Chapter 5: Integration

Welcome to Last Minute Lecture.

This free chapter overview is designed to help students review and understand key concepts.

These summaries supplement not replaced the original textbook and may not be redistributed or resold.

For complete coverage, always consult the official text.

Imagine, you are driving your car down a perfectly straight, empty highway.

Okay, setting the scene.

Right.

And you lock the cruise control at exactly 60 miles per hour and you drive for exactly two hours.

Well, figuring out how far you went is literally like middle school math at that point.

Exactly.

60 times 2 is 120 miles.

Yes.

It's clean.

It's elegant.

But, you know, it relies on one massive assumption.

That your speed never ever changes.

Right.

But real life isn't a straight highway with cruise control.

What happens when you hit traffic?

You slam on the brakes.

Yeah.

Or you accelerate to pass a semi truck or slow down for a sharp curve.

Your speed is fluidly changing every single millisecond.

So how on earth do you mathematically calculate the exact distance you just traveled when you can't, you just can't multiply two static numbers together anymore?

That seemingly simple question, how to measure accumulation when the rate is constantly changing that is the engine that drives the entire second half of calculus.

It really is.

It's the problem that stumped brilliant minds for centuries and it is exactly what we are going to conquer today.

So if you're a college student staring down one of the most notorious legendary milestones in all of mathematics, which is integration, you are in exactly the right place.

Absolutely.

Our mission for this deep dive is to build the concept of integration from the ground up.

We aren't just going to like memorize formulas.

Oh, memorizing won't save you here.

We're going to look at the underlying mechanics so you can actually see the gears turning behind the math.

Yeah, we'll start with a very intuitive geometrical problem,

literally approximating areas with tiny rectangles and then we'll push that idea to its absolute limit.

And eventually we'll uncover this magical hitting connection between derivatives and integrals.

Okay, let's unpack this starting with that car analogy.

If we graph your cruise control drive where the vertical y -axis is your velocity, your speed and the horizontal x -axis is time, what does that shape actually look like?

Well, since your velocity is locked at a constant 60 miles per hour, the graph of your velocity is just a perfectly flat horizontal line at the a value of 60.

Right, because the speed isn't going up or down.

Exactly.

Now, if you look at the time interval from hour zero to hour two on the x -axis, the space trapped under that flat line bounded by the axis, it forms a perfect geometric rectangle.

Okay, so the height of that rectangle is your velocity, which we can call v, and the width of the rectangle is the time elapsed, which we'll call delta t.

Precisely.

And the area of any basic rectangle is just height times width.

So v multiplied by delta t gives you the area.

60 times 2 is 120.

Right.

That physical area trapped under the velocity graph is mathematically identical to the distance you traveled.

Distance traveled equals the area under the graph of velocity over that specific time interval.

That makes total sense for a flat line.

But we just established that in the real world, velocity changes.

It does.

So the graph wouldn't be a flat line.

It would be a curve, you know, swooping up and down as you break and accelerate.

Right, it gets messy.

I can't just find the area under a squiggly curve using basic geometry.

I mean, there's no area of a squiggly shape formula I learned in high school.

No, there definitely isn't.

So when mathematicians are faced with a shape they can't measure,

they cheat a little bit.

They cheat.

They use shapes they can measure to build an approximation.

Our core strategy to estimate the area under a continuous positive curve,

let's call our function f of x.

Okay, f of x.

On a specific interval from a starting point A to an ending point B, we chop that curve space up into a whole bunch of vertical rectangles.

Let's say we use n number of rectangles.

I always think of this like early digital graphics.

Oh, that's a good way to look at it.

You know, if you look at an old 8 -bit video game, the developers were trying to draw a smooth circle, but all they had were chunky blocky square pixels.

Right, they didn't have the resolution.

If you look closely, the edges of the circle are actually just jagged stair steps made of rectangles.

But the smaller the pixels like, the thinner the rectangles, the better and smoother the picture looks.

It starts to actually resemble a circle.

That is a brilliant analogy.

We are essentially pixelating the area under the curve.

We can't measure the curve itself, but we can't measure the pixels.

Exactly.

Let's break down exactly how we build these mathematical pixels.

First, we need to know how wide to make each rectangle.

For our initial approach, we want them all to be exactly the same width.

So we take the total width of our interval, which is B minus A, and we divide it by the number of rectangles we want, which is n.

Makes sense.

That simple division gives us our uniform width, which we denote with a delta.

So delta x equals B minus A, all divided by n.

Okay, let's ground this with real numbers for everyone listening.

If we are looking at a time interval from one second to three seconds, and we want to pixelate this with four rectangles.

Okay, so n is four.

Right.

The total width of our interval is three minus one, which is two.

We divide that total width of two by our four rectangles, and we find that each rectangle has a width, a delta x, of one half.

That's straightforward.

But how do we decide how tall they should be?

Because the curve is constantly changing height.

That is the pivotal choice.

We have to sample the curve at specific points to set the height of our flat -topped rectangles.

We call the points along the horizontal x -axis, where our rectangles sit, x sub j.

X sub j.

Okay.

To find any specific point along the base, you start at the very beginning of your interval A, and you take j number of steps forward, where each step is your width, delta x.

So the formula for your position is x sub j equals A plus j times delta x.

Exactly.

Once you are standing at that point on the x -axis, you look straight up, plug that x value into your original function f of x, and whatever the y value is, that becomes the permanent height of your rectangle.

But wait.

Where exactly inside the rectangle's base do we stand to look up?

Because if I stand on the left side of the base, the curve might be lower than if I stand on the right side.

That is very true.

And this is where we get three different flavors of approximation, right?

Like, right, left, and midpoint approximations.

Exactly right.

The choice of where you sample dictates the shape of your approximation.

If you stand on the right side edge of the rectangle's base, go straight up to touch the curve,

and use that y value as the flat height for the entire rectangle.

That would be the right endpoint approximation.

Yes.

And we denote it as r sub n.

Now if you use the left edge of the base, it's a left endpoint approximation, l sub n.

And if you walk exactly halfway between the left and right edges.

Right.

If you stand in the dead center of the base, and look up to the curve, that is a midpoint approximation, m sub n.

Let's actually walk through what this sounds like conceptually so you, the listener, can hear how the mechanics operate without getting lost in the weeds.

