Appendix A∗

Into How Many Parts Do n Lines Divide the

Plane?

Consider n distinct lines in the real projective plane. They divide it into (con-
vex) parts. The question is this: how many parts can be obtained (under all
possible arrangements of the lines)?
For small n, the answer is clear; the possible number M of parts is given
by the following table:

n 1 2 3 4 5
M 1 2 3 4 5
4 6 8
7 9
10
11

If we consider one of the lines to be the line at infinity, then we obtain

n − 1 lines in R2 . The values of M in the table are provided by the following
configurations of n − 1 lines:
n=2: n=3: n=4: n=5:
M =2 M =3 M =4 M =5

M =4 M =6 M =8

M =7 M =9

M = 10

M = 11

∗
This appendix is a translation of the author’s article in Mat. Prosveshchenie, Ser. 3 12 (2008),
95–104. It was not part of the original Russian edition of this book. It has been appended
here at the author’s request.

V.I. Arnold, Real Algebraic Geometry, UNITEXT – La Matematica per il 3+2 66, 85
DOI 10.1007/978-3-642-36243-9, © Springer-Verlag Berlin Heidelberg 2013
86 A Into How Many Parts Do n Lines Divide the Plane?

The number of connected components of the complement of the collec-

tion of lines in the affine plane R2 is studied in similar fashion. This is the
same problem, since one can declare one of the n lines to be the line at infin-
ity and study the complement of the n − 1 lines in the affine plane (which
coincides with the complement of the n lines in the projective plane).
Looking at the previous examples, we observe that the smallest number
of parts of the complement of n lines in RP2 is M = n, while the largest is

n( n − 1)
M = 1 + (1 + 2 + 3 + · · · + (n − 1)) = 1 + .
2
However, only some (and not all) of the intermediate numbers between
these limits can be attained. Starting from some point, all sufficiently large
values of M are attained, but the beginning of the list of attainable values
contains gaps (or holes).
The aim of this paper is to describe these holes. Here is the first hole.

Theorem 1. The value M = 2(n − 1) is attainable, but no other value of M in the

interval
n < M < 2( n − 1)
is attainable.
For any choice of n > 2 lines in RP2 , their complement cannot consist of such a
number M of connected components.

Proof. Denote by k the greatest number of lines (among our n lines) passing
through one point.

Lemma 1. If k = n, then M = n.

Proof. We assume that one of these n lines is the line at infinity. Then the re-
maining lines are parallel. They divide the affine plane R2 (complementary
to the first line) into n parts, since there are n − 1 such parallel lines.

Lemma 2. If k = n − 1, then M = 2(n − 1).

Proof. We assume that the remaining nth line is the line at infinity. Its com-
plement is R2 . The family of n − 1 lines passing through one point divides
the plane R2 into 2(n − 1) parts, as required.

Lemma 3. If k ≤ n − 1, then M ≥ 2(n − 1).

Proof. Note that for n > 1 we have k ≥ 2 (since any two lines in RP2 inter-
sect). We choose one of the k lines passing through one point as the line at
infinity. Its complement is the affine plane R2 , which contains k − 1 parallel
lines of the chosen pencil together with a further n − k ≥ 1 remaining lines.
These parallel lines divide the plane RP2 into k parts. By adding the re-
maining lines one by one, we shall gradually increase the number of parts.
Here, if the sth line to be added is intersected by the lines already existing in
A Into How Many Parts Do n Lines Divide the Plane? 87

xs points, then it is divided by them into x s segments, each of which divides

one of the parts that existed before the sth line was added in two. There-
fore, when the sth line is added, the number of parts of the complement is
increased by exactly xs .
We now observe that xs ≥ k (since the line being added intersects all k
parallel lines of the original pencil in k distinct points). Therefore the total
number of added parts is

x1 + · · · + x n− k ≥ k ( n − k ) .

By adding these parts to the k parts that we already had before dealing
with the n − k “additional” lines, we conclude that

M ≥ k + k ( n − k ) = k ( n − k + 1) . (1)

The sum of both factors on the right-hand side is equal to n + 1. When two
positive numbers whose sum is equal to n + 1 are multiplied together, the
greater the smaller factor is, the greater their product:
! k+!=n+1

k! = const
2 k! = 2(n − 1) = min{k! : 2 ! k ! n − 1}
k

If 2 ≤ k ≤ n − 1, then
min (k, l ) ≥ 2,
k + l = n +1

so that
M ≥ 2( n − 2 + 1) = 2( n − 1).
This completes the proof of Lemma 3
Theorem 1 now follows from Lemmas 1, 2, and 3 (since any number k ≤ n
is either equal to n or equal to n − 1 or less than n − 1).
The first hole has now been described. Next we describe the second hole.
Let n ≥ 3.
Theorem 2. The value M = 3n − 6 is attainable, but no other value of M in the
interval
2( n − 1) < M < 3( n − 2)
is attainable.
For any choice of n > 2 lines in RP2 , their complement cannot consist of such a
number M of connected components.
Proof. As before, we denote by k the maximum number of lines (among the
given n lines) passing through one point, and we call one of them the line
88 A Into How Many Parts Do n Lines Divide the Plane?

at infinity. We regard the remaining n − 1 lines in the affine plane R2 , which

is the complement of the chosen line, as the pencil of k − 1 parallel lines
supplemented by the n − k remaining lines not parallel to these k − 1 lines
of the pencil.
If k = n − 1, then M = 2(n − 1) by Lemma 2. If, on the other hand, k ≤
n − 2, then from formula (1) in the proof of Lemma 3 we obtain

M ≥ k(n − k + 1) ≥ (n − 2)((n + 1) − (n − 2)) = 3(n − 2) ,

provided that k ≥ 3 (for which mink+l =n+1(k, l ) ≥ 3).

