Separation of Variables

⚡ Key Takeaways

Separation of Variables is a technique for solving first-order differential equations of the form $\frac{dy}{dx} = f(x)g(y)$ by algebraically gathering all $y$ -terms on one side and all $x$ -terms on the other, then integrating both sides. Key result: a separable ODE is solved by $\int \frac{dy}{g(y)} = \int f(x)\,dx + C$ . Applies to: population growth, radioactive decay, Newton's cooling, RC circuits, and any process whose rate depends on the product of a function of the input and a function of the output. Who/when: rooted in the work of Leibniz, the Bernoullis, and Euler in the late 17th–18th centuries.

💡 The Core Insight

If a rate of change factors cleanly into "a part about $x$ " times "a part about $y$ ," then the two variables can be untangled. Once untangled, each side involves only one variable, so ordinary integration — a tool we already possess — finishes the job. The entire difficulty of the differential equation collapses into two separate, independent integrals.

DefinitionSeparable First-Order ODE

A first-order ordinary differential equation is separable if it can be written in the form

1

\quad \frac{dy}{dx} = f(x)\,g(y)

Derivation hint: This is the defining form — no derivation, but note that the right side must factor into one pure- $x$ function and one pure- $y$ function with nothing mixing them. where:

$x$ = the independent variable,
$y = y(x)$ = the unknown function we seek,
$\frac{dy}{dx}$ = the derivative (rate of change of $y$ with respect to $x$ ),
$f(x)$ = a function depending only on $x$ ,
$g(y)$ = a function depending only on $y$ .

Equivalently, in differential form,

2

\quad \frac{1}{g(y)}\,dy = f(x)\,dx

Derivation hint: Divide both sides of [1] by $g(y)$ (assuming $g(y)\neq 0$ ) and multiply by $dx$ , treating $dy/dx$ formally as a quotient of differentials.

TheoremSolution by Integration

Given the separable equation [1] with $g(y)\neq 0$ , the general solution is obtained implicitly by

3

\quad \int \frac{1}{g(y)}\,dy = \int f(x)\,dx + C

Derivation hint: Integrate both sides of [2]; the single arbitrary constant $C$ absorbs the two constants of integration from the two indefinite integrals. where $C$ is an arbitrary constant determined by an initial condition $y(x_0)=y_0$ .

ProofJustification via the Chain Rule

The "multiply by $dx$ " move looks like illegal fraction abuse. Here is the rigorous version that vindicates it.

Step 1. Start from the separable equation, dividing by $g(y)$ :

4

\quad \frac{1}{g(y)}\,\frac{dy}{dx} = f(x)

Derivation hint: Divide both sides of [1] by $g(y)$ ; this is valid wherever $g(y)\neq 0$ .

Step 2. Define an antiderivative $G$ of $\frac{1}{g}$ , so that $G'(y) = \frac{1}{g(y)}$ . By the chain rule,

5

\quad \frac{d}{dx}\,G(y(x)) = G'(y)\,\frac{dy}{dx} = \frac{1}{g(y)}\,\frac{dy}{dx}

Derivation hint: Apply the chain rule to the composite function $G(y(x))$ and substitute $G'=1/g$ .

Step 3. Combining [4] and [5], the left side of [4] is the $x$ -derivative of $G(y(x))$ :

6

\quad \frac{d}{dx}\,G(y(x)) = f(x)

Derivation hint: Substitute the result of [5] into the left-hand side of [4].

