52.4. EXTENSIONS OF MAXIMUM MODULUS THEOREM 1659

52.4.2 Hadamard Three Circles TheoremLet 0 < R1 < R2 and suppose f is analytic on {z ∈ C : R1 < |z|< R2} . Then letting R1 <a < b < R2, note that g(z)≡ exp(z) satisfies

g : {z ∈ C : lna < Rez < b}→ {z ∈ C : a < |z|< b}

is onto and that in fact, g maps the line lnr+ iy onto the circle reiθ . Now let M (x) be definedas above and m be defined by m(r)≡maxθ

∣∣ f (reiθ)∣∣ .Then for a < r < b, Corollary 52.4.4

implies

m(r) = supy

∣∣∣ f (elnr+iy)∣∣∣= M (lnr)≤M (lna)

lnb−lnrlnb−lna M (lnb)

lnr−lnalnb−lna

= m(a)ln(b/r)/ ln(b/a) m(b)ln(r/a)/ ln(b/a)

and so m(r)ln(b/a) ≤ m(a)ln(b/r) m(b)ln(r/a) . Taking logarithms, this yields

ln(

ba

)ln(m(r))≤ ln

(br

)ln(m(a))+ ln

( ra

)ln(m(b))

which says the same as r→ ln(m(r)) is a convex function of lnr.The next example, also in Rudin [113] is very dramatic. An unbelievably weak as-

sumption is made on the growth of the function and still you get a uniform bound in theconclusion.

Corollary 52.4.5 Let Ω ={

z ∈ C : |Im(z)|< π

2

}. Suppose f is analytic on Ω, continuous

on Ω, and there exist constants, α < 1 and A < ∞ such that

| f (z)| ≤ exp(Aexp(α |x|)) for z = x+ iy

and ∣∣∣ f (x± iπ

2

)∣∣∣≤ 1

for all x ∈ R. Then | f (z)| ≤ 1 on Ω.

Proof: This time let φ (z) = [exp(Aexp(β z))exp(Aexp(−β z))]−1 where α < β < 1.Then φ (z) ̸= 0 on Ω and for η > 0

|φ (z)|η =1

|exp(ηAexp(β z))exp(ηAexp(−β z))|Now

exp(ηAexp(β z))exp(ηAexp(−β z))

= exp(ηA(exp(β z)+ exp(−β z)))

= exp[ηA(

cos(βy)(

eβx + e−βx)+ isin(βy)

(eβx− e−βx

))]and so

|φ (z)|η =1

exp[ηA(cos(βy)

(eβx + e−βx

))]Now cosβy > 0 because β < 1 and |y|< π

2 . Therefore, limsupz→∞ | f (z)| |φ (z)|η ≤ 0 ≤ 1and so by Theorem 52.4.1, | f (z)| ≤ 1.

52.4. EXTENSIONS OF MAXIMUM MODULUS THEOREM 165952.4.2 Hadamard Three Circles TheoremLet 0 < R; < R» and suppose f is analytic on {z € C: R; < |z| < Ro}. Then letting R; <a<b< Ro, note that g(z) = exp (z) satisfiesg:{zEC: Ina < Rez <b} {z€ C:a< |z| <b}is onto and that in fact, g maps the line Inr+‘y onto the circle re'®, Now let M (x) be definedas above and m be defined by m(r) = maxe | f (re!) | .Then for a <r <b, Corollary 52.4.4impliesInr—Inasup | f(er) | =M (Inr) <M (Ina) fo=i8@ M (In) w=ym/(r)— m (a)n(@/")/In(b/a) m (p)n(r/a)/Inb/a)and so m(r)!"2/9) < m(a)*/) m(b)"’/® . Taking logarithms, this yieldsIn (7) In(m(r)) < In (°) In(m(a)) +1n (<) In (m(b))which says the same as r — In(m(r)) is a convex function of Inr.The next example, also in Rudin [113] is very dramatic. An unbelievably weak as-sumption is made on the growth of the function and still you get a uniform bound in theconclusion.Corollary 52.4.5 Let Q= {z € C: |Im(z)| < $}. Suppose f is analytic on Q, continuouson Q, and there exist constants, & <1 and A < © such thatIf (z)| < exp (Aexp(@|x])) forz=x+iy1ji \l<(et) |s!for all x €R. Then |f (z)| <1 onQ.andProof: This time let @ (z) = [exp (Aexp (Bz)) exp (Aexp(—Bz))| | where a < B <1.Then @ (z) 40 on Q and for n > 017‘9 (N= Texp(pAexp (Ba) exp (nAexp(—Ba))Nowexp (MA exp (Bz)) exp (MA exp (—Bz))= exp(NA (exp (Bz) + exp(—Bz)))= exp nA (cos (By) Ga +e#*) +isin(By) (e — eb)and so1n19 (I = Sp [nA (cos (By) (ete B))]Now cos By > 0 because B < 1 and |y| < ¥. Therefore, limsup,. ,.. |f (z)||@ (z)|"" <0 < 1and so by Theorem 52.4.1, |f (z)| < 1.