Thus, the theorem is proved for all arrangements of n lines such that k > 2.
We now prove that M ≥ 3(n − 2) in the remaining case, in which k = 2.
If k = 2, that is, no three lines are concurrent, then all our n lines divide the
plane into the maximum possible number of parts for n lines (and attainable
for n lines in general position), namely

n( n − 1)
M=1+ .
2

Lemma. We have the inequality

n( n − 1)
+ 1 ≥ 3( n − 2) .
2
Proof. This inequality has the form

n2 − n − 6 ( n − 2 ) + 2 ≥ 0 ,

that is,
n2 − 7n + 14 ≥ 0 ,
which is true because the discriminant of the quadratic trinomial on the left-
hand side,
49 − 4 · 14,
is negative.

Thus the lemma is proved, so that the inequality M ≥ 3(n − 2) holds in

the case k = 2 as well.

Having proved Theorem 2, we have described the second hole. It first

appears when n = 6 (when in the interval defining the hole there are integer
points: 3(n − 2) − 2(n − 1) > 1 for n > 5).
The subsequent holes that we now study are under the assumption that
the number n of lines is sufficiently large (in comparison with the number
of the hole).
A Into How Many Parts Do n Lines Divide the Plane? 89

Theorem 3. Suppose that the greatest number of lines (among our n lines) passing
through one point is k. Then these n lines divide the plane RP2 into M parts, where
M lies in the interval
r (r − 1)
k(n + 1 − k) ≤ M ≤ k(n + 1 − k) + , where r = n − k .
2
(Here all the numbers M in this interval are attainable for a suitable choice of n
lines, provided that n is sufficiently large.)
Proof. Choose a pencil of k lines. The lines of this pencil divide the plane
RP2 into k parts. The remaining n − k = r lines add to k the following num-
ber of parts:
M $ = x1 + x2 + · · · + x n−k ,
where xs is the number of points of the sth added line intersected by the
previous lines.
Among these previous lines there are k lines of the chosen pencil (and
s − 1 added lines). The points of intersection with the lines of the pencil are
all distinct (since the only common point of two lines of the pencil is the
point of intersection of the k lines of the pencil originally chosen, so that no
other of our n lines can pass through this point).
Hence x s ≥ k, M $ ≥ k(n − k), M ≥ k(n − k + 1), which proves the first
inequality in Theorem 3.
On the other hand, xs ≤ k + (s − 1). Consequently,

r (r − 1)
M $ ≤ k ( n − k ) + (0 + 1 + · · · + n − k − 1) = k ( n − k ) + ,
2
r (r − 1)
M ≤ k(n + 1 − k) + .
2
This proves the second inequality in Theorem 3.
All the values of M in the interval described by both inequalities in the
above theorem are attained (for sufficiently large n) for the following reason.
The greatest number of parts is provided by choosing the r additional
lines in general position. For them all the points of intersection (with the
lines of the pencil and with each other) are distinct; as a result, this gives us
k(n + 1 − k) + r (r − 1)/2 parts.
If n is sufficiently large in comparison with r (for example,27 if n ≥ r
(r + 1)/2), then one can choose the additional lines so that any chosen points
in the pairwise intersection of the lines lie on the lines of the pencil (so that
x s = k for the corresponding s).
In fact, we can, for example, start with r lines in general position in the
affine plane R2 and draw, through any collection of r (r − 1)/2 − S points
of their pairwise intersection, lines that are parallel to each other but not
parallel to these r lines. By including these parallel lines along with the line
at infinity in the pencil of k = n − r parallel lines, we obtain a collection of
90 A Into How Many Parts Do n Lines Divide the Plane?

lines for which M $ , the sum of xs over all s = 1, . . . , r, exceeds kr precisely

by S. In this case,28 M $ = kr + S, M = k(n − k + 1) + S, and Theorem 3 is
proved.

Theorem 4. Suppose that the greatest number of lines (among our n lines) passing
through one point is k > 2. Then these n lines divide the plane RP2 into M parts,
where
n( n − 1)
M≥ .
2( k − 1)
Here it is important that the numerator increases with the number n of
lines at the same rate as n2 , while the denominator is independent of the
number n. As a result, the right-hand side becomes larger than any linear
function of n for sufficiently large n (when k is fixed).
To prove Theorem 4 we arrange the given n lines in some order. By an
“event” we mean the intersection of some line with the lowest-numbered
line. Thus, the number of events is 0 + 1 + 2 + · · · + (n − 1) = n(n − 1)/2
(independently of how many distinct points of intersection there are).
By a “partition” we mean a division into parts of some line (say the sth)
by lines with lower numberings. We denote by xs the number of points in
the partition of the sth line. These xs points divide that projective line into
xs parts.
We add the lines one by one, and at each stage, we increase the number of
components of the complement of the lines by the number x s of parts added
by the sth line (which divides each of the already existing xs components
intersected by it into two parts).
Therefore the total number of components of the complement of the
union of the n lines in the projective plane RP2 is given by
n
M= ∑ xs ,
s =1

supposing formally that x1 = 1: although the first line does not divide any
“preceding” lines, one needs to take into account the (single) component of
the complement of one line in the projective plane.
At each point of the partition, the maximum number of occurring events
is k − 1 (the intersection of the sth line with the preceding ones), since there
cannot be more than k lines of our collection passing through one point.
Therefore, the number of all events does not exceed M(k − 1). And since the
latter is equal to n(n − 1)/2, we conclude that

n( n − 1)
≤ M ( k − 1),
2
that is,
n( n − 1)
M≥ ,
2( k − 1)
A Into How Many Parts Do n Lines Divide the Plane? 91

which completes the proof of Theorem 4.

