Differential Equations › Section B · First order differential equations › Topic B.4

Bernoulli equations

A Bernoulli equation looks like a linear equation spoiled by one nonlinear term. A single change of variable removes the damage and turns it into a linear equation you already know how to solve.

Nonlinear, but only just

The logistic equation \(P' = rP - \frac{r}{K}P^2\) is nonlinear because of the \(P^2\). The Calculus topic "The logistic equation and population models" solved it by separation and partial fractions, which works but takes some algebra. There is a second route. The equation is linear except for a single term containing a power of the unknown, and equations of that shape can always be converted into linear ones. They are named after Jacob Bernoulli, who posed the problem of solving them in about 1695, and the conversion is the first of several substitutions in this section that turn an unfamiliar equation into a familiar one.

The Bernoulli form

A Bernoulli equation is a first-order equation that can be written as

y' + P(t)y = Q(t)y^n

where \(n\) is any real number. Two values are excluded because they need no new method: \(n = 0\) gives a linear equation, and \(n = 1\) gives \(y' + (P - Q)y = 0\), which is linear and separable. For every other \(n\), including fractions such as \(\frac{1}{2}\) and negative powers, the equation is nonlinear. Examples: \(y' + \frac{1}{t}y = ty^2\) has \(n = 2\), \(y' + y = \sqrt{y}\) has \(n = \frac{1}{2}\), and \(y' - 3y = ty^{-1}\) has \(n = -1\).

The substitution

Divide the equation by \(y^n\):

y^{-n}y' + P(t)y^{1-n} = Q(t)

Now set \(v = y^{1-n}\). By the chain rule, \(v' = (1 - n)y^{-n}y'\), so the first term is \(\frac{v'}{1 - n}\). Substituting and multiplying by \(1 - n\) gives

v' + (1 - n)P(t)v = (1 - n)Q(t)

which is linear in \(v\). Solve it with an integrating factor, as in the topic "Linear equations and integrating factors", and then return to \(y\) through \(y = v^{1/(1-n)}\). Rather than memorising the final formula, it is safer to repeat the division and substitution each time.

Worked example: n = 2

Solve \(y' + \frac{1}{t}y = ty^2\) with \(y(1) = \frac{1}{2}\). 1. Divide by \(y^2\): \(y^{-2}y' + \frac{1}{t}y^{-1} = t\). 2. Set \(v = y^{-1}\), so \(v' = -y^{-2}y'\), and the equation becomes \(-v' + \frac{v}{t} = t\), that is \(v' - \frac{1}{t}v = -t\). 3. The integrating factor is \(\frac{1}{t}\) for \(t > 0\): \(\left(\frac{v}{t}\right)' = -1\), so \(v = Ct - t^2\). 4. \(v(1) = 2\) gives \(C = 3\), so

y = \frac{1}{3t - t^2}

5. Interval: the denominator \(t(3 - t)\) vanishes at 0 and 3, so \(0 < t < 3\). Check: \(y(1) = \frac{1}{2}\), and substituting \(y\) back into the original equation makes both sides \(\frac{t}{(3t - t^2)^2}\).

Worked example: the logistic equation again

Solve \(P' = 0.4P - 0.0004P^2\) with \(P(0) = 100\), a population with \(r = 0.4\) per year and \(K = 1000\). 1. This is Bernoulli with \(n = 2\): \(P' - 0.4P = -0.0004P^2\). 2. Divide by \(P^2\) and set \(v = \frac{1}{P}\): \(-v' - 0.4v = -0.0004\), so \(v' + 0.4v = 0.0004\). 3. Solve: \(v = 0.001 + Ce^{-0.4t}\), and \(v(0) = 0.01\) gives \(C = 0.009\). 4. Invert:

P = \frac{1000}{1 + 9e^{-0.4t}}

the familiar S-curve, with \(P(5) \approx 451\). No partial fractions were needed, and the result matches the logistic solution formula from Calculus.