Good idea.

Let's say we want to calculate r sub 4, a right endpoint approximation using four rectangles, for a classic simple curve.

Let's use a parabola, f of x equals x squared.

Classic choice.

And we're looking at the interval from 1 to 3.

Let's build it.

Step one is determining the width of our mathematical pixels, our delta x, and locating our right endpoints.

Right.

And as we calculated a moment ago, the width delta x for four rectangles over this interval is one half.

Since we are doing a right endpoint approximation, we don't sample the height at our starting point of 1.

We take our first step forward by our width, one half, to find the right edge of the very first rectangle.

So the right edge of rectangle 1 sits at 1 .5, or 3 over 2.

We take another step of a half, and rectangle 2's right edge is at 2.

Another step, rectangle 3, is at 2 .5, or 5 over 2.

And one final step means the fourth rectangle's right edge lands exactly at the end of our interval, 3.

Perfect.

Step two is calculating the total area of these four rectangles.

We know the area of one rectangle is its width times its height, so r sub 4 is the sum of the area.

That's going to be a lot of adding.

Well, because every single rectangle shares the exact same width, delta x, we could actually factor that width out to the front of our calculation to save a massive amount of arithmetic.

Oh, nice.

We can just say,

let's add up all four heights and then multiply that grand total by the single width at the end.

Okay, that makes the algebra so much cleaner.

So we have our width, delta x, multiplied by a big bracket.

Inside the bracket, we have the heights.

F of 3 over 2 plus F of 2 plus F of 5 over 2 plus F of 3.

And since our specific function is F of x equals x squared, finding the height just means squaring those x values.

Right.

We're looking at one half and multiplied by the sum of 3 over 2 squared plus 2 squared plus 5 over 2 squared plus 3 squared.

If we expand those squares, we get one half times the sum of 9 over 4 plus 4 plus 25 over 4 plus 9.

Exactly.

I'll spare you the fraction addition, but if you find a common denominator and add all the stuff inside the brackets together, you get 43 over 2.

Finally, we multiply by that width of one half we left waiting outside and we get 43 over 4.

Which converts perfectly to 10 .75 in decimal form.

That is our approximation of the area under the curve using four rectangles.

But here is where we need to start calculating and start visualizing.

If you're listening to this, close your eyes for a second and actually picture this.

Sketch the curve of x squared in your mind.

It starts at zero and swoops upward, getting steeper as it moves right.

Right.

Now draw those four right endpoint rectangles we just built.

Because the curve is constantly rising going uphill and we are using the right side of the base to set the height,

the flat top of each rectangle is going to extend backward from the highest point on that stretch of the curve.

This means the top left corner of every single rectangle is going to poke out above the actual curve.

What you are visualizing is a profound geometric truth.

For any strictly increasing function,

any curve that goes uphill, a right endpoint approximation r sub n will always be an overestimate.

Because of those corners.

Exactly.

Those corners poking out mean you are calculating slightly more area than actually exists physically under the curve.

Your mathematical pixels are bleeding over the lines.

And if we flip it, what if we used a left endpoint approximation, l sub n, on that exact same uphill curve?

If you use the left edge, you are sampling the lowest point of the curve for that interval.

The flat top extends forward, which means the entire rectangle gets tucked completely underneath the rising curve.

Ah, so it doesn't poke out at all.

Right.

You will have empty, unmeasured white space between the top of your rectangles and the actual curve.

Therefore, l sub n for an increasing function will always be an underestimate.

Okay, so our result of 10 .75 is an overestimate.

The true area is somewhat less than that.

Correct.

If we wanted a better estimate, we'd use more rectangles.

Like if we use six rectangles, the rectangles get thinner, the blocky pixels get smaller, and the textbook notes the estimate drops to roughly 10 .037.

The rectangles are hugging the curve tighter.

But I have to be honest.

Adding up four things was annoying.

Adding up six things is tedious.

If I want a really good approximation, I might need a hundred rectangles.

And writing out a sum with a hundred terms is a fast track to madness.

You are definitely not the first person to feel that despair.

Which is exactly why mathematicians developed summation notation.

Oh, the sigma!

Yes.

We use the Greek capital letter sigma to write these massive, sprawling sums compactly.

It looks like an intimidating jagged symbol, but it is literally just a loop of computer code.

A set of instructions.

Break the code down for us.

What are the pieces of the sigma?

At the bottom of the sigma symbol, you have what's called the index of summation.

It's usually a dummy variable, like the letter j, set equal to a lower limit, like j equals 1.

So that tells you where to start counting.

Right.

At the very top of the sigma is your upper limit.

Let's say n.

This tells you where to stop counting.

And to the right of the sigma is your general term, your actual mathematical formula.

So it's basically saying, take the number 1, plug it into the formula, write down the result.

Now take the number 2, plug it in, write down the result, keep doing this, stepping up by integers until you reach the top number n.

Yes, exactly.

And once you have all those results written out, add them all together.

That is precisely what it means.

And because it's just a compact way of writing addition, it follows some incredibly helpful linearity rules.

Well, for example, if your formula has a constant number multiplied inside of it, like our uniform with delta x, you don't have to keep multiplying it in every single step.

You can pull that constant completely out to the front of the sigma sign,

evaluate the entire massive sum first, and then multiply by the constant at the very end.

Just like we factored out the one half earlier.

Exactly.

Another rule is that if you have two distinct terms being added or subtracted inside your formula, you can literally split the giant sigma into two separate smaller sigmas.

You can tackle the problem in pieces.

But even with sigma notation organizing things, eventually you still have to physically add up all the numbers, right?

Like if n is 1000, you still have to add up 1000 numbers.

That's still impossible by hand.

It would be, if not for power sum formulas.

These are established algebraic shortcuts for adding up long sequences of integers, or squares of integers, or cubes.

Instead of doing the raw addition, you just plug your top number n into a shortcut formula.

For instance, if you just want to add up the first n integers 1 plus 2 plus 3, all the way to n, the shortcut formula is simply n times the quantity n plus 1, all divided by 2.

That's way easier.

There are similar, slightly more complex, proven formulas for adding up a sequence of squares and a sequence of cubes.

And there's an amazing piece of mathematical lore here.

The textbook mentions the Swiss mathematician Jacob Bernoulli.

