.8)
Indeed, by Greens formula we have that
Z Z
hA(z)(T ((z))), (z)i dmz = div [A(z)(T ((z)))](z) dmz .

(3.9)
On the other hand, applying again Greens formula after the change of variables
w = (z), we get
Z
J (z)T ((z))(z) dmz =

Z
hT ((z)), [Dt ]1 (z)(z)iJ (z) dmz . (3.10)

 10

From the linear algebra it follows that

hT ((z)), J (z)[Dt ]1 (z)(z)i = hJ (z)D1 (z)T ((z)), (z)i. (3.11)

Indeed, let (z) = a(z) + ib(z), z = x + iy, then

 
ax ay
D = , (3.12)
bx by
 
ax b x
Dt = , (3.13)
ay b y
 
1 by ay
D J = (3.14)
bx ax
and  
by bx
[Dt ]1 J = . (3.15)
ay ax
Setting T = (u, v), and = (, ), we get that

hT, J [Dt ]1 i = h(u, v), (by bx , ay + ax )i =

u(by bx ) + v(ax ay ), (3.16)

hJ D1 T, i = h(by u ay v, bx u + ax v), (, )i =

u(by bx ) + v(ax ay ), (3.17)

and we arrive to formula (3.11).

Thus,
Z
J (z)T ((z))(z) dmz =

Z
hD1 (z)T ((z)), (z)iJ (z) dmz . (3.18)

Finally, since C01 () is dense in W01,2 (), we come to the equality (3.3) for
almost all points in . However, since all functions are smooth, the equation
holds at every point.
 11

4 Semi-linear model elliptic equations

Let be a bounded domain in C and let f : R R be a continuous function.
In this section we study a model semi-linear equation

div [A(z)u(z)] = f (u(z)), z , (4.1)

as well as its Laplaces counterpart:

T (w) = J(w)f (T (w)), w G = (), (4.2)

where : G is a quasiconformal mapping agreed with A(z) and J(w)

stands for the Jacobian of the inverse mapping 1 : G .
Under a weak solution to the equation (4.1) we understand a function u
1,2
C Wloc () such that
Z Z
hA(z)u(z), (z)i dmz + f (u(z))(z) dmz = 0 C W01,2 () .

(4.3)
The equation (4.2) contains a weight function J(w) L1loc (G). In spite of
this, we may define a notion of a weak solution to the equation (4.2) in the
similar way.
1,2
We say that T is a weak solution to the equation (4.2) if T C Wloc (G)
and
Z Z
hT (w), (w)i dmw + J(w)f (T (w))(w) dmw = 0 C W01,2 (G) .
G G
(4.4)
Since J(w) is the Jacobian of the mapping 1 (w), it is easy to verify, by
performing the change of variable by the formula w = (z), that the second
integral in (4.4) is well-defined. Here we again made use of the fact that the
1,2 1,2
composed mapping u(z) = T ((z)) is in C Wloc () if T C Wloc (G) and
is quasiconformal.
The existence of weak solutions to the equations (4.1) and (4.2 ), under
specified conditions on the right hand sides, as well as fundamental properties
of solutions can be found, e.g. in [24], see also [22].
 12

The following Factorization Theorem is a principal result of this paper.

Theorem 4.1. Let be a domain in C, A M 22 () and let f : R R be

a continuous function. Then every weak solution u of the semi-linear equation

div [A(z) u(z)] = f (u(z)), z , (4.5)

can be represented as the composition

u(z) = T ((z)), (4.6)

where : G is a quasiconformal mapping agreed with A and T is a weak

solution to the equation

T (w) = J(w) f (T (w)), w G . (4.7)

Here J(w) stands for the Jacobian of the inverse mapping 1 (w).

Proof. Let u be a weak solution of the semi-linear equation (4.5) and T = u 1.

1,2
Then by Proposition 2.1 T C Wloc (G) and we have that
Z Z
hA(z)(T ((z))), (z)i dmz + f (T ((z)))(z) dmz = 0 (4.8)

for all C W01,2 (). Next, by Proposition 3.1,

Z Z
hA(z)(T ((z))), (z)i dmz = hD1 (z)T ((z)), (z)iJ (z) dmz ,

(4.9)
and therefore
Z Z
hJ (z)D1 (z)T ((z)), (z)i dmz + f (T ((z)))(z) dmz = 0 (4.10)

for all C W01,2 ().

Given an arbitrary function (w) C W01,2 (G), we can set in (4.8) and
(4.9) (z) = ((z)), because, by Proposition 2.1, such C W01,2 ().
 13

Performing in (4.10) the change of variable by the formula z = 1 (w), we

obtain
Z
hJ ( 1 (w))D1 ( 1 (w))T (w), Dt ( 1 (w))(w)iJ(w) dmw
G
Z
+ J(w)f (T (w))(w) dmw = 0.
G

Since, by elementary algebraic arguments,

hJ ( 1 (w))D1 ( 1 (w))T (w), Dt ( 1 (w))(w)i =

= J ( 1 (w))hT (w), (w)i,

and
J ( 1 (w)) = 1/J(w),

we see that the identity

Z Z
hT (w), (w)i dmw + J(w)f (T (w))(w) dmw = 0 (4.11)
G G

holds for all C W01,2 (G). Thus, T is a weak solution to the equation
(4.7).

