Stability of equilibriums and bifurcation analysis of

two-dimensional autonomous competitive

Lotka-Volterra dynamical system
arXiv:2410.18119v1 [math.GM] 10 Oct 2024

Danijela Branković and Marija Mikić

Abstract. A detailed analysis of the stability of equilibriums and bi-

furcations of the two-dimensional autonomous competitive Lotka-Volterra
dynamical system is performed. Necessary and sufficient conditions are de-
termined for equilibriums (without the origin) to be asymptotically stable
or unstable on [0, +∞)2 . Necessary and sufficient conditions are determined
so that the observed dynamical system has no equilibriums in (0, +∞)2 .
All results are presented in five tables and five figures. We also found that
four transcritical bifurcations occur in the observed dynamical system if it
is analyzed on R2 .
2020 Mathematics Subject Classification: 34A34; 34C23; 37C75.
Keywords and phrases: Lotka-Volterra; dynamical systems; stability;
equilibriums; bifurcations.

1 Introduction
In this paper, we analyze the stability of equilibriums and bifurcations of a
two-dimensional autonomous competitive Lotka-Volterra dynamical system

x˙1 = x1 (b1 − a11 x1 − a12 x2 ) ,

(1.1)
x˙2 = x2 (b2 − a21 x1 − a22 x2 ) ,

where xi = xi (t) is the density of species i at a given time t, bi is the

inherent growth rate of species i, aij is the effect that species j has on
species i, for i, j ∈ {1, 2} and i ̸= j, while for i = j we have self-interacting
terms aii , i ∈ {1, 2}. In the absence of competition, the observed dynamical
system (1.1) consists of two logistic equations (one for each species), with
carrying capacity bi /aii , for i ∈ {1, 2}. Since the dynamical system (1.1) is
competitive, we assume that aij > 0 and bi > 0, for every i, j ∈ {1, 2}. The
reason for this is the assumed harmfulness of interspecies interactions, which
directly implies aij > 0, for every i, j ∈ {1, 2} and i ̸= j. Moreover, the self-
regulation of each species implies aii > 0, for every i ∈ {1, 2}. Moreover
bi > 0, because it is also assumed that the inherent growth rate of species i

1
is positive in the absence of competition, unless xi has the value bi /aii (its
carrying capacity), for i ∈ {1, 2}.
Competitive dynamical systems were introduced by Lotka and Volterra (see
[6, 9]). The stability of equilibriums of the autonomous Lotka-Volterra com-
petitive model is analyzed in various literature (see, e.g. [4, 8, 5]). In [10]
relationship between the coefficients aij and bi can be found so that the
system (1.1), which has no equilibrium in (0, +∞)2 , has an asymptotically
stable equilibrium on one axis, while an unstable equilibrium lies on the other
axes. These conditions are also generalized for the n-dimensional case of the
system (1.1) in [10], and [2]. Other papers dealing with suitable conditions
for the two-dimensional and n-dimensional non-autonomous Lotka-Volterra
competition model are [1] and [3], respectively.
The importance of this work lies in the fact that all possible cases concerning
the stability of equilibriums of the autonomous Lotka-Volterra competitive
model are discussed here in detail, which to our knowledge has not been
done anywhere else in the literature in this way. The new results of this
paper concern the improvement of several results from [10], specifically the
determination of necessary and sufficient conditions for equilibriums (with-
out the origin) to be asymptotically stable or unstable in [0, +∞)2 , as well
as the determination of necessary and sufficient conditions for the observed
dynamical system to have no equilibriums in (0, +∞)2 . Furthermore, we
have concluded that four transcritical bifurcations occur in the observed
dynamical system if it is analyzed on R2 , and we have determined the con-
ditions under which these bifurcations occur. To our knowledge, these are
also new results of this paper.
In Section 2 we analyze in detail the stability of equilibriums of the sys-
tem (1.1). We assume that xi ≥ 0 and that the main determinant or the
minor determinants of the system (1.1) can be zero. This leads to some
non-hyperbolic equilibriums and bifurcations. We note that in [7] only the
stability of equilibriums of the system (1.1) for xi > 0 and for all non-zero
determinants of this system was studied. This analysis can also be found in
our paper in Section 2.1.1, but with a new notation that is more suitable for
generalization to higher dimensions. In Section 3 we have improved several
two-dimensional versions of the theorems from the paper [10], in particular
with respect to sufficient and necessary conditions for equilibriums of the
system (1.1) lying on the axes to be asymptotically stable or an unstable
equilibriums. In Section 4 we present a bifurcation analysis of the dynamical
system (1.1).
First we introduce the following notation. For the determinant of the matrix
of the system (1.1) and the corresponding minors, we introduce the following
notation
a11 a12 b a a b
d12 = , d122 = − 1 12 , and d112 = 11 1 . (1.2)
a21 a22 b2 a22 a21 b2

