Multivariable calculus: The planimeter

The planimeter #

written and developed by Thomas Pietraho

Suppose that you are nineteenth-century land surveyor and are tasked with finding the area of a plot of land, say something like New Zealand:

Estimate the acreage of these islands.
Image from pngwing.com

How would you proceed? Until things like maps and blueprints became digitized sometime late in the twentieth century, determining the area of an irregularly-shaped region drawn on a piece of paper was a common and an important task in a variety of settings, including land surveying, architecture, and engineering. This lab will introduce you to a decidedly nineteenth-century solution to this problem that uses the then-newly-developed technology of vector fields and path integrals.

Exercise: Before we start, how could you approach this problem? Is there a way to estimate the area of an irregularly-shaped region without resorting to newfangled nineteenth century mathematics?
Solution
One approach is to divide the region into small squares and count them. The size of each square will affect the accuracy of your estimate. Another more destructive, but also more fun, approach is to cut the region our of paper, weigh it, and compare to the weight of a square of paper whose area is known.

1. Background #

Computing area with Green’s Theorem #

Our solution will come from a surprising application of Green’s Theorem and a nineteenth-century mechanical device. But first let us set the stage with some mathematics. Consider the vector field \(\vec{F}(x,y) = x \vec{j}\). According to Green’s Theorem, for any simple closed curve \(C\) that is oriented counter-clockwise and encloses a region \(R\), we have \[\int_\mathcal{C} \vec{F} \cdot d\vec{r} = \int \hspace{-.1in} \int_R curl \; \vec{F} \; dx \; dy.\]

For our particular choice of vector field, we have \(curl \; \vec{F} = 1\), which gives Green’s Theorem a rather unique interpretation: the right-hand-side of the equation above represents the area of the region \(R\) and leads us to the following:

Observation: Path integrals can be used to compute areas of regions in the plane. More specifically, if \(\vec{F}\) is a vector field for which \( curl \; \vec{F} = 1\), then the path integral \[ \int_\mathcal{C} \vec{F} \cdot d\vec{r} = \text{Area of } R\] where \(R\) is the region enclosed by the simple curve \(\mathcal{C}\).
Exercise: The vector field \(\vec{F}(x,y) = x \vec{j}\) is not unique with this special property. Find three additional vector fields \( \vec{F} \) whose \(curl\) is equal to \(1\). Each one of them can be used to find the area of a region by computing a path integral.
Solution

The following vector fields all have \(curl \; \vec{F} \) equal to 1:

\[ \vec{F}(x,y) = -y \vec{i} \; \; \; \; \vec{F}(x,y) = -\tfrac{y}{2} \vec{i} + \tfrac{x}{2} \vec{j} \; \; \; \; \vec{F}(x,y) = 2y\vec{i} + 3x \vec{j} \]

There are many others, as we will find out.

An upshot of this observation is the following. Suppose that we are able to construct a mechanical device that computes the path integral of a vector field with \(curl\) equal to one. Then our measurement problem will be solved! In fact, one such device was proposed in 1854 by the mathematician Jakob Amsler-Laffon. We will study how well it works in this lab.

The polar planimeter #

A polar planimeter, as shown below, is a simple mechanical device consisting of two arms, a graduated wheel, a counter and vernier scales, a tracing point, and an anchor point.

My planimeter. There are many like it, but this one is mine.

Detailed use instructions appear in this video, but here is a quick guide on how to use it:

  1. Begin by placing the planimeter on the same piece of paper as the region whose area is to be measured. Make sure the anchor point is outside the region and set the counter and the vernier scales to zero.

  2. Place the tracing tip at a point on the region’s circumference and begin tracing the curve with the tracing point of the planimeter. As you do this, the graduated wheel will turn and the counter and vernier scales will keep track of the number of the rotations made the planimeter’s graduated wheel. The graduated wheel will sometimes turn forwards, sometimes backwards, and sometimes it will stay still. The scales will keep track of the signed number of rotations.

