Lecture 9
As logzis a multi-valued function, we must interpret (9.1) and (9.2) to mean that if particular values are assigned to any two of their terms, then one can find a value of the third term so that the equation is satisfied.
Example 9.2.
Let z1 = z2 = −1. Then, z1z2 = 1. Thus, logz1 = (2k1+ 1)πi, logz2 = (2k2+ 1)πi, and log 1 = 2k3πi where k1, k2, and k3 are integers. If we select πi to be the value of logz1 and logz2, then equation (9.1) will be satisfied if we use 2πifor log 1.If we select 0 andπi to be the values of log 1 and logz1,respectively, then equation (9.1) will be satisfied if we use−πifor logz2.Abranchof a multi-valued function is a single-valued function analytic in some domain. The principal value or branch of the logarithm, denoted by Logz,is the value inherited from the principal value of the argument:
Logz = Log|z|+iArgz.
The value of Argz jumps by 2πiaszcrosses the negative real axis. There- fore, Logz is not continuous at any point on the nonpositive real axis.
However, at all points off the nonpositive real axis, Logz is continuous.
Theorem 9.1.
The function Logzis analytic in the domainD∗consist- ing of all points of the complex plane except those lying on the nonpositive real axis; i.e.,D∗=C−(−∞,0).Furthermore,d
dzLogz = 1
z for z in D∗.
Figure 9.1
x y
0
D∗
Proof.
Setw= Logz.Letz0∈D∗ andw0= Logz0. We have to show that thezlim→z0
Logz−Logz0 z−z0
exists and is equal to 1/z0. Since Logz is continuous,w = Logz →w0 =
Elementary Functions II 59 Logz0as z→z0.Thus,
zlim→z0
Logz−Logz0
z−z0 = lim
w→w0
w−w0
ew−ew0 = lim
w→w0
1 ew−ew0
w−w0
= 1
ew0 = 1
eLogz0 = 1 z0.
(9.3)
Note that (9.3) is meaningful since, for z =z0, w will not coincide with w0. This follows from the fact that z =eLogz =ew. Thus, w = Logz is differentiable at every point inD∗,and hence is analytic there.
A line used to create a domain of analyticity is called abranch line or branch cut. Any point that must lie on a branch cut-no matter what branch is used-is called abranch pointof a multi-valued function. For example, the nonpositive real axis shown inFigure 9.1is abranch cutfor Logz,and the pointz= 0 is abranch point.
Ifαis a complex constant andz= 0,then we definezαbyzα=eαlogz. Powers ofzare, in general, multi-valued.
Example 9.3.
Find all values of 1i.Since log 1 = Log 1 + 2kπi= 2kπi, we have 1i=eilog 1=e−2kπ,wherek= 0,±1,±2,· · ·.Example 9.4.
Find all values of (−2)i. Since log(−2) = Log 2 + (π+ 2kπ)i, we have (−2)i = eilog(−2) = eiLog2e−(π+2kπ), where k = 0,±1,±2,· · ·.Thus, (−2)i has infinitely many values.Example 9.5.
Find all values ofi−2i.Since logi= Log 1+2k+12 πi= 2k+12
πi, we have i−2i = e−2ilogi = e−2i(2k+1/2)πi = e(4k+1)π, where k= 0,±1,±2,· · ·.
Since logz= Log|z|+iArgz+ 2kπi,we can write
zα = eα(Log|z|+iArgz+2kπi) = eα(Log|z|+iArgz)eα2kπi, (9.4) where k = 0,±1,±2,· · ·. The value of zα obtained by taking k =k1 and k = k2(=k1) in equation (9.4) will be the same when eα2k1πi = eα2k2πi. This occurs if and only ifα2k1πi=α2k2πi+ 2mπi(m is an integer); i.e., α=m/(k1−k2). Hence, formula (9.4) yields some identical values of zα only when α is a real rational number. Consequently, if α is not a real rational number, we obtain infinitely many different values of zα, one for each choice of the integerkin (9.4).
On the other hand, ifα=m/n,wheremandn >0 are integers having no common factor, then there are exactly n distinct values (branches) of zm/n, namely
zm/n = e(m/n)Log|z|e(m(Argz+2kπ)i)/n, k= 0,1,· · ·, n−1.