2
From (1.2) we can conclude that the number of different indices of the de-
terminant indicates the order of this determinant as well as the elements on
its main diagonal. The indices of the determinant d12 of the matrix of the
system (1.1) show that it is the second-order determinant whose elements on
the main diagonal are a11 and a22 . Furthermore, the indices of the minors
dijk , where i, j, k ∈ {1, 2} denote the second-order determinant whose ele-
ments on the main diagonal are aij (position (1, 1)) and bk (position (2, 2)).
This notation is very useful for calculating the coordinates of the equilib-
riums of the system and their eigenvalues, as well as for generalizations to
higher dimensions.
The following notation applies to Tables 1-5: U denotes an unstable equi-
librium, AS represents an asymptotically stable equilibrium, SS stands for
a semi-stable equilibrium and NI stands for a non-isolated equilibrium.
In all figures in our paper, we have colored the points representing unstable
equilibriums in yellow, asymptotically stable equilibriums are colored red,
semi-stable equilibriums are colored orange and non-isolated equilibriums
are colored pink.

2 Stability analysis of the equilibriums

We consider two cases, depending on whether d12 is zero or not.

2.1 Case d12 ̸= 0

The number of equilibriums and their stability depend on the signs of d12 ,
d112 and d122 . We analyze the following cases.

2.1.1 Case d112 ̸= 0 and d122 ̸= 0

If sgn (d12 ) = sgn (d112 ) = −sgn (d122 ), we have four different equilibri-
ums of system (1.1), labelled E0 (0, 0), E1 (b1 /a11 , 0), E2 (0, b2 /a22 ) and
E12 (−d122 /d12 , d112 /d12 ). Otherwise we would have three equilibriums of
system (1.1), E0 , E1 and E2 , since equilibrium E12 would not lie in the first
quadrant. We note that the conditions d112 ̸= 0 and d122 ̸= 0 ensure that
the eigenvalues of the Jacobian matrix in our equilibriums Ei , i ∈ {1, 2, 12}
have non-zero real parts, as we will see in the following text. Consequently,
our equilibriums are hyperbolic and therefore we can discuss their stability
in the framework of the corresponding linearization of the dynamical sys-
tem (1.1), because the Hartman–Grobman theorem provides us topological
equivalence between a nonlinear dynamical system and its linearization in
the neighbourhood of the hyperbolic equilibrium. Now we investigate the
stability of all equilibriums.
The equilibrium E0 . The eigenvalues of the Jacobian matrix at E0 are
λi = bi , for i ∈ {1, 2}. Since bi > 0 for every i ∈ {1, 2}, we conclude that E0

3
is an unstable node (unstable equilibrium).
The equilibrium E1 . The eigenvalues of the Jacobian matrix at E1 are
λ1 = −b1 , λ1 < 0 and λ2 = d112 /a11 . If d112 < 0, then E1 is a stable
node (asymptotically stable equilibrium). Otherwise, i.e. if d112 > 0, E1 is
therefore a saddle (unstable equilibrium).
The equilibrium E2 . We note that the eigenvalues of the Jacobian matrix
for E2 are λ1 = −d122 /a22 and λ2 = −b2 , λ2 < 0. We conclude that E2 is
a stable node (asymptotically stable equilibrium) if d122 > 0. If d122 < 0,
then E2 is a saddle point (unstable equilibrium).
The equilibrium E12 . The situation is a little complicated in this case.
As we have already mentioned, the coordinates of E12 (−d122 /d12 , d112 /d12 )
must be positive in order to E12 lie in the first quadrant, i.e. it must be
sgn (d12 ) = sgn (d112 ) = −sgn (d122 ). If we denote the coordinates of E12 by
x1 := −d122 /d12 and x2 := d112 /d12 , the Jacobian determinant at E12 is

b1 − 2a11 x1 − a12 x2 −a12 x1

det (J (E12 )) =
−a21 x2 b2 − a21 x1 − 2a22 x2
(2.1)
a11 a12
= (−1)2 x1 x2 = x1 x2 d12 ,
a21 a22

since b1 − 2a11 x1 − a12 x2 = −a11 x1 and b2 − a21 x1 − 2a22 x2 = −a22 x2 .

Furthermore
λ1 λ2 = det (J (E12 )) = x1 x2 d12 ,
λ1 + λ2 = tr (J (E12 )) = − (a11 x1 + a22 x2 ) , (2.2)

from which we obtain that the characteristic polynomial P12 (λ) of the ma-
trix J (E12 ) is P12 (λ) = λ2 + (x1 a11 + x2 a22 ) λ + x1 x2 d12 . From (2.2) we
conclude that E12 is a stable node (asymptotically stable equilibrium) if
d12 > 0 and d112 > 0 and d122 < 0. If d12 < 0 and d112 < 0 and d122 > 0,
E12 is a saddle (unstable equilibrium).
To summarize, E0 is an unstable node (unstable equilibrium), while the
stability of equilibrium E1 depends solely on the sign of d112 , the stability
of equilibrium E2 depends solely on the sign of d122 , while the existence and
stability of equilibrium E12 depend on the signs of d12 , d112 and d122 . The
results are shown in Table 1.
The case from Table 1, in which d12 > 0, d112 < 0, d122 > 0 is not possible.
If this were possible, then we would have

a11 a22 > a12 a21 , a11 b2 < a21 b1 , a12 b2 > a22 b1 . (2.3)

If we multiply the first inequality from (2.3) by b2 /a22 and use the third
inequality from (2.3), we get a11 b2 > a12 b2 a21 /a22 > a21 b1 , which contradicts
the second inequality from (2.3). The case that d12 < 0, d112 > 0, d122 < 0
is also not possible.

