Differential Equations › Section B · First order differential equations › Topic B.2

Separable equations

When the variables in a first-order equation can be pulled apart, two ordinary integrations solve it. The real skill lies in what comes after: handling implicit answers, choosing the right branch, and finding exactly where the solution lives.

The second general tool

Linear equations yield to integrating factors; separable equations, many of them nonlinear, yield to separation. The Calculus topic "Separation of variables" introduced the method and its standard applications, cooling and flushing tanks, along with the constant solutions lost by dividing. This topic adds three things a differential equations course needs: a justification of the method that does not rely on cancelling \(dx\) like a fraction, the habit of leaving answers implicit when that is all the algebra allows, and the interval of validity, which for nonlinear equations can only be found from the solution itself.

Why separation works

An equation is separable if it can be written as

N(y)\,y' = M(x)

with all the \(y\) on one side. The Calculus form \(y' = g(x)h(y)\) is this with \(N = \frac{1}{h}\) and \(M = g\). Suppose \(H(y)\) is an antiderivative of \(N(y)\) and \(G(x)\) one of \(M(x)\). By the chain rule, \(\frac{d}{dx}H(y(x)) = N(y)\,y'\), so the equation says \(\frac{d}{dx}H(y(x)) = G'(x)\). Two functions with equal derivatives differ by a constant, so

H(y) = G(x) + C

That is the whole method: integrate each side in its own variable and add one constant. Writing \(N(y)\,dy = M(x)\,dx\) is shorthand for exactly this argument.

Intervals of validity

The interval of validity of an initial value problem is the largest interval containing the starting point on which the solution exists and satisfies the equation. For a separable equation you find it after solving, by asking where the solution breaks down: a denominator that reaches zero, a logarithm of something that is not positive, an even root of a negative number, or a point where the formula for \(y'\) itself is undefined. The interval must contain \(x_0\) and must not cross any such point. Different initial conditions for the same equation can give very different intervals.

Worked example: a solution that blows up

Solve \(y' = 3t^2y^2\) with \(y(0) = \frac{1}{8}\). 1. Separate: \(y^{-2}\,dy = 3t^2\,dt\). 2. Integrate: \(-\frac{1}{y} = t^3 + C\). 3. The condition gives \(-8 = C\), so \(-\frac{1}{y} = t^3 - 8\) and

y = \frac{1}{8 - t^3}

4. Interval: the denominator vanishes at \(t = 2\), so the interval containing 0 is \(-\infty < t < 2\). 5. Check: \(y' = \frac{3t^2}{(8 - t^3)^2} = 3t^2y^2\). Note also that dividing by \(y^2\) discarded the constant solution \(y = 0\), which is valid for all \(t\).

Worked example: an implicit answer

Solve \(y' = \frac{2x + 1}{3y^2 + 1}\) with \(y(0) = 1\). 1. Separate: \((3y^2 + 1)\,dy = (2x + 1)\,dx\). 2. Integrate: \(y^3 + y = x^2 + x + C\). 3. At \(x = 0\), \(y = 1\): \(1 + 1 = C\), so

y^3 + y = x^2 + x + 2

4. Solving this cubic for \(y\) is possible but ugly, so the implicit form is the answer. 5. Interval: the left side \(y^3 + y\) is strictly increasing and takes every real value, so each \(x\) gives exactly one \(y\), and \(3y^2 + 1\) never vanishes. The solution is valid for all \(x\). Check: implicit differentiation gives \((3y^2 + 1)y' = 2x + 1\).

Worked example: a logarithm sets the interval

Solve \(\frac{dy}{dx} = e^{y}\cos x\) with \(y(0) = 0\). 1. Separate: \(e^{-y}\,dy = \cos x\,dx\). 2. Integrate: \(-e^{-y} = \sin x + C\). 3. The condition gives \(-1 = C\), so \(e^{-y} = 1 - \sin x\). 4. Take logarithms: \(y = -\ln(1 - \sin x)\). 5. Interval: the logarithm needs \(1 - \sin x > 0\), which fails exactly where \(\sin x = 1\), at \(x = \frac{\pi}{2} + 2k\pi\). The interval containing 0 is

-\frac{3\pi}{2} < x < \frac{\pi}{2}

and \(y \to \infty\) at both ends. Check: \(y' = \frac{\cos x}{1 - \sin x}\), and \(e^{y}\cos x\) is the same.

Choosing the branch

When an implicit solution is solved for \(y\), a square root often produces two candidates, and only the initial condition can choose. Take \(y' = -\frac{x}{4y}\) with \(y(0) = -1\). Separating gives \(4y\,dy = -x\,dx\), so \(2y^2 = -\frac{x^2}{2} + C\), that is \(x^2 + 4y^2 = K\), and the condition gives \(K = 4\): an ellipse. Solving, \(y = \pm\frac{1}{2}\sqrt{4 - x^2}\), and \(y(0) = -1\) demands the minus sign. The interval is \(-2 < x < 2\), open, because at \(x = \pm 2\) the solution reaches \(y = 0\), where the equation's right side is undefined.

Spotting separable equations

Some separable equations are disguised. Exponentials split: \(y' = e^{x - 2y}\) is \(e^{x}e^{-2y}\). Grouping can expose a product: \(y' = xy + 2x - y - 2\) factors as \((x - 1)(y + 2)\), which you can confirm by expanding. A quotient \(y' = \frac{\sin x}{y\cos y}\) is already \(M(x)\) over \(N(y)\). But a genuine sum, such as \(y' = x + y\) or \(y' = \sin(xy)\), never separates. When the right side will not factor into an \(x\)-part times a \(y\)-part, look instead to the linear, exact or substitution methods of this section.

Common mistakes to avoid

First, adding a constant on each side and keeping both; one constant is enough. Second, taking a logarithm, as in \(y = -\ln(1 - \sin x)\), without checking where its argument stays positive. Third, keeping both signs of a square root, or the wrong one. Fourth, reporting an interval that crosses a point where the solution or \(y'\) is undefined, or using a closed interval whose endpoints break the equation. Fifth, forgetting constant solutions removed by the division, such as \(y = 0\) for \(y' = 3t^2y^2\).

Key takeaways

A separable equation can be written \(N(y)\,y' = M(x)\). By the chain rule, integrating each side in its own variable gives \(H(y) = G(x) + C\) with a single constant. Implicit solutions are complete answers; solve for \(y\) only when the algebra allows, and let the initial condition choose the branch. The interval of validity is the largest interval containing \(x_0\) on which the solution and its derivative make sense; nonlinear solutions can blow up or end where a root or logarithm fails. Constants at zeros of the divisor are solutions too.

Flashcards

Flashcard

Separable equation

N(y) y′ = M(x)

All y on one side, all x on the other.

Test your understanding

Question 1 of 5

Which equation is separable?

Question 2 of 5

Which function solves y′ = t/y with y(0) = −3?

Question 3 of 5

The IVP y′ = 2ty², y(0) = 1, has solution y = 1/(1 − t²). What is its interval of validity?

Question 4 of 5

Separating y′ = (y − 2) cos t by dividing by y − 2 can lose which solution?

Question 5 of 5

Separating y′ = 2x/(3y² + 2) and integrating gives which implicit solution?

← Previous topicNext topic →