How to Solve Differential Equations: First Order Methods

Imagine you're hiking down a mountain in thick fog. You can't see the peak. You can't see the valley. You can't even see ten feet ahead.

All you have is a little device strapped to your boot that tells you exactly one thing: how steep the ground is, right where you're standing, right now.

Could you still find your way down?

That's a differential equation. Not a map of where you are, but a rule for how steep things are at every possible point. And "solving" it just means finding the path that actually matches that rule, everywhere, all at once.

What is a differential equation, actually?

Most equations you've solved before ask "what number makes this true?" A differential equation asks something stranger: "what function makes this true, given a rule about its own rate of change?"

Here's the simplest version. Say a population grows at a rate proportional to how big it already is. Bigger population, faster growth. We'd write that as:

That's not telling you $y$. It's telling you the slope of $y$, at every single point, in terms of $y$ itself. The actual curve, the function $y(x)$, is hiding somewhere behind that rule.

Your job: find it. And unlike a normal equation, there isn't one number to find, there's an entire curve that has to satisfy the rule everywhere along its length.

The slope field: a map of directions, not points

Forget solving anything for a second. What if you just drew the slope at a bunch of points, using little tick marks? At every point $(x, y)$, you compute the slope the equation demands, and draw a tiny line segment tilted that way.

Do this at enough points and a pattern emerges, like iron filings around a magnet. This is called a slope field, and here's the payoff: a solution curve is just a path that's tangent to every tick mark it passes through. Nothing more mysterious than that.

Drag that dot anywhere on the field. Watch the curve snap into place, hugging the tiny slope-ticks the whole way. Different starting points give different curves, but they're all "correct", because they all follow the same underlying rule. That's what a solution to a differential equation actually is: a path that never disagrees with the field it's sitting in.

And notice something else: however you drag the point, the curve eventually settles onto the same dashed line. The starting point stops mattering. That's a real feature of this particular equation, not a coincidence, and it's the kind of thing a slope field shows you instantly that a formula would hide.

When the slope only depends on you: separable equations

A full slope field is great for building intuition, but walking it point by point isn't exactly solving anything. So when can we get an exact formula?

Look back at $dy/dx = ky$. Notice the right side only involves $y$, not $x$. That's a huge clue. It means we can gather all the $y$ stuff on one side and all the $x$ stuff on the other:

That looks like algebra, but it's really integration in disguise. We write it as $dy/y$ because we're about to integrate both sides independently, as if they were two separate area-under-a-curve problems:

Every step just used integration rules you already know. The only new idea was noticing the equation could be split apart in the first place. That's why this whole family is called separable equations.

Slide $k$ and watch the curve tilt from decay to growth. Notice the doubling time (or halving time) shrinks fast as $|k|$ grows, that's the whole "compound interest" and "radioactive decay" story, both riding on the exact same equation with a different sign on $k$.

When you can't solve it by algebra: walk the field instead

Separable equations are lucky cases. Most differential equations that show up in the real world, weather models, orbital mechanics, chemical reactions, don't split apart so nicely. There's no clean formula waiting to be found.

So what do we do? We go back to the fog.

If all you can do is feel the slope where you're standing, here's the obvious plan: take one small step in that direction, then check the slope again from your new spot, then take another small step. Repeat. You'll never land exactly on the true path, but you'll stay close to it, as long as your steps are small.

This is called Euler's method, and it's really nothing more than that mountain hike, formalized:

Here $h$ is your step size, and $f(x_n, y_n)$ is just "the slope where you currently are."

Hit play. Watch the orange polygon get built one step at a time, always turning to match the local slope before advancing. Now shrink $h$ with the slider. The polygon hugs the true curve tighter, and the final error shrinks along with it. You've just built a numerical solver out of nothing but arithmetic.

Why it's called a "first order" method

Here's a question worth sitting with: if smaller steps mean less error, how much less?

You might guess that halving the step size halves the error. That guess is exactly right for Euler's method, and that's not a coincidence, it's the definition of the term first order. A numerical method is "order $p$" if its error shrinks proportionally to $h^p$. Euler's method is order 1, so error scales with $h^1$: halve $h$, halve the error.

Compare that to a slightly smarter method (called Heun's method) that averages the slope at the start of a step and the slope at the estimated end of a step before committing to a direction. It costs a little more arithmetic per step, but its error scales with $h^2$: halve $h$, and the error drops to a quarter.

Slide the step size down and watch both error curves fall, but at different rates. Euler's orange line runs parallel to the dashed slope-1 reference. Heun's teal line runs parallel to the dashed slope-2 reference. That slope, on a log-log plot of error versus step size, is exactly what "order" means. It's not a name someone picked for flavor, it's a literal exponent you can read straight off a graph.

What about equations that don't separate?

One quick note before we wrap up. Some equations aren't separable but are still solvable by algebra, specifically the linear ones, which look like:

The trick here is multiplying both sides by a cleverly chosen function $\mu(x) = e^{\int p(x)\,dx}$, called an integrating factor. It's built so that the left side collapses into the derivative of a single product, $\frac{d}{dx}[\mu(x) y]$, which you can then just integrate directly. It's the same spirit as separation, find some transformation that turns a differential equation into an ordinary integral, just aimed at a slightly wider class of equations.

The short version

A first order differential equation is a rule for slopes, not a formula for points. A slope field shows you every one of those rules laid out at once, and a solution curve is simply a path that never fights the field it's in.

When the slope only depends on the output variable, you can separate it and integrate both sides directly. When it depends on both variables in a linear way, an integrating factor can still crack it open. And when neither trick applies, you fall back on numerical stepping, Euler's method being the simplest version, where "first order" isn't just jargon, it's a literal description of how fast the error shrinks as you take smaller steps.

Every method here, algebraic or numerical, is doing the same thing at heart: standing somewhere on the mountain, reading the slope, and finding a way forward that respects it.