ENCOMPLEXING THE WRITHE

arXiv:math/0005162v1 [math.AG] 16 May 2000

OLEG VIRO

Uppsala University, Uppsala, Sweden;

POMI, St. Petersburg, Russia

Abstract. For a nonsingular real algebraic curve in 3-dimensional pro-

jective space or 3-sphere, a new integer-valued characteristic is intro-
duced. It is invariant under rigid isotopy and multiplied by −1 under
mirror reflections. In a sense, it is a Vassiliev invariant of degree 1 and
a counterpart of a link diagram writhe.

1. Introduction
This paper is a detailed version of my preprint [10], which was written
about five years ago. Here I do not discuss results that have appeared since
then. I plan to survey them soon in another paper. The subject is now
evolving into a real algebraic knot theory.
This paper is dedicated to the memory of my teacher Vladimir Abramovich
Rokhlin. It was V. A. Rokhlin, who suggested to me, a long time ago, in
1977, to develop a theory of real algebraic knots. He suggested this as a
topic for my second dissertation (after PhD, like habilitation). Following
this suggestion, I moved then from knot theory and low-dimensional topol-
ogy to the topology of real algebraic varieties. However, in the topology of
real algebraic varieties, problems on spatial surfaces and plane curves were
more pressing than problems on spatial curves, and my second dissertation
defended in 1983 was devoted to the constructions of real algebraic plane
curves and spatial surfaces with prescribed topology.
The change in the topic occured mainly because I managed to obtain
decent results in another direction, on plane curves. There was also a less
respectable reason: I failed to relate the traditional techniques of classical
knot theory to real algebraic knots. One of the obstacles was a phenomenon
which became the initial point of this paper. A large part of the traditional
techniques in knot theory uses plane knot diagrams, i.e., projections of knots
to the plane. The projection of an algebraic curve is algebraic, and one could
try to apply results on plane real algebraic curves. However, the projection
contains extra real points, which do not correspond to real points of the
knot. These points are discussed below. In the seventies they ruined my
weak attempts to study real algebraic knots. Now they allow us to detect
crucial differences between topological and real algebraic knots.

1991 Mathematics Subject Classification. 57M25, 14G30, 14H99.

Key words and phrases. classical link, real algebraic link, linking number, self-linking
number, writhe, framing, Vassiliev invariant, isotopy, rigid isotopy.
1
2 OLEG VIRO

I am grateful to Alan Durfee, Tobias Ekholm, and V. M. Kharlamov for

stimulating conversations.
The lengthy informal introduction, which follows, is intended to explain
the matter prior to going into details. I cannot resist the temptation to
write in the style of popular mathematics and apologize to the reader whom
this style may irritate.

1.1. Knot theory and algebraic geometry. In classical knot theory,

by a link one means a smooth closed 1-dimensional submanifold of the 3-
dimensional sphere S 3 , i.e., the union of several disjoint circles smoothly
embedded into S 3 . A link may be equipped with various additional struc-
tures such as orientation or framing and considered up to various equivalence
relations like smooth (or ambient) isotopy, PL-isotopy, cobordism or homo-
topy. See, e.g., [5] or [1].
In algebraic geometry classical links naturally appear as links of singular
points of complex plane algebraic curves. Given a singular point p of a
complex plane algebraic curve C, the intersection of C with the boundary
of a sufficiently small ball centered at p is called the link of the singularity.
It provides a base for a fruitful interaction between topology and algebraic
geometry with a long history and lots of important results.
Another obvious opportunity for interaction between algebraic geometry
and knot theory is based on the fact that a classical link may emerge as the
set of real points of a real algebraic curve. This opportunity was completely
ignored, besides that a number of times it was proved that any classical
link is approximated by (and hence isotopic to) the set of real points of a
real algebraic curve. There are two natural directions in which algebraic
geometry and knot theory may interact in the study of real algebraic links:
first, the study of relationships between invariants which are provided by
link theory and algebraic geometry, second, developing a theory parallel to
the classical link theory, but taking into account the algebraic nature of the
objects. From the viewpoint of this second direction, it is more natural to
consider real algebraic links up to isotopy consisting of real algebraic links,
which belong to the same continuous family of algebraic curves, rather than
up to smooth isotopy in the class of classical links. I call an isotopy of
the former kind a rigid isotopy, following the terminology established by
Rokhlin [4] in a similar study of real algebraic plane projective curves and
the likes (see, e.g., the survey [9]). Of course, there is a forgetting functor:
any real algebraic link can be regarded as a classical link and a rigid isotopy
as a smooth isotopy. It is interesting to see how much is lost under that
transition.
In this paper I point out a real algebraic link invariant which is lost. It is
unexpectedly simple. In an obvious sense it is a nontrivial Vassiliev invariant
of degree 1 on the class of real algebraic knots (recall that a knot is a link
consisting of one component). In classical knot theory the lowest degree of a
nontrivial Vassiliev knot invariant is 2. Thus there is an essential difference
between classical knot theory and the theory of real algebraic knots. Of
course this difference has a simple topological explanation: a real algebraic
link is more complicated topologically, besides its set of real points contains
 ENCOMPLEXING THE WRITHE 3

