Existence, Uniqueness & Slope Fields

Two questions to ask before you solve anything

The earlier guides in this rung taught you to solve the solvable: separable equations, the integrating factor for linear ones, and a few special tricks. But most first-order equations dy/dx = f(x, y) you will ever write down have no formula for their solution at all — the integral that would solve them is non-elementary, or the equation is genuinely nonlinear and no method touches it. So before you reach for a technique, two more basic questions deserve an honest answer. Given a starting point — an initial value problem, the equation dy/dx = f(x, y) together with a condition y(x0) = y0 — does a solution curve through that point even exist? And if it does, is it the only one, or could two different curves both pass through the same point obeying the same law? See initial value problem.

These are not pedantic worries. Uniqueness is what makes a differential equation a usable law of nature: if you know the state of a system now, a unique future is what lets you predict it. Existence is what tells you the question is even well-posed before you waste an afternoon hunting for a formula. And the surprising fact — the one this guide exists to make vivid — is that you can settle both questions, and even draw the entire family of solution curves, without ever finding a single solution formula. The equation itself, read the right way, hands you the picture.

The slope field: solving with your eyes

Here is the central trick, and it is almost embarrassingly simple. The equation dy/dx = f(x, y) is a recipe that, at every single point (x, y) of the plane, tells you the slope a solution curve must have if it passes through that point. You do not need to solve anything to read this off — you just plug the point in. So go to a point, compute f(x, y), and draw a tiny line segment with that slope right there. Do this at a whole grid of points and you get a direction field (also called a slope field): a meadow of little tilted dashes covering the plane, each one a local instruction that says "a solution coming through here must head off in this direction." See direction field.

Now the magic. A solution curve is nothing more than a curve that, at every point it visits, is tangent to the little dash sitting there — because its slope dy/dx is exactly f(x, y) by the very definition of solving the equation. This is just the Volume I idea that a derivative is the slope of the tangent line, read in reverse: instead of computing slopes from a known curve, you are growing a curve to match prescribed slopes. So you can literally trace a solution by hand. Drop your pencil at any starting point, look at the dash under it, slide a tiny step in that direction, look at the new dash, slide again — and the smooth curve you sweep out IS the solution through that point. Different starting points give different curves, and together they fill the plane with the whole family of solutions.

Reading a real field: dy/dx = y

Let us make a concrete one in words. Take dy/dx = y, the equation of exponential growth. The slope at any point is just the height y itself: on the x-axis (y = 0) every dash is flat, well above the axis the dashes tilt up steeply, well below they tilt down steeply, and the whole field is symmetric to left and right because x never enters the formula. Trace a curve starting just above the axis and it creeps up gently, then steepens and rockets skyward — the dashes get steeper as it climbs, feeding its own growth. Start just below and it plunges to minus infinity in a mirror image. We happen to know the answer here is y = C e^{x}, but notice you can see the entire shape of that solution family in the field before writing the formula down.

The field also reveals structure that a formula can hide. The flat line of dashes along y = 0 is itself a solution — the constant zero solution — and no other curve ever crosses it, because crossing would force two different slopes at the meeting point. That dividing line is an equilibrium, and it splits the plane into an upper world of curves blasting up and a lower world plunging down. When f depends only on y like this, the equation is called an autonomous equation, and its field looks the same in every vertical strip; that special structure is exactly what the next guide on the phase line exploits. For now the lesson is plain: the slope field shows you equilibria, symmetries, and the long-run fate of every solution — all read straight off f without a single integration.

Picard-Lindelof: when the picture is trustworthy

Tracing a solution by following dashes feels foolproof — but it quietly assumes that exactly one curve emerges from your starting point. The Picard-Lindelof theorem (also called the Picard existence-uniqueness theorem) states precisely when that assumption is earned. The clean version: if f(x, y) is continuous in a box around your starting point (x0, y0), AND f does not change too violently in the y-direction — technically it is Lipschitz in y, which is guaranteed whenever the partial derivative partial f / partial y exists and stays bounded near the point — then the initial value problem has exactly one solution, living on at least some small interval around x0. Existence and uniqueness, both delivered, with no formula in sight. See Picard-Lindelof theorem.