4
 d112 > 0 d112 > 0 d112 < 0 d112 < 0
d12 eq.
d122 > 0 d122 < 0 d122 > 0 d122 < 0
E0 U (u. node) U (u. node) U (u. node)
not
E1 U (saddle) U (saddle) AS (s. node)
d12 > 0 possible
E2 AS (s. node) U (saddle) U (saddle)
case
E12 not exist AS (s. node) not exist
E0 U (u. node) U (u. node) U (u. node)
not
E1 U (saddle) AS (s. node) AS (s. node)
d12 < 0 possible
E2 AS (s. node) AS (s. node) U (saddle)
case
E12 not exist U (saddle) not exist

Table 1. Stability analysis of the equilibriums of the system (1.1).

We conclude from Table 1 that cases with d12 > 0, d112 > 0, d122 > 0 and
d12 < 0, d112 > 0, d122 > 0, have the same qualitative dynamical properties,
as well as the cases d12 > 0, d112 < 0, d122 < 0 and d12 < 0, d112 < 0,
d122 < 0. Therefore, in this section we have four qualitatively different
phase portraits of the system (1.1). They are shown in Figures 1 and 2.

4 4

3 3

2 2

1 1

0 0

0 1 2 3 4 0 1 2 3 4

(a) Case d12 > 0, d112 > 0, d122 < 0. (b) Case d12 > 0, d112 > 0, d122 > 0.
x˙1 = x1 (3 − x1 − x2 ) , x˙1 = x1 (2 − x1 − x2 ) ,
x˙2 = x2 (4 − x1 − 2x2 ) . x˙2 = x2 (6 − x1 − 2x2 ) .

Figure 1: Phase portraits for the cases d12 > 0, d112 > 0, d122 < 0 and
d12 > 0, d112 > 0, d122 > 0, respectively.

In Figures 1a and 1b, the equilibrium E0 at the origin is an unstable node,

while the equilibrium E1 is a saddle. In Figures 1a and 2a, equilibrium E2
is a saddle, while in Figures 1b and 2b it is a stable node. Equilibrium E12
in Figure 1a is a stable node, while in Figure 2b is a saddle.

5
3.0 1.0

2.5
0.8

2.0
0.6

1.5

0.4

1.0

0.2
0.5

0.0 0.0

0 1 2 3 4 0.0 0.5 1.0 1.5

(a) Case d12 > 0, d112 < 0, d122 < 0. (b) Case d12 < 0, d112 < 0, d122 > 0.
x˙1 = x1 (6 − 2x1 − x2 ) , x˙1 = x1 (1 − x1 − 2x2 ) ,
x˙2 = x2 (2 − x1 − x2 ) . x˙2 = x2 (3 − 4x1 − 5x2 ) .

Figure 2: Phase portraits for the cases d12 > 0, d112 < 0, d122 < 0 and
d12 < 0, d112 < 0, d122 > 0, respectively.

2.1.2 Case d112 ̸= 0 and d122 = 0

Since d122 = 0, we directly conclude that the first coordinate of E12 is
zero and that a12 = a22 b1 /b2 . Substituting a12 = a22 b1 /b2 into d12 , we
conclude that d12 = a22 d112 /b2 . Furthermore, the second coordinate of E12
is d112 /d12 = b2 /a22 , b2 /a22 > 0, from which it follows that the equilibrium
E12 coincides with E2 and that sgn (d112 ) = sgn (d12 ). The equilibriums E0
and E1 are hyperbolic and their nature is the same as in Section 2.1.1 (see
Table 1), considering only the cases where sgn (d112 ) = sgn (d12 ). Since the
equilibrium E2 is not hyperbolic (λ1 = −d122 /a22 = 0), we cannot determine
its nature using the Hartman-Grobman theorem. Instead, we will determine
its stability by analyzing the phase trajectories around this equilibrium. For
this purpose, we first determine the nullclines of the system (1.1).
We obtain that x1 -nullclines are the lines x1 = 0 and x2 = (b1 − a11 x1 ) /a12 ,
while x2 -nullclines are x2 = 0 and x2 = (b2 − a21 x1 ) /a22 . Note that x1 -
nullcline x1 = 0 and x2 -nullcline x2 = 0 are invariant. The vector field
through x1 -nullcline x1 = 0 is vertical. If we also substitute x1 = 0 into
the second equation in (1.1), we obtain x˙2 = x2 (b2 − a22 x2 ), from which we
conclude that the direction of the vector field through the nullcline x1 = 0 is
vertically upwards (↑) for 0 < x2 < b2 /a22 , while for x2 > b2 /a22 the vector
field is vertically downwards (↓).
The vector field is horizontal through x2 -nullcline x2 = 0. If we insert x2 = 0
into the first equation of the system (1.1), we obtain x˙1 = x1 (b1 − a11 x1 ).
Consequently, the vector field is directed horizontally to the right (→) by
the nullcline x2 = 0 for 0 < x1 < b1 /a11 , while for x1 > b1 /a11 the vector

