Introduction

Up to this point, every measure we have studied satisfies:

$$\mu(A)\ge0$$

for every measurable set:

$$A$$

This makes perfect sense when measure represents:

• length
• area
• volume
• probability
• mass

All of these quantities are naturally nonnegative.

However, many areas of mathematics require a more flexible notion.

For example:

• Differences of probability measures
• Differences of distributions
• Linear functionals in functional analysis
• Spectral theory
• Operator algebras

In such situations, it is useful to allow a measure to take negative values.

This leads to the concept of a signed measure.

Surprisingly, once negative values are allowed, entirely new mathematical phenomena appear.

---

Motivation

Suppose:

$$\mu_1$$

and:

$$\mu_2$$

are ordinary measures.

Define:

$$\nu=\mu_1-\mu_2$$

For a measurable set:

$$A$$

we have:

$$\nu(A)=\mu_1(A)-\mu_2(A)$$

This quantity may be:

• positive
• negative
• zero

Thus:

$$\nu$$

is generally no longer an ordinary measure.

Yet it still behaves very much like one.

This observation motivates the definition.

---

Definition of a Signed Measure

Let:

$$\left(X,\mathcal{F}\right)$$

be a measurable space.

A function:

$$\nu:\mathcal{F}\to[-\infty,\infty]$$

is called a signed measure if:

1. Empty Set Property

$$\nu(\emptyset)=0$$

2. Countable Additivity

Whenever:

$$A_1,A_2,\ldots$$

are pairwise disjoint measurable sets,

$$\nu\left(\bigcup_{n=1}^{\infty}A_n\right)=\sum_{n=1}^{\infty}\nu(A_n)$$

provided the series is well defined.

Notice that nonnegativity has disappeared.

That is the major difference.

---

Example 1

Let:

$$\lambda$$

denote Lebesgue measure on:

$$\mathbb{R}$$

Define:

$$\nu(A)=2\lambda(A)-5\lambda(A)$$

Then:

$$\nu(A)=-3\lambda(A)$$

This is a signed measure.

For example:

$$\nu([0,1])=-3$$

Although negative, countable additivity still holds.

---

Example 2

Suppose:

$$\delta_0$$

is the Dirac measure at:

$$0$$

and:

$$\delta_1$$

is the Dirac measure at:

$$1$$

Define:

$$\nu=\delta_0-\delta_1$$

Then:

$$\nu({0})=1$$

$$\nu({1})=-1$$

This is one of the simplest nontrivial signed measures.

---

Positive and Negative Regions

Once negative values are possible, a natural question arises:

Which parts of the space contribute positively?

and

Which parts contribute negatively?

For ordinary measures this question never appears because everything is positive.

For signed measures it becomes fundamental.

---

Positive Sets

A measurable set:

$$P$$

is called positive if every measurable subset:

$$A\subseteq P$$

satisfies:

$$\nu(A)\ge0$$

In a positive set, every measurable piece has nonnegative signed measure.

---

Negative Sets

A measurable set:

$$N$$

is called negative if every measurable subset:

$$A\subseteq N$$

satisfies:

$$\nu(A)\le0$$

Every measurable piece carries nonpositive measure.

---

An Important Observation

A positive set does not necessarily have positive measure.

It only means that all of its measurable subsets have nonnegative measure.

Likewise, a negative set means all measurable subsets have nonpositive measure.

This distinction is subtle but important.

---

The Hahn Decomposition Problem

Suppose:

$$\nu$$

is a signed measure.

Can we split the space into:

• a positive region
• a negative region

in a clean way?

In other words:

Can we find sets:

$$P$$

and:

$$N$$

such that:

$$P\cup N=X$$

and:

$$P\cap N=\emptyset$$

with:

$$P$$

positive and:

$$N$$

negative?

Remarkably, the answer is yes.

This is one of the most beautiful theorems in measure theory.

---

Hahn Decomposition Theorem

For every signed measure:

$$\nu$$

there exist measurable sets:

$$P$$

and:

$$N$$

such that:

$$P\cup N=X$$

$$P\cap N=\emptyset$$

and:

$$P$$

is positive while:

$$N$$

is negative.

This decomposition is called a Hahn decomposition.

---

Why Hahn Decomposition Is Important

The theorem tells us that every signed measure can be separated into:

• a purely positive component
• a purely negative component

The decomposition is not necessarily unique, but any two Hahn decompositions differ only by sets of measure zero.

Thus the positive and negative regions are essentially unique.

---

Geometric Intuition

Think of a signed measure as a landscape containing:

• hills (positive mass)
• valleys (negative mass)

The Hahn decomposition separates the landscape into its hill region and valley region.

Once separated, each region behaves like an ordinary measure.

---

Example

Suppose:

$$X={1,2,3}$$

and define:

$$\nu({1})=2$$

$$\nu({2})=-5$$

$$\nu({3})=1$$

A Hahn decomposition is:

$$P={1,3}$$

$$N={2}$$

Every subset of:

$$P$$

has nonnegative measure.

Every subset of:

$$N$$

has nonpositive measure.

---

Toward the Jordan Decomposition

The Hahn decomposition allows us to define two ordinary measures:

A positive measure:

$$\nu^+$$

and a negative measure:

$$\nu^-$$

such that:

$$\nu=\nu^+-\nu^-$$

This result is called the Jordan decomposition theorem.

We will study it in the next lesson.

---

Why Signed Measures Matter

At first glance, signed measures may seem like a technical generalization.

In reality they are everywhere.

They appear naturally in:

• Functional analysis
• Harmonic analysis
• Probability theory
• Spectral theory
• Operator algebras

Many deep theorems are easiest to understand using signed measures.

---

Connection to Integration

Suppose:

$$f$$

is integrable.

Then:

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

defines a signed measure.

If:

$$f$$

sometimes becomes negative, then:

$$\nu(A)$$

may also become negative.

Thus signed measures naturally arise from integration.

This observation is one of the key ideas leading to the Radon–Nikodym Theorem.

---

Connection to Functional Analysis

Signed measures can be viewed as generalized linear functionals.

Later, when studying dual spaces and Hilbert spaces, this viewpoint becomes extremely important.

Many representation theorems ultimately show that certain linear functionals are equivalent to integration against signed measures.

---

Connection to Alain Connes

The move from ordinary measures to signed measures is the first example of a recurring theme in advanced analysis:

Generalize familiar objects by relaxing constraints.

First:

$$\mu(A)\ge0$$

Then:

$$\nu(A)$$

may be positive or negative.

Later:

• functions become operators
• measures become traces
• spaces become algebras

Connes’ work repeatedly follows this pattern of abstraction and generalization.

---

Key Concepts Learned

By the end of this lesson you should understand:

• Ordinary measures are always nonnegative.
• Signed measures may take positive or negative values.
• Countable additivity remains valid.
• Positive sets contain only nonnegative measurable subsets.
• Negative sets contain only nonpositive measurable subsets.
• Every signed measure admits a Hahn decomposition.
• Signed measures often arise from integration.
• Signed measures prepare the way for the Jordan decomposition and the Radon–Nikodym Theorem.

---

Looking Ahead

In the next lesson:

Lesson 16: The Jordan Decomposition Theorem

we will prove that every signed measure can be written uniquely as the difference of two ordinary positive measures:

$$\nu=\nu^+-\nu^-$$

This decomposition is one of the most important structural results in measure theory and forms the foundation for the Radon–Nikodym Theorem.