Building Intuition

What do $\partial u/\partial t$ and $\partial^2 u/\partial x^2$ actually mean — physically?

What is $\partial u / \partial t$?

Imagine you are holding a thermometer inside a metal bar that is being heated from one end. The temperature at your exact location, at this exact moment, is $u$. You wait a short time $\Delta t$. The temperature is now $u + \Delta u$. The rate at which temperature is changing at your location is:

$\displaystyle\frac{\partial u}{\partial t} \approx \frac{u(t + \Delta t) - u(t)}{\Delta t}$

This is the time derivative: how fast the temperature is changing right here, right now. If $\partial u / \partial t > 0$, this location is warming. If it is negative, it is cooling. If it is zero, the temperature at this point is momentarily steady.

The partial derivative symbol $\partial$ (rather than the ordinary $d$) signals that $u$ depends on multiple variables — here, both position $x$ and time $t$. When we write $\partial u / \partial t$, we hold $x$ fixed and ask: how does $u$ change with $t$ at this one location? The rest of the bar does not enter the picture.

In discrete form — which is also how computers compute it — this becomes:

∂u/∂t ≈ (u[n+1] - u[n]) / Δt

where u[n] is the temperature at time step $n$. This approximation gets more accurate as $\Delta t \to 0$.

What is $\partial u / \partial x$?

Now freeze time. You are looking at a snapshot of the temperature distribution along the bar — a curve $u(x)$ that tells you the temperature at every position. The quantity $\partial u / \partial x$ is the slope of that curve: how steeply the temperature rises or falls as you move in the $x$-direction.

This derivative has a direct physical consequence. In 1822, Jean-Baptiste Joseph Fourier proposed that the rate of heat flow through a material is proportional to the temperature gradient. This is Fourier's law of heat conduction:

$q = -k\,\dfrac{\partial u}{\partial x}$

Here $q$ is the heat flux (energy per unit area per unit time) and $k$ is thermal conductivity. The minus sign is load-bearing: it says that heat flows opposite to the temperature gradient. Heat flows downhill. Where the temperature is falling in the $x$-direction ($\partial u / \partial x < 0$), heat flows in the positive $x$-direction. Where temperature is rising ($\partial u / \partial x > 0$), heat flows backward.

The gradient alone tells you the direction and intensity of heat flow, but not whether a given point will heat up or cool down. For that, you need the second derivative.

What is $\partial^2 u / \partial x^2$?

This is the central concept of the chapter. Take a moment with it.

Consider three neighbouring points on the bar: $x_{i-1}$, $x_i$, and $x_{i+1}$, spaced $\Delta x$ apart. We want to find a formula for the second spatial derivative at the middle point $x_i$.

Expand $u(x_i + \Delta x)$ and $u(x_i - \Delta x)$ using Taylor series:

$u_{i+1} = u_i + u'_i\,\Delta x + \tfrac{1}{2}u''_i\,\Delta x^2 + O(\Delta x^3)$

$u_{i-1} = u_i - u'_i\,\Delta x + \tfrac{1}{2}u''_i\,\Delta x^2 + O(\Delta x^3)$

Add the two equations. The first-derivative terms cancel:

$u_{i+1} + u_{i-1} = 2u_i + u''_i\,\Delta x^2 + O(\Delta x^4)$

Solve for $u''_i$:

Now rearrange. Factor a 2 from the numerator:

$\displaystyle\frac{\partial^2 u}{\partial x^2}\bigg|_i = \frac{2}{\Delta x^2}\left(\frac{u_{i-1} + u_{i+1}}{2} - u_i\right)$

The term $(u_{i-1} + u_{i+1})/2$ is simply the average of the two neighbours. So the second derivative is proportional to:

Read this carefully. The second derivative at a point is positive when the point is cooler than its neighbours (the average exceeds $u_i$), and negative when the point is hotter than its neighbours.

A concrete example. Suppose at three adjacent grid points we measure:

x_{i-1} = 0.30,  u_{i-1} = 0.60
x_i     = 0.40,  u_i     = 1.00   ← the peak
x_{i+1} = 0.50,  u_{i+1} = 0.60

Δx = 0.10

Neighbor average = (0.60 + 0.60) / 2 = 0.60
Deviation        = 0.60 − 1.00     = −0.40   ← cooler than you

∂²u/∂x² ≈ (0.60 − 2×1.00 + 0.60) / (0.10)² = −0.80 / 0.01 = −80

Negative second derivative at the peak: this point is hotter than its neighbours. It will cool down. This is not a mathematical curiosity — it is exactly what you would expect from physical experience. A hot spot in a bar does not stay hot. Heat diffuses outward until the temperature equalises.

The interactive simulation below makes this visible in real time.

The heat equation — it clicks

We now have all the pieces. Assemble them.

• $\partial u / \partial t$ — how fast the temperature at a point is changing.
• $\partial^2 u / \partial x^2$ — how different the temperature at a point is from its neighbours.

The heat equation says these two quantities are proportional:

In plain language: if you are hotter than your surroundings, you cool down; if you are cooler, you warm up. The rate is determined by how large the difference is and by the thermal diffusivity $\alpha$ of the material.

This is not a counter-intuitive result. It is exactly what you would expect heat to do based on everyday experience. The equation is not imposing something arbitrary on nature — it is capturing something you already knew, in the precise language of mathematics.

Look at the simulation above with this in mind. Wherever the Laplacian (green curve) is negative, the temperature is above the local average — the red fill marks those regions as ones that will cool. Wherever the Laplacian is positive — blue fill — the temperature is below average and will rise. And the time evolution (the space–time map below) shows exactly that process playing out.

One more thing to notice: as time progresses, the peaks flatten and the valleys fill in. The temperature distribution smooths out. Eventually it flattens to zero everywhere (because we fixed the bar's ends at zero temperature). This process is diffusion, and the heat equation is the mathematical statement of how it works.

What comes next

We can now stare at the heat equation and understand what it says. But can we solve it? Can we find an exact formula for $u(x,t)$ given any initial condition?

In the early 1800s, Joseph Fourier sat with precisely this problem. He tried to understand heat conduction in a solid body — a problem that Napoleon's Institut de France had offered a prize for. The approach he developed to solve it was so unexpected, and so powerful, that it reshaped nearly every branch of mathematics and physics that came after.

Fourier's key insight: any initial temperature profile can be written as a sum of sine and cosine waves, each of which the heat equation solves exactly. The complicated problem decomposes into infinitely many simple ones — and the simple ones can be solved by inspection.

This idea — decomposing a function into its frequency components — is now called the Fourier series. It underlies signal processing, image compression, quantum mechanics, and the very concept of a differential equation's spectrum. The next chapter tells the story of how Fourier found it, and why the mathematical establishment initially rejected it.