For the investigation of “stable” holes (the jth stable hole D j will be in-
vestigated under the assumption that the number n exceeds some constant
depending on j), we introduce the following notation:

(n − j)( j + 1) + j( j − 1)
α j = (n − j)( j + 1), βj = .
2
For sufficiently large n, the first terms of these two sequences are ar-
ranged in the following order:

α0 = β 0 < α1 = β 1 < α2 < β 2 < α3 < β 3 < · · · < α j −1 < β j −1 < α j .

P0 D1 P1 D2 P2 D3 P3 ... Pj−1 Dj Pj M
α0 β0 α1 β1 α2 β2 α3 β3 . . . αj−1 βj−1 αj βj

We denote by P0 , P1 , . . . the closed intervals

P0 = [α0 ≤ M ≤ β 0 ], Pj = [α1 ≤ M ≤ β 1 ], ..., P1 = [α j ≤ M ≤ β j ] ,

and by D1 , D2 , . . . , D j the complementary open intervals

D1 =] β 0 < M < α1 [, D2 =] β 1 < M < α2 [, ..., D j =] β j−1 < M < α j [ .

The stable hole D j is described as follows.

Theorem 5. If the number of lines n is sufficiently large, then the number M of

components of their complement in the projective plane RP2 cannot take values in
the intervals D j . In other words, no value of M for which

( j − 1)( j − 2)
β j −1 = j ( n + 1 − j ) + < M < ( j + 1)(n − j) = α j
2
is possible.

Proof. Denote by k the greatest number of lines (among our n lines) passing
through one point. We shall prove that M cannot lie in an interval D j for any
k, but this proof will be based on different considerations in the following
three cases:
I. k > n − j;

II. II j + 1 ≤ k ≤ n − j;

III. k ≤ j.
Here we suppose that n − j ≥ j + 1 (which holds for n sufficiently large).
92 A Into How Many Parts Do n Lines Divide the Plane?

Case I. Suppose that k takes one of the values {n, n − 1, . . . , n − j + 1}, and
put r = n − k.29
According to Theorem 3, the number M lies in the interval

αr = (n − r )(r + 1) ≤ M ≤ (n − r )(r + 1) + r (r − 1)/2 = β r ,

that is, M ∈ Pr , 0 ≤ r ≤ j − 1.
None of these r segments intersects the interval D j , so Theorem 5 is
proved for case I.
Case II. Suppose that j + 1 ≤ k ≤ n − j. Then ! = n + 1 − k also satisfies
the inequalities j + 1 ≤ ! ≤ n − j. In this case, mink+!=n+1 (k, !) ≥ j + 1:

n−j

j+1
k! = (j + 1)(n − j) = αj
III II I
k

Therefore, again by Theorem 3,

M ≥ (n − j)( j + 1) = α j .

However, M < α j in the interval D j . Thus Theorem 5 is proved for Case II

as well.
Case III. Suppose that 2 < k ≤ j. According to Theorem 4,

n( n − 1) n( n − 1)
M≥ ≥ .
2( k − 1) 2( j − 1)

The right-hand side of this inequality is greater than α j if n is sufficiently

large. Indeed,
n( n − 1)
≥ (n − j)( j + 1)
2( j − 1)
for sufficiently large n, because then

n( n − 1)
> 2 ( j2 − 1 ) .
n−j

For example,
n( n − 1)
≥ n,
n−j
A Into How Many Parts Do n Lines Divide the Plane? 93

so that the condition n > 2( j2 − 1) suffices for the inequailty M > α j , which
does not allow M to lie in the interval D j (between β j−1 and α j ).
In the only case not yet disposed of, k = 2, our n lines in general position
divide the projective plane RP2 into M = 1 + n(n − 1)/2 parts.
This number M is greater than the limit α j = (n − j)( j + 1), since for j +
1 < n/2 (which we have supposed), the inequalities

n( n − 1)
αj < <M
2
hold for j > 0.
Thus Theorem 5 is proved for k = 2 as well, and it is therefore proved for
all k (the case k = 1 for n > 1 is not realized, since every pair of lines in the
projective plane intersect).

Thus Theorem 5 is completely proved, so that (for a sufficiently large num-

ber of lines n) all the stable holes

D1 , D2 , . . . , D j , . . .

exist in the sequence of numbers M of components into which n lines divide

the real projective plane.

Remark 1. I do not know whether the unstable holes (for values of n smaller
than those indicated above) are given by the same formulas (β j−1 < M <
α j ) as the stable holes. To begin with, instability does not manifest itself for
small j.
The first unclear case is the third hole for n = 9. In this case, the formulas
give α3 = 24, β 2 = 22.
Nine lines can divide the projective plane into 22 regions and into 24 re-
gions. Whether they can divide it into 23 regions or, on the contrary, whether
M = 23 for n = 9 is the third hole, is unknown. Such an arrangement of lines
(if it is possible) would be possible only in the case that no four of these nine
lines pass though a single point (case k = 3 in the proof of Theorem 5).

