You are on page 1of 15

FROM PATHS OF FIRE TO WONDERFUL SPIRALS

Notes to a lecture on selected subjects from the


classical differential geometry of plane curves

    

Artwork by Sven Geier, retrieved from his website


http://www.sgeier.net. Flame fractal Dusty .

,  , 


All images used in these notes were either created by ourselves, or released by their respective
owners into the public domain.


This set of lecture notes, like the lecture itself, consists of two more or less independent
parts.
In the first part, we start with a brief introduction to certain basic notions from
the classical differential geometry of plane curves (essentially repeating some of
the things Prof. Siersma said earlier). After that, we would like to discuss certain
phenomena of geometric optics (viz., those which are collectively known as optical
caustics) from the differential geometer’s point of view. Due to time constraints, we
shall probably only be able to treat in some detail the so-called plane catacaustics of
the first order during the lecture, leaving a few of the variations and generalizations
to the exercises.
In the second part, we discuss several other transformational constructs of classi-
cal differential geometry (evolutes, involutes, pedal curves, etc.), using as our sole
example a special curve of particular beauty. This curve, the so-called logarithmic
spiral, is amazing in the sense that it remains unchanged (up to Euclidean isometries)
under many of the aforementioned transformations. We conclude the notes with a
series of exercises.

[]
 :   

Plane curves: a very brief re-introduction. To us, a plane curve C is simply a one-
dimensional subset of R which we assume to be the image of some function C  R
R  t ( (x(t), y(t)). The function, which need by no means be unique, is called a
parametrization for C. We will assume that the component functions x(t), y(t) have
continuous first and second derivatives, except at certain isolated points of ‘quirkyness’
[allowing phenomena such as cusps to occur, but excluding really wacky fractal-like
behaviour]. While the purist could say that our concept of curve is ridiculously
narrow, we reply that we’re not here today to please purists.
Associated to every ‘reasonable’ plain curve [or only to a particular parametriza-
tion to it] are a number of scalar and vector quantities. Their significance and precize
definition has already been explained by Prof. Siersma. We merely state some of
them here to refresh your memory. Often we will act as if these quantities exist always
everywhere, while waiving our magic wand at every nasty person who says that they
don’t. Again, purists should have stayed at home.
. The arc length between two points C(t ) and C(t) is an intuitively obvious
concept. It is given by

s(t; t ) = ∫ xœ (u) + yœ (u) du.
t

For a curve given in polar co-ordinates r = f(θ) it is given by


θ¼
s= ∫θ
f(ϕ) + (∂ϕ f(ϕ)) dϕ,

where θ and θ are angles corresponding to Cœ (t ) and C(t). The curve is
said to be parametrized by arc length (or to have a natural parametrization) if
YCœ (s(t; t ))Y = .
. The curvature κ(t) is a scalar quantity whose magnitude equals the reciprocal
of the radius of an osculating circle [a circle that closely touches the curve at
the given point C(t)]. It is given by
xœ (t)yœœ (t) − xœœ (t)yœ (t)
κ(t) = .
(xœ (t) + yœ (t))~
. The unit normal vector N(t) is a unit vector perpendicular to the curve at the
point C(t). It is given by
(−yœ (t), xœ (t))
N(t) = .
SCœ (t)S
. Finally, there is associated to any reasonable curve another curve, called the
evolute. Its meaning has already been discussed at length by Prof. Siersma. We
use here the following parametrization to define it:
N(t)
E(t) = C(t) + .
κ(t)