Why should a bound on partial f / partial y buy you uniqueness? The honest answer is a beautiful self-improving loop hiding inside the theorem. Rewrite the problem as an integral: y(x) = y0 + the integral from x0 to x of f(t, y(t)) dt — solving the equation means being a fixed point of that integral. Now start with a crude guess for y, feed it into the right-hand side to get a better guess, feed that back in, and repeat. The Lipschitz condition is exactly the gear that makes each pass shrink the error, so the guesses close in on one and only one limiting curve. The proof is a machine you could run, and it is the same fixed-point idea that powers numerical solvers.

Picard iteration for  dy/dx = y,  y(0) = 1   (so f = y):

  y_0(x) = 1
  y_1(x) = 1 + integral_0^x y_0 dt = 1 + x
  y_2(x) = 1 + integral_0^x (1 + t) dt = 1 + x + x^2/2
  y_3(x) = 1 + x + x^2/2 + x^3/6
  ...
  y_n(x) -> 1 + x + x^2/2 + x^3/6 + ... = e^x   (the unique solution)

The iterates are exactly the partial sums of the Taylor series of e^x.

The honest fine print

The theorem is generous but it is not a blank check, and three pieces of fine print keep you honest. The first and most important: the guarantee is LOCAL. It promises a unique solution on some small interval around x0, and says nothing about how far that interval reaches. A solution can blow up to infinity in finite time and simply cease to exist beyond there — dy/dx = y^2 with y(0) = 1 has the perfectly nice solution y = 1/(1 - x), which races to infinity as x approaches 1 and has no solution at all past x = 1, even though f = y^2 is as smooth as can be. The smoothness of f does not buy you a solution that lasts forever; it only buys you one that starts.

The second piece: when the Lipschitz hypothesis genuinely fails, uniqueness can genuinely fail too — this is not a technicality you can wave away. The classic example is dy/dx = square root of |y| with y(0) = 0. Here f is continuous, so a solution exists, but partial f / partial y blows up at y = 0, so the Lipschitz condition is broken right at the start. And uniqueness shatters: the constant y = 0 is a solution, and so is y = x^2/4, and so is a whole infinite family that sits on the axis for a while and then peels off at any moment you like. Infinitely many solution curves leave the same point. The slope field cannot show you this — at y = 0 every dash is flat and looks innocent — which is the whole reason you need the theorem and not just the picture.

Putting the two ideas together

The slope field and the existence-uniqueness theorem are two halves of one understanding, and they certify each other. The theorem is what guarantees that the curves you trace through the field are real, single, and well-defined: where the Lipschitz condition holds, exactly one solution threads through each point, so the dashes assemble into a clean family of non-crossing curves with no ambiguity about which way to go. And the field is what makes the theorem visible: solution curves that never cross is the geometric face of uniqueness, while a point where the field smears and curves fan out is the geometric face of a failed hypothesis. One gives you rigor, the other gives you sight.

1. Write the equation as dy/dx = f(x, y) and read f as a slope-prescription: at every point of the plane it dictates the slope a solution must have there.
2. Check the hypotheses near your starting point: is f continuous there, and is partial f / partial y bounded? If yes, Picard-Lindelof gives you one solution on a small interval — trust the picture locally.
3. Sketch the field via its isoclines (start with f = 0), mark equilibria, then trace solution curves tangent to the dashes from the starting points you care about — that is the family, drawn without a formula.
4. Stay honest about the limits: the guarantee is local (a solution can blow up in finite time), and where the hypotheses fail (look for partial f / partial y exploding), watch for non-uniqueness or a hidden singular solution.

Carry this forward as the foundation of everything in this rung and the next. Most differential equations cannot be solved in closed form, yet you are far from helpless: the slope field shows you the shape, the long-run behavior, and the equilibria of a solution you can never write down, while Picard-Lindelof tells you, honestly and with its limits stated, when that picture deserves your trust. The next guide takes the autonomous case and squeezes the entire two-dimensional field down to a single line — the phase line — to read the fate of every solution at a glance.