On the general equation of the

second degree

S. Kesavan
The Institute of Mathematical Sciences,
C.I.T. Campus, Taramani,
Chennai - 600 113.
e-mail:kesh@imsc.res.in

Abstract

We give a unified treatment of the general equation of the second degree in

two real variables in terms of the eigenvalues of the matrix associated to the
quadratic terms and describe the solution sets in all cases.

1
2
1 Introduction
The study of the general equation of the second degree in two variables used
to be a major chapter in a course on analytic geometry in the undergraduate
mathematics curriculum for a long time. The equation usually represents
a pair of straight lines or a conic. In the latter case the method of tracing
a conic was to compute the trigonometric ratios of the angle that the axes
of the conic make with the coordinate axes and then rotate the coordinate
axes to reduce the equation to the normal form. These computations could
be tedious. Further in most classical text books the treatment is rather
incomplete and the cases when the solution set is degenerate (especially when
it contains a single point or is empty) are not carefully explained.
The aim of this note is to study all cases of the equation in a unified
manner. By just computing the eigenvalues and eigenvectors of the 2 × 2 real
symmetric matrix associated to the quadratic terms, we can just read off the
properties of the solution set and also write down the equations of various
features of the set very easily.
This approach neatly brings out some of the connections between linear
algebra and geometry.

2 Some linear algebra

Consider the following 2 × 2 real symmetric matrix:
 
a h
A = .
h b

Its characteristic equation is

λ2 − (a + b)λ + (ab − h2 ) = 0.

The discriminant is

(a + b)2 − 4ab + 4h2 = (a − b)2 + 4h2 ≥ 0

and so both its eigenvalues are real and are given by

p
(a + b) ± (a − b)2 + 4h2
.
2

3
These eigenvalues are coincident if, and only if, a = b and h = 0.
Let us denote these eigenvalues by λ1 and λ2 . Then we can find an
eigenvector u = (u1 , u2 ) associated to λ1 and an eigenvector v = (v1 , v2 )
associated to λ2 . If (., .) denotes the usual euclidean scalar product in R2 ,
we have
λ1 (u, v) = (λ1 u, v) = (Au, v)
= au1 v1 + h(u2 v1 + u1 v2 ) + bu2 v2
= (u, Av) = λ2 (u, v).
If λ1 6= λ2 , it then follows that

(u, v) = 0. (2.1)

If λ1 = λ2 , then, as we already observed, A = aI, where I is the 2×2 identity

matrix, and so every vector in R2 is an eigenvector and we can always choose
the two eigenvectors u = (1, 0) and v = (0, 1) so that (2.1) is still valid.
Thus, we always have a pair of orthogonal eigenvectors. Let us normalize
them so that their euclidean norms are unity. Thus let u and v satisfy

u21 + u22 = 1 = v12 + v22 .

Now define  
u1 v1
P = .
u2 v2
Then we have P P T = P T P = I (where by B T we denote the transpose of a
given matrix B). Thus P is an orthogonal matrix. Now
    
λ1 u1 λ2 v1 u1 v1 λ1 0
AP = = .
λ1 u2 λ2 v2 u2 v2 0 λ2

Thus, if  
λ1 0
D = ,
0 λ2
we have
AP = P D, or equivalently A = P DP T .
Remark This is a particular case (when n = 2) of the following general re-
sult: If A is an n×n real symmetric matrix, then there exists an orthonormal
basis of eigenvectors; if P is the orthogonal matrix whose columns are these
eigenvectors and if D is the diagonal matrix whose diagonal entries are the

4
eigenvalues of A (in the same order corresponding to the column vectors of
P ), then A = P DP T . The same result is true if A is a complex hermitian
(i.e. self-adjoint) matrix, in which case we replace P T in the preceding rela-
tion by P ∗ , the conjugate transpose of P (and P will be a unitary matrix). 