Step 4. Integrate both sides with respect to $x$ . Let $F$ be an antiderivative of $f$ :

7

\quad G(y(x)) = F(x) + C

Derivation hint: Integrate [6] over $x$ ; the fundamental theorem of calculus undoes the $\frac{d}{dx}$ on the left and produces $F(x)+C$ on the right. Since $G(y)=\int \frac{1}{g(y)}\,dy$ and $F(x)=\int f(x)\,dx$ , equation [7] is exactly the claimed result [3]. The informal "separate and integrate" recipe is therefore a shorthand for this chain-rule argument. $\blacksquare$

∎

mathematician · Gottfried Wilhelm Leibniz | 1646–1716

Leibniz, co-inventor of calculus, introduced the very notation $\frac{dy}{dx}$ and the integral sign $\int$ that make separation of variables so visually natural. His differential notation was deliberately designed so that the chain rule and substitution look like fraction arithmetic — the "multiply by $dx$ " step that students perform is a direct legacy of Leibniz's vision that differentials could be manipulated almost algebraically. He treated $dy$ and $dx$ as genuine infinitesimal quantities, a viewpoint later made rigorous by the chain rule and, much later, by non-standard analysis.

🧠 Intuition

Imagine $\frac{dy}{dx} = f(x)g(y)$ as a recipe for the steepness of a curve at every point of the plane. The factor $g(y)$ says "how strongly does the current value of $y$ influence the slope?" and $f(x)$ says "how strongly does our horizontal position influence the slope?". When these two influences multiply cleanly — never tangling into terms like $xy + \sin(x+y)$ — you can ask the two questions separately. Accumulating all the tiny $y$ -changes (weighted by $1/g(y)$ ) on the left must equal accumulating all the tiny $x$ -pushes (weighted by $f(x)$ ) on the right. Integration is just "totalling up" these tiny contributions, and the constant $C$ records where you started.

Figure

Slope field of

dy/dx = -y

with one separated solution curve

y=e^{-x}

ExampleRadioactive Decay

Given: A sample of carbon-14 decays at a rate proportional to the amount present: $\frac{dN}{dt} = -kN$ , with $k = 1.21\times 10^{-4}\ \text{yr}^{-1}$ and initial amount $N_0 = 100\ \text{g}$ . Find $N(t)$ and the amount after $5000$ years.

Step 1 — Separate. Move all $N$ to one side, all $t$ to the other:

8

\quad \frac{1}{N}\,dN = -k\,dt

Derivation hint: Divide both sides of $dN/dt=-kN$ by $N$ and multiply by $dt$ .

Step 2 — Integrate both sides.

9

\quad \ln|N| = -kt + C

Derivation hint: $\int \frac{1}{N}dN = \ln|N|$ and $\int -k\,dt = -kt$ ; combine constants into $C$ .

Step 3 — Solve for $N$ and apply $N(0)=N_0$ .

10

\quad N(t) = N_0\,e^{-kt}

Derivation hint: Exponentiate [9] to get $N=e^C e^{-kt}$ ; set $t=0$ to identify $e^C=N_0$ .

Step 4 — Plug in numbers.

N(5000) = 100\,e^{-(1.21\times10^{-4})(5000)} = 100\,e^{-0.605}

\boxed{N(5000) \approx 54.6\ \text{g}}

ExampleNewton's Law of Cooling

Given: A cup of coffee at $90^\circ\text{C}$ cools in a room at $20^\circ\text{C}$ according to $\frac{dT}{dt} = -k(T - 20)$ with $k = 0.05\ \text{min}^{-1}$ . Find the temperature after $30$ minutes.

Step 1 — Separate. Let $u = T-20$ so $du = dT$ :

11

\quad \frac{1}{T-20}\,dT = -k\,dt

Derivation hint: Divide the cooling equation by $(T-20)$ and multiply by $dt$ .

Step 2 — Integrate.

12

\quad \ln|T-20| = -kt + C

Derivation hint: $\int \frac{dT}{T-20}=\ln|T-20|$ via the substitution $u=T-20$ .

Step 3 — Solve and apply $T(0)=90$ .

13

\quad T(t) = 20 + 70\,e^{-kt}

Derivation hint: Exponentiate [12] to get $T-20 = A e^{-kt}$ ; at $t=0$ , $T=90$ gives $A=70$ .

Step 4 — Plug in numbers.