Remark 2. The motivation for this article was the publication in Berkeley
of A.B. Givental’s translation into English of A.P. Kiselev’s book Geometry.
When in April 2007 in California I looked at this translation, I was unable to
solve one of the problems in this book (I had resolved them all in my youth).
This problem was as follows: how many lines divide the plane R2 into
five convex parts?
I asked Givental how this question was formulated in Kiselev’s original
book; he confessed that this was by no means the case: the problem was
added by the translator (who improved Kiselev in other places as well).
Every mathematical problem admits two versions: the Russian version,
which nobody can simplify (without losing the essence of the problem), and
the French version, which nobody can generalize any further (since it has
94 A Into How Many Parts Do n Lines Divide the Plane?

already been formulated in such a general form that it contains all possible
generalizations).
On arriving from Paris to Berkeley I decided to formulate the French ver-
sion of Givental’s problem. With this in mind, I changed five regions to any
number M of regions. This is how the present article came about.
I have been unable to solve this general problem. One should be able to
describe all holes for all values of n; but even the third hole was calculated
by me explicitly only for n ≥ 14 (when it becomes stable). By lecturing local
school pupils in Berkeley, Stanford, San José, and Santa Clara (where heroic
leaders of the Moscow-style mathematical circles have taught the pupils
how to solve difficult problems better than I), I hoped that they would come
up with a description of the unstable holes, which is still awaited.
It seems that the question whether 9 lines can divide the projective plane
into 23 regions remains unsolved.30
Editors’ Comments on Gudkov’s Conjecture

It is difficult to overestimate Arnold’s role in the revolutionary explosive de-

velopments in the study of topological properties of real algebraic varieties
(and in particular, in solving Hilbert’s sixteenth problem). A whole series of
works was inspired by Arnold’s foundational paper “On the Arrangements
of Ovals of Real Plane Algebraic Curves, Involutions of Four-Dimensional
Smooth Manifolds, and the Arithmetic of Integer Quadratic Forms” (Funk.
Anal. i Prilozh. 5 (1971), no. 3, 1–9). The very title of this important paper
indicated a new direction of investigation in this field. A number of new re-
sults in this paper were immediately named after Arnold, such as Arnold’s
congruence, Arnold’s inequalities, and so on.
Arnold’s paper was stimulated by (private and public) communications
with D.A. Gudkov and V.A. Rokhlin. Gudkov (in his 1971 note in Doklady
AN SSSR 200 (1971), no. 6, 1269–1272, and a 1970 talk in Moscow) had for-
mulated as a conjecture the congruence modulo 8 for maximal curves ∗ of
even degree. Rokhlin (in his 1971 paper Funk. Anal. i Prilozh. 5 (1971), no. 1,
48–60) had obtained his genus bounds in 4-dimensional topology by means
of the Atiyah–Singer–Hirzebruch formula for signatures of ramified cover-
ings (he presented these results at a 1970 talk in Moscow; it may also be
worth mentioning a slightly earlier talk given by Rokhlin in Moscow about
his attempts to prove the Whitney conjecture using congruences modulo 16).
Arnold came to his foundational discoveries through his detailed knowl-
edge of Gudkov’s habilitation thesis, in which Gudkov had completed the
classification of real nonsingular plane projective curves of degree 6, and his
awareness of both Gudkov’s conjecture and Rokhlin’s results. (Arnold him-
self, in the abstract of his talk given at the Moscow Mathematical Society
meeting of April 6, 1971, mentioned both Gudkov’s authorship in stating
the conjecture and the role of Rokhlin’s results on signatures of ramified
coverings.)

∗
A nbyjnbm! dvswf (also called an N.dvswf) is a smooth real algebraic curve whose
number of real components is equal to h + 1, where g is the genus of the curve.

V.I. Arnold, Real Algebraic Geometry, UNITEXT – La Matematica per il 3+2 66, 95
DOI 10.1007/978-3-642-36243-9, © Springer-Verlag Berlin Heidelberg 2013
96 Editors’ Comments on Gudkov’s Conjecture

As to the history of Gudkov’s conjecture, Rokhlin told us, and Gudkov

confirmed, the following story (letters corroborating this story were found
by G. M. Polotovsky in Gudkov’s personal archive). During the preparation
of his habilitation thesis, Gudkov showed a preliminary version of part of it
to V. V. Morozov, a professor at Kazan University. Thi s version contained
detailed proofs of a complete classification of smooth degree-6 curves, which,
as was discovered later, contained a few errors at this stage. Morosov (who
later was one of the official referees of Gudkov’s habilitation) pointed out
to Gudkov a strange—and doubtful in Morosov’s opinion—asymmetry in
a diagram expressing the classification obtained. It is in rectifying this
asymmetry that Gudkov came to the correct classification and thence to
his conjecture.
The first traces of the conjecture (in its complete form) are found in a
letter of Gudkov to Morosov (winter 1969–1970, Gudkov’s archive, commu-
nicated by Polotovsky), in the verbatim record of the habilitation defense
(1970, Gudkov’s archive, communicated by Polotovsky), and in his paper
“Construction of New Series of M-Curves,” Doklady AN SSSR 200 (1971),
no. 6, 1269–1272.
Additional documentary confirmation that Gudkov was the author of the
conjecture for M-curves is a letter from Gudkov to Arnold dated October 15,
1972 (Gudkov’s personal archive, communicated by Polotovsky), in which
Gudkov proposes an additional conjecture and writes to Arnold that “J bm.
sfbez ibe uijt jo njoe xifo J gpsnvmbufe uif dpokfduvsf gps N.dvswft- cvu!qfsgjej.
ously concealed it from you and Rokhlin.”
Let us conclude by observing that as Arnold wrote in “On the Arrange-
ments of Ovals of Real Plane Algebraic Curves, Involutions of Four-Dimen-
sional Smooth Manifolds, and the Arithmetic of Integer Quadratic Forms,”
that paper would not exist if Gudkov had not communicated his conjecture
to Arnold.
Notes 97