[]
Envelopes. The concept of envelope formalizes the intuitive notion that certain sets
of curves tend to ‘accumulate’ or ‘concentrate’ along some other curve that somehow
wraps itself around the entire set. Here, we will deal exclusively with sets of curves
that are generated by varying a single parameter, the so-called one-parameter families
of curves.
D . The envelope (Dutch: omhullende) of a one-parameter family of curves
is a curve which is tangent to each member of the familiy of curves at some point.
With regards to envelopes, we have the following useful fact.
T . If a family of curves is defined (implicitly) by F(x, y, t) =  (at each t we
have a curve) then one obtains the envelope by adding the restriction
∂F(x, y, t)
= .
∂t
Proof. Let r(t) = (x(t), y(t)) be a parametrization of the envelope. Then r(t) must
lie on the t-curve for each t and hence F(x(t), y(t), t) =  for all t. Therefore also
d
dt F(x(t), y(t), t) = Fx ẋ + Fy ẏ + Ft =  for all t. Now use the fact that the envelope
and the t-curve have the same tangent. With some vector calculus this implies
(ẋ, ẏ) ċ (Fx , Fy) = Fx ẋ + Fy ẏ = . Altogether this leads to Ft = .
We shall now proceed to give a few interesting exam-
ples of envelopes. Since these are not really essential to our
story, you needn’t study them in full detail. [that is to say:
look at all the pretty pictures, and enjoy].
Example . A family of circles with radius t ~ centered at
(t , t ~) is given by
F = (x − t ) + (y − t ~) − (t ~) = .
For the envelope we have, by our theorem,
Ḟ = t(−x − ty + t ) = .
Figure : A one-para-
meter family of circles, Solving these two equations yields the desired result. Be-
together with its enve-
lope.
cause the resulting formulas are a bit too messy to write
down here, let it suffice that we present you with a nice
picture.
Example . We have already encountered this example earlier in the course: the family
of straight lines through (t, ) and (, A−t). In this case, take F = (A−t)(x−t)+ty =
. Then we obtain for the envelope
Ḟ = t − A − x + y = .
Substituting this back into the original equation, we get
y − (A + x)y + (A − x) = ,
which is (as expected) a parabola. See Figure .

[]
During the remainder of this lecture we will only con-
sider families of straight lines. The general description of
such families is F(x, y, t) = f(t)x + g(t)y + h(t) = , so
its envelope is the solution of the system

fx + gy + h = 
œ Figure : The evolute of
ḟx + ġy + ḣ = .
the parabola exhibited
The envelope of a family of rays (straight lines) is called a as the envelope of the
normals.
caustic (Dutch: brandlijn) .
The next theorem, which is illustrated for the case of
the parabola, points us to a close relation between the evolute (which was defined
earlier in the course) and the concept of envelope.

T . The envelope of all the normals to a given curve is the evolute of the curve.

Proof. We most obligingly leave this for the reader to work out.

Figure : The parabola as an envelope.

Enter geometric optics. As many of you will probably remember from elementary
physics classes, the reflection of light (moving in a single medium) on a specular
curve or surface is governed by a few very simple rules.

. Light always moves along a straight line, at a constant velocity. When reflected
it merely changes direction.

. In two dimensions, the change of direction upon reflection is such that the
angle of incidence θ i and the angle of reflection θr are equal when measured
from their respective sides of the normal [at the point of reflection] to the
curve.

This terminology is a bit confusing, to say the least. Most people, but physicists in particular, use
the word ‘caustic’ as a generic term for what we call either ‘catacaustics’ or ‘dicaustics’.

[]
Figure : Catacaustic of a circle, with respect Figure : Optical caustic inside an annular
to parallel rays. object.

. In three dimensions, there is the additional rule that the reflected ray should
always lie in the same plane as the incident ray and the unit normal vector to
the surface.

These rules are neatly summarized in Figure .


One interesting question to ask [especially for the two-
dimensional case] is whether in certain cases the reflected
rays are somehow concentrated along some ‘path of fire’ [i.e.,
a sort of curvilinear analogon to the focal point of a parabolic
mirror]. The existence of so-called optical caustics seems to
indicate that the answer is often affirmative [see Figure ].
Figure : The law We shall soon see how we can apply the envelope concept and
of reflection. For a some of the rest of differential geometry to such phenomena.
general curve or sur-
face the straight line
on which the inci- Catacaustics and orthotomics. A catacaustic (Dutch: weer-
dent ray falls repre- kaatsingsbrandlijn) is the envelope of rays reflected from a
sents the tangent or certain point E [the so-called radiant point; which may be
tangent plane. infinitely distant, in which case the rays are parallel] by some
curve C. In the example in Figure  we have a bundle of parallel rays [like rays of
sunlight], and a semi-circle as the reflective curve. This heart- or kidney-shaped
catacaustic (already encountered in Figure ) is the one you typically encounter when
the sun shines in your cup of coffee.
Another kind of curve we would like to introduce is
the so-called orthotomic curve. This curve is the transfor-
mation of a given curve C with respect to a given point E,
obtained by taking the reflections D of E with respect to
all tangent lines t of C. See the example on the right.
The beautiful thing about this orthotomic is its close
relation to the catacaustic, as shown in the following the-
orem.