Back in the late 1600s, hundreds of years before a computer existed, he was obsessed with finding the patterns hidden inside these power sums.

He didn't just stop at squares or cubes.

He generalized the pattern and figured out the shortcut formula for the sum of the first 10 powers.

Meaning, he could calculate 1 to the 10th plus 2 to the 10th plus 3 to the 10th, all the way up to 1000 to the 10th.

That's unbelievable.

And he famously boasted that using his formula, he calculated that monstrous result in less than half a quarter of an hour, seven and a half minutes.

Just raw, terrifying, genius pattern recognition.

It's a phenomenal demonstration of human intellect.

And we absolutely need those power sum formulas.

We need those shortcuts for the next vital step in our logical journey.

Why is that?

Because approximations are nice, but they aren't calculus.

Right.

If you are listening, here is where we cross the threshold.

We don't want a pretty close approximation of the area.

We want the exact mathematically perfect area.

Exactly.

To get that, we don't use four rectangles or a hundred or a billion.

We take the limit as the number of rectangles n approaches infinity.

This is a profound philosophical and conceptual leap.

We are imagining an infinite number of rectangles.

As the number of rectangles goes to infinity, the width of each individual rectangle shrinks towards zero.

They become like lines.

Infinitely thin slices of area, perfectly conforming to every microscopic swoop of the curve.

The exact area under the curve is mathematically defined as the limit as n approaches infinity of r sub n.

This sounds elegant in theory, but the textbook shows us what happens when you actually try to execute this limit using raw algebra.

And frankly, it is a nightmare.

It really is.

They walk through finding the exact area under a parabola.

f of x equals 2x squared minus x plus 3 over the interval from 2 to 4.

And they do it the hard way, constructing the infinite limit of a right end point sum.

It is intentionally terrifying.

Let's trace the shape of the labor involved without getting bogged down in every number.

First, you have to express your width delta x abstractly.

Since the interval is two units wide, your width is 2 over n.

Okay, because n is going to infinity.

Right.

You can't use a clean fraction like one half anymore.

It stays a variable.

Okay, so your width has an n in it, and then you have to find your sample points.

Right.

Your end points are x sub j equals 2 plus j times 2 over n.

Now to find the heights of your rectangles, you have to take that entire messy expression and substitute it into every single x in the original parabola equation.

Which means I have to take 2 plus 2j over n, and I have to square it.

I have to physically multiply out all those binomials, keeping track of the j's and the n's.

Then I multiply it all by 2, subtract the original expression, and add 3.

You are generating a massive,

sprawling algebraic polynomial filled with fractions where n and n squared are sitting in the denominators.

Exactly.

And once you have that massive polynomial, you have to apply the sigma notation to it.

You have to distribute the sigma across the polynomial using the linearity rules.

Then you have to replace every single sigma with those power sum shortcut formulas we just talked about.

This introduces even more n's into the numerators and denominators.

Right.

You spend an entire page of scratch paper just expanding fractions, finding common denominators, and canceling terms.

And after all of that algebraic torture, you haven't even done the calculus yet.

You still have to take the limit of this massive algebraic monstrosity as n goes to infinity.

Now, spoiler alert, the final exact answer they get is 50 over 3.

But the journey to get there is an agonizing slog.

It screams that there has to be a better way to do this.

I mean, if we are engineers or physicists, we can't be doing infinite limit sums on two pages of scratch paper every single time we want to find the area under a curve.

And that profound frustration is the perfect bridge into our next conceptual leap.

We've been forcing ourselves to use rectangles that are all exactly the same width.

But what if we didn't have to?

Wait, if the rectangles aren't the same width, how do we even set up the math?

This is where we introduce the concept of Riemann sums, named after Bernhard Riemann, a brilliant 19th century mathematician who formalized this chaos.

Okay.

A Riemann sum completely relaxes the strict rules we just spent 20 minutes setting up.

Instead of neatly dividing our interval into n equal sub -intervals, Riemann says, let's just create a partition.

We'll call it P.

A partition?

This is simply a random choice of points that chops the interval into sub -intervals that can be wildly different widths.

So I could have a massive wide rectangle on the left and a microscopic skinny rectangle right next to it.

Exactly.

And the freedom doesn't stop there.

Instead of strictly forcing ourselves to use the right endpoint or the left endpoint or the midpoint to set the height, Riemann says we can pick a sample point.

Let's call it C sub.

Absolutely anywhere within that specific sub -intervals base.

So for the wide rectangle, I could sample the height on the far left edge.

But for the skinny rectangle next to it, I could sample the height at some random spot slightly off center.

This sounds like mathematical anarchy.

How can you find a reliable limit if everything is randomized?

It sounds like chaos, yes.

But it provides incredible theoretical flexibility, which is necessary for advanced proofs.

Because Riemann figured out that we don't need to control every rectangle.

We only need to control the largest one.

The largest one.

We define the most important metric of this chaotic partition P as the norm, denoted by double vertical bars like absolute value signs.

The norm is simply defined as the width of the widest rectangle in your entire collection.

Okay, the widest rectangle is the norm.

Why does that matter?

Because it leads us to the formal universal definition of the definite integral.

The definite integral of a function f from a to b is defined as the limit of the Riemann sum as the norm approaches zero.

Let me think about this.

If the width of the absolute widest rectangle in the chaotic pile is shrinking down to zero, then the widths of all the smaller rectangles must also be shrinking to zero.

Right.

They are being crushed.

And if every single rectangle is shrinking to a width of zero, but they still have to cover the whole interval from a to b, the number of rectangles must explode to infinity.

Precisely.

By forcing the widest rectangle to shrink to zero, you naturally force the number of rectangles to infinity, regardless of how chaotic the initial setup was.

We arrive back at our exact area.

But because of Riemann's generalization, we know this limit works no matter how you chop up the space.

And because this concept is so fundamental to the universe, we don't write limit of a Riemann sum every time.

We use one of the most iconic, elegant symbols in all of mathematics.

Right.

It looks like an elongated, stretched -out letter S.

Gottfried Wilhelm Leibniz, one of the founders of calculus, introduced it.

And the shape isn't an accident.

It originally stood for summa, the Latin word for sum.

Let's make sure you, the listener, have the vocabulary down for this symbol because you will be looking at it for the rest of your mathematical career.