Notes

1. Questions concerning the topology of real algebraic varieties form the first part of
Hilbert’s sixteenth problem. Below we present an English translation of the text of this
part. (Hilbert’s lecture was translated by Mary Winston Newson for the Bulletin of the
American Mathematical Society 8 (1902), 437–479.) It is exactly this part, under the ti-
tle “Problem der Topologie algebraischer Curven und Flächen,” that was presented by
Hilbert in 1900 as the sixteenth problem in his renowned lecture at the second Interna-
tional Congress of Mathematicians. For the lecture he selected 10 problems. The famous
list of 23 problems appeared in its final form only later and was published in 1901 in
Archiv der Mathematik und Physik:

The maximum number of closed and separate branches which a plane algebraic
curve of the nth order can have has been determined by Harnack (Mathematische
Annalen 10). There arises the further question as to the relative position of the
branches in the plane.
As to curves of the 6th order, I have satisfied myself—by a complicated process,
it is true—that of the eleven branches which they can have according to Harnack,
by no means all can lie external to one another, but that one branch must exist in
whose interior one branch and in whose exterior nine branches lie, or inversely. A
thorough investigation of the relative position of the separate branches when their
number is the maximum seems to me to be of very great interest, and not less so the
corresponding investigation as to the number, form, and position of the sheets of an
algebraic surface in space. Till now, indeed, it is not even known what is the maxi-
mum number of sheets which a surface of the 4th order in three-dimensional space
can really have. (cf. Rohn, “Flächen vierter Ordnung,” Preisschriften der Fürstlich
Jablonowskischen Gesellschaft, Leipzig 1886).

Hilbert’s questions on curves of degree 6 in the real projective plane and surfaces of
degree 4 in three-dimensional real projective space have found their complete answers
through the work of D. A. Gudkov, V. I. Arnold, V. A. Rokhlin, and V. M. Kharlamov in the
breakthrough of 1969–1976; see http://www.pdmi.ras.ru/˜ olegviro/H16-e.pdf
for a detailed discussion.
2. In principle, one could add to this list some other real forms of curves described by a
second-degree equation, namely pairs of intersecting complex conjugate imaginary lines
(such as the pair given by the equation x2 + y2 = 0), pairs of parallel complex conjugate
imaginary lines (such as the pair given by the equation x2 + 1 = 0), and empty “ellipses”
(such as the one given by the equation x2 + y2 + 1 = 0). A possible reason not to include
them in the list is the fact that they do not represent a curve in the usual purely real sense,
but either a point or the empty set.
3. Arnold is not very precise in this comment. Perhaps he is intentionally forcing his read-
ers to think further and recall that every point that can be constructed with straightedge
and compass can be also constructed with compass alone.
4. Presumably, Arnold has here in mind the following beautiful lines from Goethe’s poem
“Gott, Gemüt und Welt” (God, Soul, and World): “Willst du ins Unendliche schreiten; Geh
nur im Endlichen nach allen Seiten.” This can be translated literally as “If you would step
into the infinite, you have only to walk in the finite in all directions.” Or more poetically,
“If to the Infinite you want to stride, Just walk in the Finite to every side.”
5. A semicubic cusp (or an ordinary cusp) is an isolated singularity that in appropriate local
coordinates is given by the equation y2 = x3 .
6. Even if the precise meaning of such a philosophical statement is unclear, it may push
us to think about a kind of asymptotic real algebraic geometry. Such a field does not
exist yet, but see V. M. Kharlamov, S. Yu. Orevkov, “The Number of Trees Half of Whose
Vertices Are Leaves and Asymptotic Enumeration of Plane Real Algebraic Curves,” J.
98 Notes