6
field is directed horizontally to the left (←).
Using that d122 = 0 and x1 > 0, we can easily see that the x1 -nullcline
x2 = (b1 − a11 x1 ) /a12 is above the x2 -nullcline x2 = (b2 − a21 x1 ) /a22 if
and only if d12 < 0. We determine that the intersection point of these two
nullclines is E2 (0, b1 /a12 ). First, we analyze the nature of the equilibrium
E2 when d12 < 0.
Since we conclude that x˙1 = 0 for x2 = (b1 − a11 x1 ) /a12 , it follows that
the vector field through the nullcline x2 = (b1 − a11 x1 ) /a12 is vertical. If
we substitute x2 = (b1 − a11 x1 ) /a12 into the second equation in system
(1.1) (the substitution only applies to the term x2 in brackets), we get
x˙2 = x1 x2 d12 /a12 . Therefore, the direction of the vector field through the
nullcline x2 = (b1 − a11 x1 ) /a12 is vertically downwards (↓).
The vector field through the nullcline x2 = (b2 − a21 x1 ) /a22 is horizontal. If
we insert x2 = (b2 − a21 x1 ) /a22 into the first equation of the system (1.1),
we derive x˙1 = −x1 2 d12 /a22 , from which we conclude that the direction of
the vector field is directed horizontally to the right (→) by the nullcline
x2 = (b2 − a21 x1 ) /a22 .
The sketch of the direction of the vector field through the nullclines implies
that the equilibrium E2 is unstable, which we will prove.
Consider the subset

G1 = {(x1 , x2 ) ∈ R2 | (b2 − a21 x1 ) /a22 < x2 < (b1 − a11 x1 ) /a12 , x2 > 0}.

We note that G1 is invariant, i.e. any trajectory that starts at some point
of G1 remains in G1 . Let M (x1 , x2 ) be an arbitrary point from the area G1 .
Therefore, x1 > 0 and b1 − a11 x1 − a12 x2 > 0 and from the first equation
of the system (1.1) we conclude that x˙1 > 0, i.e. x1 = x1 (t) increases over
time for any point in G1 . Consequently, there is a neighbourhood OE2 of
the equilibrium E2 such that for every point M (x1 , x2 ) ∈ OE2 ∩ G1 the
trajectory starting at M (x1 , x2 ) never returns to OE2 ∩ G1 . We conclude
that E2 (0, b2 /a22 ) is an unstable equilibrium for d12 < 0.
For d12 > 0 we will prove that the equilibrium E2 is semi-stable. More
precisely, we will prove that E2 is asymptotically stable in (0, +∞)2 using
Lyapunov stability theorem and that it is unstable in (−∞, 0) × (0, +∞) by
analyzing trajectories around E2 .
Let suppose that the Lyapunov function has the form

V (x1 , x2 ) = x1 a22 x2 −a12 ,

(2.4)
V̇ (x1 , x2 ) = −d12 x1 a22 +1 x2 −a12 .

From (2.4) we conclude that V is differentiable on (0, +∞)2 , V (E2 ) = 0

and V (x1 , x2 ) > 0 for every (x1 , x2 ) ∈ (0, +∞)2 . Since d12 > 0, then
V̇ (x1 , x2 ) < 0 for every (x1 , x2 ) ∈ (0, +∞)2 . Therefore, the function V
from (2.4) is a Lyapunov function for E2 and according to the Lyapunov
stability theorem, E2 is an asymptotically stable equilibrium on (0, +∞)2 .

7
Consider the subset

G2 = {(x1 , x2 ) ∈ R2 | (b2 − a21 x1 ) /a22 < x2 < (b1 − a11 x1 ) /a12 }.

We note that, since here d12 > 0 and x1 < 0, the line x2 = (b2 − a21 x1 ) /a22
is indeed below the line x2 = (b1 − a11 x1 ) /a12 . Moreover, we notice that
x2 > b2 /a22 , where b2 /a22 = d112 /d12 is exactly the x2 -coordinate of the
equilibrium E2 .
Similarly to the case d12 < 0, we find that the direction of the vector field of
(1.1) by the nullcline x2 = (b1 − a11 x1 ) /a12 is vertically upwards (↑) if and
only if x1 x2 > 0. The vector field by the nullcline x2 = (b2 − a21 x1 ) /a22 is
directed horizontally to the left (←).
The directions of the vector field through the nullclines suggest that the
equilibrium E2 is unstable in G2 , which we will prove.
Let M (x1 , x2 ) be an arbitrary point from the area G2 . Therefore, x1 < 0
and x2 < (b1 − a11 x1 ) /a12 , so we conclude from the first equation of the
system (1.1) that x˙1 < 0, i.e. x1 = x1 (t) decreases with time for every point
in G2 . Consequently, there exists neighbourhood OE2 of the equilibrium E2
such that for every point M (x1 , x2 ) ∈ OE2 ∩ G2 a trajectory starting at
M (x1 , x2 ) never returns to OE2 ∩ G2 . We conclude that E2 (0, b2 /a22 ) is
an unstable equilibrium in (−∞, 0) × (0, +∞). Overall, E2 is a semi-stable
equilibrium.
In this section we have two cases that are not possible. The first is for
d12 > 0, d112 < 0, d122 = 0. If we assume that it is possible, then

a11 a22 > a12 a21 , a11 b2 < a21 b1 , a12 b2 = a22 b1 . (2.5)