Let's break down the anatomy of the notation.

The elongated S itself is the integral sign.

The mathematical function sitting inside it, the thing you are actually trying to find the area under, is called the integrand.

Right.

At the very bottom and very top of the stretched S, you have your A and B numbers.

Those are your limits of integration, marking where you start and stop measuring area on the x -axis.

And finally, at the very end of the expression, you have a dx.

That dx is crucial, though it often confuses students.

It represents the infinitely small width of your rectangles.

But more practically, the x is just a dummy variable.

It tells you which variable in your function is the one changing along the horizontal axis.

Could be det, if your axis is time, or do, or die.

It simply tells you what language the integral is speaking.

Now, setting all this notation up raises a very strange, almost philosophical question.

We keep talking about finding area.

In the physical world, area is a positive thing.

You can't go to the hardware store and buy negative three square feet of carpet.

True.

But what happens if the graph of our function dips below the horizontal x -axis?

If the curve dips below the x -axis, the function values we use to determine the height of our rectangles,

those are negative numbers.

Right.

And since we are calculating the area of our pixels by multiplying height times width, if the height is a negative number, the resulting area of that rectangle is mathematically negative.

Correct.

In cure geometry, negative area is nonsense.

But in calculus, the definite integral computes what we call the net signed area.

Net signed area.

This is a vital distinction.

It means the integral takes all the physical geometric area sitting above the x -axis and treats it as positive.

Then it takes all the geometric area hanging down below the x -axis and treats it as negative.

It subtracts the bottom area from the top area.

So if a curve swoops up, creating 10 units of area above the axis and then swoops down, creating 10 units of area below the axis, the positive and negative areas perfectly cancel each other out.

Your definite integral would be exactly zero, even though there is a lot of physical space trapped by the curve.

Exactly.

This behavior treating area algebraically rather than just geometrically gives the definite integral some very logical, very useful properties.

Let's rapid fire through the core properties you need to know.

Let's do it.

First, if you integrate a plane constant number, let's call it k from a to b, the answer is just k multiplied by the quantity b minus a.

Because a constant function is just a flat horizontal line.

You aren't dealing with curves.

You're just finding the area of a single simple rectangle with height k and with b minus a.

Second is linearity.

If you have a definite integral of two functions added together, f of x plus g of x, you can just break it apart.

It's the exact same as taking the integral of f of x, taking the integral of g of x and adding the two answers together.

Very handy.

And if there is a constant number multiplying the function, you can factor it entirely out to the front of the integral sign, just like we did with the sigma notation.

What if you reverse the limits of integration?

Normally we read left to right integrating from a to b.

What if the problem asks you to integrate backwards from b down to a?

Reversing the limits simply flips the sign of the answer.

So the integral from b to a is the negative version of the integral from a to b.

The reason is that if you go backwards, your delta x widths become negative.

Speaking of limits, what if your upper and lower limits are the exact same number?

What is the interval from a to a?

That evaluates to zero.

Think about the geometry.

Your starting point and ending point are identical.

Your interval has zero width.

You're trying to find the area of a line, and a line has no width.

Therefore, it traps new area.

And finally, additivity for adjacent intervals.

If you integrate a function from a to b, and then you add the integral of that exact same function from b onward to c.

It's mathematically identical to just doing one massive integral all the way from a to c.

You're just gluing two puzzle pieces of area together at the seam b.

To ground this strange concept of net signed area, in reality, the textbook provides a fantastic practical example involving an economics concept.

Example four talks about a grid -connected energy system.

Oh, the solar panel example.

We have a family, the Molinas, who have solar panels on their roof.

Let's define a function E of t that represents their net energy use at any given time t.

Okay, so when the sun is shining brightly at noon, the panels produce way more energy than the house is using.

So E of t is negative.

They're literally pushing energy backward out of the house, feeding it into the Apollo power company grid.

Right, but when it's dark outside, the panels do nothing.

The house is drawing power from the grid to run the fridge and the lights, so E of t is positive.

If you take the definite integral of E of t over a full 24 -hour period,

the math does something incredible.

It adds up all the positive area, which represents the total power the family pulled from the grid at night.

And it ends up all the negative area, which represents the total power the family generated and gave to the grid during the day.

And the positive and negative cancel each other out where applicable.

The final result of that definite integral is a single number representing the total kWh of energy the family owes the power company at the end of the month.

The integral effortlessly tallies the net change of a fluctuating system.

Which perfectly positions us to ask the most important question of the day.

Up until now, we've been hyper -focused on finding the area under a curve.

We know what it represents, and we know we can approximate it with limits of Riemann sums.

But we also agreed that calculating those infinite limits algebraically is a nightmare, so we need to flip the script.

We need to look at the inverse problem.

We spent all of calculus eye learning how to take derivatives, finding the instantaneous slope of a function.

The question is, if someone hands us the derivative of a function, can we work backward to figure out what the original function was?

This process of working backward is called finding an antiderivative.

And we represent this general family of antiderivatives using a new concept.

The indefinite integral.

An indefinite integral.

It looks exactly like the definite integral symbol we just learned, but notice the crucial visual difference.

There are no limits of integration.

There is no a at the bottom, and no b at the top of the elongated s.

It's just the integral of f of x dx.

Because we aren't finding a specific numerical area trapped between two hard boundaries anymore, we are finding a general algebraic formula.

We are undoing a derivative.

But there is a massive trap here that catches almost every calculus student at least once on an exam.

Ah yes, the dreaded constant of integration.

You must always, always add plus c at the end of an indefinite integral.

If you forget it, your answer is technically incorrect.

Let me explain exactly why this happens, because it's fascinating.

Think back to derivatives.

What is the derivative of x squared?

It's 2x, right?

The 2 comes down, the power drops by 1.

Simple.

Now, what is the derivative of x squared plus 5?

It's also 2x, because the derivative of any constant number is 0, so the 5 vanishes.

What is the derivative of x squared minus 100?

It's 2x, the negative 100 vanishes.

The act of taking a derivative destroys information.

It wipes out any constant flat baseline the function had.

Therefore, if I hand you the function 2x and ask you to integrate it to find its antiderivative, how do you know which original function I started with?

You don't.

Was my original function x squared?

Was it x squared plus 5?

