Subtitles section Play video Print subtitles OK, let's get started. Now... I'm assuming that, A, you went recitation yesterday, B, that even if you didn't, you know how to separate variables, and you know how to construct simple models, solve physical problems with differential equations, and possibly even solve them. So, you should have learned that either in high school, or 18.01 here, or... yeah. So, I'm going to start from that point, assume you know that. I'm not going to tell you what differential equations are, or what modeling is. If you still are uncertain about those things, the book has a very long and good explanation of it. Just read that stuff. So, we are talking about first order ODEs. ODE: I'll only use three ... two acronyms. ODE is ordinary differential equations. I think all of MIT knows that, whether they've been taking the course or not. So, we are talking about first-order ODEs. which in standard form, are written, you isolate the derivative of y with respect to, x, let's say, on the left-hand side, and on the right-hand side you write everything else. You can't always do this very well, but for today, I'm going to assume that it has been done and it's doable. So, for example, some of the ones that will be considered either today or in the problem set are things like oh... y' = x / y That's pretty simple. The problem set has y' = ...let's see... x - y^2. And, it also has y' = y - x^2. There are others, too. Now, when you look at this, this, of course, you can solve by separating variables. So, this is solvable. This one is-- and neither of these can you separate variables. And they look extremely similar. But they are extremely dissimilar. The most dissimilar about them is that this one is easily solvable. And you will learn, if you don't know already, next time next Friday how to solve this one This one, which looks almost the same, is unsolvable in a certain sense. Namely, there are no elementary functions which you can write down, which will give a solution of that differential equation. So, right away, one confronts the most significant fact that even for the simplest possible differential equations, those which only involve the first derivative, it's possible to write down extremely looking simple guys. I'll put this one up in blue to indicate that it's bad. Whoops, sorry, I mean, not really bad, but recalcitrant. It's not solvable in the ordinary sense in which you think of an equation is solvable. And, since those equations are the rule rather than the exception, I'm going about this first day to not solving a single differential equation, but indicating to you what you do when you meet a blue equation like that. What do you do with it? So, this first day is going to be devoted to geometric ways of looking at differential equations and numerical. At the very end, I'll talk a little bit about numerical ways. And you'll work on both of those for the first problem set. So, what's our geometric view of differential equations? Well, it's something that's contrasted with the usual procedures, by which you solve things and find elementary functions which solve them. I'll call that the analytic method. So, on the one hand, we have the analytic ideas, in which you write down explicitly the equation, y' = f(x,y). And, you look for certain functions, which are called its solutions. Now, so there's the ODE. And, y1 of x, notice I don't use a separate letter. I don't use g or h or something like that for the solution because the letters multiply so quickly, that is, multiply in the sense of rabbits, that after a while, if you keep using different letters for each new idea, you can't figure out what you're talking about. So, I'll use y1 means, it's a solution of this differential equation. Of course, the differential equation has many solutions containing an arbitrary constant. So, we'll call this the solution. Now, the geometric view, the geometric guy that corresponds to this version of writing the equation, is something called a direction field. And, the solution is, from the geometric point of view, something called an integral curve. So, let me explain if you don't know what the direction field is. I know for some of you, I'm reviewing what you learned in high school. Those of you who had the BC syllabus in high school should know these things. But, it never hurts to get a little more practice. And, in any event, I think the computer stuff that you will be doing on the problem set, a certain amount of it should be novel to you. It was novel to me, so why not to you? So, what's a direction field? Well, the direction field is, you take the plane, and in each point of the plane-- of course, that's an impossibility. But, you pick some points of the plane. You draw what's called a little line element. So, there is a point. It's a little line, and the only thing which distinguishes it outside of its position in the plane, so here's the point, (x,y), at which we are drawing this line element, is its slope. And, what is its slope? Its slope is to be f(x,y). And now, You fill up the plane with these things until you're tired of putting then in. So, I'm going to get tired pretty quickly. So, I don't know, let's not make them all go the same way. That sort of seems cheating. How about here? Here's a few randomly chosen line elements that I put in, and I putted the slopes at random since I didn't have any particular differential equation in mind. Now, the integral curve, so those are the line elements. The integral curve is a curve, which goes through the plane, and at every point is tangent to the line element there. So, this is the integral curve. Hey, wait a minute, I thought tangents were the line element there didn't even touch it. Well, I can't fill up the plane with line elements. Here, at this point, there was a line element, which I didn't bother drawing in. And, it was tangent to that. Same thing over here: if I drew the line element here, I would find that the curve had exactly the right slope there. So, the point is the integral, what distinguishes the integral curve is that everywhere it has the direction, that's the way I'll indicate that it's tangent, has the direction of the field everywhere at all points on the curve, of