Example 2.1 Let  

5 −3
A = .
−3 5
Then, its characteristic equation is λ2 − 10λ + 16 = 0 and its eigenvalues are,
therefore, λ1 = 2 and λ2 = 8. Corresponding to λ1 = 2, we get the equation
5x − 3y = 2x or, equivalently, x = y. Thus the eigenvectors corresponding
to λ1 are scalar multiples of (1,1). Normalizing, we get
 
1 1
(u1 , u2 ) = √ , √ .
2 2
Corresponding to the eigenvalue λ2 = 8, we get the equation 5x−3y = 8x, or,
equivalently, x = −y. Thus all eigenvectors corresponding to this eigenvalue
are scalar multiples of (1, −1) and, normalizing, we get
 
1 1
(v1 , v2 ) = √ , − √ .
2 2
Thus " #
√1 √1
2 2
P = √1
2
− √12

and it is easy to verify that A = P DP T where

 
2 0
D = .
0 8

3 The homogeneous equation

Consider the following homogeneous equation of the second degree in two
real variables:
ax2 + 2hxy + by 2 = 0. (3.1)
Let us denote by S the set of all points (x, y) in the plane which satisfy this
equation. Our aim is to determine this set.

5
 We will set  
a h
A =
h b
and use the notations developed in the preceding section. The above equation
can be written in matrix form as
  
a h x
[ x y ] = 0.
h b y

Writing A = P DP T as in the previous section, let us define

 0   
x T x
= P .
y0 y
Thus
x0 = u1 x + u2 y,
y 0 = v1 x + v2 y.
Then the equation (3.1) reduces to
 0 
x 2 2
0 0
[ x y ]D 0 = λ1 x0 + λ2 y 0 = 0. (3.2)
y
We can now easily determine S from (3.2).
• If λ1 > 0 and λ2 > 0 or if λ1 < 0 and λ2 < 0, then the only solution
will be x0 = y 0 = 0. Then it follows that x = y = 0 and so S = {(0, 0)}.
• Let λ1 > 0 and λ2 < 0, or λ1 < 0 and λ2 > 0. Then − λλ12 > 0 and we
get from (3.2) that r
λ2
x = ± − y0.
0
λ1
Thus S consists of a pair of lines passing through the origin given by
q
u1 x + u2 y = − λλ21 (v1 x + v2 y),
q
u1 x + u2 y = − − λλ21 (v1 x + v2 y).

• If λ1 = 0 and λ2 6= 0 or λ1 6= 0 and λ2 = 0, we get the solution as

y 0 = 0 or x0 = 0, respectively. Thus we have S to consist of a single
line given by
v1 x + v2 y = 0, if λ1 = 0, λ2 6= 0,

6
 and
u1 x + u2 y = 0, if λ1 6= 0, λ2 = 0.

• The case λ1 = λ2 = 0 is excluded since in this case we have a = b =

h = 0 and so the equation is vacuous.

• Finally, if λ1 = −λ2 6= 0, the equation (3.2) reduces to

2 2
x0 − y 0 = 0.

In this case we get two perpendicular lines x0 + y 0 = 0 and x0 − y 0 = 0

which are given in the original coordinates by

(u1 x + u2 y) + (v1 x + v2 y) = 0,
(u1 x + u2 y) − (v1 x + v2 y) = 0.

Thus by computing the eigenvalues and normalized eigenvectors of the matrix

A, we can immediately explicitly describe the solution set S.
The results derived above can be summarized in the following figure.

λ2
2 lines ←one line

λ1 > 0, λ2 > 0
⊥r lines→ S = {(0, 0)}

2 lines
one line↓ one line↓ λ1

2 lines

S = {(0, 0)} ⊥r lines→

λ1 < 0, λ2 < 0
2 lines
←one line
Figure 1

7
4 The inhomogeneous equation
Let us now consider the inhomogeneous equation

ax2 + 2hxy + by 2 = 1. (4.1)

With the notations established in the preceding sections, this equation re-
duces to
2 2
λ1 x0 + λ2 y 0 = 1. (4.2)
If S is the solution set, then we have the following cases.
• Once again the case λ1 = λ2 = 0 is excluded since then we have that
a = b = h = 0 and the equation (4.1) is meaningless.
• If λ1 = λ2 > 0, i.e if a = b > 0 and h = 0, then S is a circle centered
at the origin with radius √1a .

• If λ1 = λ2 < 0, i.e. a = b < 0, h = 0, then S = ∅.