the set of complex points invariant under the complex conjugation and a
rigid isotopy induces an equivariant smooth isotopy of this set.
The invariant of real algebraic links which is defined below is very similar
to the self-linking number of a framed knot. In [10] I call it also the self-
linking number. Its definition looks like a replacement of an elementary
definition of the writhe of a knot diagram, but taking into consideration the
imaginary part of the knot.
1.2. The word ‘encomplex’. Here I propose to change this name (i.e.,
self-linking number) to encomplexed writhe, and, in general, since many
other characteristics can also be enhanced in a similar way, I suggest a
new verb encomplex for similar enhancements by taking into consideration
additional imaginary ingredients. This would agree with the general usage of
the prefix en- which is described in the Oxford Dictionary of Current English
as follows: “en- prefix . . . forming verbs . . . 1 from nouns, meaning ‘put into
or on’ (engulf ; entrust; embed), 2 from nouns or adjectives, meaning ‘bring
into the condition of’ (enslave) . . . ”.
The word complexification does not seem to be appropriate for what we
do here with the writhe. A complexification of the writhe should be a
complex counterpart for the writhe, it should be a characteristic of complex
objects, while our enhancement of the writhe is defined only for real objects
possessing complexification.
1.3. Self-linking and writhe of nonalgebraic knots. The linking num-
ber is a well-known numerical characteristic of a pair of disjoint oriented
circles embedded in three-dimensional Euclidean space. Roughly speaking,
it measures how many times one of the circles runs around the other. It is
one of the most classical topological invariants, introduced in the nineteenth
century by Gauss [3].
In the classical theory, a self-linking number of a knot is defined if the knot
is equipped with an additional structure like a framing or just a vector field
nowhere tangent to the knot.1 The self-linking number is the linking number
of the knot oriented somehow and its copy obtained by a small shift in the
direction specified by the vector field. It does not depend on the orientation,
since reversing the orientation of the knot is compensated by reversing the
induced orientation of its shifted copy. Of course, the self-linking number
depends on the homotopy class of the vector field.
A knot has no natural preferable homotopy class of framings, which would
allow us to speak about a self-linking number of the knot without a special
care on the choice of the framing.2 Some framings appear naturally in
geometric situations. For example, if one fixes a generic projection of a knot
to a plane, the vector field of directions of the projection appears. The
corresponding self-linking number is called the writhe of the knot. However,
it depends on the choice of the projection and changes under isotopy.
1
A framing is a pair of normal vector fields on the knot orthogonal to each other.
There is an obvious construction that makes a framing from a nontangent vector field
and establishes a one-to-one correspondence between homotopy classes of framings and
nontangent vector fields. The vector fields are more flexible and relevant to the case.
2
Moreover, the self-linking number is used to define a natural class of framings: namely,
the framings with self-linking number zero.
4 OLEG VIRO

The linking number is a Vassiliev invariant of order 1 of two-component

oriented links. This means that it changes by a constant (in fact, by 2)
when the link experiences a homotopy with the generic appearance of an
intersection point of the components. Whether the linking number increases
or decreases depends only on the local picture of orientations near the double
point: when it passes from through to , the linking number increases
by 2. Generalities on Vassiliev invariants see, e.g., in [8].
In a sense the linking number is the only Vassiliev invariant of degree 1
of two-component oriented links: any Vassiliev invariant of degree 1 of two-
component oriented links is a linear function of the linking number. Simi-
larly, the self-linking number is a Vassiliev invariant of degree 1 of framed
knots (it changes by 2 when the knot experiences a homotopy with a generic
appearance of a self-intersection point) and it is the only Vassiliev of degree
1 of framed knots in the same sense. The necessity of a framing for the def-
inition of self-linking number can now be formulated more rigorously: only
constants are Vassiliev invariants of degree 1 of (non-framed) knots.
The diagrammatical definition of the writhe, which is imitated below,
runs as follows: for each crossing point of the knot projection, one defines
a local writhe equal to +1 if near the point the knot diagram looks like
and −1 if it looks like . Then one sums the local writhes over all double
points of the projection. The sum is the writhe.
A continuous change of the projection may cause the vanishing of a cross-
ing point. This happens under the first Reidemeister move shown in the left
hand half of Figure 1. This move changes the writhe by ±1.

1.4. Algebraicity enhances the writhe. If a link is algebraic, then its

projection to a plane is algebraic, too. A generic projection has only ordinary
double points and the total number of its complex double points is constant.3
The number of real double points can vary, but only by an even number.
A real double point cannot turn alone into an imaginary one, as it seems
to happen under the first Reidemeister move. Under an algebraic version
of the first Reidemeister move, the double point stays in the real domain,
but becomes solitary, like the only real point of the curve x2 + y 2 = 0. The
algebraic version of the first Reidemeister move is shown in the right hand
half of Figure 1.
It is not difficult to prove that the family of spatial curves that realizes
this move can be transformed by a local diffeomorphism to the family of
affine curves defined by the following system of equations
(
xz + y = 0,
x + z 2 + τ = 0,

3
Here by a generic projection we mean a projection from a generic point. When one
says that a generic projection has some properties, this means that for an open everywhere
dense set of points the projection from any point of this set has these properties. The whole
set of undesirable points is closed nowhere dense although it depends on the properties
under consideration. A proof is an easy exercise either on Sard’s Lemma, or Bertini’s
Theorem.
 ENCOMPLEXING THE WRITHE 5

Figure 1. Topological (left) and real algebraic (right) ver-

sions of the first Reidemeister move. At the solitary crossing
point, which is on the right hand side of the picture, the con-
jugate imaginary branches are indicated by dashed segments,
according to an outdated tradition of Analytic Geometry.

where τ is the parameter of the deformation. These are rational curves,

admitting a rational parametrization
 2
 x = −t − τ,

y = −t(t2 + τ ),

z = −t.