If we multiply the first inequality from (2.5) by b2 /a22 and apply the third
inequality from (2.5), we obtain, as before, a11 b2 > a12 b2 a21 /a22 = a21 b1 ,
which contradicts the second inequality from (2.5). The case
d12 < 0, d112 > 0, d122 = 0 is also not possible.

d12 eq. d112 > 0 and d122 = 0 d112 < 0 and d122 = 0
E0 U (u. node)
d12 > 0 E1 U (saddle) not possible case
E2 SS
E0 U (u. node)
d12 < 0 E1 not possible case AS (s. node)
E2 U

Table 2. Stability analysis of the equilibriums of the system (1.1).

All results are listed in Table 2. To conclude this chapter, we have two
different phase portraits of the dynamical system (1.1), shown in Figure 3.

8
 4
1.5

1.0

0.5

-2

-4 0.0

-4 -2 0 2 4 0.0 0.5 1.0 1.5 2.0 2.5 3.0

(a) Case d12 > 0, d112 > 0, d122 = 0. (b) Case d12 < 0, d112 < 0, d122 = 0.
x˙1 = x1 (2 − x1 − x2 ) , x˙1 = x1 (2 − x1 − 2x2 ) ,
x˙2 = x2 (4 − x1 − 2x2 ) . x˙2 = x2 (1 − x1 − x2 ) .

Figure 3: Phase portraits for the cases d12 > 0, d112 > 0, d122 = 0 and
d12 < 0, d112 < 0, d122 = 0, respectively.

In both figures, the equilibrium E0 at the origin is an unstable node, while

in Figure 3a the equilibrium E1 is a saddle point and the equilibrium E2 is
a semi-stable, while in Figure 3b the equilibrium E1 is asymptotically stable
and the equilibrium E2 is an unstable equilibrium.

2.1.3 Case d112 = 0 and d122 ̸= 0

Similarly as in the Section 2.1.2, from d112 = 0 we conclude that the equilib-
rium E12 coincides with E1 and that sgn (d122 ) = −sgn (d12 ). The equilibri-
ums E0 and E2 are hyperbolic and their nature is the same as in Section 2.1.2
(Table 1), considering only the cases when sgn (d122 ) = −sgn (d12 ). The
equilibrium E1 is nonhyperbolic (λ2 = d112 /a11 = 0), and we will determine
its stability by analyzing the phase trajectories around this equilibrium.
However, we will only present the final results here, since the procedure is
similar to the one presented in Section 2.1.2.
If d12 < 0, then equilibrium E1 will be unstable (proof of this statement
can be obtained using nullclines, similarly as in Section 2.1.2). If d12 > 0,
then the equilibrium E1 is semi-stable. The equilibrium E1 is asymptoti-
cally stable on (0, +∞)2 . To prove this statement, the reader can use the
Lyapunov function V (x1 , x2 ) = x1 −a21 x2 a11 . The equilibrium E1 is unstable
in (0, +∞)×(−∞, 0). To prove this statement, the reader can use nullclines,
similar to Section 2.1.2. The results are shown in the Table 3. Therefore, in
this section we have two different phase portraits of the dynamical system
(1.1), which are shown in Figure 4.

9
 d12 eq. d112 = 0 and d122 > 0 d112 = 0 and d122 < 0
E0 U (u. node)
d12 > 0 E1 not possible case SS
E2 U (saddle)
E0 U (u. node)
d12 < 0 E1 U not possible case
E2 AS

Table 3. Stability analysis of the equilibriums of the system (1.1).

0.6

1.0

0.4

0.8

0.2

0.6

0.0

0.4

-0.2

0.2

-0.4

0.0

-0.6
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5

(a) Case d12 > 0, d112 = 0, d122 < 0. (b) Case d12 < 0, d112 = 0, d122 > 0.
x˙1 = x1 (1 − x1 − 2x2 ) , x˙1 = x1 (1 − x1 − 4x2 ) ,
x˙2 = x2 (1 − x1 − 4x2 ) . x˙2 = x2 (1 − x1 − 2x2 ) .

Figure 4: Phase portraits for the cases d12 > 0, d112 = 0, d122 < 0 and
d12 < 0, d112 = 0, d122 > 0, respectively.

In both figures, the equilibrium E0 at the origin is an unstable node, while

in Figure 4a, the equilibrium E1 is a semi-stable and the equilibrium E2 is
a saddle, while in Figure 4b the equilibrium E1 is an unstable equilibrium
and the equilibrium E2 is a stable node.