T . The catacaustic of a curve C (generated by


the rays from some point E) is the evolute of the orthotomic Figure : The ortho-
tomic C of a circle,
(resulting from E). with respect to point E.

[]
Proof. Let E be the origin of coordinates and let the curve
be parametrized by arc length s: C(s) = (x(s), y(s)), and let T(s) and N(s) be the
unit tangent and normal vector respectively, and let κ(s) be the curvature of the
curve. Then the distance from the origin to the tangent line is C(s) ċ N(s). Now it is
easy to see that the orthotomic is given by

ρ(s) = (C(s) ċ N(s))N(s).

Differentiating with respect to s yields

ρ̇(s) = −κ(s)ŠC(s) ċ T(s)N(s) + ŠC(s) ċ N(s)T(s)

and the expression in square brackets must be tangent to the orthotomic. Therefore

(C(s) ċ T(s))T(s) − (C(s) ċ N(s))N(s) ()

is a normal vector of the orthotomic. The envelope of these lines is by Theorem 


equal to its evolute. On the other hand: the vectors in () exactly coincide with the
direction of the ray reflected at C(s): the ‘normal’ component of C(s) is just inverted.
The envelope of this line is by definition the catacaustic.

Further developments. The catacaustics presented in the previous section may be


generalized in a number of ways. A few words about these seem appropriate.
The first of these generalizations is furnished by so-called catacaustics of higher
order. In the formation of ordinary catacaustics (also called catacaustics of first order)
only one reflection on the curve is taken into account. It is assumed that after this
first reflection the rays somehow simply pass through the curve. But it would be far
more realistic to allow the rays to be reflected two, three, or even infinitely many
times, giving rise to catacaustics of order any natural number, or infinite order. It
needs no comment that the attendant calculations become increasingly complicated.
The second generalization comes into existence when you allow the rays not
to be reflected, but to be refracted. Then you get so-called dicaustics or caustics of
refraction. The analysis of these things is often quite messy, because the law governing
refractions, Snell’s law, is not very easy to state in vector analytic terms. Also the
shape of the dicaustic tends to vary non-continuously with the refractive index.
Finally, the third generalization seems most interesting to us. Place the radiant
point in three-dimensional space and replace the specular curve by a specular surface.
What do you get, given that the reflections are governed by the three rules given
above? In general, the reflected rays will envelope another surface, which is called
the catacaustic surface [of first order].
Of course combinations of the three aforementioned variations are also possible,
giving almost endless opportunities for playing around with differential geometry.
[See also the exercises.]

[]
 :  ,     

Figure : A logarithmic spiral, with a = , b = ., plotted for −π B θ B π.

Say hello to the logarithmic spiral. . . Arguably one of the prettiest objects in math-
ematics is the so-called logarithmic spiral, also known as equiangular spiral, or (as
Jakob Bernoulli lovingly called it) spira mirabilis. Actually, it is wrong to speak of the
logarithmic spiral, since it’s just a generic name for an entire family of curves having
polar equations of the form
r = aebθ ,
with a, b real numbers [but we will often ignore this, and act as if a and b were fixed.].
Such polar equations are of course equivalent to the cartesian parametrizations
¢̈
¨x(t) = aebt cos t
¦
¨ y(t) = aebt sin t,
¤̈
with the range of t the same as that of θ. See Figure 
It is of course called logarithmic spiral because the angle
associated to a point p on the spiral depends logarithmi-
cally on the distance of p from the origin. The reason for
calling it equiangular is also rather obvious. Consider

`x(t), xœ (t)e a bebt


arccos arccos º
Yx(t)YYxœ (t)Y
=
a ebt  + b
b
= arccos º
Figure : The spiral
galaxy M (‘whirlpool .
galaxy’) and its compan-
 + b
ion NGC . Notice
Because this is clearly a constant, the angle between the
how each of the arms
of M clearly is an ap- radial and tangent vectors is constant. Hence the name
proximate logarithmic equiangular spiral.
spiral. To make life a bit easier [especially when drawing pic-
tures], let us agree to take b A . This causes the width of
the spiral to increase without bounds with t ª. To see that this must be true, ob-
serve that r(θ + πk) = aeπk ebθ . So, if we let k run through the integers (ascending),
this merely tells us that any ray from the origin meets the spiral at distances which
grow according to a geometric progression.