Remark 4.1. Since the arguments given above are invertible, we see that
if T is a weak solution to the equation (4.7), then the function u(z) = T ((z))
is a weak solution to the semi-linear equation (4.5)

Assuming that the function f is non-negative, we arrive at the following

statement.

Corollary 4.1. Let be a domain in C, A M 22 () and let f : R R be

a continuous function. If f (u) 0, then every weak solution u of the semi-linear
equation
div [A(z)u(z)] = f (u(z)), z , (4.12)

can be represented as the composition

u(z) = T ((z)), (4.13)

 14

where : G is quasiconformal mapping agreed with A and T is a subhar-

monic function, being a weak solution of the equation

T = J(w) f (T (w)) in G . (4.14)

Here J(w) stands for the Jacobian of the inverse mapping 1 (w).

Among the quasiconformal mappings : G there are a variety of the

so-called volume-preserving maps, for which J (z) 1, z . Many examples
of such maps will be given in the next section. The following statement may
have of independent interest.

Corollary 4.2. Let A M 22 () be a matrix function that generates

a volume-preserving quasiconformal mapping : G. Then every weak
solution u of the semi-linear equation

div [A(z)u(z)] = f (u(z)), z , (4.15)

can be represented as the composition

u(z) = T ((z)), (4.16)

where T is a weak solution to the equation

T = f (T (w)), a.e. in G. (4.17)

Remark 4.2. By the Measurable Riemann mapping theorem, see e.g.

Theorem V.B.3 in [1] and Theorem V.1.3 in [25], given (z), z , agreed
with the matrix function A M 22 (), there exists a quasiconformal mapping
: G with the complex dilatation given by (2.6). Here if is finitely
connected, then G is either the canonical circular domain, or the complex plane,
see e.g. Theorem V.6.2 in [12]. It allows us to choose in the Factorization The-
orem 4.1 as the domain G one from the above canonical domains. If is simply
 15

connected with a nondegenerate boundary then we can set G = D. The latter

makes it possible to remove the restrictions on the regularity of the boundary
in the study of boundary value problems for equation (4.7), see Section 8.

5 Some Applications
We start with a few examples of the application of Theorem 4.1 to the study of
boundary blow-up solutions for some classical model semi-linear elliptic equa-
tions.
Let be a bounded domain in C and let denote its boundary. In this
section we study the problem

div [A(z)u(z)] = f (u(z)), (5.1)

u(z) as d(z) := dist (z, ) 0, (5.2)

as well as its Laplace counterpart:

u(z) = f (u(z)), (5.3)

u(z) as d(z) := dist (z, ) 0. (5.4)

Solutions to these problems are called boundary blow-up solutions, or large so-
lution.
The existence of a large solution to (5.3) is related to the existence of a
maximal solution u of (5.3) in , which in turn depends on the so called Keller-
Osserman condition, see [20] and [29]. For example, in the paper [20] a simple
upper bound was obtained for any solution, in any number of variables, of semi-
linear equation (5.3). The bound depends on the function f which must be
positive and satisfy the well-known Keler-Osserman condition. Recall that a
function f C(R+ ) satisfies the Keler-Osserman condition if there exists a
positive non-decreasing function h such that
Z Z t 1/2
f (t) h(t), t R+ and h(s)ds dt < for all t0 > 0. (5.5)
t0 0
 16

It is known that if f is non-decreasing and satisfies the Keller-Osserman condi-

tion, then a large solution exists in every bounded smooth domain. Uniqueness
in smooth domains was established under some additional conditions on f, see
e.g. [3]. It is easy to check that the functions f (t) = et and f (t) = tp , p > 1,
satisfy (5.5). The Gauss-Bieberbach-Rademacher semi-linear equation

u = eu , (5.6)

as far as we know, was first investigated by Bieberbach in his pioneering work

[5] related to the study of automorphic functions in the plane. It is this work
has stimulated numerous studies in the field of semi-linear differential equations
in Rn , n 1, and the equation (5.6) continues to play the role of one of the
fundamental model equations of the theory. It is important to note that in
simply connected planar domains the large solutions for the equation (5.6)
are represented explicitly by means of the conformal map : D := {z
C : |z| < 1} :
8| (z)|2
u(z) = log . (5.7)
(1 |(z)|2 )2
Let us also remark that, given an arbitrary analytic function (z) in , the
formula (5.7) generates a solution, generally speaking with singularities, of the
equation (5.6).
For the model case of equation

u = eau , a > 0, (5.8)

the following result holds, see [23], Theorem 5.3.7.

Theorem 5.1 Let be a bounded domain in C such that = Int . Then

there exists one and only one blow-up solution to (5.8).