2.1.4 Case d122 = 0 and d112 = 0

This case is not possible. This can be shown in a similar way as in Sec-
tion 2.1.2 (the proof differs if d12 > 0 or d12 < 0).

2.2 Case d12 = 0

We distinguish between two cases, when at least one minor is not equal to
zero and when both minors are equal to zero.

10
2.2.1 Case d122 ̸= 0 or d112 ̸= 0
The equilibrium E12 does not exist in this case, since the system

a11 x1 + a12 x2 = b1 ,
(2.6)
a21 x1 + a22 x2 = b2 ,

has no solutions. Accordingly, the system (1.1) has three equilibriums, E0 ,

E1 and E2 . If d122 ̸= 0 and d112 ̸= 0, then all three equilibriums are
hyperbolic and the conclusions regarding their stability are given in Table 4.

d12 = 0 d12 = 0 d12 = 0

d12 = 0
eq. d112 > 0 d112 d122 = 0 d112 < 0
d112 d122 < 0
d122 > 0 d112 2 +d122 2 ̸= 0 d122 < 0
E0 U (u. node) not not U (u. node)
E1 U (saddle) possible possible AS (s. node)
E2 AS (s. node) cases cases U (saddle)

Table 4. Stability analysis of the equilibriums of the system (1.1).

Here the cases for d12 = 0 and d112 d122 < 0 are impossible and the cases for
d12 = 0, d112 d122 = 0 and d112 2 + d122 2 ̸= 0 are also impossible. The proofs
are similar to those in Section 2.1.2 and so we omit them here.
We note that this section does not provide us with any qualitatively new
phase portraits of the dynamical system (1.1), since we already had these
qualitatively equal phase portraits before. For d12 = 0, d112 > 0, d122 > 0,
we have the same phase portrait as for d12 > 0, d112 > 0, d122 > 0 and
d12 < 0, d112 > 0, d122 > 0 from Section 2.1.1. Furthermore, for d12 = 0,
d112 < 0, d122 < 0, we have the same phase portrait as for the d12 > 0,
d112 < 0, d122 < 0 and d12 < 0, d112 < 0, d122 < 0 from Section 2.1.1.

2.2.2 Case d122 = 0 and d112 = 0

If both minors are equal to zero, the system (2.6) has infinitely many
solutions (x1 , x2 ) = ((b1 − a12 α) /a11 , α), where α ≥ 0 and b1 ≥ a12 α.
Consequently, the system (1.1) has infinitely many equilibriums, E0 and
E α = ((b1 − a12 α) /a11 , α), where 0 ≤ α ≤ b1 /a12 . We note that the equi-
libriums E α are non-isolated and that for α = 0 we conclude E 0 = E1 ,
while for α = b1 /a12 we obtain E b1 /a12 = E b2 /a22 = E2 . Therefore we
exclude these values for α in the following discussion.
The Jacobian matrix calculated for E α is given by
(b1 − a12 α) a12
 
− (b1 − a12 α) −
J (E α ) =  a11
a11 (b2 − 2a22 α) − a21 (b1 − a12 α)  . (2.7)
 
−a21 α
a11

11
From (2.7), using d112 = 0 and d12 = 0, we deduce that

λ1 λ2 = det (J (E α )) = 0,
λ1 + λ2 = tr (J (E α )) = a12 α − a22 α − b1 , (2.8)

from which we conclude that λ2 = − (b1 − a12 α) − a22 α, λ2 < 0, λ1 = 0,

since (b1 − a12 α) /a11 = x1 , x1 > 0 and α = x2 , x2 > 0. Thus, E α is
nonhyperbolic equilibrium for every 0 < α < b1 /a12 . E α is a nonhyperbolic
equilibrium for every 0 ≤ α ≤ b1 /a12 , i.e. x2 = (b1 − a11 x1 ) /a12 is a line
of equilibriums. E0 is a hyperbolic equilibrium, which is an unstable node,
just as in the previous sections. The phase portrait is shown in Figure 5.

0.8

0.6

●
●
●
●
0.4 ●
●
●
●
●
●
●
●
0.2 ●
●
●
●
●
●
●
●
0.0 ●

0.0 0.5 1.0 1.5

Figure 5: Case d12 = d112 = d122 = 0.

x˙1 = x1 (1 − x1 − 2x2 ), x˙2 = x2 (2 − 2x1 − 4x2 ).

The equilibrium E0 at the origin is unstable, while the equilibriums E α ,

where 0 ≤ α ≤ b1 /a12 , are non-isolated.

3 Necessary and sufficient conditions for stability

of equilibriums
Section 2 shows that there are nine qualitatively different phase portraits of
the dynamical system (1.1) which are shown in Figures 1-5, when we consider
the stability of equilibriums of the system (1.1) on their full neighbourhoods.
The signs of d12 , d112 and d122 as well as the equilibriums of the dynamical
system (1.1) are shown in Table 5. After analyzing Figures 1-5 and the data
in Table 5, we conclude that the phase portraits under serial numbers 2, 3
and 6 that are shown in Figures 3a, 1b and 4b have qualitatively the same
dynamical properties in [0, +∞)2 , i.e. that in [0, +∞)2 the equilibrium E1
is an unstable equilibrium and that the equilibrium E2 is asymptotically
stable.