[]
A further interesting property of the logarithmic spiral
is the following. Because r(θ + ϕ) = ebϕ r(θ), we see that
rotating the spiral is essentially equivalent to scaling it. This
remarkable behaviour of the logarithmic spiral under two
important Euclidean transformations will serve us well in
what follows.
As a side note: the logarithmic spiral (or at least approx-
Figure : A shell left imations to it) tends to show up in all sorts of unexpected
behind by the lovely
places, like the sea, or the depths of space. Look, for ex-
sea creature Nautilus (a
species of squid). ample, at the seashell in Figure , or the galaxy in Figure

. . . its evolute . . . From Prof. Siersma’s lecture a few weeks ago, we already thoroughly
know what the evolute of a given curve is, and also how to parametrize it, given
a parametrization for the original curve. Plugging the standard parametrization
C(t) of a logarithmic spiral into the formulas we already know, and a fair bit of
computation, yields

xœ (t)yœœ (t) − xœœ (t)yœ (t) e−bt


κ(t) = º ,
(xœ (t) + yœ (t) )~
=
a  + b
(−yœ (t), xœ (t)) (−[cos t + b sin t], b cos t − sin t)
N(t) = » = º ,
x (t) + y (t)
œ  œ   + b
N(t)
E(t) = C(t) + = (−abebt sin t, abebt cos t)
κ(t)
So, the evolute of a logarithmic spiral is just a scaled version of the original. [Work out
the details of this remark for yourself. Remember that scaling an logarithmic spiral
is equivalent to rotating it.]

. . . its involute . . . The concept of involute, though it can


be defined in terms of the evolute, is best explained in a
‘kinemetic’ way. Imagine you have a curved wall [of a castle,
for example] with somewhere, at arm height, a big metal
ring built into it. Attached to this ring is an enormous yet
flexible metal chain. Now grab the free end of the chain
and keep the thing taut against the wall. If you then wind
the chain around the wall, keeping it as tight as possible,
which path are you forced to follow?
As you may well know, the answer to this question is Figure : A string
that your feet [at least approximately] trace out an invo- wrapping itself around
lute of the wall [which is viewed from above as a plane a curve γ traces out an
evolute of γ.
curve]. While each curve has a unique evolute [proof?], it
in general has an infinity of evolutes, each corresponding
to different lengths of the chain. The point of attachment of the chain also appears

[]
to add a degree of freedom, but we can remove this by demanding that point of
attachment be ‘infinitely far away’, i.e., we look only at the limit of the involute as the
distance to the point of attachment increases without bounds [or until the parameter
runs out.]
Given a parametrization C(t) of the original curve C, a parametrization of
the involute is easily obtained. [We assume that the chain is attached at point
(x(t ), y(t )).] For any t, take the tangent vector at C(t), make it unit length (divide
by its length), then multiply that by the arc length from t to t [because the length of
the wound-up portion of the chain is exactly that], negate it, then translate the result
back to the location C(t). Therefore we get for the involute I

Cœ (t) t»
I(t) = C(t) −
YCœ (t)Y ∫
t
xœ (t) + yœ (t) dt.

For the arc length of the logarithmic spiral we obtain, by first reverting to polar
co-ordinates,
º
º

t a  + b bt
s(t; t ) = a  + b bt
e dt = (e − ebt ).
t b
Let us now then take a chain so lang that we can wind all the way to the point
(x(−ª), y(−ª)), i.e., to the origin. In the real world this is of course a dubious thing
to do, since we are human beings with only finite amounts of time, but mathematicians
º in which just about anything goes. The arc
seem to live in some sort of fairy world
length then collapses to s(t) = ae bt  + b ~b, and for the involute we get [after some
laborious calculation]

(aebt sin t, −aebt cos t)


I(t) = ,
b
which is easily recognizable as yet another logarithmic spiral.
For other choices of chain length the calculations get considerably messier.

. . . its catacaustic . . . As was said in the first part of the lecture, the catacaustic of
a curve C with respect to a radiant point E is actually equal to the evolute of the
orthotomic curve of C with respect to E. We could calculate a parametrization for the
orthotomic of the logarithmic spiral here, but since in the first part we have already
given a construction for the orthotomic, we leave it up to the reader to do so. [See
exercises.]