  3. Continue and trace one complete revolution around the region and note the final reading from the counter and vernier scales. Each individual device comes with a scaling factor. Multiply your reading by this scaling factor. The result purports to be the area of the region.

We will dissect what is going on, and whether the polar planimeter actually does what it claims to do, in the sections below.

2. A mathematical model #

First, let’s create a mathematical model of the planimeter and the problem it is trying to solve. Let \(\mathcal{C}\) be a simple closed curve in the plane that is oriented counter-clockwise and write \(\mathcal{R}\) for the region it encloses. Our task is to compute the area of \(\mathcal{R}\).

Tracing arm and pivot point #

Let us set the anchor point at the origin, and introduce the following conventions:

  • let \((x,y)\) be the coordinates of the tracing point which will circumnavigate the curve \(\mathcal{C}\),
  • and let \((a,b)\) be the coordinates of the pivot point where the arms of the planimeter join.

Note that both \(a\) and \(b\) will change as \(x\) and \(y\) change, so we can think of both as functions of \(x\) and \(y\). Finally, let us assume that both arms of the planimeter have unit length.

The animation above shows the relative positions of the two arms as the planimeter traces out the circumference of the region and how \((a,b)\) changes as we change \((x,y)\).

Exercise: Find a formula for the coordinates \((a,b)\) of the pivot point in terms of the coordinates of the tracing point \((x,y)\). The computation is easier if you use this notebook. Hold on to your result, we will need it later.
Hint

If we treat \(x\) and \(y\) as known, there are two unknowns: \(a\) and \(b\), so it would be good to have two equations to work with. Derive them from the fact that you know the lengths of both arms of the planimeter are equal to one.

Once you have the two equations, use Mathematica to solve the resulting system of equations as it does get quite messy. You can adapt the following syntax, which shows how to solve a system of equations:

	Solve[{a+b==1, a-b==17},{a,b}]

Question: Once you have solved the system, can you explain why there are two different solutions? Think about the possible physical arrangements of this device.

Graduated wheel #

The planimeter’s graduated wheel is mounted perpendicular to its tracing arm; the exact location depends on the device manufacturer. Since the planimeter ultimately counts how many times this wheel rotates, it seems appropriate for us to figure out an answer to the following:

Question: As the tracing arm of the planimeter traces a curve, under what circumstances does the wheel spin, and when does it stay still?

To begin, we consider two special cases. In the first, suppose that the motion of the tracing point is exactly in the direction of the tracing arm, as in the animation below:

It should be clear that in this case the wheel will not turn at all. On the other hand, if the motion of the tracing tip is perpendicular to the tracing arm, the wheel will turn by an amount proportional to the distance traveled by the tracing point; the proportion will depend on exactly where the wheel is mounted along the tracing arm:

If the wheel is halfway between the pivot and the tracing point, the constant of proportionality will be one-half. If it is a tenth of the way, the constant will be one tenth.

What if the tracing point is being moved some other direction than the two above? How much will the graduated wheel turn? Read on.

Planimeter vector field #

Things will be a little easier if we introduce a new vector. Let \(\vec{F}\) be a unit vector perpendicular to the tracing arm and recall that:

  • if the tracing point moves perpendicular to \(\vec{F}\) then the graduated wheel does not spin, and
  • if the tracing point moves exactly in the direction of \(\vec{F}\), then the graduated wheel spins by an amount proportional to the speed of the tracing point.

In fact, as the coordinates \((x,y)\) of the tracing point change, so does the perpendicular vector \(\vec{F}\), so in fact what we have is not a single vector, but a vector field!

Definition:

To define a vector at a point in the plane, first arrange the planimeter so that its tracing point is at \((x,y)\). Then draw a unit vector starting at the tracing point perpendicular to the tracing arm. Since there are two possibilities, pick one, just make sure to do so in a consistent manner. This process defines a vector field \(\vec{F}(x,y)\) whose graph appears below:

Plot of the vector field F(x,y)

We are now ready to return to the question of what happens to the graduated wheel if the tracing point moves in a general direction, say \(\vec{v}\). How quickly will the graduated wheel spin?

