1

Consider the two curves $y=e^x\cdot x^2$ and $y=e^{-x}\cdot x^2$ both have $y=0$ as their tangent line at $x=0$.

Another example is the curve pair $y=e^{-ex}$ and $y=\log_{e^{-e}} x$.

Image Link 1: $y=e^x\cdot x^2$ and $y=e^{-x}\cdot x^2$

Image Link 2: $y=e^{(x/e)}$ and $y=\log_{e^{1/e}}x$

Do we consider these two curves tangent to each other? If not, what is the best mathematical keyword to describe their behavior? If yes, how do we separate this behavior from touch of the two curves? Is there a better way to describe 'touch' mathematically?

I learn from Wikipedia that the tangent line does not necessarily touch the curve without crossing it. A point where the tangent (at this point) crosses the curve is called an inflection point (e.g. $y=x^3$ at $x=0$). I am not sure how to generalize this to tangency between two curves.

  • 1
    I'm not familiar with the phrase "external touch of the two curves". Could you add a diagram of what you consider to be tangency of two curves and what you consider to be an "external touch of two curves" so that it is clear what distinction you intend to draw? – Eric Towers Aug 22 '21 at 15:07
  • Yes: the usual definition of tangency between two curves just requires them to have a common tangent line at the common tangency point. – Intelligenti pauca Aug 22 '21 at 15:16
  • @EricTowers I just added image links to the post. The first one is the one that two curves don't touch but have the same tangent line. The second one is the one that two curves touch with the same tangent line. – mathephobia Aug 22 '21 at 15:25
  • @Intelligentipauca If that's the case, how do we separate between two scenarios (touch vs not touch) under the same tangent line condition?

    I'm not sure I'm using the word 'touch' correctly.

    – mathephobia Aug 22 '21 at 15:27
  • "How do we separate this behavior from external touch of the two curves?" If you really need to separate them, then you'll have to give a definition of "external". – Intelligenti pauca Aug 22 '21 at 15:27
  • @Intelligentipauca Sorry I have edited my message. I try to avoid 'external' and 'internal' keyword. – mathephobia Aug 22 '21 at 15:28
  • Both are examples of tangent curves, tough the second example do not cross one another, while the first one does. – user3257842 Aug 22 '21 at 15:28
  • As I wrote, those two curves you graphed ARE tangent at $(0,0)$. – Intelligenti pauca Aug 22 '21 at 15:28
  • @Intelligentipauca I see. Is there a way to mathematically describe the difference between the two scenarios? – mathephobia Aug 22 '21 at 15:30
  • When you find two points of common slope between two smooth plane curves, you are finding two points (one on each curve) where (locally) the two curves are at minimum distance from each other. This is one possible way you can describe both situations. In the case of a common tangent line at a point, the minimum distance is zero. – DOUGLAS BRUNSON Aug 22 '21 at 15:38

1 Answers1

3

Let $f:S\subseteq\mathbb{R}\to\mathbb{R}^2$ be a curve differentiable at $t_0\in S$. The tangent line of $f$ at point $f(t_0)$ is defined as the best linear approximation of $f$ at $t_0$, namely $$L(t)=f(t_0)+\frac{df}{dt}(t_0)(t-t_0)$$ for all $t\in\mathbb{R}$.

If two curves $f,g$ have the same tangent line at point $(x,y)$, then we say that $f,g$ are tangent to each other at $(x,y)$.

For example, if $f(t)=(t, e^tt^2)$ and $g(t)=(t, e^{-t}t^2)$ as given by OP, then $\frac{df}{dt}(t)=(1,e^t(t^2+2t))$ and $\frac{dg}{dt}(t)=(1, e^{-t}(-t^2+2t))$. We have that $f(0)=(0,0)=g(0)$ and $\frac{df}{dt}(0)=(1,0)=\frac{dg}{dt}(0)$. Thus, the tangent line $L$ at $(0,0)$ of both curves is the same: $$L(t)=(0,0)+(1,0)t$$ for all $t\in\mathbb{R}$. We say that the $f,g$ are tangent to each other at point $(0,0)$.

As for internal and external tangencies, I am not able to find any reference to those notions on general curves. However, since we do have such notions for circles, we can intuitively extend it to general curves using osculating circle. If the two osculating circles of the curves at some point $(x,y)$ are internally tangent, then we say that the two curves are internally tangent to each other at $(x,y)$, and similarly for external tangent.


Edit:

I found an answer here which inspires a perhaps more general definition for internal and external tangencies. Let $f,g$ be $\mathbb{R}^2$-curves having a common tangent line $L$ at point $(a,b)$. Let $C_f$, $C_g$, $C_L$ each be the respective range of $f$, $g$, $L$.

Suppose that there exists an open disk $B$ centered at $(a,b)$ such that the following is true:

  1. $C_L$ divides $B$ into two disconnected regions $B_1$ and $B_2$.
  2. $B_1\cap C_f=\emptyset$ and $B_1\cap C_g=\emptyset$.
  3. $B_2\cap C_f\neq\emptyset$ and $B_2\cap C_g\neq\emptyset$.

Then we say that $f$ and $g$ are internally tangent to each other at $(a,b)$.

On the other hand, if there exists an open disk $B$ centered at $(a,b)$ such that the following is true instead:

  1. $C_L$ divides $B$ into two disconnected regions $B_1$ and $B_2$.
  2. $B_1\cap C_f=\emptyset$ and $B_1\cap C_g\neq\emptyset$.
  3. $B_2\cap C_f\neq\emptyset$ and $B_2\cap C_g=\emptyset$.