. . . its pedal and contrapedal curves . . . Given a plane curve C and a fixed point o,
called the pedal point, the pedal curve (Dutch: voetpuntskromme) P of C with respect
to o is defined to be: the locus of the intersection points p of the tangents to C and
the perpendicular line segments from o to these tangents. Formulated in somewhat

This is what people usually mean when they speak of the involute of a given curve. It is determined
up to point of attachment.

[]
less awkward prose: if c is a point on C, draw a tangent T to C. Then draw a line L
through o perpendicular to T, and let p be the intersection point of T and L. If we
let c move along the curve, the point p generally traces out another curve, which is
the pedal curve of C with respect to o. See also figure.
The pedal curve has lots of interesting properties, but
due to the limited time and space available, we’ll have to
leave them for someone else to discuss. Instead let’s just
derive a formula for the pedal curve and apply it, as usual,
to our beloved logarithmic spiral.
Let the curve C be parametrized as C(t). Then, for
each t, P(t) is simply, almost by definition, the orthogonal
projection of o − C(t) on Cœ (t), shifted to C(t). Using Figure : Construction
of a pedal curve for the
elementary vector geometry, we therefore obtain ellips.

`Cœ (t), o − C(t)e œ


P(t) = C(t) + C (t),
YCœ (t)Y

where we’ll just waive our hands to make points where Cœ (t) is zero or undefined
magically disappear.
Let’s now try and see what comes out if we take as our pedal point the origin
and plug into the previous equation the standard parametrization of the logarithmic
spiral. We then get
b
P(t) = (ebt cos t, ebt sin t) − (aebt (b cos t − sin t), aebt (cos t + b sin t))
 + b
a
= (ebt (cos t + b sin t), −ebt (b cos t − sin t)).
 + b
Next, we revert to polar co-ordinates [remember this trick] and obtain
» aebt
r(t) = x(t) + y(t) = º ,
 + b
which clearly represents a scaled version of the original spiral. We’ve just discovered
yet another self-replicating property of the logarithmic spiral.
As for the second part of this section, the contrapedal curve CP of a curve C with
respect to a pedal point o is defined to be the locus of the feet of the perpendiculars
let down from o on the normal lines of C. So, in a sense, the contrapedal is a ‘dual’ to
the pedal, with ‘tangents’ replaced by ‘normals’. Given a parametrization C(t) of C, it
is once again easy to derive one for the contrapedal, viz..

`Cœ (t), o − C(t)e œ


CP(t) = o − C (t).
YCœ (t)Y
We also state without the proof the following fact: the pedal points being the same,
the contrapedal of a curve C is the pedal of the evolute of C. By what we’ve already
discovered about the logarithmic spiral, we immediately realize that the contrapedal
of a logarithmic spiral is also a similar spiral.

[]
Figure : A detail from Jakob Bernoulli’s grave monument in the Basler Münster church. It’s
quite obvious that the stonemason simply goofed up.

. . . and its inverse curve. We’ve encountered inversion in a circle before, but when
we did deal with it we were mainly interested in inverting lines and circles. But of
course nothing stops us from inverting other things, like logarithmic spirals, as well.
In fact, if a curve C is parametrized by (x(t), y(t)), and is inverted in a circle with
center (x , y ) and radius k, then it is not very difficult to see that
¢̈ k (x(t) − x )
¨
¨
¨
¨
X(t) x
¨ (x(t) − x ) + (y(t) − y )
=  +
¨
¨
¦
¨
¨
¨
¨
¨ k (y(t) − y )
¨
¨ Y(t) y
¤̈ (x(t) − x ) + (y(t) − y )
=  +

parametrizes the result of the inversion [except possibly at one point, if the curve
passes through the inversion center].
What happens, then, if we place the inversion center at the origin and invert the
logarithmic spiral? Of course you can all dream this answer by now: we get a similar
logarithmic spiral. The computation needed to check this is rather straightforward,
and is therefore left to you.