The projection corresponds to the standard projection (x, y, z) 7→ (x, y) to

the coordinate xy-plane. It maps these curves to the family of affine plane
rational cubic curves defined by y 2 + x2 (τ + x) = 0 with τ ∈ R.
A solitary double point of the projection is not the image of any real point
of the link. It is the image of two imaginary complex conjugate points of the
complexification of the link. The preimage of the point in the 3-space under
the projection is a real line. It is disjoint from the real part of the link, but
intersects its complexification in a couple of complex conjugate imaginary
points.
In the model of the first Reidemeister move above, (0, 0) is the double
point of the projection for each τ 6= 0. If τ < 0,√it is a usual crossing
√ point.
Its preimage consists of two real points (0, 0, −τ ) and (0, 0, − −τ ). If
τ > 0, it is a solitary double
√ point. Its preimage
√ consists of two imaginary
conjugate points (0, 0, i τ ) and (0, 0, −i τ ), which lie on a real line x =
y=0
Below, in Section 2.2, with any solitary double point of the projection, a
local writhe equal to ±1 is associated. This is done in such a way that the
local writhe of the crossing point vanishing in the first Reidemeister move
is equal to the local writhe of the new-born solitary double point. In the
case of an algebraic knot, the sum of local writhes of all double points, both
solitary and crossings, does not depend on the choice of projection and is
invariant under rigid isotopy. This sum is the encomplexed writhe.
1.5. Encomplexed writhe for nonoriented and semi-oriented links.
A construction similar to the construction of the encomplexed writhe number
of an algebraic knot can be applied to an algebraic link. However in this
case there are two versions of the construction.
In the first of these, we define an encomplexed writhe number generalizing
the encomplexed writhe number defined above for knots. We consider a link
diagram and the sum of local writhes at solitary double points and crossing
points where the branches belong the same connected component of the set
6 OLEG VIRO

of real points. At these crossing point, to define a local writhe, we need

orientations of the branches. As above, we choose an orientation on each of
the components. If we make another choice, at a crossing point for which
the branches belong the same component, either both orientations change
or none. Hence the local writhe numbers at crossing points of this kind do
not depend on the choice. In Section 2 below, we prove that the whole sum
of local writhes over crossing points of this kind and solitary double points
does not depend on the projection and is invariant under rigid isotopy. We
call this sum the encomplexed writhe number of the link A and denote by
Cw(A).
In the second version of the construction, we consider a real algebraic
link which is equipped with an orientation of the set of real points, use
these orientations to define local writhe numbers at all crossing points and
sum the local writhe numbers over all crossing points and all solitary double
points. The result is called the encomplexed writhe number of an oriented
real algebraic link. This encomplexed writhe number does not change when
the orientation reverses. An orientation considered up to reversing is called
a semi-orientation. Thus the encomplexed writhe number depends only on
the semi-orientation of the link.
The (semi-)orientation may be an artificial extra structure, but it may also
appear in a natural way, say, as a complex orientation, if the set of real points
divides the set of real points, see [4]. In fact, the complex orientation is
defined up to reversing, so it is indeed a semi-orientation. Another important
class of semi-oriented algebraic links appears as transversal intersections of
two real algebraic surfaces of degrees p and q with p ≡ q mod 2.
The encomplexed writhe number of (semi-)oriented real algebraic link dif-
fers from the encomplexed writhe number of the same link without orienta-
tion by the sum of all pairwise linking numbers of the components multiplied
by 2: let A be a real algebraic link, let Ā be the same link equipped with an
orientation of its set of real points and Ā1 , . . . , Ān the (oriented) connected
components of this set, then
X
Cw(Ā) = Cw(A) + 2 lk(Āi , Āj ).
16i6j6n

1.6. Encomplexed writhe and framings. In the case of a knot, the

encomplexed writhe number defines a natural class of framings, since homo-
topy classes of framings are enumerated by their self-linking numbers and
we can choose the framing having the self-linking number equal to the al-
gebraic encomplexed writhe number. I do not know any direct construction
of this framing. Moreover, there seems to be a reason for the absence of
such a construction. In the case of links, the construction above gives a sin-
gle number, while framings are enumerated by sequences of numbers with
entries corresponding to components.