course, where it doesn't go. It doesn't have any mission to fulfill. Now, I say that this integral curve is the graph of the solution to the differential equation. In other words, writing down analytically the differential equation is the same geometrically as drawing this direction field, and solving analytically for a solution of the differential equation is the same thing as geometrically drawing an integral curve. So, what am I saying? I say that an integral curve, all right, let me write it this way. I'll make a little theorem out of it, that y1(x) is a solution to the differential equation if, and only if, the graph, the curve associated with this, the graph of y1 of x is an integral curve. Integral curve of what? Well, of the direction field associated with that equation. But there isn't quite enough room to write that on the board. But, you could put it in your notes, if you take notes. So, this is the relation between the two, the integral curves of the graphs or solutions. Now, why is that so? Well, in fact, all I have to do to prove this, if you can call it a proof at all, is simply to translate what each side really means. What does it really mean to say that a given function is a solution to the differential equation? Well, it means that if you plug it into the differential equation, it satisfies it. Okay, what is that? So, how do I plug it into the differential equation and check that it satisfies it? Well, doing it in the abstract, I first calculate its derivative. And then, how will it look after I plugged it into the differential equation? Well, I don't do anything to the x, but wherever I see y, I plug in this particular function. So, in notation, that would be written this way. So, for this to be a solution means this, that that equation is satisfied. Okay, what does it mean for the graph to be an integral curve? Well, it means that at each point, the slope of this curve, it means that the slope of y1 of x should be, at each point, (x1,y1). It should be equal to the slope of the direction field at that point. And then, what is the slope of the direction field at that point? Well, it is f of that particular, well, at the point, (x,y1). If you like, you can put a subscript, one, on there, send a one here or a zero there, to indicate that you mean a particular point. But, it looks better if you don't. But, there's some possibility of confusion. I admit to that. So, the slope of the direction field, what is that slope? Well, by the way, I calculated the direction field. Its slope at the point was to be x, whatever the value of x was, and whatever the value of y1(x) was, substituted into the right-hand side of the equation. So, what the slope of this function of that curve of the graph should be equal to the slope of the direction field. Now, what does this say? Well, what's the slope of y1(x)? That's y1'(x). That's from the first day of 18.01, calculus. What's the slope of the direction field? This? Well, it's this. And, that's with the right hand side. So, saying these two guys are the same or equal, is exactly, analytically, the same as saying these two guys are equal. So, in other words, the proof consists of, what does this really mean? What does this really mean? And after you see what both really mean, you say, yeah, they're the same. So, I don't how to write that. It's okay: same, same, how's that? This is the same as that. Okay, well, this leaves us the interesting question of how do you draw a direction from the, well, this being 2003, mostly computers draw them for you. Nonetheless, you do have to know a certain amount. I've given you a couple of exercises where you have to draw the direction field yourself. This is so you get a feeling for it, and also because humans don't draw direction fields the same way computers do. So, let's first of all, how did computers do it? They are very stupid. There's no problem. Since they go very fast and have unlimited amounts of energy to waste, the computer method is the naive one. You pick the point. You pick a point, and generally, they are usually equally spaced. You determine some spacing, that one: blah, blah, blah, blah, blah, blah, blah, equally spaced. And, at each point, it computes f(x, y) at the point, finds, meets, and computes the value of f of (x, y), that function, and the next thing is, on the screen, it draws, at (x, y), the little line element having slope f(x, y). In other words, it does what the differential equation tells it to do. And the only thing that it does is you can, if you are telling the thing to draw the direction field, about the only option you have is telling what the spacing should be, and sometimes people don't like to see a whole line. They only like to see a little bit of a half line. And, you can sometimes tell, according to the program, tell the computer how long you want that line to be, if you want it teeny or a little bigger. Once in awhile you want you want it narrower on it, but not right now. Okay, that's what a computer does. What does a human do? This is what it means to be human. You use your intelligence. From a human point of view, this stuff has been done in the wrong order. And the reason it's been done in the wrong order: because for each new point, it requires a recalculation of f(x, y). That is horrible. The computer doesn't mind, but a human does. So, for a human, the way to do it is not to begin by picking the point, but to begin by picking the slope that you would like to see. So, you begin by taking the slope. Let's call it the value of the slope, C. So, you pick a number. C is two. I want to see where are all the points in the plane where the slope of that line element would be two? Well, they will satisfy an equation. The equation is f(x,y) = C, in general. So, what you do is plot this, plot the equation, plot this equation. Notice, it's not the differential equation. You can't exactly plot a differential equation. It's a curve, an ordinary curve. But which curve will depend; it's, in fact, from the 18.02 point of view, the level curve of C, sorry, it's a level curve of f of (x, y), the function f of x and y corresponding to the level of value C But we are not going to call it that because this is not 18.02. Instead, we're going to