Combin. Theory Ser. A 1051 (2004), no. 1, 127–142, where it is shown that the only known
asymptotically significant results on the topology of real plane algebraic curves are some
consequences of Bézout’s theorem and certain Arnold inequalities (and not the Gudkov
congruence, contrary to the impression given by Arnold’s remarks on p. 45).
7. The discovery of the Möbius strip is usually cited as originating with a paper by
Möbius on another subject: “On the Determination of the Volume of a Polyhedron.”
Drawings representing such a shape date back to ancient times; it is for, example, ev-
ident in the famous Ouroboros drawing from the early alchemical text the Chrysopoeia of
Cleopatra from the second century.
8. If by Möbius’s assertion one understands the statement that for a smooth projective
curve obtained from a projective line by a deformation in the class of embedded circles,
the number of inflection points is odd and greater than or equal to 3, then such a statement
is proved, for example, by S. Sasaki in Tôhoku Math. J. 2 (1957), no. 9, 113–117 (another
proof and a bit more geometric information can be found in R. Pignoni, Manuscripta Math
72 (1991) 223–249). Indeed, even in a stronger form, such an inflection property holds for
every embedded smooth curve that is not contractible to a point. In all these statements,
it is more convenient to define an inflection point as a point where locally the curve goes
from one side of the tangent line to the other.
On the other hand, there is another question, which was raised by Arnold himself and
which is still open. Namely, Arnold conjectured a statement that reinforces Möbius’s the-
orem. It is based on the notion of dangerous self-tangencies of an immersed circle. Follow-
ing Arnold, a self-tangency is called dangerous if an orientation of the circle induces, by
means of the tangent branches, the same direction on the tangent at the tangency point.
The conjecture states that every projective curve obtained from a projective line by de-
formation in the class of immersed circles without dangerous self-tangencies has at least
three inflection points. Some partial results in this direction can be found in D. Panov,
Funct. Analysis and Its Appl. 32 (1998), no. 1, 23–31.
9. In fact, in the same volume of Mathematische Annalen that contains the paper by A.
Harnack, there appeared a paper by F. Klein (at that time editor of Mathematische Annalen).
Harnack’s paper was submitted in January, and Klein’s in April. Formally, Klein’s paper
concerns a related but different topic. Klein, recognizing that the topological technique
from his paper allowed him to give another proof of Harnack’s bound for the number
of connected components of a real algebraic curve by g + 1 (this gives half of Harnack’s
theorem; the other half is the realizability of this bound by plane curves in any degree),
inserted such a proof as an additional section.
Let us observe also that Harnack’s proof of the inequality is different. He never explic-
itly manipulates the Riemann surface of the curve, using instead Bézout’s theorem. As
to the genus, it appears in Harnack’s proof as the standard expression ( d − 1)( d − 2) /2
diminished by the number of nodes and cusps (Harnack does not consider more compli-
cated singularities).
10. This discussion is in fact postponed to Chapter 5.
11. Neither Newton nor Descartes carried out or even intended a “full” investigation of
real plane curves of degree 4. Such an investigation was launched by A. Cayley and con-
tinued, in what concerns at least the nonsingular quartics, by F. Klein and H.-G. Zeuthen.
Stratifications (of the same type as Newton’s stratifications in the case of degree 3) of the
space of real plane curves of degree 4 were produced in a series of papers by D. A. Gud-
kov and his collaborators; see references and further information in A. B. Korchagin and
D. A. Weinberg, Rocky Mountain J. Math. 32 (2002), no. 1, 255–347.
12. This is probably a misprint, and Arnold meant the opposite statement. The topology
of nonsingular real curves of degree 5 and 7 is somehow less complicated than that of
curves of degree 6. It seems to be a general rule: the real curves of degree 2k + 1 are, in a
sense, not more complicated than the real curves of degree 2k.
13. This sentence is a bit confusing. The situation was in fact as follows. It is in his PhD
thesis that Gudkov “proved” Hilbert’s statement on the arrangements of 11 ovals of real
Notes 99

sextics. A little while later, he found a mistake in the proof, and after correcting the mis-
take, completed the classification of all possible arrangements of nonsingular sextics (even
with any number of ovals). This correct classification (which refutes Hilbert’s theorem)
became the core of Gudkov’s habilitation thesis.
14. Gudkov’s result on M-sextics excludes only a small part of these billion possibilities.
A majority is in fact excluded by Bézout’s-theorem arguments. For curves of higher de-
grees as well, Bézout’s-theorem arguments forbid many more oval arrangements than
Gudkov’s congruence (cf., endnote 6).
15. Arnold’s discussion of Gudkov’s conjecture may give the impression that it was
Arnold and not Gudkov who conjectured the congruence. We refer the reader to the com-
ments on Gudkov’s conjecture at the end of this book, where we present a short review
that refutes such an attribution of the conjecture.
16. Rokhlin’s proof is based on different tools, but it uses the same 4-manifold as in
Arnold’s proof.
17. At present, 83 configurations have been realized, and 6 remain uncertain; see S. Yu.
Orevkov, Funct. Analysis and Appl. 36 (2002), 247–249, and Geometric and Functional Anal-
ysis 12 (2002), no. 4, 723–755. The questionable arrangements are shown in the following
figures:

...
14

...
14

...
18

...
15

... ...
7 12

... ...
9 10
100 Notes

18. The expression “the numbers of points of intersection with lines” means the restric-
tions that arise from intersecting the curves with lines (as explained a few pages above in
the case of quartics), as well as with conics.
19. In what follows, this definition is understood in such a way that it is necessary to
assume, in addition, that g0 = h0 = id.
20. In an equivalent form, that is, regarding K ( n) as the number of up–down sequences,
this theorem is an old result of D. André (“Développement de sec x et de tan x,” C. R. Acad.
Sci. Paris 88 (1879), 965–967).
21. Additional information and references can be found in J. Millar, N. J. A. Sloane, and
N. E. Young, J. Combinatorial Theory, Series A 76 (1996), 44–54.
22. The lower bound given by Ortiz-Rodriguez was improved by 2 in E. Brugallé and B.
Bertrand, C. R. Math. Acad. Sci. Paris 348 (2010), no. 5–6, 287–289; compare the next note.
23. This lower bound, similar to that on p. 52, was also improved by 2 in E. Brugallé and
B. Bertrand, C. R. Math. Acad. Sci. Paris 348 (2010), no. 5–6, 287–289.
24. In fact, it has already been established above that “complex projective circles” are
homeomorphic (and even diffeomorphic) to a 2-dimensional sphere, which gives the
same result for complex projective lines, since they are diffeomorphic to complex circles.
Arnold gives below an independent and more straightforward explanation.
25. This is an illustration of the “Arnold principle”: If a mathematical notion bears a personal
name, then that name is not the name of the discoverer.
26. More information on these problems, as well as principal proofs (based on the Picard–
Lefschetz monodromy theory), can be found in V. A. Vassiliev, Ramified Integrals, Singu-
larities, and Lacunas, Kluwer Academic Publishers, 1995.
27. In the original, Arnold requires the inequality n ≥ r (r − 1) /2, but in fact, he uses
n ≥ r (r + 1) /2.
28. Here the proof is slightly corrected in order to cover all possible situations, while in
the original, Arnold presents only one situation.
29. In the original, Arnold writes k = n − r, but we have written it as r = n − k to stress
the fact that this formula actually defines the parameter r.
30. R. Cordovil showed in “Sur l’evaluation t( M; 2, 0) du polynôme de Tutte d’un ma-
troı̈de et une conjecture de B. Grünbaum relative aux arrangements de droites du plan,”
European J. Combin. 1 (1980), 317–322, that 9 lines cannot divide the projective plane into
23 regions. A complete answer to the general problem was discovered by N. Martinov:
“Classification of Arrangements by the Number of Their Cells,” Discrete Comput. Geom. 9
(1993), 39–46. An induction step in his proof is apparently questionable. For a new proof,
see I. Shnurnikov, “Into How Many Regions Do n Lines Divide the Plane If at Most k of
Them Are Concurrent?” Moscow Univ. Math. Bull. 65 (2010), no. 5, 208–212.
Collana Unitext – La Matematica per il 3+2