2. Real algebraic projective links

Let A be a nonsingular real algebraic curve in 3-dimensional projective
space. Then the set RA of its real points is a smooth closed 1-dimensional
 ENCOMPLEXING THE WRITHE 7

submanifold of RP 3 , i.e., a smooth projective link. The set CA of its complex

points is a smooth complex 1-dimensional submanifold of CP 3 .
Let c be a point of RP 3 . Consider the projection pc : CP 3 rc → CP 2 from
c. Assume that c is such that the restriction to CA of pc is generic. This
means that it is an immersion without triple points and at each double point
the images of the branches have distinct tangent lines. It follows from well-
known theorems that those c’s for which this is the case form an open dense
subset of RP 3 (in fact, it is the complement of a 2-dimensional subvariety).
The real part pc (CA) ∩ RP 2 of the image consists of the image pc (RA) of
the real part and, maybe, several solitary points, which are double points of
pc (CA).
2.1. The local writhe of a crossing. There is a purely topological con-
struction which assigns a local writhe equal to ±1 to a crossing belonging
to the image of only one component of RA. This construction is well-known
in the case of classical knots. Here is its projective version. I borrow it
from Drobotukhina’s paper [2] on the generalization of Kauffman brackets
to links in projective space.

Figure 2. Construction of the frame v, l, w′ .

Let K be a smooth connected one-dimensional submanifold of RP 3 , and

c be a point of RP 3 r K. Let x be a generic double point of the projection
pc (K) ⊂ RP 2 and L ⊂ RP 3 be the line which is the preimage of x under
the projection. Denote by a and b the points of L ∩ RP 3 . The points a and
b divide the line L into two segments. Choose one of them and denote it by
S. Choose an orientation of K. Let v and w be tangent vectors of K at a
and b respectively directed along the selected orientation of K.
Let l be a vector tangent to L at a and directed inside S. Let w′ be a
vector at a such that it is tangent to the plane containing L and w and is
directed to the same side of S as w (in an affine part of the plane containing
S and w). See Figure 2. The triple v, l, w′ is a base of the tangent space
Ta RP 3 . Define the local writhe of x to be the value taken by the orientation
of RP 3 on this frame.
The construction of the local writhe of x contains several choices. Here
is a proof that the result does not depend on them.
We have chosen an orientation of K. Had the opposite orientation been
selected, then v and w′ would be replaced by the opposite vectors −v and
−w′ . This would not change the result, since −v, l, −w′ defines the same
orientation as v, l, w′ .
We have chosen the segment S. If the other half of L was selected, then
l and w′ would be replaced by the opposite vectors. But v, −l, −w′ defines
the same orientation as v, l, w′ .
8 OLEG VIRO

The construction depends on the order of points a and b. The other choice
(with the same choice of the orientation of K and segment S) gives a triple
of vectors at b. It can be moved continuously without degeneration along S
into the triple w′ , −l, v, which defines the same orientation as v, l, w′ .
2.2. Local writhe of a solitary double point. Let A, c, and pc be as in
the beginning of Section 2 and let s ∈ RP 2 be a solitary double point of pc .
Here is a construction assigning ±1 to s. I will also call the result a local
writhe of s.
Denote the preimage of s under pc by L. This is a real line in RP 3
connecting c and s. It intersects CA in two imaginary complex conjugate
points, say, a and b. Since a and b are conjugate, they belong to different
components of CL r RL.
Choose one of the common points of CA and CL, say, a. The natural
orientation of the component of CL r RL defined by the complex structure
of CL induces an orientation on RL as on the boundary of its closure. The
image under pc of the local branch of CA passing through a intersects the
plane of the projection RP 2 transversally at s. Take the local orientation of
the plane of projection such that the local intersection number of the plane
and the image of the branch of CA is +1.
Thus the choice of one of two points of CA ∩ CL defines an orientation of
RL and a local orientation of the plane of projection RP 2 (we can speak only
of a local orientation of RP 2 , since the whole RP 2 is not orientable). The
plane of projection intersects4 transversally RL in s. The local orientation
of the plane, the orientation of RL and the orientation of the ambient RP 3
determine the intersection number. This is the local writhe.
It does not depend on the choice of a. Indeed, if one chooses b instead,
then both the orientation of RL and the local orientation of RP 2 would be
reversed. The orientation of RL would be reversed, because RL inherits
opposite orientations from the different halves of CL r RL. The local orien-
tation of RP 2 would be reversed, because the complex conjugation involu-
tion conj : CP 2 → CP 2 preserves the complex orientation of CP 2 , preserves
RP 2 (point-wise) and maps one of the branches of pc (CA) at s to the other
reversing its complex orientation.
2.3. Encomplexed writhe and its invariance. Now for any real alge-
braic projective link A, choose a point c ∈ RP 3 such that the projection of A
from c is generic and sum the writhes of all crossing points of the projection
belonging to the image of only one component of RA and the writhes of all
solitary double points. This sum is called the encomplexed writhe number
of A.
I have to show that it does not depend on the choice of projection. The
proof given below proves more: the sum is invariant under rigid isotopy of A.
By rigid isotopy we mean an isotopy consisting of nonsingular real algebraic
curves. The effect of a movement of c on the projection can be achieved by a
rigid isotopy defined by a path in the group of projective transformations of
4
We may think on the plane of projection as embedded into RP 3 . If you would like
to think on it as on the set of lines of RP 3 passing through c, please, identify it in a
natural way with any real projective plane contained in RP 3 and disjoint from c. All such
embeddings RP 2 → RP 3 are isotopic.
 ENCOMPLEXING THE WRITHE 9

RP 3 . Therefore the following theorem implies both the independence of the

encomplexed writhe number from the choice of projection and its invariance
under rigid isotopy.
2.A. Theorem. For any two rigidly isotopic real algebraic projective links
A1 and A2 whose projections from the same point c ∈ RP 3 are generic, the
encomplexed writhe numbers of A1 and A2 defined via c are equal.
This theorem is proved in Section 2.5.
2.B. Corollary 1. The encomplexed writhe number of a real algebraic pro-
jective link does not depend on the choice of the projection involved in its
definition.
Proof of 2.B. A projection depends only on the center from which it is done.
The effect on the projection of a movement of the center can be achieved by
a rigid isotopy defined by a path in the group of projective transformations
of RP 3 .