• More generally, if λ1 ≤ 0 and λ2 ≤ 0, we have S = ∅.
• If λ1 > 0 and λ2 > 0, then the equation represents an ellipse with
centre at the origin. The lengths of the semi-axes of the ellipse are √1λ1
and √1λ2 . The equations of the respective axes are y 0 = 0 and x0 = 0,
which can be written in the original coordinates as
v1 x + v2 y = 0,
u1 x + u2 y = 0.

• If λ1 > 0 and λ2 < 0 or if λ1 < 0 and λ2 > 0, then we have a hyperbola.

The lengths and equations of the axes are given as in the case of the
ellipse above.
• If λ1 = −λ2 6= 0, then the equation (4.2) reduces to
2 2 1
x0 − y 0 =
λ1
which is a rectangular hyperbola with axes given by x0 = ±y 0 . In the
original coordinates, this reduces to

u1 x + u2 y = ±(v1 x + v2 y).

8
 • Finally if λ1 = 0, λ2 > 0 or if λ1 > 0, λ2 = 0, then the equation (4.2)
reduces to
2 2
λ2 y 0 = 1 or λ1 x0 = 1
respectively. In these cases, each time we get a pair of parallel lines
given respectively by
1 1
y 0 = ± √ , or x0 = √ .
λ2 λ1
Thus the equations to the lines in the original coordinates are
1 1
v1 x + v2 y = ± √ , or u1 x + u2 y = ± √ .
λ2 λ1

Thus, once again, by computing the eigenvalues and eigenvectors of the

matrix A, we can explicitly describe the solution set S of (4.1). In this case
also we can summarize the results obtained above in the following figure.

λ2
hyperbola ←pair of k` lines

←rect.hyp. ellipse
←circle

hyperbola ellipse
pair of k` lines↓ λ1

hyperbola

S=∅
λ1 ≤ 0, λ2 ≤ 0 rect.hyp.→

hyperbola
Figure 2

9
5 The general equation of the second degree:
straight lines
Let us now consider the general equation of the second degree in two variables
given by
ax2 + 2hxy + by 2 + 2gx + 2f y + c = 0. (5.1)
We will try to completely describe the solution set S of this equation.
Certain computations will repeatedly occur and so it will be useful for us
to do them once and for all. Set x = X + α and y = Y + β. Then (5.1)
becomes 
aX 2 + 2hXY + bY 2 
+2(aα + hβ + g)X + 2(hα + bβ + f )Y = 0. (5.2)
+(aα2 + 2hαβ + bβ 2 + 2gα + 2f β + c)


Notice that the constant term in the last line on the left-hand side of the
above equation can also be rewritten as

α(aα + hβ + g) + β(hα + bβ + f ) + (gα + f β + c). (5.3)

We will also set  

 a h g 
 
∆ =  h b f  .
 g f c 
We will denote by A the symmetric matrix associated to the quadratic terms,
i.e.  
a h
A = .
h b
We will use the notations of the previous sections. In particular λ1 and λ2 will
stand for its two eigenvalues. An associated pair of normalized eigenvectors
will be denoted respectively by (u1 , u2 ) and (v1 , v2 ) as before. Notice that

ab − h2 = det(A) = λ1 λ2 .

Theorem 5.1 The general equation of the second degree in two variables
given by (5.1) defines a pair of intersecting lines if, and only if, ab − h2 < 0
and ∆ = 0.
Proof: Let us assume that the equation (5.1) represents a pair of intersecting
lines and that the point of intersection is (α, β). Then, if we set x = X +

10
α, y = Y +β, the equation represents a pair of lines intersecting at the origin
in the XY -plane. Since (α, β) obviously satisfies the equation, we have

aα2 + 2hαβ + bβ 2 + 2gα + 2f β + c = 0. (5.4)

Further since the equation in the XY variables must be homogeneous, we

must have
aα + hβ + g = 0,
(5.5)
hα + bβ + f = 0,
and (5.2) reduces to
aX 2 + 2hXY + bY 2 = 0.
Since this represents a pair of lines intersecting at the origin, we deduce that
ab − h2 < 0. This then implies that the equations (5.5) admit a unique
solution (α, β) which is the point of intersection of these lines in the xy-
plane. Now it follows from (5.4), the formulation of the left-hand side of this
equation given in (5.3) and the system of equations in (5.5) that we also have

gα + f β + c = 0. (5.6)

Thus (α, β, 1) is a non-trivial solution to the system of equations

aα + hβ + g = 0,
hα + bβ + f = 0, (5.7)
gα + f β + c = 0.