T(30) = 20 + 70\,e^{-(0.05)(30)} = 20 + 70\,e^{-1.5} = 20 + 70(0.2231)

\boxed{T(30) \approx 35.6^\circ\text{C}}

⚠️ Do Not Forget the Constant — Or the Lost Solutions

Two classic errors: 1. Dropping $C$ . The constant of integration is not optional; it encodes the initial condition. Without it you get one specific curve instead of the whole family of solutions. 2. Dividing by $g(y)$ without checking $g(y)=0$ . When you divide [1] by $g(y)$ you silently assume $g(y)\neq 0$ . Any constant $y$ with $g(y)=0$ is an equilibrium (singular) solution that the integration procedure may discard. Always check these separately.

CounterexampleA Solution Lost by Careless Division

Consider $\frac{dy}{dx} = y^{2/3}$ . Separating gives $\int y^{-2/3}\,dy = \int dx$ , hence $3y^{1/3} = x + C$ , so $y = \left(\frac{x+C}{3}\right)^3$ . But $g(y)=y^{2/3}$ vanishes at $y=0$ , and indeed $y\equiv 0$ is also a perfectly valid solution that our separation discarded by dividing by $y^{2/3}$ . Worse, this equation violates uniqueness: through the point $(0,0)$ both $y\equiv 0$ and $y=(x/3)^3$ pass, demonstrating that separation alone does not guarantee a unique solution where $g(y)=0$ .

connection

Separation of variables is the gateway technique in the study of Ordinary Differential Equations. It relies entirely on Integration (the inverse of differentiation) and is rigorously justified by the Chain Rule. When an equation is not separable, one turns to neighbouring methods such as Integrating Factors for linear first-order equations or Exact Differential Equations. In physics, the equations it solves are instances of Exponential Growth and Decay. The same name, "separation of variables," also denotes a far more advanced technique for partial differential equations — see Separation of Variables (PDEs) — where a solution is sought as a product $X(x)T(t)$ ; the conceptual spirit of splitting variables is shared, though the machinery differs.

🔬 Where this appears

⏳ History

- 1671 — Newton, in Method of Fluxions (circulated privately), solves rate equations using series and informal "fluxional" arguments anticipating separation.

- 1684 — Leibniz publishes the first calculus paper, introducing

d

and the differential notation that makes the method visually natural.

- 1691 — Leibniz explicitly uses the phrase and procedure of separating variables in correspondence solving geometric rate problems.

- 1694 — Johann Bernoulli systematizes the separation technique while solving the catenary and brachistochrone-related curves.

- 1696 — Jacob Bernoulli applies separation-type reasoning to the equation now bearing his name (the Bernoulli equation), reducible to separable form.

- 1740s — Leonhard Euler organizes first-order methods, distinguishing separable, linear, and exact types in his foundational analysis texts.

- 1768–1770 — Euler's Institutiones Calculi Integralis presents separation of variables as a standard, named classification of integrable ODEs.

- 1820s–1830s — Cauchy formalizes the notion of an initial-value problem and existence of solutions, providing the rigorous backdrop for when separation is valid.

- 1890s — Picard and Lindelöf prove the existence–uniqueness theorem, clarifying exactly when separation's discarded equilibrium solutions matter.

open-problem

Separation of variables is a recognition method, not a universal algorithm: most differential equations are simply not separable, and there is no mechanical procedure to transform an arbitrary equation into separable form. Even when separation succeeds, the resulting integrals $\int \frac{dy}{g(y)}$ and $\int f(x)\,dx$ may have no closed-form antiderivative in terms of elementary functions (for instance, $\int e^{-x^2}\,dx$ ), forcing numerical or special-function treatment. Deeper still, the solution emerges implicitly via [7] and may not be solvable for $y$ explicitly; deciding when an implicit relation $G(y)=F(x)+C$ defines a genuine function, and over what domain, connects to the implicit function theorem and to questions of solution existence and uniqueness that remain central in the theory of differential equations.

Formal Statement

Why It Holds / Derivation

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also