Thus the encomplexed writhe number is a characteristic of a real algebraic

link.
2.C . Corollary 2. The encomplexed writhe number of a real algebraic pro-
jective link is invariant under rigid isotopy.

Figure 3.

2.4. Algebraic counterparts of Reidemeister moves. As in the purely

topological situation of an isotopy of a classical link, a generic rigid isotopy of
a real algebraic link may be decomposed into a composition of rigid isotopies,
each of which involves a single local standard move of the projection. There
are 5 local standard moves. They are similar to the Reidemeister moves.
The first of these 5 moves is shown in the right hand half of Figure 1. The
other moves are shown in Figure 3. The first two of these coincide with
the second and third Reidemeister moves. The fourth move is similar to
the second Reidemeister move: also two double points of projection come to
each other and disappear. However the double points are solitary. The fifth
move is similar to the third Reidemeister move: a triple point also appears
for a moment. But at this triple point only one branch is real, the other two
are imaginary conjugate to each other. In this move a solitary double point
traverses a real branch.
10 OLEG VIRO

2.5. Reduction of Theorem 2.A to Lemmas. To prove Theorem 2.A,

first replace the rigid isotopy by a generic one and then decompose the
latter into local moves described above, in Section 2.4. Only in the first,
fourth and fifth moves solitary double points are involved. The invariance
under the second and the third move follows from the well-known fact of
knot theory that the topological writhe is invariant under the second and
third Reidemeister moves. Cf. [2]. Thus the following three lemmas imply
Theorem 2.A.
2.D. Lemma. In the fifth move the writhe of the solitary point does not
change.

2.E . Lemma. In the fourth move the writhes of the vanishing solitary
points are opposite.

2.F . Lemma. In the first move the writhe of vanishing crossing point is
equal to the writhe of the new-born solitary point.
2.6. Proof of Lemmas 2.D and 2.E. Proof of Lemma 2.F is postponed
to Section 2.7. Note that although Lemma 2.F is the most difficult to prove,
it is the least significant: here its only role is to justify the choice of sign made
in the definition of local writhe in solitary double point of the projection. It
is clear that the writhes of vanishing double points involved in the first move
are related, and if they were opposite to each other, then the definition of the
encomplexed writhe number should be changed, but would not be destroyed
irrecoverably.
Proof of Lemma 2.D. This is obvious. Indeed, the real branch of the pro-
jection does not interact with the imaginary branches, it just passes through
their intersection point.
Proof of Lemma 2.E. At the moment of the fourth move take a small ball B
in the complex projective plane centered in the solitary self-tangency point
of the projection of the curve. Its intersection with the projection of the
complex point set of the curve consists of two smoothly embedded disks
tangent to each other and to the disk B ∩ RP 2 . Under the move each of the
disks experiences a diffeotopy. Before and after the move the intersection the
curve with B is the union of the two disks meeting each other transversally
in two points, but before the move the disks do not intersect RP 2 , while
after the move they intersect RP 2 in their common points.
To calculate the writhe at both vanishing solitary double points, let us
select the same imaginary branch of the projection of the curve passing
through the points. This means that we select one of the disks described
above. The sum of the local intersection numbers of this disk (equipped with
the complex orientation) and B ∩ RP 2 (equipped with some orientation) is
zero since under the fourth move the intersection disappears, while in the
boundary of B no intersection happens.
Therefore the local orientations of the projective plane in the vanishing
solitary double points defined by this branch define opposite orientations
of B ∩ RP 2 . (Recall that the local orientations are distinguished by the
condition that the local intersection numbers are positive.)
 ENCOMPLEXING THE WRITHE 11

On the other hand, under the move the preimages of the vanishing solitary
double points come to each other up to coincidence at the moment of the
move and their orientations defined by the choice of the same imaginary
branch are carried to the same orientation of the preimage of the point of
solitary self-tangency. Indeed, the preimages are real lines and points of
intersection of their complexifications with the selected imaginary branch
of the curve also come to the same position. Therefore the halves of the
complexifications containing the points come to coincidence, as well as the
orientations defined by the halves on the real lines.
It follows that the intersection numbers of B with the preimages of the
vanishing solitary double points equipped with these orientations are equal.
Since the local orientations of the projective plane in the vanishing solitary
double points define distinct orientations of B∩RP 2 , the writhes are opposite
to each other.

2.7. Proof of Lemma 2.F. It is sufficient to consider the model family

of curves described in Section 1.4. Recall that the curves of this family are
defined by the following system of equations
(
xz + y = 0,
x + z 2 + τ = 0,
where τ is the parameter of the deformation. These curves admit a rational
parametrization
 2
 x = −t − τ,

y = −t(t2 + τ ),

z = −t.