Exercise: If the tracing point is moving with velocity \(\vec{v}\) explain why the speed at which the graduated wheel rotates is proportional to the dot product: \[\vec{F}(x,y) \cdot \vec{v} = ||\vec{F}(x,y)|| \; ||\vec{v}|| \; \cos \theta = ||\vec{v}|| \; \cos \theta. \]
Hint

Consider the quantity \( \vec{F}(x,y) \cdot \vec{v}\). If \(\vec{v}\) is perpendicular to \( \vec{F}(x,y)\), then \(\vec{v}\) is parallel to the tracing arm, and by our first observation, the graduated wheel will not spin. Note that in this case, \(\vec{F}(x,y) \cdot \vec{v} = 0\), since they are perpendicular.

If \(\vec{v}\) is parallel to \( \vec{F}(x,y)\), that is, motion of the tracing tip is perpendicular to the tracing arm, then

\[ \vec{F}(x,y) \cdot \vec{v} = ||\vec{F}(x,y)|| \; ||\vec{v}|| \cos \theta = ||\vec{v}|| \]

To get the speed of the graduated wheel, we have to multiply this by 10%. Now what happens when \( \theta \) is something else?

The final exercise justifies our introduction of the vector field \(\vec{F}(x,y)\).

Exercise:

Suppose that \(\vec{r}(t)\) for \(a \leq t \leq b\) is a parametrization of the motion of the tracing tip as it circumnavigates the curve \(\mathcal{C}\). Explain why the planimeter wheel will have turned a total number of turns proportional to:

\[\int_\mathcal{C} \vec{F} \cdot d\vec{r} = \int_a^b \vec{F}(\vec{r}(t)) \cdot \vec{r}’(t) \; dt.\]

Hint

The key is the right hand side of the expression: \[\int_a^b \vec{F}(\vec{r}(t)) \cdot \vec{r}’(t) \; dt.\] By the previous exercise, the rotation speed of the graduated wheel is \(\vec{F}(x,y) \cdot \vec{v}\) where \(\vec{v}\) is the velocity of the tracing point. If we a tracing \(\mathcal{C}\) according to the parametrization \(\vec{r}(t)\), then:

  • the velocity of the tracing point at time \(t\) is \(\vec{r}’(t)\), and
  • the position of the tracing point at time \(t\) is \( \vec{F}(\vec{r}(t))\).

So the speed of rotation of the graduated wheel is proportional to \(\vec{F}(\vec{r}(t)) \cdot \vec{r}’(t)\). To get the number of rotations, we integrate rotation speed.

3. A formula for \(\vec{F}(x,y)\) #

We are ready to address the main goal of this lab: does the planimeter actually work? In the previous section, we found that the planimeter does in fact compute a path integral. It is a path integral of a very particular vector field \(\vec{F}(x,y)\). Follow the outline of these problems to complete our work.

Problem 1: Use the fact that \(\vec{F}(x,y)\) is perpendicular to the tracing arm to write down a formula for this vector field in terms of the coordinates \(a\), \(b\), \(x\), and \(y\) of the tracing point and the pivot.
Problem 2: From our work in Section 2, we know that \(a\) and \(b\) are in fact functions of \(x\) and \(y\). Incorporate the formulas you found there into your expression for \(\vec{F}(x,y)\) you found above. You should now have an expression for \(\vec{F}(x,y)\) entirely in terms of \(x\) and \(y\). Make sure that the vector field consists of unit vectors.
Problem 3: Compute \(curl \; \vec{F}\). Mathematica is great at computing derivatives and simplifying complicated expressions. With an appropriate scaling factor, does the planimeter compute area, or is the device another example of nineteenth-century snake oil?

Finally, to make the planimeter usable, we need to know the constant of proportionality: what number should be multiply our answer by to obtain the area of the region. This turns out to be quite simple. Each device would come with a constant that was calibrated at the factory. The calibration sheet for my planimeter is a small piece of paper with different values for different settings for the device:

Planimeter calibration sheet