Integration is one of the two most important operations in calculus (the other being, of course, differentiation). It is important both because it has countless applications both inside and outside of mathematics, and also because it is extraordinarily versatile.
Some applications of integration include:
- deriving formulas for calculating the area of a circle, volume of a sphere, or in general any measure (weight, density, center of mass, rotational inertia) of a real object of any geometric shape
- using information about forces to derive the path of motion of an object, or calculating an object’s position at any time given its velocity
- running the mathematical machinery responsible for Laplace transforms, which allow complex circuitry and mechanical systems to be modeled easily, and Lagrangian mechanics, which allow computing the behavior of physical systems for which a direct approach is intractable
As you might guess from its many applications, integration can be used in many different situations, so each use can at first seem quite different. However, every application of integration follows the same general outline.
Introduction to integration with pizza
One might say that integration is a general form of rescaling. Let me demonstrate what I mean by that using an analogy. Say we have a pizza:
We can cut up that pizza in many different ways. For example, into four pieces like so:
The exact way doesn’t matter. Now suppose we measure how large each of these pieces is. For example, maybe in square inches the sizes are:
Of course, you can get the size of the whole pizza by adding together the sizes of all the slices. In this case,
But let’s say instead of the size, we want to know how many pepperonis are on the pizza. (Bear with me, we’ll get to integration eventually.) We can still take the same approach, namely to find out how many pepperonis are on each slice individually, and add them all together.
But we don’t want to count the pepperonis by hand. Instead, we can use our knowledge of how common pepperonis are on different types of pizza.
- The red piece is cheese, so there are no pepperonis. In other words, zero pepperonis per square inch.
- The green and blue pieces are pepperoni, so they have an average of 0.15 pepperonis per square inch.
- The purple piece is mixed, so it has an average of 0.05 pepperonis per square inch.
We can then calculate the total number of pepperonis (p):
What we have done here is essentially rescaled each slice of pizza. Now instead of a size for each slice, we have a number of pepperonis. The red piece has been scaled to zero, since its scaling factor, 0 p/in^2^, is zero. Both the green and blue pieces have been scaled by a factor of 0.15 p/in^2^, which is three times greater than the scaling factor for the purple piece (0.05 p/in^2^). Note that this particular scaling operation also changes the units, from area to pepperoni count.
How can we describe the scaling operation? One way is to just use English, like I did in the bullet-point list above. However, for examples with more complex mathematics than pepperoni counting, it is useful to use a function. Sort of like this, but more mathy:
Generalizing the integration idea
With our pizza example in mind, here is the outline for every form of integration:
- We have some object, either concrete or abstract (e.g. the pizza). Usually this object is called the domain of integration, because we define a function whose domain is the object (e.g. the pepperoni density function).
- We have some way to split up the object into pieces (e.g. the four slices for the pizza). This is called a partition.
- We have some way to calculate a size for each piece in the partition (e.g. the area of each pizza slice).
- We have some way of assigning a scaling factor to each point in the domain of integration (e.g. the aforementioned pepperoni density function).
Then the procedure is simple:
- split the object (domain) into different pieces, according to the partition
- figure out what the scaling factor is for each piece
- multiply the size of each piece by its corresponding scaling factor
- add up the results.
In the pizza example, we couldn’t calculate the number of pepperonis directly. Splitting the pizza into different slices, where each slice had a single pepperoni density, was a crucial step. This is true in general, and it is why integration is so useful.
The reason this is a useful technique is that if the scaling factor is different for different parts of the object, you may not be able to calculate the answer directly. But by splitting the object apart, the calculation for each piece individually is quite easy: just multiplication.
You may ask why it would ever actually useful to find out what number you get when you scale a bunch of different parts of the same object in different ways. Let’s see why.
Application: using integration to calculate area
We all learn in geometry class how to calculate the areas of some simple shapes: squares, rectangles, triangles, circles, trapezoids, and so on. But in the real world, shapes are much more complicated. Think of your pen, your car’s engine, the wing of an airplane, and so on. But calculating the areas of irregular shapes is important if we want to know things like the stress experienced by your pen, the heat transfer through your car engine, and the lift on the airplane’s wing. Often all we have to go on is either a mathematical function describing the shape of the object or just numerical measurements of its outline. But with integration, we can still find its area.
It goes like this. Let’s say we want to find the area of the red region shaded in the figure above. Then to integrate, we will split the red region into many small regions, calculate the area of such small region, and add together all the answers.
“But wait,” you object, “you haven’t split the area at all! There are little bits that aren’t covered by the rectangles, and the rectangles stick out in other places.”
My answer to that problem? Use more rectangles. You can see that the difference in area between the rectangles and the curve is not that large, because the rectangles are pretty narrow. If we used more rectangles, the approximation would be even better. If we want a numerical approximation of the area under the curve, we can just use “enough” rectangles – where “enough” means the approximation is good enough for whatever we’re trying to do. We’ll see later how, paradoxically, we can use this approximation technique and some additional cleverness to get an exact answer, with no error at all.
