4.2 Direction Fields and Numerical Methods

Direction Fields and First-Order Differential Equations

Direction fields give you a visual way to understand first-order differential equations. Instead of solving the equation algebraically, you draw short slope lines at grid points to see how solutions behave. This lets you spot patterns of growth, decay, or oscillation at a glance.

Solution curves can then be sketched by following those slope lines from a starting point. Each curve represents a specific solution for a given initial condition, and because of the uniqueness theorem, these curves never cross each other.

Direction Fields for Differential Equations

A direction field is a graphical tool for a first-order differential equation of the form $\frac{dy}{dx} = f(x, y)$. You build one by picking a grid of $(x, y)$ points, plugging each into the equation to get the slope, and drawing a short line segment with that slope at each point.

The result is a field of tiny arrows or dashes that show the "flow" of solutions across the plane. Even without finding an explicit formula, you can read off important behavior: Where are solutions increasing? Decreasing? Leveling off toward an equilibrium?

To construct a direction field:

1. Write the equation in the form $\frac{dy}{dx} = f(x, y)$.
2. Choose a set of grid points $(x, y)$ across the region you care about.
3. At each point, evaluate $f(x, y)$ to get the slope.
4. Draw a short line segment with that slope at the point.

For example, given $\frac{dy}{dx} = x - y$, at the point $(2, 1)$ the slope is $2 - 1 = 1$, so you'd draw a small segment tilted at 45°. At $(0, 3)$ the slope is $0 - 3 = -3$, so the segment is steeply downward.

Solution Curves from Direction Fields

A solution curve (also called an integral curve) is a specific solution to the differential equation that passes through a chosen initial point $(x_0, y_0)$. You sketch one by starting at that point and tracing a path that stays tangent to the nearby line segments, extending in both directions.

Different initial conditions produce different solution curves on the same direction field. This is exactly the setup of an initial value problem: a differential equation plus a starting condition.

One important constraint: solution curves cannot cross each other. The existence and uniqueness theorem guarantees that through any point where $f(x, y)$ is well-behaved, there is exactly one solution curve. If two curves crossed at a point, that point would belong to two different solutions, violating uniqueness.

Numerical Methods for Solving First-Order Differential Equations

Euler's Method for Approximating Solutions

Many differential equations can't be solved with a neat formula. Euler's method is the simplest numerical technique for approximating a solution: you step forward in small increments, using the slope at your current point to estimate the next point.

Here's the procedure:

1. Start with the initial condition $(x_0, y_0)$ and choose a step size $h$.

2. Compute the slope at the current point: $m = f(x_0, y_0)$.

3. Step forward to the next point using:

   • $x_1 = x_0 + h$
   • $y_1 = y_0 + h \cdot f(x_0, y_0)$

4. Repeat, using each new point as the starting point for the next step, until you reach the desired endpoint.

The general formula at step $n$ is:

$y_{n+1} = y_n + h \cdot f(x_n, y_n)$

Concrete example: Suppose $\frac{dy}{dx} = y$, with $y(0) = 1$ and step size $h = 0.5$. The exact solution is $y = e^x$.

• Step 1: At $(0, 1)$, slope $= 1$. So $y_1 = 1 + 0.5(1) = 1.5$.
• Step 2: At $(0.5, 1.5)$, slope $= 1.5$. So $y_2 = 1.5 + 0.5(1.5) = 2.25$.
• The exact value at $x = 1$ is $e^1 \approx 2.718$, so our estimate of 2.25 undershoots. A smaller step size would get closer.

Step size matters a lot. Smaller $h$ gives more accurate results but requires more iterations. Larger $h$ is faster but drifts further from the true solution. Euler's method is a first-order method, meaning the error at each step is roughly proportional to $h$. This error accumulates over many steps, so the global error over a fixed interval is proportional to $h$ as well.

Beyond Euler's Method

Euler's method is a starting point, but its accuracy is limited. A few ideas worth knowing at this level:

• Truncation error is the error introduced at each step because you're approximating a smooth curve with straight-line segments. Smaller steps reduce this but never eliminate it entirely.
• Predictor-corrector methods improve on Euler's approach by first estimating the next point (the "predictor" step), then refining that estimate using slope information at the predicted point (the "corrector" step). The improved Euler method (also called Heun's method) is a common example.
• Runge-Kutta methods (especially the classical fourth-order version, RK4) sample the slope at multiple points within each step to achieve much better accuracy without needing an extremely small $h$.
• Stability becomes a concern over long intervals. Some differential equations cause certain methods to produce wildly growing errors even with small step sizes. Choosing a method and step size that remain stable for your particular equation is important in practice.

For Calc II, you'll mostly work with Euler's method directly. The key takeaway is that numerical methods trade exact answers for flexibility: they can approximate solutions to equations that have no closed-form solution at all.