The projection corresponds to the standard projection (x, y, z) 7→ (x, y) to

the coordinate xy-plane. It maps these curves to the family of affine plane
rational cubic curves defined by y 2 + x2 (τ + x) = 0 with τ ∈ R.
We must prove that the local writhe at (0, 0) for τ < 0 coincides with the
local writhe at (0, 0) for τ > 0. √
Let us calculate the local writhe for τ < 0. Denote −τ by ρ. The preim-
age of (0, 0) consists of points a = (0, 0, ρ) and b = (0, 0, −ρ) corresponding
to the values −ρ and ρ of t, respectively, see Figure 4. The tangent vectors to
the curve at these points are v = (2ρ, −2ρ2 , −1) and w = (−2ρ, −2ρ2 , −1).
The vector l connecting a and b is (0, 0, −2ρ). By definition, the writhe is
the value taken by the orientation of RP 3 on the frame v, l, w′ . This value is
equal to the value of this orientation on the frame (1, 0, 0), (0, 1, 0), (0, 0, 1)
multiplied by the sign of
2ρ −2ρ2 −1
 

det  0 0 −2ρ = −16ρ4 < 0.

−2ρ −2ρ 2 −1
√
Let us calculate the local writhe for τ > 0. Denote τ by ρ. The preimage
of (0, 0) consists of points a′ = (0, 0, iρ) and b′ = (0, 0, −iρ) corresponding
to the values −iρ and iρ of t. Choose the branch which passes through a′ .
It belongs to the upper half of the line x = y = 0, which induces the positive
orientation of the real part directed along (0, 0, 1). At a′ the branch of the
12 OLEG VIRO

RL
a RA

RA B

Figure 4. Real algebraic version of the first Reidemeister move.

curve has tangent vector v = (2iρ, 2ρ2 , −1) and the real basis consisting of
v and iv = (−2ρ, 2iρ2 , −i) positively oriented with respect to the complex
orientation of this branch. The projection maps this basis to the positively
oriented basis (2iρ, 2ρ2 ), (−2ρ, 2iρ2 ) of the projection of the branch. The
intersection number of this projection and R2 in C2 is the sign of
2ρ 2ρ2 0
 
0
−2ρ 0 0 2ρ2   = −4ρ3 < 0.
det 
 1 0 0 0 
0 0 1 0
Hence the orientation of R2 such that its local intersection number with
the selected branch of the projection does not coincide with the orientation
defined by the standard basis. The intersection number of the line x = y = 0
with the standard orientation and the xy-plane with the standard orientation
is the value of the orientation of the ambient space R3 taken on the standard
basis (1, 0, 0), (0, 1, 0), (0, 0, 1). Therefore the local writhe is opposite to this
value.
Remark. There is a more conceptual proof of Lemma 2.F. It is based on a
local version of the Rokhlin Complex Orientation Formula, see [4] and [9].
In fact, the original proof was done in that way. However, the Complex
Orientation Formula is more complicated than the calculation above.
2.8. Encomplexed writhe of an algebraic link as a Vassiliev invari-
ant of degree one. To speak about Vassiliev invariants, we need to fix
a connected family of curves, in which links under consideration comprise
the complement to a hypersurface. In the case of classical knots one could
include all knots in such a family by adjoining knots with self-intersections
and other singularities. A singular knot is a right equivalence class of a
smooth map of the circle to the space (recall that two maps from a circle
are right equivalent if one of them is a composition of a self-diffeomorphism
of the circle with the other one).
In the case of real algebraic knots, such a family including all real alge-
braic knots does not exist. Even the space of complex curves in the three-
dimensional projective space consists of infinitely many components. It is
impossible to change the homology class realized by the set of complex points
 ENCOMPLEXING THE WRITHE 13

of an algebraic curve in CP 3 by a continuous deformation. Recall that the

homology class belongs to the group H2 (CP 3 ) = Z and is a positive multiple
d[CP 1 ] of the natural generator of [CP 1 ] ∈ H2 (CP 3 ) realized by a line. The
coefficient d is called the order of the curve. The genus is another numerical
characteristic of a complex curve which takes the same value for all nonsin-
gular curves in any irreducible family. As is well known, the nonsingular
complex curves of given order and genus in three-dimensional projective
space are parametrized by a finite union of quasi-projective varieties. For
each of these varieties, one can try to build a separate theory of Vassiliev
invariants on a class of nonsingular real algebraic curves whose complexifi-
cations are parametrized by points of this variety. (A similar phenomenon
takes place in topology: links with different numbers of components can-
not be included into a reasonable connected family, and therefore for each
number of components there is a separate theory of Vassiliev invariants.)
Among the varieties of algebraic curves in three-dimensional projective
space, there are two special families: for each natural number d there is an
irreducible variety of rational curves of order d (recall that a an algebraic
curve is called rational if it admits an algebraic parametrization by a line),
and for each pair of natural numbers p and q there is an irreducible variety
of curves which can be presented as intersection of surfaces of degrees p and
q.
In the class of real algebraic rational curves of order d, singular curves
comprise a discriminant hypersurface in which a generic point is a rational
curve such that it has exactly one singular point and this point is an ordinary
double point. An ordinary double point may be of one of the following two
types: either it is an intersection point of two real branches, or two imaginary
conjugate branches.
Any two real algebraic rational nonsingular curves of order d can be con-
nected by a path in the space of real rational curves of degree d that inter-
sects the discriminant hypersurface only transversally at a finite number of
generic points. Such a path can be regarded as a deformation of a curve to
the other one. When it intersects the discriminant hypersurface at a point,
which is a curve with singularity on real branches, the set of real points
of the curve behaves as in classical knot theory: two pieces of the set of
real points come to each other and pass through each other. As in classical
knot theory, at the moment of intersection, the generic projection of the
curve experiences an isotopy. Nothing happens besides that one crossing
point becomes for a moment the image of a double point and then changes
back into a crossing point, but with the opposite writhe. When the path
intersects the discriminant hypersurface at a point, which is a curve with sin-
gularity on imaginary branches, two complex conjugate imaginary branches
pass through each other. At the moment of passing, they intersect in a real
isolated double point. At this moment the set of real points of a generic
projection experiences an isotopy. No event happens besides that a solitary
double point becomes for a moment the image of a solitary real double point
of the curve and then changes back into an ordinary solitary double point
of the projection (which is not the image of a real point of the knot), but
with the opposite writhe number.
14 OLEG VIRO