The Gauss-Bieberbach-Rademacher semi-linear equation is one of the prin-

cipal model equations in the theory of non-linear partial differential equations
and their applications, see e.g. [26], [23], [27] and the references therein. Note
 17

that the equation appears also as a model one in problems of differential geom-
etry in relation with existence of surfaces of negative Gaussian curvature [35]
and in studying the equilibrium of a charged gas.
In the general case for |(z)| k < 1 the solution of the Beltrami equation
can be written as an infinite series of singular integral transforms of Hilbert
and Cauchy type for complex dilatations, see, e.g. [7], p. 33. Here we give
the explicit solutions of the Beltrami equation for the cases where the complex
dilatation is a measurable function that depends on a single real variable x =
Re z, y = Im z, arg z or |z|, see [7], 5.10, [17]. We make use of the corresponding
explicit formulae to write down a number of explicit solutions for the counterpart
(5.1) of the Gauss-Bieberbach-Rademacher equation in the unit disk and the
upper half-plane.
We start with the following statement, see, e.g., [7], p. 82.

Proposition 5.2. Let the complex dilatation (z) has the form
z
(z) = k(|z|) , (5.9)
z
where k( ) : R C is a measurable function, kkk < 1. Then the formula

Z1
z 1 + k( ) d
(z) = exp (5.10)
|z|
1 k( )
|z|

represents a unique quasiconformal mapping of the unit disk onto itself with the
complex dilatation and the normalizations (0) = 0 and (1) = 1.

Analyzing formula (5.10), we arive at the following statement.

Corollary 5.1. Let D be the unit disk in the complex plane C centered at the
origin and let the matrix function A(z) be generated by the Beltrami coefficient
z
(z) = k(|z|) (5.11)
z
in accordance with formula (2.8) where
p
k(t) = 2 (t) i (t) 1 2 (t) (5.12)
 18

and (t), 0 t < 1, stands for an arbitrary measurable real-valued function

such that |(t)| q < 1. Then there exists one and only one boundary blow-up
solution to the semi-linear equation

div [A(z)u] = eu , z D, (5.13)

which is written explicitly by the Liouville-Bieberbach formula

8
u(z) = log . (5.14)
(1 |z|2 )2

Proof. Let A(z) be a matrix function generated by the complex Beltrami coef-
ficient (5.11), satisfying (5.12). By the Measurable Riemann mapping theorem,
see e.g. Theorem V.B.3 in [1] and Theorem V.1.3 in [25], there exists the unique
normalized quasiconformal self-homeomorphism of the unit disk D with the
Beltrami coefficient (z) = k(|z|)z/z and this homeomorphism, by Proposition
5.2, is written explicitly,
( Z )
1
z 1 + k(t) dt
(z) = exp . (5.15)
|z| |z| 1 k(t) t

We see that
z k 1
= ,
1 k z
z 1 1
=
1kz
and therefore, for the Jacobian J (z) = |z |2 |z |2 we have

1 |k|2 ||2
J (z) = . (5.16)
|1 k|2 |z|2

In order to apply Theorem 4.2, we have to verify that J (z) 1 and |(z)| = |z|
for z D. Indeed, noting that Re k = |k|2 and substituting k(t) given by
formula (5.12) into (5.15) we see that |(z)| = |z|, z D. The later implies that
J (z) 1.
 19

Making use of Corollary 4.2, we see that u(z) = T ((z)) and T satisfies the
Gauss-Bieberbach-Rademacher equation

T = eT in D. (5.17)

By the mentioned above Bieberbachs result and referring also to Theorem 5.3.7
from [23], we see that the unique boundary blow-up solution to the equation
(5.17) in the unit disk D is given by the explicit formula
8
T (w) = log . (5.18)
(1 |w|2 )2
Since |(z)| = |z| for z D, the required solution to the equation (5.13) also
has the explicit representation
8
u(z) = log . (5.19)
(1 |z|2 )2

Thus, there is an infinite set of matrix functions A(z) such that the corre-
sponding anisotropic semi-linear equations of the form (5.13) have a solution
defined by the explicit formula (5.14).
Let us give a nontrivial example, based on the well-known logarithmic spiral
quasiconformal mapping
(z) = ze2i log |z|

which plays important role in the study of different problems of contemporary

analysis, see, e.g., [7], 13.2, [13], [14]. This function maps the unit disk
D onto itself and transforms radial lines into spirals, infinitely winding about
the origin, and it is just the volume preserving. The mapping satisfies the
Beltrami equation with
z 1 z
(z) = = (1 + i)
z 2 z
and the Jacobian J (z) = |z (z)|2 |z (z)| 1. We see that corresponds to

(5.12) with (t) 1/ 2. Since
x2 y 2 2xy x2 y 2 + 2xy
Re (z) = , Im (z) = , z = x + iy,
2(x2 + y 2 ) 2(x2 + y 2 )
 20

we see that generates by the formula (2.8) the matrix function Asp (x, y) with
the following entries:

|1 |2 x2 y 2 2xy
a11 = = = 3 2 ,
1 ||2 (x2 + y 2 )

|1 + |2 x2 y 2 2xy
a22 = = = 3 + 2 ,
1 ||2 (x2 + y 2 )
2Im x2 y 2 + 2xy
a12 = a21 = = 2
= 2 .
1 || (x2 + y 2 )
By Corollary 5.1 the semi-linear equation

div [Asp (x, y)u] = eu , z D, (5.20)

must have the function

u(x, y) = 2 log(1 x2 y 2 ) + log 8

as the blow-up solution in the disk D. Let us verify this conclusion by means of
direct computation.
We see that
4x 4y
ux = , uy =
1 x2 y 2 1 x2 y 2
and therefore

x + 2y y 2x
ux + uy = 4 2 2
, ux + uy = 4 .
1x y 1 x2 y 2

Then

8
div (Asp (x, y)u) = 4(ux + uy )x + 4(ux + uy )y = = eu ,
(1 x2 y 2 )2

and thus we successfully complete the verification.

