TRACTRIX : Exploration

As we have seen from the ‘Generation’ tractrix page, a branch of the tractrix arises in connection with a physical problem where a weight is pulled along. In fact, the name ‘tractrix’ comes from the Latin verb trahere ’to pull, drag‘.

Suppose we have a weight W, placed (say) at the cusp point of the tractrix (see graph below). A string of length 1 is attached to the weight, and initially reaches down to the origin. The free end T is now steadily pulled in a vertical direction along the y-axis. The weight W then traces out a branch of the tractrix. To see why this occurs, first modify the graph ... .

Notice that TW = 1, the length of the string. Also, we expect TW to be tangent to the curve at W, since the wight W will always be pulled directly towards T.

Using the notation of the (modified) figure, and in particular the triangle TVW, we see that the slope of the tangent at W is given by

We also have y(0) = 1.

Solving this differential equation gives the rather unpleasant Cartesian equation of the tractrix as

The signs allow for the pulling point T to move up or down the y-axis, generating both branches of the tractrix. This equation has various other equivalent forms, some involving the hyperbolic cosine.

The parametric equations are not much better :

x = log (sec t + tan t) – sin t ,

y = c cos t , where 0 t < 2.

Fortunately the programmer has only to enter in the formulae – not do any hard manipulation!

We notice that in this position the tractrix has a right pointing cusp on the x-axis, and the y-axis is an asymptote to the curve: as y gets infinitely large positive or negative, the tractrix curve approaches the axis more and more closely. You might like to think of it as a double tangent ‘at infinity’.





Tangent length

The diagram at right shows one branch of the tractrix, a point P on the branch, and a tangent to the curve at P meeting the asymptote at point T.

Play with the applet. Moving Q manually is probably more helpful than using the ‘Animate’ button. Do you notice anything about the length of segment PT? Can you conjecture a value here? (Notice that OC = 1.) Why would you expect this property to be true? Now click on the ‘Show Length’ button to check your result.

In spite of the measured lengths(!) we expect the length here to be 1. It is of course the length of the string in the original setting up of the problem. We would also expect the applet to fail at the cusp point C itself, since no tangent is defined at this point.

We have then the result:

The length of the intercept of a tangent to the tractrix from the contact point P to the asymptote is constant, and equals OC.











Area between curve and asymptote

It is known that the region bounded by the tractrix curve and the asymptote has area /2. We can illustrate this property using the applet linked at right.

Run the applet linked at right. The radius of the circle is equal to length OC. What is the area of a quadrant of this circle? The segments which trace out the region between the tractrix and the axes are translated so that the point P in each case moves to the centre of the circle. What appears to be true? What area do you think the region between the whole tractrix and the aymptote has?

A quadrant of the circle has area /4. The quadrant appears to correspond to the region traced out between half the tractrix and its asymptote. Hence we conjecture the value /2 between the whole tractrix and its asymptote. The same argument can be carried out using small sectors of the circle, making the conclusion more convincing. You might like to check out the web site

http://demonstrations.wolfram.com/

and do a site search for ‘tractrix’. Here is a delightful (and elaborate) applet which demonstrates this property.


BIBLIOGRAPHY

A Book of Curves, Lockwood, E. H. (C.U.P. 1967)

Mamikon, http://www.its.caltech.edu/~mamikon/calculus.html

Wikipedia : http://en.wikipedia.org/wiki/Tractrix

Wolfram MathWorld : http://mathworld.wolfram.com/Tractrix.html