Introduction

Throughout calculus, differentiation and integration are presented as opposite operations.

For a function:

$$F(x)=\int_a^x f(t),dt$$

the Fundamental Theorem of Calculus states:

$$F’(x)=f(x)$$

Differentiation recovers the original function from its integral.

One of the deepest achievements of measure theory is the realization that a similar phenomenon exists for measures.

This naturally leads to the question:

Can a measure be differentiated?

At first glance this sounds strange.

Measures assign sizes to sets.

Derivatives apply to functions.

These seem like completely different objects.

Remarkably, measure theory shows that measures possess a genuine notion of differentiation.

This idea culminates in the Lebesgue Differentiation Theorem, one of the central results of modern analysis.

---

Motivation

Consider a density function:

$$f(x)\ge0$$

Define a measure:

$$\nu(A)=\int_A f,d\lambda$$

for measurable sets:

$$A$$

Suppose we know:

$$\nu$$

but not:

$$f$$

Can we recover:

$$f$$

from:

$$\nu$$

This is exactly analogous to asking:

Can we recover a function from its integral?

The answer is yes.

---

Average Density

Fix a point:

$$x$$

and consider a small interval:

$$I_r(x)=(x-r,x+r)$$

The measure assigned to this interval is:

$$\nu(I_r(x))$$

Its length is:

$$\lambda(I_r(x))=2r$$

Define the average density:

$$\frac{\nu(I_r(x))}{\lambda(I_r(x))}$$

This quantity measures:

Mass per unit length near x.

---

Example

Suppose:

$$f(x)=3x^2$$

and:

$$\nu(A)=\int_A3x^2,d\lambda$$

Then:

$$\frac{\nu(I_r(x))}{\lambda(I_r(x))}=\frac{1}{2r}\int_{x-r}^{x+r}3t^2,dt$$

As:

$$r\to0$$

the interval shrinks toward:

$$x$$

and the average density approaches:

$$3x^2$$

Thus:

$$\lim_{r\to0}\frac{\nu(I_r(x))}{\lambda(I_r(x))}=3x^2$$

The density has been recovered.

---

The Central Idea

A derivative measures local behavior.

For functions:

$$\frac{F(x+h)-F(x)}{h}$$

examines increasingly small neighborhoods.

For measures:

$$\frac{\nu(I_r(x))}{\lambda(I_r(x))}$$

does the same thing.

Both concepts investigate what happens near a point.

---

Radon–Nikodym Revisited

Recall the Radon–Nikodym Theorem.

If:

$$\nu\ll\mu$$

then there exists:

$$f=\frac{d\nu}{d\mu}$$

such that:

$$\nu(A)=\int_Af,d\mu$$

The notation:

$$\frac{d\nu}{d\mu}$$

was intentionally chosen to resemble a derivative.

Today we begin seeing why.

---

Local Ratios

Suppose:

$$\mu=\lambda$$

Then:

$$\frac{\nu(I_r(x))}{\lambda(I_r(x))}$$

compares:

• mass under ν
• mass under λ

inside a tiny neighborhood.

As:

$$r\to0$$

this ratio becomes the Radon–Nikodym derivative.

This is the measure-theoretic analogue of differentiation.

---

A Simple Example

Let:

$$\nu(A)=5\lambda(A)$$

Then:

$$\frac{\nu(I_r(x))}{\lambda(I_r(x))}=5$$

for every interval.

Thus:

$$\frac{d\nu}{d\lambda}=5$$

The derivative is constant.

---

Another Example

Let:

$$\nu(A)=\int_Ax^2,d\lambda$$

Then:

$$\frac{\nu(I_r(x))}{\lambda(I_r(x))}=\frac{1}{2r}\int_{x-r}^{x+r}t^2,dt$$

As:

$$r\to0$$

the average approaches:

$$x^2$$

Thus:

$$\frac{d\nu}{d\lambda}=x^2$$

---

Why This Is Difficult

The previous examples involve smooth functions.

But measure theory must also handle:

• discontinuous densities
• measurable functions
• irregular measures

The theorem must work even when classical calculus fails.