Note also that the matrix function Asp in the polar coordinates z = rei has
the form
 
3 2 2 cos(2 + /4) 2 2 sin(2 + /4)
A= . (5.21)
2 2 sin(2 + /4) 3 + 2 2 cos(2 + /4)
 21

Corollary 5.2. Let be the annulus r < |z| < 1 in the complex plane C
and let the matrix function A(z) be generated by the Beltrami coefficient

z
(z) = k(|z|) (5.22)
z

in accordance with formula (2.8) where

p
k(t) = 2 (t) i (t) 1 2 (t) (5.23)

and (t), 0 t < 1 stands for an arbitrary measurable function. If |(t)| q <
1, then there exists one and only one boundary blow-up solution to the classical
model semi-linear equation

div [A(z)u] = eu , in the annulus r < |z| < 1, (5.24)

and which is written explicitly by the formula

2 2
u(z) = log (5.25)
|z|2 (log2 r) sin2 ( log r log |z|).

Proof. Arguing in the same way as in the proof of Corollary 5.1, we see that
u(z) = T ((z)), r < |z| < 1, where quasiconformal self-homeomorphism (z) of
the annulus is given by (5.15), (note that under the condition (5.23) |(z)| = |z|),
and T is the unique boundary blow-up solution to the semi-linear equation

T = eT (5.26)

in the annulus r < |w| < 1. By Bieberbachs result, the unique boundary blow-
up solution to the equation (5.26) is given by the formula

8|F ()|2
T () = log (5.27)
(1 |F ()|2 )2

where F () stands for a conformal mapping of the annulus r < || < 1 onto
the unit disk. It remains to find the corresponding conformal mapping F. We
see that: 1) w = log maps the annulus onto the strip log r < Re w < 0; 2)
 22

= ie log r iw maps the strip onto the right half-plane; 3) t = ( 1)/( + 1)

maps the right half-plane onto the unit disk. Composing the above mappings
we get the required formula

ie log r i log + 1

F () = . (5.28)
ie log r i log 1

Next, if we set () = ie log r i log , then we see that

16Re2 ()
(1 |F ()|2 )2 = (5.29)
| 1|4

where  

Re () = | | sin log || .
log r
On the other hand,
2
4 4 2 1
|F ()|2 = |
()| 2
= | | 2 ||2 . (5.30)
| 1|4 | 1|4 log r

Thus
2 2
T () = log 2 (5.31)
||2 (log r) sin2 ( log r log ||)
and since |(z)| = |z|, we complete the proof of Corollary 5.2.

Making use of the limit in (5.25) as r 0 we get the following result which
may have of independent interest.

Corollary 5.3 The semi-linear equation

div [A(z)u] = eu (5.32)

for each matrix function A from Corollary 5.2 as well as the Gauss-Bieberbach-
Rademacher equation
u = eu (5.33)

admit the following boundary blow-up solution in the punctured unit disk 0 <
|z| < 1
2
u(z) = log . (5.34)
|z|2 log2 |z|
 23

Remark 5.1. The approach given above to the construction of a boundary

blow-up solution to the Gauss-Bieberbach-Rademacher equation in the unit disk
D with a singularity at the origin can be extended to the case of a finite number
of singular points zk , |zk | < 1, k = 1, 2, ..., n. Indeed, let r > 0 be such that all
the disks dk = {z : |z zk | r} belong to D and do not intersect each other.
Denote by Fr (z) a conformal mapping of the circular multi-connected domain
D\ nk=1 dk onto the unit disc D, see Theorem VI.1 in [12] due to Poincare. Then
the required solution with prescribed singularities at the points zk is given by
8|Fr (z)|2
u(z) = lim log . (5.35)
r0 (1 |Fr (z)|2 )2
In order to give more examples of applications of our Factorization Theorem
and its corollaries, recall the following statement, which provides us with the
explicit solutions of the Beltrami equation for the cases where the complex
dilatation is a measurable function that depends on a single real variable
x = Re z or y = Im z, see, [7], p. 78, [17].

Proposition 5.3. Let : C D be an arbitrary measurable function with

kk q < 1 that depends on x = Re z only and let
Zx
1 + (t)
(x) = dt . (5.36)
1 (t)
0

Then the formula

(z) = (x) + iy (5.37)

represents a unique quasiconformal mapping of C onto itself with complex di-

latation and normalizations:

(0) = 0, (i) = i, () = . (5.38)

Indeed, from formulas (5.36) and (5.37) it follows that

K 1 |x1 x2 | |(z1 ) (z2 )| K|z1 z2 | . (5.39)

 24

Thus, is a lipschitz homeomorphism of the plane and hence ACL. The nor-
malization (5.38) is obvious. Furthermore,

1 + (x)
x = (x) = , (5.40)
1 (x)

y = i . (5.41)

Therefore,
1 (x)
z = (x + iy ) = , (5.42)
2 1 (x)
1 1
z = (x iy ) = , (5.43)
2 1 (x)
and hence satisfies the Beltrami equation

z = (x)z . (5.44)

The Jacobian of
1 |(x)|2
J (z) = K 1 > 0 (5.45)
|1 (x)|2
is positive, that is, is orientation-preserving. This completes the proof of
Proposition 5.3.