It is clear that the encomplexed writhe number of an algebraic curve

changes under a modification of each of these kinds by ±2, with the sign
depending only on the local structure of the modification near the double
point. This means that the encomplexed writhe number on the family of real
rational curves under consideration is a Vassiliev invariant of degree 1.
This is true also for any space of nonsingular real algebraic curves that
can be included into a connected family of real algebraic curves by adjoining
a hypersurface, penetration through which at a generic point looks as in the
family of rational curves described above.
There are many families of this kind besides the families of rational knots.
However, in many families of algebraic curves a transversal penetration
through the discriminant hypersurface in a generic point looks differently.
In particular, for intersections of two surfaces it is a Morse modification
of the real part of the curve. At the moment, the old double points of
the projection, both solitary and crossing, do not change. An additional
double point appears just for a moment. However the division of crossing
points to self-crossing points of a single component and crossing points of
different components may change. Therefore the encomplexed writhe num-
ber changes in a complicated way. If the degrees of the surfaces defining
the curve are of the same parity, the real part of the curve has a natu-
ral semi-orientation. The Morse modification respects this semi-orientation.
Therefore the encomplexed writhe number of the semi-oriented curve does
not change.
2.G. Theorem. The encomplexed writhe number of any nonsingular semi-
oriented real algebraic link which is a transversal intersection of two real
algebraic surfaces whose degrees are of the same parity is zero.
Proof. Any two nonsingular real curves of the type under consideration can
be connected by a path as above. Hence their self-linking numbers coincide.
On the other hand, it is easy to construct, for any pair of natural numbers
p and q of the same parity, a pair of nonsingular real algebraic surfaces of
degrees p and q transversal to each other in three-dimensional projective
space such that their intersection has zero self-linking number.
In contrast to this vanishing result, one can prove that the encomplexed
writhe number of a real algebraic rational knots of degree d can take any value
in the interval between −(d − 1)(d − 2)/2 and (d − 1)(d − 2)/2 including these
limits and congruent to them modulo 2.

3. Generalizations
3.1. The case of an algebraic link with imaginary singularities. The
same construction may be applied to real algebraic curves in RP 3 having
singular imaginary points, but no real singularities. In the construction we
can eliminate projections from the points such that some singular point is
projected from them to a real point. Indeed, for any imaginary point there
exists only one real line passing through it (the line connecting the point
with its complex conjugate), thus we have to exclude a finite number of real
lines.
 ENCOMPLEXING THE WRITHE 15

This gives a generalization of encomplexed writhe numbers with the same

properties: it is invariant with respect to rigid isotopies (i.e., isotopies made
of curves from this class), and is multiplied by −1 under a mirror reflection.

3.2. Real algebraic links in the sphere. The construction of this pa-
per can be applied to algebraic links in the sphere S 3 . Although from the
viewpoint of knot theory this is the most classical case, from the viewpoint
of algebraic geometry the case of curves in the projective space is simpler.
The three-dimensional sphere S 3 is a real algebraic variety. It is a quadric
in four-dimensional real affine space. The stereographic projection is a bi-
rational isomorphism of S 3 onto RP 3 . It defines a diffeomorphism between
the complement of the center of the projection in S 3 and a real affine space.
Given a real algebraic link in S 3 , one may choose a real point of S 3 from
the complement of the link and project the link from this point to an affine
space. Then include the affine space into the projective space and apply the
construction above. The image has no real singular points, therefore we can
use the result of the previous section.
This construction blows up the center of projection, making a real projec-
tive plane out of it, and maps the complement to the center of the projection
in the set of real points of the sphere isomorphically onto the complement
of the projective plane. In the imaginary domain, it contracts each gen-
eratrix of the cone which is the intersection of the sphere with its tangent
plane at the center of projection. The image of the cone is an imaginary
quadric curve contained in the projective plane which appeared as the result
of blowing up of the central point.