12
 Serial
d12 d112 d122 E0 E1 E2 E12 Eα
No.
1 d12 >0 d112 >0 d122 <0 U U U AS /
2 d12 >0 d112 >0 d122 =0 U U SS / /
3 d12 >0 d112 >0 d122 >0 U U AS / /
4 d12 >0 d112 =0 d122 <0 U SS U / /
5 d12 >0 d112 <0 d122 <0 U AS U / /
6 d12 <0 d112 =0 d122 >0 U U AS / /
7 d12 <0 d112 <0 d122 =0 U AS U / /
8 d12 <0 d112 <0 d122 >0 U AS AS U /
9 d12 =0 d112 =0 d122 =0 U NI NI / NI

Table 5. Stability analysis of equilibriums of the dynamical system (1.1).

Moreover, the phase portraits under serial numbers 4, 5 and 7 that are shown
in Figures 4a, 2a and 3b have qualitatively the same dynamical properties in
[0, +∞)2 , i.e. the equilibrium E1 is asymptotically stable and the equilibrium
E2 is unstable.
In summary, we have nine qualitatively different phase portraits of the dy-
namical system (1.1) when we consider the stability of equilibriums of the
system (1.1) on their full neighbourhoods, while we have five qualitatively
different phase portraits when we analyze the stability of its equilibriums
on [0, +∞)2 . We note that it is interesting from mathematical point of view
to analyze the stability of the equilibriums of the system (1.1) on their full
neighbourhoods, although it lacks an appropriate physical interpretation
since the density of the species cannot be negative.
To summarize all the results from the Section 2, we formulate the following
theorems.

Theorem 3.1 The equilibrium E1 (b1 /a11 , 0) of the system (1.1) is asymp-
totically stable on [0, +∞)2 if and only if one of the following conditions is
satisfied:

(1) d112 < 0 and d122 < 0 (regardless of the sgn (d12 ));

(2) d12 < 0 and d112 < 0 and d122 = 0;

(3) d12 > 0 and d112 = 0 and d122 < 0;

(4) d12 < 0 and d112 < 0 and d122 > 0.

Theorem 3.1 is an extension of Theorem 2.1. from [10] for n = 2, where it

states (with a different notation than the one used here) that if d122 < 0
and d112 < 0, then E1 (b1 /a11 , 0) is asymptotically stable on (0, +∞)2 .
Now we formulate a similar theorem for the equilibrium E2 .

13
Theorem 3.2 The equilibrium E2 (0, b2 /a22 ) of the system (1.1) is asymp-
totically stable on [0, +∞)2 if and only if one of the following conditions is
satisfied:

(1) d112 > 0 and d122 > 0 (regardless of the sgn (d12 ));

(2) d12 > 0 and d112 > 0 and d122 = 0;

(3) d12 < 0 and d112 = 0 and d122 > 0;

(4) d12 < 0 and d112 < 0 and d122 > 0.

As a consequence of Theorem 3.1 and Theorem 3.2 and the fact that E0 is
an unstable node (unstable equilibrium) regardless of the signs of d12 , d112
and d122 , we formulate the following two theorems.

Theorem 3.3 The equilibrium E1 (b1 /a11 , 0) of the system (1.1) is unstable
if and only if one of the conditions (1), (2), (3) from Theorem 3.2 is satisfied,
or d12 > 0 and d112 > 0 and d122 < 0 is satisfied.

Theorem 3.4 The equilibrium E2 (0, b2 /a22 ) of the system (1.1) is unstable
if and only if one of the conditions (1), (2), (3) from Theorem 3.1 is satisfied,
or d12 > 0 and d112 > 0 and d122 < 0 is satisfied.

Moreover, the following can be concluded.

Theorem 3.5 The system (1.1) has no equilibriums in (0, +∞)2 if and
only if one of the conditions (1), (2), (3) from Theorem 3.1 or one of the
conditions (1), (2), (3) from Theorem 3.2 is satisfied.

Theorem 3.5 is an extension of Lemma 7.2. from [10] for n = 2, which states
(with a different notation than the one used here) that if the system (1.1)
satisfies the inequalities d122 < 0 and d112 < 0, then there is no equilibrium
in (0, +∞)2 . The next theorem considers the stability of equilibrium E12 .

Theorem 3.6 The equilibrium E12 (−d122 /d12 , d112 /d12 ) of the system (1.1)
is:

(1) asymptotically stable on (0, +∞)2 if and only if d12 > 0 and d112 > 0
and d122 < 0;

(2) unstable if and only if d12 < 0 and d112 < 0 and d122 > 0.

Proofs for all theorems can be found in discussions in the Section 2 and so
are omitted here. We note that in [7] are considered only the cases when
appropriate determinants are non-zero, while here we investigated those
cases too.