Was it x squared minus 1000?

You literally don't know.

The information is gone forever.

So to cover all our mathematical bases, we say the indefinite integral is x squared plus c, where c represents any possible constant number in the universe.

Think of it like a map.

The derivative tells you your velocity, how fast you are driving.

The integral works backward to tell you your position where you are on the map.

Right.

If I am blindfolded and you tell me I just drove north at 60 miles an hour for two hours, I can use an integral to tell you that you moved exactly 120 miles north.

I have the displacement.

But if you ask me where am I on the map, I can't tell you, because I don't know where you started.

Did you start in New York?

Did you start in Miami?

That starting point is your plus c.

It anchors the floating relative distance onto a specific absolute reality.

Without the plus c, you just have a distance, not a location.

Okay, so how do we actually calculate these indefinite integrals?

We don't use limits and rectangles for this.

We start by literally running the rules we learned for derivatives in reverse.

The most foundational is the power rule for integrals.

Let's recall the power rule for derivatives first.

To take the derivative of x to the n, you bring the power n down to the front and multiply it.

And then you decrease the exponent by one, multiply, then subtract.

So logic dictates that to integrate, to work backward, we just do the exact opposite operations in the exact reverse order.

Instead of subtracting one from the power at the end, we add one to the power first.

And instead of multiplying by the old power at the beginning, we divide by the new power at the end.

Precisely.

To integrate x to the n, you increase the power by one and then divide by that new power.

The formal rule is the integral of x to the n is x to the n plus one divided by n plus one plus c.

Let's test it.

If I want to integrate x cubed, step one, increase the power by one, so it becomes x to the fourth.

Step two, divide by the new power, four.

So my answer is one fourth x to the fourth plus c.

Perfect.

And the beauty of indefinite integrals is that you can always check your own work.

Just take the derivative of your answer.

If I take the derivative of one fourth x to the fourth, the four comes down to the front, cancels out the one fourth, the power drops back down to three, and I am left with x cubed.

It perfectly matches what I started with.

But the textbook explicitly points out a massive fatal restriction here.

This power rule for integrals works beautifully for almost every exponent.

Except one.

The power n cannot equal negative one.

Why?

What happens if you try to use the rule on x to the negative one?

Follow the steps.

Step one, increase the power by one.

Negative one plus one is zero, so you get x to the zero.

Step two, divide by the new power.

You have to divide by zero.

Oh, right.

And as we know, dividing by zero catastrophically breaks mathematics.

It's undefined.

The power rule simply collapses.

But x to the negative one is just another way of writing the fraction one over x.

That is a very common function.

It curves gracefully along the axis.

It clearly has area underneath it.

So what is the integral of one over x?

To find that, we have to abandon the power rule and look through our mental catalog of derivative rules.

We have to ask ourselves, what function, when I take its derivative, results in one over x?

Oh, the natural logarithm.

The derivative of natural log of x is one over x.

Therefore, working backward, the integral of one over x must be the natural log.

Specifically, the integral of one over x dx is the natural log of the absolute value of x plus c.

Why the absolute value?

The absolute value bars around the x are not optional.

They are crucial because the domain of the original fraction one over x includes negative numbers, but you cannot take the logarithm of a negative number.

Putting absolute value bars around the x ensures the domains match and the math doesn't break.

We can reverse our trigonometric and exponential derivatives just as easily.

We know the derivative of sine of x is cosine of x.

Therefore, the integral of cosine of x is sine of x plus c.

And because the derivative of cosine of x is negative sine of x, we have to balance the sines.

The integral of positive sine of x must be negative cosine of x plus c.

Yeah, you have to be incredibly careful with those negative sines.

Students mix them up constantly when their brain has to switch rapidly between taking derivatives and taking integrals on an exam.

It happens all the time.

Furthermore, if you recall your classic trig identities, the derivative of tangent of x is second square root of x.

Thus, the integral of second square root of x is tangent of x plus c.

The easiest one to remember is always the exponential function.

The derivative of e to the x is just e to the x.

It's immune to differentiation.

So working backward, the integral of e to the x is also just e to the x plus c.

If there's a constant multiplier tucked into the exponent like e to the kx, you simply divide by that constant.

The integral is 1 over k times e to the kx plus c.

The overarching conceptual insight here, the thing you absolutely cannot mix up in your mind, is the profound difference between definite and indefinite integrals.

They use the same elongated s symbol, but they're fundamentally different creatures.

Yes.

If you take away nothing else from this deep dive, take this.

Let's clarify it.

A definite integral, the one with the numbers a and b sitting at the bottom and top of the integral sign, is a calculation of physical geometry.

It evaluates to a specific singular number, an area.

An indefinite integral, the one with no limits, is a calculation of algebra.

It evaluates to a family of functions, characterized by that plus c.

One outputs a number, the other outputs an equation.

But wait.

If they're fundamentally different creatures, one geometrically calculating net area using infinite limits of crushed rectangles, and the other algebraically undoing a derivative, why on earth do they share the exact same integral symbol?

Why did mathematicians lump them together?

That brilliant question is answered in the most important theorem in all of mathematics.

We arise at section 5 .4, the fundamental theorem of calculus, part one.

This is the big reveal.

This is the ultimate crossover event.

We spent the first half of this deep dive realizing that calculating the limits of Riemann sums to find an exact numerical area is exhausting, tear -inducing algebra.

We desperately needed a shortcut.

And this theorem is the shortcut.

It is breathtaking in its elegance.

The fundamental theorem of calculus, part one, states this.

If a function f is continuous on a solid interval from a to b, and if f is any antiderivative of f, then the definite integral of f of x from a to b is simply equal to f of b minus f of a.

Wait.

Pause.

Let me make sure I'm understanding the scale of what you're saying.

You are telling me that to find the exact area under a curve, a process that normally requires me to set up an infinite limit of n shrinking rectangles, expand massive algebraic polynomials,

use power -sum formulas, and calculate a limit to infinity.

Yes.

You are telling me that I can skip all of that geometry and all of that limit algebra completely.

All I have to do is use my reverse derivative rules to find the antiderivative formula, plug the top boundary number b into that formula, plug the bottom boundary number a into that formula, and just subtract the two numbers.

That's it.

That is it.

That is the magic of calculus.

That feels entirely like magic.