Remark 5.2. From

u = Re (z) = Re (x) , (5.46)

v = Im (z) = y + Im (x) , (5.47)

we conclude that maps vertical lines onto vertical lines without contractions
or dilations, and the imaginary axis is mapped identically onto itself, since
(0) = 0. It is easy to show that these geometric properties characterize the
class of quasiconformal mappings in question, because

(x) 1
(z) = (5.48)
(x) + 1

depends only on x.
 25

Corollary 5.4. Suppose that the complex dilatation (z) of a quasiconformal

mapping g(z) of the extended complex plane, keeping the points 0, 1, , depends
only on y = Im z. Then
g(z) = x + i(y) , (5.49)

where
Zy
1 (it)
(y) = dt . (5.50)
1 + (it)
0

Indeed, let
= A g A1 , (5.51)

where
A() = ei 2 = i

(5.52)

is a counter-clockwise rotation by the angle /2. Then (0) = 0, (i) = i,

() = , and the complex dilatation of ,

(z) = (iz) , (5.53)

depends only on x = Re z. Thus, from (5.36), (5.37), and (5.51) we obtain

(5.49) and (5.50).

Let us return to the study of the blow-up solutions to the Liouville-Biberbach

type equation defined in the right half-plane.

Corollary 5.5. Let H + be the right half-plane {z : Re z > 0} in the complex

plane C and let the matrix function A(z) be generated by the formula (2.8) with
the Beltrami coefficient
p
(x) = 2 (x) i (x) 1 2 (x) (5.54)

where (x), x R, stands for an arbitrary measurable real-valued function,

such that |(x)| q < 1. Then there exist boundary blow-up solutions to the
semi-linear equation
div [A(z)u] = eu , z H + , (5.55)
 26

and which are written explicitly:

2
u(z) = log , z = x + iy, (5.56)
x2

u(z) = log 82 2x 2 log(1 e2x ), > 0. (5.57)

Proof. We prove only that the function u(z), given by the formula (5.56) solves
the equation (5.55). Let A(z) be a matrix function generated by the complex
Beltrami coefficient , satisfying (5.54). Since depends on x only, then for-
mulae (5.36), (5.37) represent a unique quasiconformal mapping w = (z) of
the right half-plane H + onto itself with complex dilatation normalized by
(0) = 0, (i) = i, () = .
In order to apply Corollary 4.2, we have first to verify that J (z) 1. Indeed,
in our case Re = ||2 and formula (5.45) implies that J (z) 1. Moreover,
Re (z) = Re (x) = x. By Corollary 4.2 a weak solution u of the semi-linear
equation
div [A(z)u(z)] = eu , z H + , (5.58)

is represented as the composition

u(z) = T ((z)), (5.59)

where T is a weak solution to the equation

T (w) = eT (w) , in H + . (5.60)

Since the function

w1
F (w) =
w+1
is a conformal mapping of H + onto the unit disk D, we see that the function

8|F (w)|2
T (w) = log = 2 log Re w + log 2
(1 |F (w)|2 )2
 27

gives us a blow-up solution to the equation (5.60) in H + . Now, by formula

(5.59), we have that the first required solution has the form

u(z) = T () = 2 log Re (z) + log 2 = 2 log Re x + log 2.

In order to get the second solution, we have to note that the function

F (w) = ew , > 0,

is also conformal mapping of H + onto the punctured unit disk D\{0}. Repeating
the above arguments and taking into account that Re (z) = x, we arrive to
solution (5.57)

82 |e(z) |2
u(z) = T () = log = log 82 2x 2 log(1 e2x ).
(1 |e(z) |2 )2

As an example, one can take the matrix function A with the following con-
stant entries a11 = 1, a12 = 2, a22 = 5. Then (z) = (1 + i)/2 and (z) =
x + i(y + 2x). It is easy to verify by straightforward computation that the func-
tions (5.56), (5.57) solve the equation (5.55). In the general case the admissible
p
matrices in Corollary 5.3 have the entries: a11 = 1, a12 = 2(x)/ 1 2 (x),
a22 = (1+3 2 (x))/(1 2 (x)). The particular case considered above corresponds

to the function (x) = 1/ 2.

Similar results can be given for the upper half-plane on the base of Corollary
5.4.

6 Free boundary
The very important applied effect of the "dead zone" for solutions of some
partial differential equations, see e.g. [9], the Introduction and 1, is that a
solution of the corresponding differential equation vanishes on some nonempty
 28

open proper subset of the domain of its definition. For example, it is well-known
that a solution of the semi-linear equation

u = uq

may have the "dead zone" only when 0 < q < 1, see e.g. [9], p. 15.
We confine ourselves to only one result in this respect, which is a simple
consequence of Proposition 5.3 and Corollary 4.2.