This is possible only if ∆ = 0.

Conversely, if ab − h2 < 0 and ∆ = 0, then choose (α, β) as the unique
solution of (5.5). Then, since ∆ = 0, it automatically follows that (5.6) is
also satisfied. Then, taking into accont (5.3), we get that (5.2) reduces to

aX 2 + 2hXY + bY 2 = 0

under the transformation x = X + α, y = Y + β. Now we know that this

represents a pair of straight lines intersecting at the origin in the XY -plane
and hence (5.1) represents a pair of straight lines intersecting at (α, β). This
completes the proof. 

11
Remark 5.1 The unique solution of (5.5) gives the point of intersection of
the two lines. From the discussion in Section 2, we can explicitly write down
the equations of the two lines. They are given by
q
(u1 (x − α) + u2 (y − β)) = − λλ21 (v1 (x − α) + v2 (y − β)),
q
(u1 (x − α) + u2 (y − β)) = − − λλ21 (v1 (x − α) + v2 (y − β)).

Remark 5.2 If, in addition, we also have that a + b = 0, then the lines will
be perpendicular to each other. 

Example 5.1 Consider the equation

x2 − y 2 + x − 3y − 2 = 0.
Then ab − h2 = −1 < 0. Further
 1

 1 0 
 2 
 
 
∆ =  0 −1 − 32  = 0.

 
 1 

2
− 32 −2 

Thus this equation represents a pair of intersecting straight lines. The system
(5.5) reduces to
α + 12 = 0,

3
−β − 2
= 0.
Thus the point of intersection is (− 12 , − 32 ).
Now
 
1 0
A = .
0 −1
Thus its eigenvalues and eigenvectors are λ1 = 1, with (u1 , u2 ) = (1, 0) and
λ2 = −1 with (v1 , v2 ) = (0, 1). Thus the pair of lines are
(x + 21 ) = (y + 23 )

(x + 12 ) = −(y + 32 ),

12
or, equivalently, x − y − 1 = 0 and x + y + 2 = 0. Notice that since a + b = 0,
we also have that these lines are perpendicular to each other. 

Theorem 5.2 With the notations established above, assume that ∆ = 0 and
ab − h2 > 0. Then the solution set S of the equation (5.1) consists of a single
point.

Proof: If ab − h2 > 0, then the system (5.5) has a unique solution (α, β).
Since ∆ = 0, this implies that (5.6) is also satisfied. Then the transformation
x = X + α, y = Y + β yields

aX 2 + 2hXY + bY 2 = 0

and we have seen that in this case the solution set consists of only the origin
in the XY -plane. Thus S = {(α, β)}. 

Let us now turn to the case where ab − h2 = 0 and ∆ = 0. In this case

the matrix A has λ1 = 0 and λ2 6= 0. (If λ2 = 0 as well, then we saw that
a = b = h = 0 and the second degree terms disappear altogether.) Let us
define
x = u1 X + v1 Y, y = u2 X + v2 Y,
or, equivalently,
X = u1 x + u2 y, Y = v1 x + v2 y.
Then, as in Section 2, the quadratic terms reduce to λ2 Y 2 . The equation
(5.1) now becomes

λ2 Y 2 + 2GX + 2F Y + c = 0

where
G = gu1 + f u2 , and F = gv1 + f v2 .
Now, consider the determinant
 
 a h g 
 
 h b f .
 
 g f c 

Developing this by the third row (or third column) and taking into account
the fact that ab − h2 = 0, we immediately observe that the determinant

13
is independent of the value of c. Thus if ∆ = 0, it follows that the vec-
tors (a, h, g) and (h, b, f ) must be linearly dependent. Since (u1 , u2 ) is an
eigenvector corresponding to λ1 = 0, we have

au1 + hu2 = 0 and hu1 + bu2 = 0.