So now let’s see how this technique fits into the general outline of integration discussed earlier:
- The object (domain) being integrated is not actually the region we’re trying to calculate the area of. Instead, it’s the line segment below the region, on the x-axis, from to We’ll see why momentarily.
- The partition is how we have split up the domain of integration. Since the domain is a line segment on the x-axis, our partition splits that segment into smaller line segments. Each small line segment is the base of one of our rectangles.
- The size of each small line segment is its length.
- The scaling factor is given by the function
Let’s see why this setup works.
From our outline of integration above, our first step is to figure out what the scaling factor is for each piece (small line segment). Since the scaling factor is given by the function we can calculate it by taking the x-coordinate of the segment and then finding Now you may object that a line segment doesn’t just have one x-coordinate, because it’s a segment rather than a point. But since these are small line segments, the different x-coordinates aren’t that different, and you can pick any one. Remember, this is an approximation, so there will be a bit of error.
So our scaling factor is What does this mean, physically? It is actually a length: the height of one of our rectangles! Pictured above is a single rectangle from the previous figure. Since it goes from the x-axis up to the curve, its height is the difference of the y-coordinates,
The next step in our outline is to multiply the size of each piece in our partition by its scaling factor. In the figure above, the piece in our partition is the line segment on the x-axis which forms the base of the rectangle. Let’s call its length because it is a change in x-coordinate (a horizontal distance).
Now we can see that the multiplication of our partition piece size by its scaling factor is actually a multiplication of a horizontal length by a vertical length: we are just calculating the area of our rectangle.
The final step in integration is add all the products of sizes and scaling factors, which in this case means combining the areas of all the rectangles to get an approximation of the area under the curve.
Symbolically, we can write this calculation as:
which means that we take all the rectangles from to for each one we multiply the height by the width and then we take all the resultant areas and add them together (indicated by the symbol In calculus, we would call this operation integrating the function over the domain to .
So this gives us a way to calculate areas, if we are so inclined. This technique is used numerically all the time. However, what is more interesting is calculating the area exactly. We will come back to calculating areas after we know how to integrate exactly.
Application: integrating slope
Let’s consider another example of integration. Here we have a generic function:
In the previous example, we integrated the function over the the domain to and this gave us the area under the curve (or at least an approximation). It is possible to guess this result, like so: in integration, we are multiplying a size by a scaling factor (the function being integrated). The size of the domain is a horizontal length, from to and the scaling factor is a vertical height, from the x-axis to the curve. We would then expect that the resulting product will be an area, and it is.
Now, we will integrate not the function itself but its derivative over the same domain to And this is what we get:
Let me explain. Our domain is still the portion of the x-axis between and so our partition is composed of small horizontal line segments. These are the horizontal lines in green. Now, the function we are integrating is the derivative which is the slope of the curve. The slope of a curve is basically its vertical displacement divided by its horizontal displacement,
so if we take the horizontal displacement and multiply it by the slope, we will then get the vertical displacement,
In this case, is the slope of the red line segments drawn tangent to the curve, is the length of the horizontal green segments, and is the length of the vertical green segments.
For each horizontal line segment, then, multiplying by the scaling factor (the slope of the curve) gives us the corresponding vertical segment. The last step in integration is to add all of these results together, and this gives (approximately) the difference Why is this?
Let’s call the y-coordinate of the left-hand side of the first red segment Then the left-hand side of the second red segment is the third one is and so on up to which is the right-hand side of the last line segment.
The vertical displacement for the first red line segment is the displacement for the second is and so on up to for the last. Adding all these displacements together, the intermediate values cancel and we just get Because the red line segments approximately follow the curve, is approximately and is approximately That means our total displacement is an approximation of
Now, this is only an approximation. Depending on the exact points we choose within each interval of x-coordinates in order to measure the slope of the curve (and thus determine the slope of the line segments), the red segments may overshoot or undershoot the curve by a little. But nevertheless, we can write:
where means the width of the appropriate horizontal line segment in the figure.
Let’s pause again to consider why this makes sense. We are integrating the function over the domain to As in the previous example, the size of the domain is a horizontal length. However, now the scaling factor (the function being integrated) is a slope: the derivative of We know that multiplying a slope by a horizontal displacement gives a vertical displacement, so it is not surprising that we get a vertical displacement, from the integration.
Integration as a limiting process
So far, it may look like integration is only useful if you have a computer, since you need to split your domain into many, many pieces to get a good approximation of the answer. But this is far from true.
Consider the result we obtained in the previous section by seeing what happens when we integrate the derivative of a function
This is an approximation, and it gets better when we split the domain (from to into more pieces. (Why? If you look at the figure, adding more line segments will allow the slope of the segments to match the slope of the curve more closely, and so the segments will follow the curve better, leading to less error.) What happens if we just keep letting the approximation get better and better?
