Integration of Complex Functions

As before, μ\mu will in this section be a positive measure on an arbitrary measurable space XX .

1.30 Definition We define L1(μ)L^1(\mu) to be the collection of all complex measurable functions ff on XX for which

Xfdμ<.\int_X |f| d\mu < \infty.

Note that the measurability of ff implies that of f|f| , as we saw in Proposition 1.9(b); hence the above integral is defined.

The members of L1(μ)L^1(\mu) are called Lebesgue integrable functions (with respect to μ\mu ) or summable functions. The significance of the exponent 1 will become clear in Chap. 3.

1.31 Definition If f=u+ivf = u + iv , where uu and vv are real measurable functions on XX , and if fL1(μ)f \in L^1(\mu) , we define

Efdμ=Eu+dμEudμ+iEv+dμiEvdμ(1)\int_E f d\mu = \int_E u^+ d\mu - \int_E u^- d\mu + i \int_E v^+ d\mu - i \int_E v^- d\mu \quad (1)

for every measurable set EE .

Here u+u^+ and uu^- are the positive and negative parts of uu , as defined in Sec. 1.15; v+v^+ and vv^- are similarly obtained from vv . These four functions are measurable, real, and nonnegative; hence the four integrals on the right of (1) exist, by Definition 1.23. Furthermore, we have u+u<fu^+ \le |u| < |f| , etc., so that each of these four integrals is finite. Thus (1) defines the integral on the left as a complex number.

Occasionally it is desirable to define the integral of a measurable function ff with range in [,][-\infty, \infty] to be

Efdμ=Ef+dμEfdμ,(2)\int_E f \, d\mu = \int_E f^+ \, d\mu - \int_E f^- \, d\mu, \quad (2)

provided that at least one of the integrals on the right of (2) is finite. The left side of (2) is then a number in [,][-\infty, \infty] .

1.32 Theorem Suppose ff and gL1(μ)g \in L^1(\mu) and α\alpha and β\beta are complex numbers. Then αf+βgL1(μ)\alpha f + \beta g \in L^1(\mu) , and

X(αf+βg)dμ=αXfdμ+βXgdμ.(1)\int_X (\alpha f + \beta g) \, d\mu = \alpha \int_X f \, d\mu + \beta \int_X g \, d\mu. \quad (1)

PROOF The measurability of αf+βg\alpha f + \beta g follows from Proposition 1.9(c). By Sec. 1.24 and Theorem 1.27,

Xαf+βgdμX(αf+βg)dμ=αXfdμ+βXgdμ<.\begin{aligned}\int_X |\alpha f + \beta g| \, d\mu &\le \int_X (|\alpha| |f| + |\beta| |g|) \, d\mu \\&= |\alpha| \int_X |f| \, d\mu + |\beta| \int_X |g| \, d\mu < \infty.\end{aligned}

Thus αf+βgL1(μ)\alpha f + \beta g \in L^1(\mu) .

To prove (1), it is clearly sufficient to prove

X(f+g)dμ=Xfdμ+Xgdμ(2)\int_X (f + g) \, d\mu = \int_X f \, d\mu + \int_X g \, d\mu \quad (2)

and

X(αf)dμ=αXfdμ,(3)\int_X (\alpha f) \, d\mu = \alpha \int_X f \, d\mu, \quad (3)

and the general case of (2) will follow if we prove (2) for real ff and gg in L1(μ)L^1(\mu) .

Assuming this, and setting h=f+gh = f + g , we have

h+h=f+f+g+gh^+ - h^- = f^+ - f^- + g^+ - g^-

or

h++f+g=f++g++h(4)h^+ + f^- + g^- = f^+ + g^+ + h^- \quad (4)

By Theorem 1.27,

h++f+g=f++g++h,(5)\int h^+ + \int f^- + \int g^- = \int f^+ + \int g^+ + \int h^-, \quad (5)

and since each of these integrals is finite, we may transpose and obtain (2).

That (3) holds if α0\alpha \ge 0 follows from Proposition 1.24(c). It is easy to verify that (3) holds if α=1\alpha = -1 , using relations like (u)+=u(-u)^+ = u^- . The case α=i\alpha = i is also easy: If f=u+ivf = u + iv , then

(if)=(iuv)=(v)+iu=v+iu=i(u+iv)=if.\int (if) = \int (iu - v) = \int (-v) + i \int u = - \int v + i \int u = i \left( \int u + i \int v \right) \\ = i \int f.

Combining these cases with (2), we obtain (3) for any complex α\alpha . ////

1.33 Theorem If fL1(μ)f \in L^1(\mu) , then

XfdμXfdμ.\left| \int_X f \, d\mu \right| \le \int_X |f| \, d\mu.

PROOF Put z=Xfdμz = \int_X f \, d\mu . Since zz is a complex number, there is a complex number α\alpha , with α=1|\alpha| = 1 , such that αz=z\alpha z = |z| . Let uu be the real part of αf\alpha f . Then uαf=fu \le |\alpha f| = |f| . Hence

Xfdμ=αXfdμ=Xαfdμ=XudμXfdμ.\left| \int_X f \, d\mu \right| = \alpha \int_X f \, d\mu = \int_X \alpha f \, d\mu = \int_X u \, d\mu \le \int_X |f| \, d\mu.

The third of the above equalities holds since the preceding ones show that αfdμ\int \alpha f \, d\mu is real. ////

We conclude this section with another important convergence theorem.

1.34 Lebesgue’s Dominated Convergence Theorem Suppose {fn}\{f_n\} is a sequence of complex measurable functions on XX such that

f(x)=limnfn(x)(1)f(x) = \lim_{n \to \infty} f_n(x) \quad (1)

exists for every xXx \in X . If there is a function gL1(μ)g \in L^1(\mu) such that

fn(x)g(x)(n=1,2,3,;xX),(2)|f_n(x)| \le g(x) \quad (n = 1, 2, 3, \dots; x \in X), \quad (2)

then fL1(μ)f \in L^1(\mu) ,

limnXfnfdμ=0,(3)\lim_{n \to \infty} \int_X |f_n - f| \, d\mu = 0, \quad (3)

and

limnXfndμ=Xfdμ.(4)\lim_{n \to \infty} \int_X f_n \, d\mu = \int_X f \, d\mu. \quad (4)

PROOF Since fg|f| \le g and ff is measurable, fL1(μ)f \in L^1(\mu) . Since fnf2g|f_n - f| \le 2g , Fatou’s lemma applies to the functions 2gfnf2g - |f_n - f| and yields

X2gdμlim infnX(2gfnf)dμ=X2gdμ+lim infn(Xfnfdμ)=X2gdμlim supnXfnfdμ.\begin{aligned}\int_X 2g \, d\mu &\le \liminf_{n \to \infty} \int_X (2g - |f_n - f|) \, d\mu \\&= \int_X 2g \, d\mu + \liminf_{n \to \infty} \left( - \int_X |f_n - f| \, d\mu \right) \\&= \int_X 2g \, d\mu - \limsup_{n \to \infty} \int_X |f_n - f| \, d\mu.\end{aligned}

Since 2gdμ\int 2g \, d\mu is finite, we may subtract it and obtain

lim supnXfnfdμ0.(5)\limsup_{n \to \infty} \int_X |f_n - f| \, d\mu \le 0. \quad (5)

If a sequence of nonnegative real numbers fails to converge to 0, then its upper limit is positive. Thus (5) implies (3). By Theorem 1.33, applied to fnff_n - f , (3) implies (4). ////