This is what makes the result profound.

---

Differentiation Bases

To study local behavior, we examine shrinking neighborhoods.

In:

$$\mathbb R$$

we use intervals:

$$I_r(x)$$

In:

$$\mathbb R^n$$

we typically use balls:

$$B(x,r)$$

defined by:

$$B(x,r)={y:|y-x|<r}$$

These balls shrink toward:

$$x$$

as:

$$r\to0$$

---

Local Density

The natural local density becomes:

$$\frac{\nu(B(x,r))}{\lambda(B(x,r))}$$

This measures:

Average mass near x.

The key question is:

Does this ratio converge?

If so, what does it converge to?

---

The Fundamental Guess

Suppose:

$$\nu(A)=\int_Af,d\lambda$$

Then intuition suggests:

$$\frac{\nu(B(x,r))}{\lambda(B(x,r))}\approx f(x)$$

for very small:

$$r$$

Thus:

$$f(x)$$

should be recoverable from local averages.

The remarkable fact is that this intuition is correct.

---

Why Analysts Care

Differentiation of measures provides:

• recovery of densities
• local geometric information
• connections between measures and functions

It is one of the foundational ideas behind:

• harmonic analysis
• PDEs
• probability theory
• geometric measure theory

---

Connection to Probability

Suppose:

$$P$$

has density:

$$p(x)$$

Then:

$$P(B(x,r))$$

represents the probability of landing near:

$$x$$

The ratio:

$$\frac{P(B(x,r))}{\lambda(B(x,r))}$$

describes local probability concentration.

As:

$$r\to0$$

this should recover:

$$p(x)$$

The Lebesgue Differentiation Theorem makes this precise.

---

Geometric Interpretation

Imagine measuring the population density of a city.

Take a neighborhood around a location:

$$x$$

Compute:

$$\frac{\text{Population in neighborhood}}{\text{Area of neighborhood}}$$

As the neighborhood becomes smaller and smaller, the average density approaches the true local density.

This everyday idea is exactly what measure differentiation formalizes.

---

The Road Ahead

Everything now points toward a major theorem.

We want to prove:

$$\lim_{r\to0}\frac{1}{\lambda(B(x,r))}\int_{B(x,r)}f(y),d\lambda(y)=f(x)$$

for almost every:

$$x$$

This theorem is called the Lebesgue Differentiation Theorem.

It is one of the deepest results in classical measure theory.

---

Why It Is Important

The theorem shows:

Integrating and then differentiating recovers the original function.

Not just for continuous functions.

Not just for smooth functions.

But for almost every integrable function.

In many ways, it is the true measure-theoretic version of the Fundamental Theorem of Calculus.

---

Connection to Alain Connes

The Radon–Nikodym derivative introduced the idea that measures can be differentiated.

The differentiation of measures explains why this terminology is justified.

Later, in noncommutative geometry, Connes develops analogues of:

• measures
• derivatives
• integration

for operator algebras.

The classical differentiation theory we are studying now becomes one of the conceptual ancestors of these far more sophisticated constructions.

---

Key Concepts Learned

By the end of this lesson you should understand:

• Measures possess a notion of local density.
• Local density is studied through ratios such as:

$$\frac{\nu(B(x,r))}{\lambda(B(x,r))}$$

• Differentiation of measures is the measure-theoretic analogue of ordinary differentiation.
• Radon–Nikodym derivatives can often be recovered from local averages.
• Shrinking intervals and balls reveal local behavior.
• The central goal is to recover a density from a measure.
• This leads directly to the Lebesgue Differentiation Theorem.

---

Looking Ahead

Measure Theory Lesson 29: The Lebesgue Differentiation Theorem

In the next lesson, we prove one of the crown jewels of classical analysis. We will show that for almost every point:

$$\lim_{r\to0}\frac{1}{\lambda(B(x,r))}\int_{B(x,r)}f(y),d\lambda(y)=f(x)$$

This theorem explains why local averages recover pointwise values and provides the rigorous foundation for interpreting the Radon–Nikodym derivative as a true derivative.