The result is called a limit. There is no such thing as splitting our domain into an infinite number of infinitely short line segments, but we can imagine doing that as a conceptual aid. What we are really doing is investigating how integration responds to the limiting process of increasing the number of segments in our partition without bound.
We can express the fact that the error in our approximation gets smaller and smaller with the following equation:
The symbol has been replaced with to indicate that we are now talking about a limiting process (or, if you will accept some hand-waving, a sum of an “infinite” number of products). And the sign has been replaced with an sign to indicate that the limiting value of the summation as we add more and more segments is exactly We may never get the exact right answer when doing a real computation, but we can still say exactly what our numbers are getting closer and closer to. Finally, has been replaced with as a matter of convention. A quantity starting with in integration, is called a differential, and the indicates that it is “infinitely small” (or, more realistically, it is part of a limiting process that makes it smaller and smaller without bound).
Integrating exactly: the trick
The equation that we derived in the previous section,
is called the fundamental theorem of calculus. Why is it fundamental? Because instead of an approximation, we have an equality – and this allows us to find the exact value of any integral (more or less).
Let’s return to the problem of finding the area under a curve. We previously said you could get an approximation of the area under the curve by splitting it into tall, thin rectangles and then evaluating the sum
As a reminder, it looks like this:
Let’s now consider applying the same limiting process as we did in the previous example (relating slope and displacement). If we split the area under the curve into more (thinner) rectangles, we’ll get a better approximation. And the value of this approximation will get closer and closer to the true area under the curve (we call that its limiting value). So, using the same notation as before, we can say that the exact area under the curve is given by
where is the height of each rectangle and is the differential corresponding to the width of each rectangle.
But wait: this looks very similar to what we have above. In fact, it’s exactly the same as the left-hand side of the fundamental theorem of calculus, except that we have instead of
The trick is to rewrite the expression for the area under the curve so that it looks like the fundamental theorem of calculus. In particular, let’s say that we can find some new function whose derivative is – in other words, This would mean that
and then the fundamental theorem of calculus would tell us that
The function is called an antiderivative of because it’s the opposite of a derivative. (The derivative of an antiderivative is just the function you started with.)
Because this is a pretty abstract bit of trickery (and because it is a very important bit of trickery), we’ll now spend some time looking at examples and discussing the intuitive meaning of our results.
Example: calculating the area of a triangle
Let’s start simple. Shown below is a triangle, with base and height equal to 1:
I’ve claimed that we can calculate the area under any curve using integration and the fundamental theorem of calculus, so let’s apply that to the curve If you look at the area under from to shaded in red on the figure above, it forms a triangle with area
Now in this case, we have and so our area integral looks like:
The interpretation is that we are splitting the area shaded in red into many thin rectangles, and then considering the limiting process as those rectangles become thinner and thinner.
As shown in the figure above, the quantity which appears inside the integral corresponds in a very direct way to the area of a rectangle. The height is and the width is The symbol indicates that we should add up the areas of all these rectangles (from to
Now we can find a function whose derivative is equal to i.e. some function such that It just so happens that we can use in this case, since
According to the fundamental theorem of calculus, we then have
So our result from integration agrees with basic geometry, which is good. Let’s discuss the intuition of our calculation. You might notice that the expression actually gives the area of the red shaded triangle in the figure below:
When we get the total area of But we can view that final answer as the accumulation of the areas of all the individual rectangles, starting at the left and moving to the right. If we interrupt that accumulation at some point in between, we get a picture like the one above. Imagine drawing rectangles over the shaded area: we’ve added the areas of all the rectangles on the left-hand side, but haven’t included any on the right-hand side yet.
So we can view the antiderivative as a function that describes how the areas of rectangles with individual areas add up from left to right. Why is it an antiderivative? Let’s look at what happens when we take the derivative. Intuitively, the derivative of this area accumulation function will reflect the rate at which new area is added as we increase So, we can calculate the derivative as:
As usual, we denote the change in by the differential In the figure below, you can see that increasing by some amount will cause the area to increase by approximately a small rectangle (the error in this approximation can be disregarded since differentiation is a limiting process, and the approximation may be made as precise as desired):
As we’ve already discussed, the area of such a rectangle is the product of its height and width, Thus, the derivative is
So this tells us that when we take the derivative of our accumulation function, we get back the function we started with. Given that the accumulation function is the value of our integral, we have found that differentiating “cancels out” integration in some sense. The formal statement, which is a second form of the fundamental theorem of calculus, is written like this:
You may ask where all the came from. It is because we want to make a distinction between the x-coordinate of any arbitrary rectangle (since we are dealing with all the rectangles that make up the triangles above) and the x-coordinate of the specific rectangle at the right-hand side. We use the variable t for the former, and x for the latter. The use of specifically isn’t important. It’s just that we’d like to pick two different letters, to avoid confusion.
… to be continued