A stonemason’s nightmare. Jakob Bernoulli discovered most of the wonderful prop-


erties of the logarithmic spiral described above. Born in  to a prosperous family
of merchants [which had fled Antwerp in the late th centery for religious reasons],
he decided to defy the will of his father, and eventually became a mathematician
instead of a spice trader. He was made a professor of mathematics at the University
of Basle, rapidly turning out to be one of the leading mathematicians of his day. It
is probably fair to say that he practically created the calculus of variations. He also
did important research in the theory of differential equations, probability, mechanics,
and, of course, in the theory of curves. Finally, he’s one of those who really made
an effort to let the Leibnitzian version of the infinitesimal calculus, with its clearer
notations, prevail over the Newtonian.
When he died in , not yet  years old, it became known that he had wanted
a picture of a logarithmic spiral [the curve with which he had been so in love] to be

Remember that if O is the center of the inversion circle and k its radius, then the inverse of a
point P is another point Pœ , such that OP  OPœ = k .

[]
engraved on his grave monument. Unfortunately [see Figure ], the stonemason did
a rather poor job, and the spiral became an Archimedean one instead of a logarithmic
one.
His epitaph was: eadem mutata resurgo, ‘though changed, I shall arise the same.’
This refers both to the properterties of the logarithmic spiral and to the Christian
resurrection of the dead. [Amen.]

[]


Here we present you with a series of eight exercises. Do not, by any means, attempt
to solve all of them [in the unlikely event you feel inclined to do so, please contact
us; then we will produce some more]. Instead, each exercise can earn you a fixed
number of points, and we would like you to solve [or at least attempt to solve] as
many exercises as are worth exactly four points. Good luck!

Exercise  (Degenerate catacaustic,  pt). Fix the radiant point at the origin. Give an example
of a curve for which the first-order catacaustic degenerates to a single point. Prove that your
example is correct.

Exercise  (Catacaustic surface,  pt). What does the first-order catacaustic surface of the unit
sphere, centered at the origin, look like, if the radiant point is at (, , )? No formal proof
is needed , but convincing intuitive reasoning is. You may use also your excellent drawing
skills.

Exercise  (Evolute as envelope of normals,  pt). Give a full proof of Theorem , using the
definition of evolute given at the beginning of the notes.

Exercise  (Identical pedal curve, pt). Fix the pedal point at the origin. Give an example of a
curve which is identical to its own pedal curve. Prove that your example is correct.

Exercise  (Orthotomic of logarithmic spiral, pt). (a) Prove that, if xœ (t) + yœ (t) never
becomes zero, then the curve with the following parametrization is the orthotomic, with
respect to the origin, of the curve parametrized by ((x(t), y(t)):

¢̈ y(t)[xœ (t)y(t) − x(t)yœ (t)]


¨
¨
¨ O (t)
¨
¨ xœ (t) + yœ (t)
x = −
¨
¨
¦
¨
¨
¨
¨
¨ x(t)[xœ (t)y(t) − x(t)yœ (t)]
¨
¨ yO (t) .
¤̈ xœ (t) + yœ (t)
=

This gives as a parametrization of the orthotomic that doesn’t require the original curve to
be parametrized by arc length. (b) Apply the result from the previous part to parametrize
the orthotomic of the logarithmic spiral. What do you get? [Hint: at first you get a bit of a
mess, but recall a certain trick from the notes. Use Mathematica, or something like it, if the
calculations are too difficult or boring for you.] (c) So, what can you say about the first-order
catacaustic of the logarithmic spiral if the radiant point is at the origin?

Exercise  (A cruel game with lions, pt). Long before the Colosseum existed, emperor
Ridiculus Maximus devised a fighting theatre in the shape of an equilateral triangle. His
favourite passtime was to set one hungry lion on the loose at each of the vertices of the triangle,
and to watch the spectacle unfold. The lions moved in such a way that each approached its
nearest neighbour in a counterclockwise direction and at a constant velocity. Describe the
paths the lions followed. Prove your assertion, or give a convincing intuitive argument.

But provide one if you wish.

Assume the curve and its parametrization to be smooth.

[]
Exercise  (Evolute of the logarithmic spiral, pt). Sometimes the evolute of the logarithmic
spiral is not only similar but identical to the orginal spiral. Investigate when this happens.
[Hint: effect the following change of variables: t = ϕ − π~  nπ, and let n run through the
natural numbers.]

Exercise  (Prize question, pt). Does there exist a curve which is similar to its own catacaustic
when the rays are parallel (i.e., the radiant point is at infinity)? If such a curve does exist, give
an example. If it doesn’t, prove it. [We don’t know the answer.]

[]

You might also like