Then we say that $f$ and $g$ are externally tangent to each other at $(a,b)$.

Examples and non-examples:

Take $g(t)=(t,t^2)$ and consider the tangent line $L$ at point $(0,0)$ with the following $f$:

  1. $f(t)=(t,0)$. In this case, $B\setminus C_L$ does not intersect $C_f$ at all, but our definitions require that there must be some nonempty intersection. Thus, we say that $f$ and $g$ are neither internally nor externally tangent to each other (which makes intuitive sense).
  2. $f(t)=(t,t^3)$. In this case, both $B_1$ and $B_2$ have some nonempty intersection with $C_f$, but our definitions require that either $B_1$ or $B_2$ does not have intersection with $C_f$. Thus, we say that $f$ and $g$ are neither internally nor externally tangent to each other (which again makes sense).
  3. $f(t)=(t, t^2\sin(1/t))$. Just as the previous case, both $B_1$ and $B_2$ have nonempty intersection with $C_f$ (because no matter how small $B$ is, there are always some $(x, y)\in C_f$ and $(-x,-y)\in C_f$ inside $B$, where $y>0$). It also makes sense to say that $f$ and $g$ are neither internally nor externally tangent to each other in this case.
  4. $f(t)=(t, |t|^{3/2})$. Note that $f$ does not take any value in the lower half-plane, so (say) $B_1\cap C_f=\emptyset$. Also it is clear that $B_2$ must contain some $(x,y)\in C_f$ where $x,y>0$. The same is true for $C_g$. Thus, we say that $f$ and $g$ are internally tangent to each other. This demonstrates that twice-differentiability is not required for internal and external tangencies (because $f$ is not twice-differentiable).
Paprika7191
  • 2,741
  • I see! I think the tangent definition is now clear. But the difference between Image Link 1 vs Image Link 2 isn't answered yet.

    Osculating circle idea looks interesting, as it works for both scenarios (except $y=x^3$ and $y=-x^3$ where the circle radius is infinite). Regarding the osculating circle, do we need the curve to be twice differentiable based on the curvature definition?

    – mathephobia Aug 22 '21 at 16:05
  • 1
    @mathephobia If we define internal and external tangency using osculating circles, we do need twice-differentiability. However, consider the functions $f(x)=|x|^{3/2}$ and $g(x)=x^2$. The function $f$ is not twice-differentiable at $0$, but $f$ and $g$ look like they are internally tangent to each other. So I'm wondering if osculating circle is general enough. – Paprika7191 Aug 22 '21 at 16:22
  • 1
    It is hard to cover all cases I believe. An example is the pair $f(x)=x^2\sin(1/x)$ and $g(x)=-x^2\sin(1/x)$. They are considered tangent as the slope at $x=0$ is $0$. But it is hard to evaluate whether they are crossing or touching (also internally vs externally). – mathephobia Aug 23 '21 at 06:52
  • So let me try to classify with the following examples:
    1. (Typical touching, able to identify internal/external touch, with osculating idea) $y=x^2$ and $y=-x^2$

    2. (Typical touching, able to identify internal/external touch, not with osculating idea) $y=|x|^{1.5}$ and $y=x^2$

    3. (Tangent but crossing; not able to identify internal/external tangent through osculating idea) $y=x^3$ and $y=-x^3$

    4. (Tangent but crossing, able to identify internal/external tangent through osculating idea) $y=x^2e^x$ and $y=2x^2e^{-x}$

    5. (Weird) $y=x^2\sin(1/x)$ and $y=-x^2\sin(1/x)$

    and so on...

    – mathephobia Aug 23 '21 at 07:04
  • @mathephobia I have updated my answer to include a more general definition. Also, I don't quite understand what you mean by "tangent but crossing". – Paprika7191 Aug 23 '21 at 18:50
  • Thanks a lot for the edit. I think the definitions of internal and external tangent are now clear. Roughly speaking, in a tiny ball, the internal tangent wants two curves to be completely on the same side and the external tangent wants each of two curves to be completely on the opposite side. – mathephobia Aug 23 '21 at 19:59
  • Let's say we have curve A and curve B. 'Cross' gives me a sense that inside the tiny ball, the curves intersect the ball in the following order: ABAB, not AABB. Sometimes this order cannot be explicitly identified as in the example $(f(x),g(x))=(x^2\sin(1/x),-x^2\sin(1/x)). This is more like a terminology question on how do we separate 'touch' case and 'cross' case into different mathematical words. – mathephobia Aug 23 '21 at 20:10
  • I shouldn't use the word 'tangent but crossing'. Since crossing and tangent are different words. Initially I was trying to separate between two cases:
    1. Two curves just 'touch'. This one includes internal and external tangent definitions you mentioned. It also includes $(f,g)=(x^3,x^2)$.
    2. Two curves crosses. Example: in a tiny ball at tangent point, $(f(x),g(x))=(x^3,-x^3)$ (two curves on both sides of tangent lines), $(f(x),g(x))=(x^3,x^2)$ (one curve on both side, another on one side), $((f(x),g(x))=(x^2e^x,x^2e^{-x})$ (both curves on one side), and $(f(x),g(x))=(x^2\sin(1/x),x^2)$.
    – mathephobia Aug 23 '21 at 20:15