Series Editors:
A. Quarteroni (Editor-in-Chief)
L. Ambrosio
P. Biscari
C. Ciliberto
G. van der Geer
G. Rinaldi
W.J. Runggaldier

Editor at Springer:
F. Bonadei
[email protected]

As of 2004, the books published in the series have been given a volume num-
ber. Titles in grey indicate editions out of print.
As of 2011, the series also publishes books in English.

A. Bernasconi, B. Codenotti
Introduzione alla complessità computazionale
1998, X+260 pp, ISBN 88-470-0020-3
A. Bernasconi, B. Codenotti, G. Resta
Metodi matematici in complessità computazionale
1999, X+364 pp, ISBN 88-470-0060-2
E. Salinelli, F. Tomarelli
Modelli dinamici discreti
2002, XII+354 pp, ISBN 88-470-0187-0
S. Bosch
Algebra
2003, VIII+380 pp, ISBN 88-470-0221-4
S. Graffi, M. Degli Esposti
Fisica matematica discreta
2003, X+248 pp, ISBN 88-470-0212-5
S. Margarita, E. Salinelli
MultiMath – Matematica Multimediale per l’Università
2004, XX+270 pp, ISBN 88-470-0228-1

A. Quarteroni, R. Sacco, F.Saleri

Matematica numerica (2a Ed.)
2000, XIV+448 pp, ISBN 88-470-0077-7
2002, 2004 ristampa riveduta e corretta
(1a edizione 1998, ISBN 88-470-0010-6)

13. A. Quarteroni, F. Saleri

Introduzione al Calcolo Scientifico (2a Ed.)
2004, X+262 pp, ISBN 88-470-0256-7
(1a edizione 2002, ISBN 88-470-0149-8)

14. S. Salsa
Equazioni a derivate parziali - Metodi, modelli e applicazioni
2004, XII+426 pp, ISBN 88-470-0259-1

15. G. Riccardi
Calcolo differenziale ed integrale
2004, XII+314 pp, ISBN 88-470-0285-0

16. M. Impedovo
Matematica generale con il calcolatore
2005, X+526 pp, ISBN 88-470-0258-3

17. L. Formaggia, F. Saleri, A. Veneziani

Applicazioni ed esercizi di modellistica numerica
per problemi differenziali
2005, VIII+396 pp, ISBN 88-470-0257-5

18. S. Salsa, G. Verzini

Equazioni a derivate parziali – Complementi ed esercizi
2005, VIII+406 pp, ISBN 88-470-0260-5
2007, ristampa con modifiche

19. C. Canuto, A. Tabacco

Analisi Matematica I (2a Ed.)
2005, XII+448 pp, ISBN 88-470-0337-7
(1a edizione, 2003, XII+376 pp, ISBN 88-470-0220-6)
20. F. Biagini, M. Campanino
Elementi di Probabilità e Statistica
2006, XII+236 pp, ISBN 88-470-0330-X