14
4 Bifurcation analysis
As a consequence of Section 2, there are five different phase portraits of
the system (1.1) in [0, +∞)2 , which can be found in Table 5 under serial
numbers 1, 2, 4, 8 and 9. They are already shown in Figures 1a, 3a, 4a, 2b
and 5 respectively. We note that it is interesting from mathematical point
of view to analyze the stability of the equilibriums of the system (1.1) on
R2 , although it lacks an appropriate physical interpretation. In that case,
four transcritical bifurcations can occur.
Comparing cases 1, 2 and 3 from Table 5 but with the stability of equilibri-
ums analyzed on R2 , we see that for d122 < 0 (case 1) all equilibriums are
hyperbolic and that E0 , E1 and E2 are an unstable equilibriums, while E12 is
an asymptotically stable equilibrium. For d122 = 0 (case 2), nonhyperbolic
equilibrium E2 collides with E12 and becomes a semi-stable equilibrium,
while the stability of other two equilibriums remains the same. Further-
more, for d122 > 0 (case 3), the equilibrium E12 appears again, but this
time as an unstable equilibrium in the second quadrant, while the equilib-
rium E2 is now an asymptotically stable equilibrium, i.e. the equilibrium
E2 have swapped stability with the equilibrium E12 at d122 = 0, while the
stability of other equilibriums have remained the same.
We notice similar situation when we compare cases 1, 4 and 5 from Table 5,
again analyzing the system (1.1) on R2 . More precisely, the equilibrium E1
have swapped stability with E12 (which transitioned from the first quadrant
to the fourth quadrant) at d112 = 0. Furthermore, when we compare cases 8
and 6 from Table 5 and the case d12 < 0, d112 > 0 and d122 > 0 from Table 1
(again on R2 ), we easily obtain that the equilibrium E1 have swapped sta-
bility with the equilibrium E12 (which transitioned from the first quadrant
to the fourth quadrant) at d112 = 0 if a11 x1 + a22 x2 > 0 (a22 d112 < a11 d122 ).
Finally, when we compare cases 8 and 7 from Table 5 and the case d12 < 0,
d112 < 0 and d122 < 0 from Table 1, we derive that the equilibrium E2 have
swapped stability with the equilibrium E12 (which transitioned from the first
quadrant to the second quadrant) at d122 = 0, again if a11 x1 + a22 x2 > 0.

5 Conclusion
We have analyzed the stability of equilibriums and bifurcations of the dy-
namical system (1.1) for various signs of the main and minor determinants
of this system, d12 , d112 and d122 . Of the total of twenty-seven different
combinations of the signs of the determinants d12 , d112 and d122 , thirteen
were meaningful, while nine of them were different from each other on full
neighbourhood of the equilibriums and their phase portraits are shown in
Figures 1-5. Only five of these nine cases had different phase portraits on
[0, +∞), which are shown in Figures 1a, 3a, 4a, 2b and 5. Four of these

15
five phase portraits can also be found in [7], while the last one is new and
very specific as it has infinite equilibriums. All results are presented in Ta-
bles 1-5. In addition, we have improved several results from [10], mainly in
terms of necessary and sufficient conditions for the equilibriums E1 , E2 and
E12 to be asymptotically stable or unstable. We noticed four transcritical
bifurcations among thirteen meaningful and different phase portraits on R2 .

References
[1] S. Ahmad, On the nonautonomous Volterra-Lotka competition equa-
tions, Proc. Amer. Math. Soc., 117(1) (1993), 199–204.
[2] S. Ahmad, A. C. Lazer, One species extinction in an autonomous com-
petition model, Proceedings of the First World Congress of Nonlinear
Analysis, Walter de Gruyter, Berlin, 1996, 359–368.
[3] S. Ahmad, A. C. Lazer, Necessary and sufficient average growth in a
Lotka–Volterra system, Nonlinear Anal., 34 (1998), 191–228.
[4] E. Gonzalez-Olivares, A. Rojas-Palma, Stability in Kolmogorov-type
quadratic systems describing interactions among two species. A brief
revision, Selecciones Matemáticas, 8(1) (2021), 131–146.
[5] J. Hofbauer, K. Sigmund, The theory of evolution and dynamical sys-
tems, Cambridge Univ. Press, Cambridge, 1988.
[6] A. J. Lotka, Elements of Physical Biology, Williams-Wilkins, Baltimore,
1925.
[7] F. Munteanu, A study of a three-dimensional competitive Lotka-Volterra
system, ITM Web Conf., 34 (2020), 03010.
[8] J. Murray, Mathematical Biology, Springer, New York, 1993.
[9] V. Volterra, Leçon sur la Theorie Mathématique de la lutte pour la vie,
Gauthier-Villars, Paris, 1931.
[10] M. L. Zeeman, Extinction in competitive Lotka-Volterra systems, Proc.
Amer. Math. Soc., 123(1) (1995), 87–96.
School of Electrical Engineering, University of Belgrade, Bulevar kralja Alek-
sandra 73, 11120 Belgrade, Serbia
[email protected]

Faculty of Mathematics, University of Belgrade, Studentski trg 16, 11158

Belgrade, Serbia
[email protected]

16