It feels like we are cheating the universe.

How can algebraic derivatives possibly shortcut geometric area?

It's not magic.

It's rigorous logic.

What's truly fascinating here is why this works.

Let's walk through the conceptual proof provided in the text.

It intimately connects everything we've built so far.

Walk us through it.

I need to see the gears turning on this one.

Imagine you have the total change of your antiderivative over the entire interval.

That total overall change is expressed mathematically as f of b minus f of a.

Imagine breaking up that one giant leap into a series of tiny incremental steps along the x -axis.

We chop the interval up into points, x sub 0, x sub 1, x sub 2, all the way to x sub n.

Instead of one big change, it's the change from step 1 to step 2, plus the change from step 2 to step 3, and so on.

Exactly.

This creates what mathematicians call a telescoping sum.

If you write it out, almost all the intermediate terms cancel each other out.

A positive f of x sub 1 cancels a negative f of x sub 1, leaving just the outer ends, f of b minus f of a.

So we haven't changed the value.

We've just chopped the journey into microscopic steps.

Now, we zoom in on just one of those microscopic steps.

Let's look at a tiny subinterval.

A few chapters ago, you learned the mean value theorem.

Applied here, it tells us that the change in the function over a tiny interval, which is f of x sub i minus f of x sub i minus 1, is equal to the derivative of the function evaluated somewhere inside that interval multiplied by the width of the interval.

Mathematically, the change equals f prime of c sub i times delta x sub i.

Wait.

Stop.

f is defined as the antiderivative of our original function f, which means, by definition, the derivative of f f prime is exactly the same thing as our original function f.

Yes.

That is the lock tumbling into place.

We substitute f in for f prime.

Now, look at what we've built.

We said the total change, f of b minus f of a, was the sum of all these tiny steps.

And we just proved that each tiny step is equal to f of c sub i times delta x sub i.

So the total algebraic change, f of b minus f of a, is equal to the sum of f of c sub i times delta x sub i.

Oh my goodness.

The sum of f of c sub i times delta x sub i.

That is literally a Riemann sum.

The function of value f of c sub i is the height of a rectangle, and delta x sub i is the width of a rectangle.

The total algebraic change in the antiderivative is mathematically structurally identical to a Riemann sum approximation of geometric area.

And as we take the limit as those rectangle widths shrink to zero,

it perfectly flawlessly matches the true definite integral.

Integration and differentiation are intimately and extricably bound together.

They're inverse operations of each other.

That is just, it's a beautiful piece of intellectual architecture.

Finding the area under a curve and finding the slope of a curve are literally the same mechanical process running in opposite directions.

But the textbook doesn't stop there.

Because next we get the fundamental theorem of calculus, part two.

Part i showed us how to evaluate a definite integral, practically using an antiderivative shortcut.

Part two is the theoretical sledgehammer.

It formally establishes the inverse relationship between the derivative and the integral by creating a brand new concept called an area function.

Let's define this area function.

It's written as a of x equals the integral from a to x of f of t dt.

What is this notation actually doing?

Look closely at the limits of integration.

The bottom limit is a, which is a hard fixed starting number.

But the top limit is x, a variable.

This means the interval we are measuring over is not fixed.

As x increases, the interval gets wider and the accumulated area inside the interval grows.

As x decreases, the interval shrinks and the area shrinks.

The output of this function a of x is the exact amount of geometric area accumulated up to whatever point x you plug in.

I picture this visually like a rubber squeegee wiping across a soapy, dirty window.

I like that.

The squeegee starts on the left side of the window at the fixed point a.

And you pull it to the right.

The squeegee handle itself is located at point x.

As you keep pulling x to the right, the area of perfectly clean glass left behind the squeegee, the accumulated area, gets larger and larger.

And notice they change the variable inside the integral to a t, making it f of t dt.

That t t is just a dummy variable for the horizontal axis, holding the place of the curve so we don't confuse it with our moving boundary line x.

So we have this function a of x that represents accumulated area.

FTC part 2 asks a very simple, profound question.

What is the rate of change of that area?

In other words, if I take the derivative of the area function, what do I get?

Mathematically, what is the derivative with respect to x of the interval from a 2x of s of t dt?

And the theorem gives us the answer.

It's just f of x, the original function.

Think about the gravity of what that means.

If you define a new function by taking the integral of f, and then you turn around and take the derivative of that new function, you get exactly f back.

They completely and utterly cancel each other out.

Integration and differentiation destroy each other, just like addition and subtraction, or multiplication and division.

The squeegee analogy works perfectly to explain the y here, too.

Think about it.

The area function a of x is the total amount of clean space on the window.

How fast is that clean space growing at any exact second?

What is the rate of growth?

Well, the rate of growth depends entirely on how tall the soapy window is, exactly where the squeegee is located right now.

If the squeegee hits a really tall part of the window, you uncover a massive amount of area very fast.

If the window tapers down and the squeegee hits a short part, you only uncover a little area slowly.

So the rate of change of the area, the derivative a prime of x, is exactly equal to the height of the window, the function f of x at that precise spot.

That physical intuition is perfectly captured by the squeeze theorem proof they provide in the text.

Let's look at a very thin sliver of newly uncovered area between x and x plus h, where h is a tiny microscopic step forward.

The actual physical area of the sliver is a of x plus h minus a of x.

Now, if we divide that total area by the width of the sliver, h, we get the difference quotient.

a of x plus h minus a of x, all divided by h, area divided by width, gives us height.

Specifically, it gives us the average height of that sliver of curve.

And we know logically that this average height must be squeezed somewhere between the absolute minimum height the curve hits in that tiny interval and the absolute maximum height the curve hits in that interval.

Now, imagine taking the limit.

Let h shrink down to zero.

The sliver becomes infinitely thin.

As it shrinks, the minimum possible height and the maximum possible height squeeze together, converging onto the exact singular height of the function at that exact point, which is f of x.

Therefore, the derivative of the accumulated area must equal the height f of x.

Okay, so we have built all this incredible theoretical machinery.

We know how to approximate area with pixels.

We know how to find exact area using infinite limits.

And we know we can bypass those limits entirely by using anti -derivatives, thanks to the fundamental theorem.

So what does this all mean for the real world?

Why should an engineering or economics student care?

That brings us to net change as the integral of a rate.