21. S. Leonesi, C. Toffalori

Numeri e Crittografia
2006, VIII+178 pp, ISBN 88-470-0331-8
22. A. Quarteroni, F. Saleri
Introduzione al Calcolo Scientifico (3a Ed.)
2006, X+306 pp, ISBN 88-470-0480-2
23. S. Leonesi, C. Toffalori
Un invito all’Algebra
2006, XVII+432 pp, ISBN 88-470-0313-X
24. W.M. Baldoni, C. Ciliberto, G.M. Piacentini Cattaneo
Aritmetica, Crittografia e Codici
2006, XVI+518 pp, ISBN 88-470-0455-1
25. A. Quarteroni
Modellistica numerica per problemi differenziali (3a Ed.)
2006, XIV+452 pp, ISBN 88-470-0493-4
(1a edizione 2000, ISBN 88-470-0108-0)
(2a edizione 2003, ISBN 88-470-0203-6)
26. M. Abate, F. Tovena
Curve e superfici
2006, XIV+394 pp, ISBN 88-470-0535-3
27. L. Giuzzi
Codici correttori
2006, XVI+402 pp, ISBN 88-470-0539-6
28. L. Robbiano
Algebra lineare
2007, XVI+210 pp, ISBN 88-470-0446-2
29. E. Rosazza Gianin, C. Sgarra
Esercizi di finanza matematica
2007, X+184 pp,ISBN 978-88-470-0610-2
30. A. Machì
Gruppi – Una introduzione a idee e metodi della Teoria dei Gruppi
2007, XII+350 pp, ISBN 978-88-470-0622-5
2010, ristampa con modifiche
31 Y. Biollay, A. Chaabouni, J. Stubbe
Matematica si parte!
A cura di A. Quarteroni
2007, XII+196 pp, ISBN 978-88-470-0675-1
32. M. Manetti
Topologia
2008, XII+298 pp, ISBN 978-88-470-0756-7
33. A. Pascucci
Calcolo stocastico per la finanza
2008, XVI+518 pp, ISBN 978-88-470-0600-3
34. A. Quarteroni, R. Sacco, F. Saleri
Matematica numerica (3a Ed.)
2008, XVI+510 pp, ISBN 978-88-470-0782-6
35. P. Cannarsa, T. D’Aprile
Introduzione alla teoria della misura e all’analisi funzionale
2008, XII+268 pp, ISBN 978-88-470-0701-7
36. A. Quarteroni, F. Saleri
Calcolo scientifico (4a Ed.)
2008, XIV+358 pp, ISBN 978-88-470-0837-3
37. C. Canuto, A. Tabacco
Analisi Matematica I (3a Ed.)
2008, XIV+452 pp, ISBN 978-88-470-0871-3
38. S. Gabelli
Teoria delle Equazioni e Teoria di Galois
2008, XVI+410 pp, ISBN 978-88-470-0618-8
39. A. Quarteroni
Modellistica numerica per problemi differenziali (4a Ed.)
2008, XVI+560 pp, ISBN 978-88-470-0841-0
40. C. Canuto, A. Tabacco
Analisi Matematica II
2008, XVI+536 pp, ISBN 978-88-470-0873-1
2010, ristampa con modifiche
41. E. Salinelli, F. Tomarelli
Modelli Dinamici Discreti (2a Ed.)
2009, XIV+382 pp, ISBN 978-88-470-1075-8
42. S. Salsa, F.M.G. Vegni, A. Zaretti, P. Zunino
Invito alle equazioni a derivate parziali
2009, XIV+440 pp, ISBN 978-88-470-1179-3
43. S. Dulli, S. Furini, E. Peron
Data mining
2009, XIV+178 pp, ISBN 978-88-470-1162-5
44. A. Pascucci, W.J. Runggaldier
Finanza Matematica
2009, X+264 pp, ISBN 978-88-470-1441-1
45. S. Salsa
Equazioni a derivate parziali – Metodi, modelli e applicazioni (2a Ed.)
2010, XVI+614 pp, ISBN 978-88-470-1645-3
46. C. D’Angelo, A. Quarteroni
Matematica Numerica – Esercizi, Laboratori e Progetti
2010, VIII+374 pp, ISBN 978-88-470-1639-2
47. V. Moretti
Teoria Spettrale e Meccanica Quantistica – Operatori in spazi di Hilbert
2010, XVI+704 pp, ISBN 978-88-470-1610-1
48. C. Parenti, A. Parmeggiani
Algebra lineare ed equazioni differenziali ordinarie
2010, VIII+208 pp, ISBN 978-88-470-1787-0
49. B. Korte, J. Vygen
Ottimizzazione Combinatoria. Teoria e Algoritmi
2010, XVI+662 pp, ISBN 978-88-470-1522-7
50. D. Mundici
Logica: Metodo Breve
2011, XII+126 pp, ISBN 978-88-470-1883-9
51. E. Fortuna, R. Frigerio, R. Pardini
Geometria proiettiva. Problemi risolti e richiami di teoria
2011, VIII+274 pp, ISBN 978-88-470-1746-7
52. C. Presilla
Elementi di Analisi Complessa. Funzioni di una variabile
2011, XII+324 pp, ISBN 978-88-470-1829-7
53. L. Grippo, M. Sciandrone
Metodi di ottimizzazione non vincolata
2011, XIV+614 pp, ISBN 978-88-470-1793-1
54. M. Abate, F. Tovena
Geometria Differenziale
2011, XIV+466 pp, ISBN 978-88-470-1919-5
55. M. Abate, F. Tovena
Curves and Surfaces
2011, XIV+390 pp, ISBN 978-88-470-1940-9
56. A. Ambrosetti
Appunti sulle equazioni differenziali ordinarie
2011, X+114 pp, ISBN 978-88-470-2393-2
57. L. Formaggia, F. Saleri, A. Veneziani
Solving Numerical PDEs: Problems, Applications, Exercises
2011, X+434 pp, ISBN 978-88-470-2411-3
58. A. Machì
Groups. An Introduction to Ideas and Methods of the Theory of Groups
2011, XIV+372 pp, ISBN 978-88-470-2420-5
59. A. Pascucci, W.J. Runggaldier
Financial Mathematics. Theory and Problems for Multi-period Models
2011, X+288 pp, ISBN 978-88-470-2537-0
60. D. Mundici
Logic: a Brief Course
2012, XII+124 pp, ISBN 978-88-470-2360-4
61. A. Machì
Algebra for Symbolic Computation
2012, VIII+174 pp, ISBN 978-88-470-2396-3
62. A. Quarteroni, F. Saleri, P. Gervasio
Calcolo Scientifico (5a ed.)
2012, XVIII+450 pp, ISBN 978-88-470-2744-2
63. A. Quarteroni
Modellistica Numerica per Problemi Differenziali (5a ed.)
2012, XVIII+628 pp, ISBN 978-88-470-2747-3
64. V. Moretti
Spectral Theory and Quantum Mechanics
With an Introduction to the Algebraic Formulation
2013, XVI+728 pp, ISBN 978-88-470-2834-0
65. S. Salsa, F.M.G. Vegni, A. Zaretti, P. Zunino
A Primer on PDEs. Models, Methods, Simulations
2013, XIV+484 pp, ISBN 978-88-470-2861-6

66. V.I. Arnold

Real Algebraic Geometry
2013, X+110 pp, ISBN 978-3-642-36242-2
The online version of the books published in this series is available at
SpringerLink.
For further information, please visit the following link:
http://www.springer.com/series/5418