Theorem 6.1. Let C be the complex plane and let the matrix function A(z),
z = x + iy, be generated by the Beltrami coefficient
p
(x) = 2 (x) i (x) 1 2 (x), (6.1)

that is
2(x)

1 2
1 (x)
A(z) = 1+3 2 (x)
, (6.2)
2(x)
2 1 2 (x)
1 (x)

where (x), x R, stands for an arbitrary measurable real-valued function, such

that |(x)| k < 1. Then the semi-linear equation

div [A(z)u] = uq , 0 < q < 1, z C, (6.3)

has in the complex plane the following solution with the "dead zone":
2
Rx 2(t)
  1q
y dt , if y > (x), x R,

u(x, y) = 0 1 2 (t) (6.4)

0 if x (x).

Here 1
(1 q)2
  1q
= ,
2(1 + q)
and
Zx
2(t)dt
y = (x) = p , < x < +,
1 2 (t)
0

stands for the corresponding free boundary parametrization.

 29

Proof. Let A(z) be a matrix function generated by the complex Beltrami coeffi-
cient , satisfying (6.1). Since depends on x only, then formulae (5.36), (5.37)
represent a unique quasiconformal mapping
Zx

2(t)
(z) = x + i y p dt (6.5)
1 2 (t)
0

of the complex plane C onto itself with complex dilatation normalize by (0) =
0, (i) = i, () = .
Since J (z) 1, one can apply Corollary 4.2 to represent solutions to the
equation (6.3) in the form
u(z) = T ((z)),

where T (w) satisfies the equation

T (w) = T q (w), w = + i.

We see that the function

2
T (w) = 1q , if > 0

and T (w) = 0, if 0, satisfy the above equation. Thus we arrive at the

required solution to the equation (6.3):
2
Zx
1q
2(t)
u(x, y) = y p dt , if y > (x), x R,
1 2 (t)
0

and u(z) = 0, if x (x) where

1 Zx
(1 q)2
  1q
2(t)dt
= , (x) = p .
2(1 + q) 1 2 (t)
0

Let us verify this solution by straightforward computation. Since

2
Zx
1q

u(x, y) = y a12 (t)dt ,

0
 30

we see that

2 1+q 2 1+q
ux = a12 (y + (x)) 1q , and uy = (y + (x)) 1q .
1q 1q

It yields that

2 1+q 2 1+q
a11 ux + a12 uy = a12 (y + (x)) 1q + a12 (y + (x)) 1q = 0.
1q 1q

On the other hand, since a22 a212 = 1, we get

2 1+q
a12 ux + a22 uy = (y + (x)) 1q .
1q

Thus,
2q
div (A(z)u) = (a12 ux + a22 uy )y = q (y + (x)) 1q = uq ,

and we complete the verification.

7 Heat equation
Factorization theorems, similar to Theorem 4.1 and its corollaries, also hold for
parabolic and hyperbolic linear and semi-linear equations in the plane, which
contain the corresponding divergence operator in the linear part. We give,
without proof, one of such results.

Theorem 7.1. Let be a domain in C, A M 22 () and let f : R R

be a continuous function. Then every weak solution of the semi-linear equation

ut a2 div [A(z) u] = f (u) in , t > 0 , (7.1)

can be represented as the composition

u(z, t) = T ( (z), t ) , (7.2)

where : G is a quasiconformal mapping agreed with A, and T is a weak

solution of the equation

Tt a2 J 1 T = f (T ) in G , t > 0 . (7.3)
 31

Here J stands for the Jacobian of the mapping .

As an example for the illustration of Theorem 6.1. we give a statement

concerning the heat conductivity equation in divergent form.

Proposition 7.1. Let D be the unit disk and let the matrix function A(z)
be generated by the Beltrami coefficient
z
(z) = k(|z|) (7.4)
z
in accordance with formula (2.8) where
p
k(t) = 2 (t) i (t) 1 2 (t) (7.5)

and (t), 0 t < 1, stands for an arbitrary measurable real-valued function,

such that |(t)| q < 1. Then the fundamental solution to the inhomogeneous
anisotropic heat equation

ut a2 div (A(z)u) = 0, z D, t > 0, (7.6)

has the representation

|z|2
 
1
u(z, t) = exp , z D, t > 0. (7.7)
4a2 t 4a2 t

Proof. We are looking for a solution in the form u(z, t) = T ((z), t), where (z)
stands for a quasiconformal automorphism of the unit disk D agreed with the
matrix function A(z). Since the mapping w = (z) is volume preserving, we see
that its Jacobian is identically equal to the unit. Then, by Theorem 7.1, T (w, t)
will satisfy the equation

Tt a2 T = 0, w D, t > 0. (7.8)

It is a classical result that the fundamental solution to the inhomogeneous heat

equation (7.8) has the representation
|w|2
 
1
T (w, t) = exp , w D, t > 0. (7.9)
(2a t)2 4a2 t
 32

Since in our case |(z)| = |z|, we arrive at the required conclusion.