The FTC isn't just about finding geometric area on a piece of graph paper.

It establishes a universal principle codified in the net change theorem.

The definite integral of a rate of change gives you the net change in the underlying quantity.

Let's bring back our car from the very beginning of the deep dive.

We know your velocity, v of t, is the rate of change of your physical position.

So if we take the definite integral of your velocity function from time a to time b, the FTC says we get the net change in your position.

In physics, this specific net change is called displacement.

And displacement is a tricky concept because velocity is a vector.

It has a direction.

If you drive 10 miles forward, realize you left your wallet at home, put the car in reverse, and drive 10 miles backward to the exact spot you started, your net displacement is zero.

The positive area of driving forward and the negative area of driving backward perfectly canceled each other out.

You haven't gotten anywhere relative to your start.

But what if I am trying to calculate how much gas I burned?

My gas tank doesn't care about displacement.

My car still burned fuel driving backward.

I need to know the total miles driven.

If I drove 10 miles out and 10 miles back, my odometer says I drove 20 miles.

How do we get the odometer reading with an integral if the positive and negative areas keep canceling each other?

To find total distance traveled, you have to mathematically force all the accumulated area to be positive, regardless of direction.

You do this by taking the definite integral of the absolute value of the velocity.

So the integral from A to B of the absolute value of V of PTT.

Ah, absolute value.

Yes.

The absolute value strips away the negative sign from the backward direction.

It turns driving backward into positive distance accumulated.

It forces the integral to act like an odometer instead of a GPS.

The textbook also applies this net change theorem to economics, which is a classic high -stakes use case.

They use the concept of marginal cost.

Let's say you run a massive factory making computer chips.

The marginal cost, represented by C prime of X, is the rate at which your total costs increase when you produce just one additional chip.

Because it's a rate of change, it's a derivative.

Exactly.

The cost of making the very first chip includes all the factory setup.

So it's high.

Yeah.

But the cost of making the 10 ,000th chip might be very low because of economies of scale.

The rate is constantly fluctuating.

So by the net change theorem, if you integrate that fluctuating marginal cost C prime of X over a specific interval, say, from 10 ,000 chips up to 15 ,000 chips, the result of that definite integral gives you the total net increase in cost to ramp up production.

Let's actually look at the numbers they provide to see how fast this works compared to Riemann sums.

They give a marginal cost formula.

C prime of X equals 300 X squared minus 4 ,000 X plus 40 ,000, where X is measured in thousands of chips.

To find the cost of increasing production from 10 to 15, we integrate that formula with a lower limit of 10 and an upper limit of 15.

Using the FTC part I shortcut, we don't draw rectangles.

We just find the antiderivative.

Yeah.

Applying the reverse power rule.

The integral of 300 X squared becomes 100 X cubed.

The integral of negative 4 ,000 X becomes negative 2 ,000 X squared.

And the integral of the constant 40 ,000 becomes 40 ,000 X.

So our complete antiderivative formula is 100 X cubed minus 2 ,000 X squared plus 40 ,000 X.

Now we just evaluate it.

We plug in the upper limit 15 and calculate that number.

Then we plug in the lower limit 10, calculate that number, subtract the bottom result from the top result, and the text shows the final value is 187 ,500.

So it will cost the company an additional $187 ,500 to bump production up by those 5 ,000 units.

The definite integral effortlessly aggregated 5 ,000 different, constantly shifting marginal costs into one solid, actionable financial figure in three lines of algebra.

It is an incredibly powerful tool.

But as we move into the final stretches of the chapter, we encounter a severe mechanical problem.

Up to this point, we've only integrated relatively simple polynomials and basic standalone trig functions.

We've essentially only reversed the basic power rule.

Right.

If I ask you to integrate x to the fifth, you know exactly what to do.

But what if you were looking at a nested composite function?

What if the problem is the integral of 2x times cosine of x squared dx?

The power rule alone fails us completely here.

We have an x squared trapped inside a cosine function, and it's being multiplied by a 2x on the outside.

In calculus I, we learned how to take the derivative of a nested mess like that using the chain rule.

We need a way to run the chain rule in reverse to integrate functions that were created by it.

And that is exactly what the substitution method provides.

It is universally known among students as u -substitution, and it is the direct structural inverse of the chain rule.

When you use the chain rule to take a derivative, you take the derivative of the outside function, leave the inside function completely alone, and then multiply the whole thing by the derivative of the inside function.

U -substitution is all about playing detective.

You have to locate that inner function and pack it away into a single clean variable areal to clean up the algebraic mess.

Let's walk through the textbook's example too carefully, step by step, because mastering this puzzle -solving process is non -negotiable for success in any future calculus course.

The problem is evaluating the indefinite integral of x times the quantity x squared plus 9 raised to the fifth power dx.

Okay, let's solve the puzzle.

Step one, choose the inside function.

What part of this equation is tucked physically inside another mathematical operation?

Here, the expression x squared plus 9 is trapped inside a fifth power parenthesis.

So we make a declaration.

We say, let u equal x squared plus 9.

We are creating a new parallel mathematical universe where u is the primary variable, substituting out the messy x stuff.

Step two, now that we have u, we need to find the differential, do.

To do this, you simply take the derivative of your new u equation.

The derivative of x squared plus 9 with respect to x is just 2x.

So we write, do equals 2x dx.

This differential equation is crucial.

It x's the translation dictionary between the u universe and the x universe, telling us how a tiny change in u relates to a tiny change in x.

Step three, now we look back at our original ugly integral.

We want to substitute everything out so there are absolutely no x's left, only u's.

We can easily replace the x squared plus 9 to the fifth with a very clean n to the fifth.

But what's left behind?

We have a lone x floating in the front and a dx at the end, so we have x dx left over.

Our translation dictionary says do is strictly equal to 2x dx.

We only have x dx in our integral, not 2x dx.

We have a mismatch.

What do we do?

This is a very common scenario.

The derivative of the inside function is off by a constant multiplier.

We simply use basic algebra to fix the dictionary.

We can multiply or divide our differential equation by a constant.

If we divide both sides of do equals 2x dx by 2, we get one half do equals x dx.

Now we have a perfect identical match to what is left in our integral.

We can replace the x dx entirely with one half do.

Step 4.

We perform the full substitution.