Worked example: a fractional power

Solve \(y' + y = \sqrt{y}\) with \(y(0) = 4\). 1. Here \(n = \frac{1}{2}\), so divide by \(y^{1/2}\): \(y^{-1/2}y' + y^{1/2} = 1\). 2. Set \(v = y^{1/2}\), so \(v' = \frac{1}{2}y^{-1/2}y'\), giving \(2v' + v = 1\), or \(v' + \frac{1}{2}v = \frac{1}{2}\). 3. Solve: \(v = 1 + Ce^{-t/2}\), and \(v(0) = \sqrt{4} = 2\) gives \(C = 1\). 4. Square:

y = \left(1 + e^{-t/2}\right)^2

valid for all \(t\), since \(v\) stays positive. 5. Check: \(y(0) = 4\), and as \(t \to \infty\), \(y \to 1\), where \(y = \sqrt{y}\) and the right side balances \(y\).

Signs, roots and lost solutions

Three details need care when returning to \(y\). First, when \(n > 0\), dividing by \(y^n\) discards the constant solution \(y = 0\), which satisfies every such equation; list it separately. Second, when \(1 - n\) is an even integer, say \(n = 3\) with \(v = y^{-2}\), recovering \(y\) needs a square root, and the initial condition decides the sign. Third, a fractional power such as \(\sqrt{y}\) makes sense only for \(y \ge 0\), so the solution must keep \(v\) non-negative. As always with nonlinear equations, the interval of validity is found from the solution itself, by locating where it blows up or stops being real.

Recognising a Bernoulli equation

Bernoulli equations often arrive in disguise. Rearrange until \(y'\) has coefficient 1 and only one term is not first power in \(y\). For instance, \(ty' = 2y + t^4y^3\) becomes \(y' - \frac{2}{t}y = t^3y^3\), Bernoulli with \(n = 3\). But \(y' + y^2 = t\) is not Bernoulli: besides the \(y^2\) term it contains the term \(t\), with no \(y\) at all, and the form allows only \(Py\) and \(Qy^n\). That equation is a Riccati equation, which needs a different approach. If an equation is separable as well, either method will work.

Common mistakes to avoid

First, using \(v = y^n\) or \(v = y^{n-1}\) instead of \(v = y^{1-n}\): for \(n = 2\) the right choice is \(v = y^{-1}\). Second, forgetting the factor \(1 - n\) in \(v'\), which changes the signs and sizes of every coefficient; for \(n = 2\) it is \(-1\). Third, applying the initial condition to \(v\) with the value of \(y\): if \(y(1) = \frac{1}{2}\) and \(v = y^{-1}\), then \(v(1) = 2\). Fourth, solving for \(v\) and stopping, or inverting carelessly. Fifth, losing \(y = 0\) or the sign of a square root.

Key takeaways

A Bernoulli equation has the form \(y' + P(t)y = Q(t)y^n\) with \(n \ne 0, 1\). Dividing by \(y^n\) and setting \(v = y^{1-n}\) gives the linear equation \(v' + (1 - n)Pv = (1 - n)Q\), which an integrating factor solves. Convert the initial condition to \(v\), solve, then return to \(y\), choosing signs from the initial condition and checking the interval of validity. For \(n > 0\), \(y = 0\) is an extra solution. The logistic equation is the best-known example, with \(n = 2\).

Flashcards

Flashcard

Bernoulli equation

y′ + P(t)y = Q(t)yⁿ, with n ≠ 0, 1

Linear apart from one power of y.

Test your understanding

Question 1 of 5

Which substitution turns y′ + 2y = ty⁴ into a linear equation?

Question 2 of 5

For n = 2, setting v = 1/y turns y′ + Py = Qy² into which linear equation?

Question 3 of 5

Which of these is a Bernoulli equation that is not linear?

Question 4 of 5

When a Bernoulli equation with n > 0 is divided by yⁿ, which solution can be lost?

Question 5 of 5

Which function solves y′ = y − y² with y(0) = 1/2?

← Previous topicNext topic →