First-Order ODEs and Direction Fields

Direction fields and the first-order framework are the entry point to modeling in nearly every quantitative discipline. In physics, Newton's second law for a particle moving in one dimension under a velocity-dependent force, $mv' = F(t,v)$, is a first-order ODE in the velocity $v$, and its direction field reveals terminal velocity as the isocline where net force vanishes. In population biology and epidemiology, the logistic and SIR-type models are first analyzed through their slope fields to identify carrying capacities, thresholds, and long-term equilibria before any closed-form solution is sought. In chemistry, first-order reaction kinetics $y' = -ky$ is literally named after this equation type, and direction-field intuition explains exponential decay visually as a field of ever-shallowing negative slopes. In economics and finance, continuously compounded interest and simple growth models are first-order ODEs, and phase-line reasoning (a direction field restricted to one dimension) underlies stability analysis of equilibrium prices. In engineering, RC and RL electrical circuits obey first-order linear ODEs for voltage or current, and control theorists use direction fields to design feedback so that all trajectories are pulled toward a desired equilibrium. In numerical computing, Euler's method and its refinements (Runge–Kutta methods) are literally algorithms for walking along the direction field with small time steps, forming the computational backbone of simulation software across science and engineering.

Ask the tutor: How does the direction field of $y' = f(t,y)$ change when $f$ is replaced by a Lipschitz approximation, and why does that restore uniqueness of solutions?

Think of $f(t,y)$ as a rule that, at every location $(t,y)$ in the plane — think of $t$ as time and $y$ as the state of a system — tells you the instantaneous slope a solution must have if it happens to pass through that point. The differential equation itself never mentions a specific function; it is a law that every solution obeys, checked pointwise. This is exactly analogous to a vector field describing wind speed and direction at every point in a room: an individual dust particle (a solution curve) does not know the whole field, it only feels the local wind and moves accordingly, moment to moment. The direction field converts this abstract law into a picture. At each grid point $(t,y)$ you draw a tiny segment of slope $f(t,y)$. Doing this at enough points reveals "streamlines" that the eye can follow, and these are strong visual approximations to actual solution curves — because an actual solution curve is required, by definition, to be tangent to the field at every single point along its length, not just at grid points. A useful organizing device is the isocline: the set of points where $f(t,y) = c$ for a fixed constant $c$. Along an isocline, every field segment has the same slope $c$, even though the isocline itself is generally not a solution curve. Isoclines let you sketch a direction field efficiently by hand: pick a few values of $c$, plot the curve $f(t,y)=c$, and draw slope-$c$ ticks along it. Consider the concrete equation $y' = t - y$. Here $f(t,y) = t-y$, so the isocline for slope $c$ is the line $y = t - c$, a family of parallel lines of slope $1$. Along the line $y=t$ (that is $c=0$), every solution passing through has a horizontal tangent. Along $y = t-1$ (that is $c=1$), every solution crosses with slope $1$, matching the line's own slope — which is a strong hint that this line is nearly, or exactly, an invariant curve.

This theorem is exactly why direction fields are trustworthy sketching tools: no matter how two hand-drawn solution curves are approximated, if the underlying $f$ is a genuine function (single-valued at each point), two true solution curves can only touch tangentially where they meet, never cross at an angle. In particular, note what the theorem does not claim: it does not say two solutions through the same point must be equal as functions on a common interval (that stronger uniqueness statement requires more than continuity of $f$, and is the content of Existence and Uniqueness (Picard-Lindelof)). It also does not assert that solutions with different initial values never touch — only that if they touch, they do so tangentially.

Example 1: Newton's law of cooling with a nonzero ambient drift. A cup of coffee at temperature $y(t)$ (in °C) cools in a room whose ambient temperature is not constant but drifts slowly according to $20 + 0.1t$ (the room warms up over the day, say from morning sun). Newton's law of cooling states the rate of heat loss is proportional to the difference between the object's temperature and the ambient temperature, with proportionality constant $k = 0.05$ per minute. This gives the first-order ODE

Here $f(t,y) = -0.05y + 0.005t + 1$, defined and continuous on all of $D=\mathbb{R}^2$. Direction field analysis without solving. Consider the isocline $f(t,y)=0$:

Notice this isocline is exactly the ambient temperature line $20+0.1t$ — of course, since the ODE says $y'=0$ precisely when $y$ equals the ambient temperature. Along this line every solution has a horizontal tangent. Just above this line ($y$ slightly greater than $0.1t+20$), $f(t,y) < 0$, so solutions decrease (coffee above ambient cools). Just below, $f(t,y)>0$, so solutions increase (coffee below ambient warms up). This confirms the physical expectation: temperature is pulled toward the moving ambient line from both sides. Checking a candidate solution. Try the ansatz $\phi(t) = 0.1t + 20 + Ce^{-0.05t}$ for constant $C$ (this guess is motivated by the fact that a particular solution should track the drifting isocline, plus a decaying correction). Differentiate:

Compute $f(t,\phi(t))$:

Wait — compare to $\phi'(t) = 0.1 - 0.05Ce^{-0.05t}$; these match only if the constant term $0.1$ is also accounted for. Re-examine: $f(t,\phi(t)) = -0.05Ce^{-0.05t}$ lacks the $+0.1$ term present in $\phi'(t)$, so the ansatz needs adjustment. Retry with $\phi(t) = 0.1t + 18 + Ce^{-0.05t}$ (adjusting the intercept). Then

which now matches $\phi'(t) = 0.1 - 0.05Ce^{-0.05t}$ exactly. So

is a genuine one-parameter family of solutions, for any constant $C$. If the coffee starts at $y(0) = 90$°C, then $\phi(0) = 18 + C = 90$, so $C = 72$, giving

As $t\to\infty$, $\phi(t) - (0.1t+20) = -2 + 72e^{-0.05t} \to -2$, so the coffee temperature approaches a curve running 2 degrees below the naive ambient line $0.1t+20$ — a subtlety only the algebra reveals, though the direction field already told us qualitatively that solutions converge toward the isocline $y=0.1t+20$ from either side.

Example 2: A logistic-type field with a non-symmetric right-hand side. Consider

modeling a population with logistic growth (carrying capacity $3$) subject to a slowly increasing harvesting/removal term $0.2t$. Here $f$ is a polynomial in $t$ and $y$, hence continuous (indeed smooth) on all of $D = \mathbb{R}^2$, so the crossing theorem applies everywhere. Equilibrium check at $t=0$ only. At $t=0$, $f(0,y) = y(3-y) = 3y - y^2$, which is zero at $y=0$ and $y=3$. But for $t>0$ these are not equilibria of the full equation, since $f(t,0) = -0.2t \neq 0$ for $t>0$ and $f(t,3) = -0.2t \ne 0$ for $t > 0$. This is an important asymmetry: the "carrying capacity" $y=3$ is only instantaneously a zero-slope isocline at the single moment $t=0$; the harvesting term immediately pulls it off equilibrium. This shows a common trap — reading off equilibria from $f(0,y)=0$ and assuming they persist for all $t$ is an error when $f$ genuinely depends on $t$. Isocline for slope $c=0$ at general $t$: solve $y(3-y) = 0.2t$, i.e. $y^2 - 3y + 0.2t = 0$, giving

This is real only for $t \le 11.25$. For $t< 11.25$ there are two branches of the zero-slope isocline (an upper branch near $y\approx 3$ and a lower branch near $y\approx 0$ for small $t$), and they merge into a single point at $t=11.25$, $y = 1.5$, where $9-0.8(11.25)=0$. For $t>11.25$, $f(t,y)<0$ for every $y$, since the discriminant of $y^2-3y+0.2t$ (as a quadratic in $y$) is negative, so $y^2-3y+0.2t>0$ always, meaning $f(t,y) = -(y^2-3y+0.2t) < 0$ for all $y$. Consequently every solution curve is eventually strictly decreasing once $t$ exceeds $11.25$ — a global qualitative conclusion about every solution, extracted entirely from analyzing the sign of $f$, without solving the ODE. This is the payoff of direction-field thinking: a hard nonlinear equation (not separable in $t$ and $y$ as given, because of the $-0.2t$ term coupling them) still yields a decisive long-term prediction. By the Tangency and non-crossing of solution curves theorem, we also know that any two solution curves of this equation which happen to touch at some $(t_0,y_0)$ must do so with matching slope $f(t_0,y_0)$ — so on a direction-field sketch, solution curves may bunch closely together and appear to touch, but a careful drawing never shows two of them crossing at a visible angle.