Our terrifying original integral transforms into a beautiful, clean, new integral in the universe.

The integral of u to the fifth times one half do.

Using our linearity rules, we can pull that constant one half all the way to the front of the integral sign.

So we are just integrating one half times the integral of u to the fifth do.

Step 5.

Now we integrate.

We are safely back to the basic power rule.

The integral of u to the fifth is just one sixth u to the sixth.

We multiply that by the one half we left out front and we get one twelfth u to the sixth plus c.

We have our antiderivative.

Step 6.

The final step.

We started the problem into x universe.

We have to deliver the final answer in the x universe.

We can't hand in an answer with a u in it.

So we take our original definition, u equals x squared plus 9, and plug it back in for you.

Our final correct answer is one twelfth times the quantity x squared plus 9 to the sixth plus c.

It is a supremely satisfying process when it clicks.

You are untying a mathematical knot.

However, there is a major common pitfall when you use u substitution with definite integrals, the ones that have boundary limits a and b attached to them.

Yes.

Warning sirens should be going off in your head right now.

If your original integral asks you to find the area from x equals one to x equals two, those boundary limits are strictly x values.

If you switch your entire equation over to the universe to make the integration easier, you cannot use one and two as your limits anymore.

Those numbers don't belong in the universe.

If you evaluate a u equation with x limits, you will get the wrong area.

To fix this, you must translate your limits just like you translated your variables.

You plug your original a and b x limits into your u equation to generate new u compatible limits.

For example, if your translation equation was u equals x squared plus 1, and your original lower limit was x equals one, you calculate your new lower limit.

One squared plus one is two.

If your upper limit was x equals two, your new upper limit is two squared plus one, which is five.

And the wildly beautiful part about doing this is that once you change your boundaries to u limits, you never have to substitute x back into the equation at the end.

You don't have to go back to the x universe.

You just evaluate your clean u antiderivative using your new u limits, and it will output the exact mathematically correct numerical area.

It saves you an entire messy algebraic step at the end.

U substitution isn't just for undoing the chain rule.

It actually unlocks entirely new foundational formulas for us.

For instance, until this exact moment in the text, we cannot integrate the tangent function.

We knew the integral of second squared zeta, but not plane tangents of the theta.

But with u sub, it becomes a simple puzzle.

We just use trig identities to rewrite tangent theta as sine theta over cosine theta.

Then we set our inner function to be the denominator, u equals cosine theta.

We find our differential do, which is negative sine theta d theta.

We happen to have a sine theta d theta sitting perfectly in the numerator, so we just move the negative sign over.

We substitute, and we're left integrating negative one over u.

The integral of one over u is the natural log.

So we find that the integral of tangent theta d theta is negative natural log of the absolute value of cosine theta plus c, which can also be written algebraically as natural log of the absolute value of secant theta plus c.

We just unpacked a whole new tool for our toolbox.

Which is the perfect segue to our final topic.

We have built a massive arsenal of integration tools today, but there are a few lingering inverse trigonometric functions we need to cover to complete the standard calculus toolkit.

Right.

Just like we derived newer integrals from known derivatives earlier, we can look at the bizarre derivatives of inverse trig functions like arcsine, arc tangent, and arc secant, and run them in reverse.

These formulas pop up constantly in advanced geometry and physics problems, and they often look incredibly intimidating.

For example, if you encounter an integral that looks like a fraction with the square root in the denominator,

the integral of one over the square root of one minus x squared dx,

that doesn't look like trigonometry at all.

But that exact expression is the established derivative of the inverse sine function.

So by definition, its antiderivative is simply inverse sine of x plus c.

And if you see an integral without a square root, the integral of one over x squared plus one dx, that is the signature of the inverse tangent.

The answer is just inverse tangent of x plus c.

And if you see a truly ugly one, the integral of one over the absolute value of x times the square root of x squared minus one dx that evaluates to the inverse secant of x plus c.

At this level of calculus, half the battle isn't doing the math.

It's pattern recognition.

It's looking at the structure of the integrand and knowing exactly which tool to pull out of your toolbox.

Exactly.

Sometimes it won't perfectly match the formula right away.

You might have to use a clever u -substitution to manipulate an ugly fraction until it perfectly matches that clean one over x squared plus one form.

And the moment it matches, you hit it with the Arkanjian formula and solve the puzzle.

And what an incredible sprawling puzzle we've put together today.

Let's synthesize this journey we've been on.

We started an hour ago by literally drawing blocky digital rectangles under a curve just to estimate an area.

We saw how tedious the sums became, so we learned sigma notation to manage the algebra.

Then we made a massive philosophical leap, pushing the number of those rectangles to the limit of infinity to create the definite integral calculating the exact net signed area.

From there, we uncovered the fundamental theorem of calculus, which revealed the beautiful hidden symmetry of the mathematical universe.

That integration and differentiation, finding physical areas and finding instantaneous slopes, are actually two sides of the exact same coin.

They're inverses.

We learned how to find the net change of physical and economic systems.

And we learned how to untangle complex composite functions by running the chain rule backward using u -substitution.

It is a towering achievement of human logic.

You now have the foundational tools to measure accumulation in a constantly changing universe.

But before we wrap up, I have one final provocative thought to leave you with.

Something to chew on as you study this material.

Oh, I love these.

Give it to us.

We spent this entire deep dive talking about finding the area under a curve between two hard fixed boundaries.

An interval from a equals one, b equals three.

You plug those specific numbers in, you do the math, you get a finite area.

But what if the top boundary, b, never stops?

What if you try to find the area under a curve starting at x equals one and going all the way to infinity?

Is it even conceptually possible for a shape that is infinitely long, a shape that never closes, to have a finite, perfectly measurable mathematical area?

Oh, that is a fantastic tease.

My brain instantly says no.

An infinitely long shape must have infinite area.

But everything we've learned today about limits says our intuition is probably wrong.

That is a massive concept waiting for you in calculus, Sess.

Improper integrals.

You'll have to keep pushing your math journey forward to find the answer to that one.

Thanks for tuning into this session.

Remember, whenever calculus feels completely overwhelming, just break it down into tiny, manageable pieces, just like a rhyme and sum.

From all of us at the Last Minute Lecture team, keep questioning, keep learning, and we'll catch you on the next deep dive.