Remark 7.1. In a similar way we can study the well-known semi-linear

combustion equation in an anisotropic medium

ut a2 div [A(z) u] = eu (7.10)

by means of reducing its weak solution to the composition of a weak solution in

the isotropic case, see e.g. [4] and [30] and the references therein,

Tt a2 J 1 T = eT (7.11)

with a quasiconformal mapping generated by A. As above, J stands for the

Jacobian of the mapping .

8 Boundary value problems

The Factorization Theorem 4.1 allows us, in particular, to reduce the study of
the boundary value problems, say the Dirichlet problem

div (A(z)u) = f (u) if z ,
(8.1)
u = , if z

in a Jordan domain with A M 22 () and continuous boundary function

to the study of the following Dirichlet problem in the unit disk

T = J(w)f (T ) if |w| < 1,
(8.2)
T = , if |w| = 1

with the corresponding boundary function = 1 where is a quasicon-

formal mapping of onto the unit disk D agreed with A, and J is the Jacobian
of 1 . First of all, note that exists by the Measurable Riemann mapping
theorem, see e.g. Theorem V.B.3 in [1] and Theorem V.1.3 in [25].
The existence and continuity of the boundary function in the case of an
arbitrary Jordan domain is a fundamental result of the theory of the boundary
behavior of conformal and quasiconformal mappings. Namely,

1 = h g
 33

where g stands for a quasiconformal automorphism of the unit disk D and h is a

conformal mapping of D onto . It is known that g can be extended to a home-
omorphism of D onto itself, see e.g. Theorem I.8.2 in [25]. Moreover, by the
well-known Caratheodory-Osgood-Taylor theorem on the boundary correspon-
dence under conformal mappings, see [8] and [28], the mapping h is extended to
a homeomorphism of D onto . Thus, the function is well defined and really
continuous on the unit circle.

References
[1] Ahlfors L.V., Lectures on quasiconformal mappings. Princeton, N.J.: Van Nos-
trand, 1966; Reprinted by Wadsworth Ink. Belmont, 1987.

[2] Astala K., Iwaniec T., Martin G., Elliptic partial differential equations and quasi-
conformal mappings in the plane. Princeton Mathematical Series, 48. Princeton
University Press, Princeton, NJ, 2009. xviii+677 pp.

[3] Bandle C., Marcus M. Asymptotic behavior of sulutions and their derivative for
semilinear elliptic problems with blow-up on the boundary, Ann. Inst. H. Poincare
Analyse Non Lineaire 12 (1995), no. 2, 155171.

[4] Barenblatt G.I., Zeldovic Ja.B., Librovich V.B., Mahviladze, G.M. Matematich-
eskaya teoriya goreniya i vzryva. "Nauka, Moscow, 1980. 479 pp. (Russian);
Transl. in "The mathematical theory of combustion and explosions." Consultants
Bureau [Plenum], New York, 1985. xxi+597 pp. ISBN: 0-306-10974-3

[5] Bieberbach, L. u = eu und die automorphen Funktionen. Math. Ann. 77 (1916),

173212.

[6] Bojarski B.V., Generalized solutions of a system of differential equations of the

first order of the elliptic type with discontinuous coefficients. Mat. Sb. 43 (85)
(1957), 451503.

[7] Bojarski B., Gutlyanskii V., Martio O., Ryazanov V. Infinitesimal geometry of
quasiconformal and bi-lipschitz mappings in the plane. EMS Tracts in Mathemat-
ics, 19. European Mathematical Society (EMS), Zurich, 2013. x+205 pp. ISBN:
978-3-03719-122-4.

[8] Caratheodory C. Uber die gegenseitige Beziehung der Rander bei der konformen
Abbildungen des Inneren einer Jordanschen Kurve auf einen Kreis. Math. Ann.
73 (1913), 305-320.

[9] Diaz J.I. Nonlinear partial differential equations and free boundaries. Vol. I, El-
liptic equations. Research Notes in Mathematics, 106, 1985.

[10] Goldshtein V., Gurov L., Romanov A. Homeomorphisms that induce monomor-
phisms of Sobolev spaces. Israel J. Math. 91 (1995), no. 13, 3160.
 34

[11] Goldshtein V., Ukhlov A. About homeomorphisms that induce composition opera-
tors on Sobolev spaces. Complex Var. Elliptic Equ. 55 (2010), no. 810, 833845.

[12] Goluzin G.M. Geometric theory of functions of a complex variable. Amer. Math.
Soc., Providence, Rhode Island, 1969.

[13] Gutlyanski V.Ya., Martio, O. Conformality of a quasiconformal mapping at a

point, J. Anal. Math. 91 (2003), 179192.

[14] Gutlyanski V., Martio, O. Rotation estimates and spirals. Conform. Geom. Dyn.
5 (2001), 620.

[15] Gutlyanskii, V., Nesmelova, O., Ryazanov, V. Semilinear equations in the plane
and quasiconformal mappings, Dopov. Nats. Akad. Nauk Ukr. Mat. Prirodozn.
Tekh. Nauki (2017), no. 1, 8-16.

[16] Gutlyanskii, V., Nesmelova, O., Ryazanov, V. On a model semilinear elliptic

equation in the plane, Ukr. Mat. Visn. 13 (2016), no. 1, 91105; transl. in J.
Math. Sci. (US) 220 (2017), no. 5, 603614.

