# Number of points at which a tangent touches a curve

My teacher told me that we are mistaken coming out of school that tangent tocuhes a point at one point. According to him, a tangent is just a special type of secant where two points share the same position but actually are just two different points.

My friend said it could happen if we think of tangent as a secant where the two points tend to each other.

I am dissatisfied with it. According to my teacher, the points share the location but are different. Since a point has only its position as its property, how are they differentiated among themselves? E.g. if we take the set of points on which tangent touches the curve (in a sufficiently small neighbourhood around the points), we get a single point since a set doesn't allow redundant entries.

(Moreover, my teacher's tried to explanation his position using quadratic formula taking a quadratic equation as example; but I fail to recognise its relevance except that a quadratic equation has at most two roots, and maybe that tangent has a solution with the curve. But isn't it the problem to begin with, how many solutions do the two of them have?)

And I could agree with my friend too, but if we take a curve to be $y = x^2$ and tangent to be $y = 0$ than there is no point except $(0,0)$ that is common to them. No point in any small neighbourhood around $(0,0)$ is common to both. That seems to only reinforce myself.

Question

How many points do a tangent and its curve actually share? And if they are more than one, how can they be differentiated? That is, what is the application of treating two positions with same position differently?

-
In Calculus, when you talk about a tangent point, you consider locally, that is, near the point. But, what does 'near' mean? And, what is the tangent line at the point $(0,0)$ for the curve $(t,t^3)$? Think about this. – Sigur Aug 22 '12 at 2:31
I really hope that "two points share the same position but actually are just two different points" is not what the teacher really said. Otherwise that sounds like one very confused teacher (or maybe two sharing the same body?). – Robert Israel Aug 22 '12 at 8:31
@RobertIsrael - He said it exactly like I said. Maybe he didn't want me to get into detail - I'll give him benefit of doubt since he was at the end correct. Plus, he is otherwise an excelent teacher! – Anurag Kalia Aug 28 '12 at 16:26

The fact that the tangent to a curve, defined as the limit of a variable secant, cuts the curve in two indistinguishable points is clear, geometric...and completely nonsensical fron a rigorous point of view!

Grothendieck in the 1950's found the solution to that centuries old charade: his scheme theory defines the intersection as the contact point plus a ring whose size reflects the degree of tangency.
In your example of the tangent $y=0$ to $y=x^2$, the ring is $\mathbb R[x]/(x^2)$ and has size (=dimension as a vector space) 2.
In the case of the tangent $y=0$ to the curve $y=x^3$, the size would be 3, indicating a higher contact of the tangent to the curve.
And of course for the curve $y=x^n$, the size would be n.
One of the revolutionary aspects of this point of view is that tangency becomes a completely static concept: no complicated calculations of limits of secants are involved.

[This answer is at a level more advanced than calculus, but less so than one might think.
It might help put things in perspective and perhaps serve as a slightly enigmatic magnet to a more sophisticated version of geometry.]

-
I recommend Abhyankar's Algebraic Geometry for Engineers. It handles double points very simply using calculus first and then lengths of artinian modules later. – Jack Schmidt Aug 22 '12 at 11:46
An excellent recommendation, @Jack. As a bonus the reader will enjoy that great mathematician's amusingly idiosyncratic style. – Georges Elencwajg Aug 22 '12 at 12:54
This seems intriguing, and though I don't understand bits of it right now, I would have just have to study more. Thanks for your answer. I was expecting something of this kind from this site. :-) – Anurag Kalia Aug 24 '12 at 4:36

Literally, your teacher is wrong. The tangent line meets the curve at the one point to which it is tangent, and as you note in your post, this point is one point. (Here I am ignoring the fact that the tangent may also intersect the curve at some other, unrelated point, if the curve is not convex --- this is an unrelated issue.)

What your teacher has in mind, though, is that the curve is a limit of secants through a pair of points, the limit being taken as the two points in the pair tend to the one point at which you are taking the tangent.

Related to this picture of taking a limit of secant lines, in some parts of mathematics, one also says that the tangent meets the curve in a "double point". But this does not mean that the tangent meets the curve in two points; rather, it is a short-hand way of expressing the particular way in which the tangent line meets the curve at the point of tangency.

-

The issue here depends on context. For example if one is counting the number of roots of an equation - say $y=(x-r)^n p(x)$ with $p(r)\neq 0$ one sometimes wants to count the root $x=r$ n times, and therefore treat the line $y=0$, which is a tangent at $x=0$ when $n>1$, as meeting the curve at $n$ points.

This kind of thing happens with orders in algebraic geometry and complex analysis, for example.

To take the quadratic example, the line $y=0$ meets the curve $y=x^2$ at the point $(0,0)$. If you want to say something like "every quadratic equation in one variable over $\mathbb C$ has two roots", you need to count this twice, and use a phrase like "roots counted with their multiplicities".

But I think the real mistake is to be dogmatic, since the "correct" answer depends on how the wider mathematical context frames the issue of intersection - and there will be some contexts where a dogmatic answer of any flavour will simply be wrong.

It is useful to think about the multiplicities idea, though, because it does turn out to be very useful.

-

The tangent to a curve at a point is usually defined as the limit (if it exists) of the secant through the point and another point on the curve as the other point approaches the first point.

For example for the parabola $y = x^2$ at $x = 0$, if the other point is on the parabola at $x = a$, it is at $(a, a^2)$. This secant is the line $y = a\ x$ (this is $0$ at $x = 0$ and $a^2$ at $x = a$).

As $a$ approaches $0$ (and this is where you need to know what "limit" means), this approaches the line $y = 0$, and this line is the tangent to the parabola at $x = 0$.

As to how many points a curve and its tangent "share" at a point, the answer is usually one if the curve is reasonably smooth and we only look close to the point. For example, with the curve $y = \cos x$, the tangent at $x = 0$ passes through $(2\pi k, 1)$ for all integers $k$, but only $(0, 1)$ is close - the others are at least $2\pi$ away.

However, if you consider the curve $y = \cos(1/x)$ as $x \to 0$, things get more interesting, and there I will stop.

If you are having problems, that is OK - it took a while before the limit definition of tangent became understood.

-
Better use $y = x^2 \cos(1/x)$ (with $y(0) = 0$) if you want the curve to have a tangent at $0$. – Robert Israel Aug 22 '12 at 8:11