This then implies that G = gu1 + f u2 = 0. Thus the equation (5.1) now
reduces to
λ2 Y 2 + 2F Y + c = 0.
Completing the square, we get
2
F2

F
λ2 Y + +c− = 0
λ2 λ2
or, equivalently  2
F 1 2
Y + = (F − λ2 c).
λ2 λ22
This leads us to the following conclusions.
• If F 2 > λ2 c, then the solution set S of the equation (5.1) consists of
two parallel lines. They are given by
gv1 + f v2 1 p
v1 x + v2 y + = ± (gv1 + f v2 )2 − λ2 c.
λ2 |λ2 |

• If F 2 = λ2 c, then S consists of a single line given by

gv1 + f v2
v1 x + v2 y = − .
λ2

• If F 2 < λ2 c, then S = ∅.
Example 5.2 Consider the equation

2x2 + 8xy + 8y 2 + 2gx + 2f y + c = 0.

Then  
 2 4 g 
 
∆ =  4 8 f .

 g f c 

14
Then ∆ vanishes if, and only if, f = 2g. Further,
 
2 4
A = .
4 8

Its characteristic equation is λ2 − 10λ = 0. The eigenvalues are, therefore,

λ1 = 0 and λ2 = 10. Correspondiong to λ1 = 0, we get 2x + 4y = 0, and so
every eigenvector is a multiple of (2, −1). Thus we can take
 
2 −1
(u1 , u2 ) = √ , √ .
5 5
Corresponding to λ2 = 10, we get 2x + 4y = 10x, and so every eigenvector is
a constant multiple of (1, 2) so that we can take
 
1 2
(v1 , v2 ) = √ , √ .
5 5
Thus g = gu1 + f u2 = g(u1 + 2u2 ) = 0. Similarly,

F = gv1 + f v2 = g(v1 + 2v2 ) = 5gv1 .

Then
F 2 − λ2 c = 25g 2 v12 − 10c = 5g 2 − 10c.
(i) Let us take g = 2, c = 1. Then F 2 > λ2 c. In this case, the equation is

2x2 + 8xy + 8y 2 + 4x + 8y + 1 = 0.

Dividing throughout by 2, we get

1 1
(x + 2y + 1)2 = i.e. x + 2y + 1 = ± √ .
2 2
It can be easily seen that this expression agrees with the abstract equations
of the lines given earlier.
(ii) Let us take g = 2, c = 2. Then F 2 = λ2 c. In this case, the equation is

2x2 + 8xy + 8y 2 + 4x + 8y + 2 = 0

which can be rewritten as

(x + 2y + 1)2 = 0.

15
Thus we get a single line x + 2y + 1 = 0 which can be seen to agree with the
abstract expression given earlier.
(iii) Let us take g = 1, c = 1. Then F 2 < λ2 c. In this case, the equation is

2x2 + 8xy + 8y 2 + 2x + 4y + 1 = 0.

This can be rewritten as

(x + 2y)2 + (x + 2y + 1)2 = 0.

It is easy to see that this equation has no solution.

Thus, all three cases can occur.

We now summarize the results of this section as follows.

• If ∆ = 0 and if ab − h2 < 0, then the equation (5.1) represents a pair
of intersecting straight lines.

• If ∆ = 0 and if ab − h2 > 0, then the solution set of the equation (5.1)

consists of a single point.

• If ∆ = 0 and if ab − h2 = 0, then the solution set is one of the following:

– a pair of parallel lines;

– a single line;
– the empty set.

Starting from the coefficients of the equation and the eigenvalues and
eigenvectors of the matrix A associated to the quadratic terms, we can
determine which of these cases occurs.

6 The general equation of the second degree:

conics
We will consider the equation (5.1) when ∆ 6= 0.

Case 1 ∆ 6= 0 and ab − h2 < 0.

Since ab − h2 < 0, there exists a unique solution (α, β) to the system (5.5).

16
However, since ∆ 6= 0, we also have that gα + f β + c 6= 0. Setting x =
X + α, y = Y + β, the equation now reads as

aX 2 + 2hXY + bY 2 + (gα + f β + c) = 0.