3.3. Other generalizations. It is difficult to survey all possible general-

izations. Here I indicate only two directions.
First, consider the most straightforward generalization. Let L be a non-
singular real algebraic (2k−1)-dimensional subvariety in the projective space
of dimension 4k −1. Its generic projection to RP 4k−2 has only ordinary dou-
ble points. At each double point either both branches of the image are real or
they are imaginary complex conjugate. If the set of real points is orientable,
then one can repeat everything with obvious changes and obtain a definition
of a numeric invariant generalizing the encomplexed writhe number defined
above.
Let M be a nonsingular three-dimensional real algebraic variety with
oriented set of real points equipped with a real algebraic fibration over a
real algebraic surface F with fiber a projective line. There is a construction
which assigns to a real algebraic link (i.e., a nonsingular real algebraic curve
in M ) with a generic projection to F an integer, which is invariant under
rigid isotopy, is multiplied by −1 under the orientation reversal in M and
is a Vassiliev invariant of degree 1. This construction is similar to the one
presented above, but uses, instead of the projection to RP 2 , an algebraic
version of Turaev’s shadow descriptions of links [7].

3.4. Not only writhe can be encomplexed. Here we discuss only one
example. However it can be easily generalized. Consider immersions of the
sphere S 2n to R4n . Up to regular homotopy (i.e., a homotopy consisting
of immersions whose differentials also comprise a homotopy), an immersion
16 OLEG VIRO

S 2n → R4n is defined by its Smale invariant [6], which is an element of

π2n (V4n,2n ) = Z. For a generic immersion, it can be expressed as the sum of
local self-intersection numbers over all double points of the immersion, see
[6].
Let us encomplex the Smale invariant. For this, first, we have to consider
a real algebraic counterpart for the notion of generic immersion S 2n →
R4n . The identification is defined via the universal covering R4n → (S 1 )4n .
Replace Euclidean space R4n by torus (S 1 )4n , which has the advantage of
being compact. The classification of immersions S 2n → (S 1 )4n up to regular
homotopy coincides with the Smale classification of immersions S 2n → R4n .
The sphere S 2n is the real part of a quadric projective hypersurface. The
torus (S 1 )4n is the real part of a complex Abelian variety. Consider real
regular maps of the quadric to the Abelian variety. A generic map defines
an immersion both for the complex and real parts. The only singularities
are transversal double points. Double points in the real part of the target
variety are of two kinds. At a double point of the first kind two sheets of
the image of S 2n meet. At a double point of the second kind the images two
complex conjugate sheets of the complexification of S 2n meet. The Smale
invariant is the sum of the local intersection numbers over the double points
of the first kind. One can extend the definition of the local intersection
number to the double points of the second kind in such a way that the total
sum of the local intersection numbers over double points of both kinds would
be invariant under continuous deformations of regular maps.
This total sum is the encomplexed Smale invariant. Notice that it is, in a
sense, more invariant than the original Smale invariant. The Smale invariant
may change under homotopy, it is invariant only under regular homotopy.
The encomplexed Smale invariant does not change under a homotopy in the
class of regular maps, which corresponds to the class of all continuous maps.

References
[1] G. Burde and H. Zieschang, Knots, de Gruyter Studies in Mathematics, vol. 5, Walter
de Gruyter, Berlin–New York, 1985.
[2] Yu. Drobotukhina, An analogue of the Jones polynomial for links in RP 3 and a
generalization of the Kauffman–Murasugi theorem, Algebra i Analiz 2 (1990), no. 3,
171–191; English transl. Leningrad Math. J. 2 (1991), no. 3, 613–630.
[3] K. F. Gauss, Zur Mathematischen Theorie der electrodynamischen Wirkungen, Man-
uscript, First published in his Werke Vol. 5, Königl. Ges. Wiss. Göttingen, Göttingen,
1877, p. 605.
[4] V. A. Rokhlin, Complex topological characteristics of real algebraic curves, Uspekhi
Mat. Nauk 33 (1978), no. 5, 77–89; English transl. Russian Math. Surveys 33 (1978),
no. 5, 85–98.
[5] D. Rolfsen, Knots and links, Publish or Perish, Inc., 1990. Mathematics Lecture
Series, vol. 7, Houston, TX, Publish Perish, 1990.
[6] S. Smale, The classification of immersions of spheres in Euclidean spaces, Ann. of
Math. (2) 69 (1959), 327–344.
[7] V. G. Turaev, Shadow links and face models of statistical mechanics, J. Differential
Geom. 36 (1992), 35–74.
[8] V. Vassiliev, Cohomology of knot spaces, Adv. Soviet Math., vol. 1, Amer. Math. Soc,
1990, pp. 23–70.
 ENCOMPLEXING THE WRITHE 17

[9] O. Ya. Viro, Progress in the topology of real algebraic varieties over the last six years,
Uspekhi Mat. Nauk 41 (1986), no. 3, 45–67; English transl. Russian Math. Surveys
41 (1986), no. 3, 55–82.
[10] , Self-linking number of real algebraic link, Preprint of Uppsala Univer-
sity, U.U.D.M. Report 1994:35, see also http://xxx.lanl.gov/list/math/9410, alg-
geom/9410030.

Matematiska institutionen, Box 480, Uppsala universitet, 751 06, Uppsala,

Sweden
E-mail address: oleg@math.uu.se