[17] Gutlyanskii, V.Ya.; Ryazanov, V.I. On the theory of the local behavior of quasi-
conformal mappings. Izv. Ross. Akad. Nauk Ser. Mat. 59 (1995), no. 3, 3158
(Russian); transl. in Izv. Math. 59 (1995), no. 3, 471498.

[18] Gutlyanskii V.Ya., Ryazanov V.I., The geometric and topological theory of func-
tions and mappings. Kiev, Naukova Dumka, 2011, 425 pp.

[19] Gutlyanskii V., Ryazanov V., Srebro U., Yakubov E., The Beltrami Equation:
A Geometric Approach. Developments in Mathematics, 26, Springer, New York,
2012. xiv+301 pp. ISBN: 978-1-4614-3190-9

[20] Keller, J. B. On solutions of u = f (u). Comm. Pure Appl. Math. 10 (1957),

503510.

[21] Kondratiev, V.A., Landis E.M. Semilinear equations of the second order with
non-negative characteristic form. Mat. Zam. 44 (1988), no. 4, 457468.

[22] Kovalevskii, A. A.; Skrypnik, I. I.; Shishkov, A. E. Singular solutions of nonlinear

elliptic and parabolic equations. De Gruyter Series in Nonlinear Analysis and
Applications, 24, 2016, 435 pp.

[23] Marcus, M., Veron, L. Nonlinear second oder elliptic equations involving mea-
sures. De Gruyter, Series in nonlinear analysis and applications, 21, Berlin
Boston, 2014. 248 pp. ISBN: 978-3-11-030515-9.

[24] Ladyzhenskaya, O.A., Uraltseva, N.N., Linear and quasilinear elliptic equations,
Nauka, Moscow, 1964, 358 pp.

[25] Lehto O., Virtanen K.I., Quasiconformal mappings in the plane, 2-nd Edition,
SpringerVerlag, Berlin - Heidelberg - New York, 1973.

[26] Neklyudov A.V. Behavior of solutions to Gauss-Bieberbach-Rademacher equation

on plane. Ufa Math. J. 6 (2014), no. 3, 8594.
 35

[27] Oleinik O.A. The equation u + k(x)eu = 0. Uspehi Mat. Nauk 33 (1978), no. 2
(200), 203204.

[28] Osgood W., Taylor E. Conformal transformations on the boundaries of their re-
gions of definition. Trans. Amer. Math. Soc. 14 (1913), 277-298.

[29] Osserman R. On the inequality u f (u). Pacic J. Math. 7 (1957), 16411647.

[30] Pokhozhaev S.I. On an equation of combustion theory. Mat. Zametki 88 (2010),

no. 1, 5362 (Russian); transl. in Math. Notes 88 (2010), no. 12, 48-56.

[31] Reimann H. M., Rychener T. Funktionen beschrankter mittlerer Oszillation. (Ger-

man) Lecture Notes in Mathematics, 487. Springer-Verlag, Berlin-New York,
1975. vi+141 pp.

[32] Reshetnyak Yu.G. Prostranstvennye otobrazheniya s ogranichennym iskazheniem.

(Russian) Nauka, Novosibirsk, 1982. 278 pp.; Space mappings with bounded dis-
tortion. Transl. Math. Mon., 73. AMS, Providence, RI, 1989. xvi+362 pp.

[33] Sobolev S.L. On some transformation groups of an n-dimensional space. Dokl.

Akad. Nauk SSSR 32 (1941), no. 6, 380382 (Russian).

[34] Ukhlov A. Mappings that generate embeddings of Sobolev spaces. Sibirsk. Mat.
Zh. 34 (1993), no. 1, 185192 (Russian); transl. in Siberian Math. J. 34 (1993),
no. 1, 165171.

[35] Vekua I.N. Some properties of solutions of Gauss equation // Trudy Mat. Inst.
Steklov. 64 (1961), 58.

[36] Vodopyanov S. K. On the regularity of mappings inverse to the Sobolev mapping.

Mat. Sb. 203 (2012), no. 10, 332 (Russian); transl. in Sb. Math. 203 (2012),
no. 910, 13831410.

[37] Vodopyanov S.K., Evseev N.A. Isomorphisms of Sobolev spaces on Carnot groups
and quasi-isometric mappings. Sibirsk. Mat. Zh. 55 (2014), no. 5, 10011039
(Russian); transl. in Sib. Math. J. 55 (2014), no. 5, 817848.

[38] Vodopjanov S. K., Goldshtein V. M. Lattice isomorphisms of the spaces Wn1 and
quasiconformal mappings. Sibirsk. Mat. Zh. 16 (1975), 224246 (Russian).

[39] Vodopyanov S.K., Ukhlov A. Sobolev spaces and (P, Q)quasiconformal mappings
of Carnot groups. // Sib. Mat. Zh., 39 (1998), no. 4, 665682 (Russian); transl.
in Siberian Math. J. 39 (1998), no. 4, 665682.

Institute of Applied Mathematics and Mechanics of National Academy

of Sciences of Ukraine, 84100, Ukraine, Slavyansk, Dobrovolskogo str. 1,
Email: [email protected], [email protected], [email protected]