Let us set C = gα + f β + c. Then as in Section 3, we can easily see that

the solution set is a hyperbola with centre at (α, β) with the lengths of the
semi-axes being s  s 
C 
  and  C .
 
 λ1   λ2 

The equations to the axes are given by

v1 (x − α) + v2 (y − β) = 0,
u1 (x − α) + u2 (y − β) = 0,

respectively. If, in addition, a + b = 0, this will be a rectangular hyperbola.

Case 2 ∆ 6= 0 and ab − h2 > 0.

Once again, there exists a unique solution (α, β) to the system (5.5) and
setting x = X + α, y = Y + β, the equation reduces to

aX 2 + 2hXY + bY 2 + C = 0

where C = gα + f β + c 6= 0. The solution set is either an ellipse or is empty,

depending on the signs of C, λ1 and λ2 . The centre, lengths of the semi-axes
and the equations of the axes are exactly as in the previous case.

Example 6.1 Consider the equation

5x2 − 6xy + 5y 2 + 22x − 26y + 29 = 0.

In this case  
 5 −3 11 
 
∆ =  −3
 5 −13  6= 0.
 11 −13 29 
Further ab − h2 > 0. The system (5.5) reads as

5α − 3β + 11 = 0
−3α + 5β − 13 = 0

17
which admits the unique solution (α, β) = (−1, 2). Further

gα + f β + c = −8.

Thus, setting x = X + α, y = Y + β, we get

5X 2 − 6XY + 5Y 2 = 8.

The eigenvalues and eigenvectors of the matrix

 
5 −3
−3 5

were computed in Example 2.1. They are

 
1 1
λ1 = 2, (u1 , u2 ) = √ ,√
2 2
and  
1 1
λ2 = 8, (v1 , v2 ) = √ , −√ .
2 2
Then setting
X 0 = u1 X + u2 Y
Y 0 = v1 X + v2 Y,
the equation reduces to
2 2
2X 0 + 8Y 0 = 8
or, equivalently,
X 02 Y 02
+ = 1.
4 1
Thus the solution set is an ellipse with centre at (−1, 2) with the lengths of
the semi-major axis being 2 and that of the semi-minor axis being 1. The
equation of the major axis is Y 0 = 0 which becomes X = Y , or,

x + 1 = y − 2, i.e. x − y + 3 = 0

and that of the minor axis is X 0 = 0, which becomes X = −Y , or,

x + 1 = 2 − y i.e. x + y − 1 = 0.

18
 If we change the value of the constant c from 29 to 39, then gα+f β+c = 2
and in this case the equation reduces to
2 2
2X 0 + 8Y 0 = −2

and clearly the solution set is empty. 

Case 3 ∆ 6= 0 and ab − h2 = 0.
Since the matrix A is singular, one of its eigenvalues will be zero. Without
loss of generality, let λ1 = 0 and let λ2 6= 0. Let us make the usual change of
coordinates using the normalized eignevectors of A:

x = u1 x0 + v1 y 0
y = u2 x0 + v2 y 0 .

The equation then transforms to

2
λ2 y 0 + 2Gx0 + 2F y 0 + c = 0

where
F = gu1 + f u2 and F = gv1 + f v2 .
Since we already have
au1 + hu2 = 0,
hu1 + bu2 = 0
if we also have G = 0, then it will imply that ∆ = 0, which is not the case.
Thus G 6= 0. We complete squares and rewrite the equation as follows:
2
F2

0 F
λ2 y + = −2Gx0 − c + .
λ2 λ2
Now set
c F2 F
X = x0 + − , Y = y0 +
2G 2Gλ2 λ2
to get
2G
Y2 = − X
λ2
which is a parabola.

We summarize the results of this section as follows.

19
• If ∆ 6= 0 and if ab − h2 < 0, the equation (5.1) represents a hyperbola.
If, in addition, a + b = 0, it represents a rectangular hyperbola.

• If ∆ 6= 0 and if ab − h2 > 0, then the equation (5.1) either represents

an ellipse or the solution set is empty. If a = b and h = 0, then the
equation represents a circle provided the solution set is non-empty.

• If ∆ 6= 0 and if ab − h2 = 0, then the equation (5.1) represents a

parabola.

20