In summary, we find:
1.zαis single-valued whenαis a real integer.
2.zαtakes finitely many values whenαis a real rational number.
3.zαtakes infinitely many values in all other cases.
If we use the principal value of logz,we obtain theprincipal branchof zα,namelyeαLogz.
Example 9.6.
The principal value of (−i)i is eiLog(−i)=ei(−πi/2) = eπ/2.Sinceez is entire and Logz is analytic in the domainD∗=C\(−∞,0], the chain rule implies that the principal branch ofzαis also analytic inD∗. Furthermore, forzin D∗,we have
d dz
eαLogz
= eαLogz d
dz(αLogz) = eαLogzα z
.
Now we shall use logarithms to describe the inverse of the trigonomet- ric and hyperbolic functions. For this, we recall that if f is a one-to-one complex function with domain S and rangeS, then theinverse function of f,denoted as f−1,is the function with domain S and rangeS defined by f−1(w) =z if f(z) = w. It is clear that if the function f is bijective, thenf−1mapsS ontoS.Furthermore, both the compositionsf◦f−1and f−1◦f are the identity function. For example, the inverse of the function f(z) =ax+b, a= 0,isf−1(z) = (z−b)/a.
The inverse sine function w = sin−1z is defined by the equation z = sinw.We shall show that sin−1z is a multi-valued function given by
sin−1z = −ilog[iz+ (1−z2)1/2]. (9.5) From the equation
z = sinw = eiw−e−iw 2i ,
we have 2iz=eiw−e−iw.Multiplying both sides bye−iw,we deduce that e2iw−2izeiw−1 = 0,
which is quadratic ineiw.Solving foreiw,we find
eiw = iz+ (1−z2)1/2, (9.6) where (1−z2)1/2is a double-valued function ofz.Taking logarithms of each side of (9.6) and recalling thatw= sin−1z,we arrive at the representation (9.5).
Elementary Functions II 61
Example 9.7.
From (9.5), we havesin−1(−i) = −ilog(1±√ 2).
However, since log(1 +√
2) = Log (1 +√
2) + 2kπi, k= 0,±1,±2,· · ·, (9.7) log(1−√
2) = Log (√
2−1) + (2k+ 1)πi, k= 0,±1,±2,· · ·, (9.8) and
Log(√
2−1) = Log 1 1 +√
2 = −Log (1 +√ 2), the lists (9.7) and (9.8) can be combined in one list as
(−1)kLog (1 +√
2) +kπi, k= 0,±1,±2,· · ·. Thus, it follows that
sin−1(−i) = kπ+i(−1)k+1Log(1 +√
2), k= 0,±1,±2,· · ·. Similar to the expression for sin−1zin (9.5), it is easy to show that
cos−1z = −ilog[z+i(1−z2)1/2] and
tan−1z = i
2logi+z
i−z. (9.9)
Clearly, the functions cos−1zand tan−1z are also multi-valued.
The derivatives of these three functions are readily obtained from the representations above and appear as
d
dzsin−1z= 1
(1−z2)1/2, d
dzcos−1z=− 1
(1−z2)1/2, d
dztan−1z= 1 1+z2. Finally, we note that the inverse hyperbolic functions can be treated in a corresponding manner. It turns out that
sinh−1z = log[z+ (z2+ 1)1/2], cosh−1z = log[z+ (z2−1)1/2], tanh−1z = 1
2log1 +z 1−z,
and d
dzsinh−1z = 1 (z2+ 1)1/2, d
dzcosh−1z = 1 (z2−1)1/2, d
dztanh−1z = 1 1−z2.
Problems
9.1. Find all values ofz such that (a).ez=−2, (b).ez= 1 +√
3i, (c). exp(2z−1) = 1, (d). sinz= 2.
9.2. If z1, z2 = (π/2) +kπ, show that tanz1 = tanz2 if and only if z1=z2+kπ,wherek is an integer.
9.3. Use (7.7) and (7.8) to prove Theorem 9.1.
9.4. Evaluate the following:
(a). Log(−ei), (b). Log(1−i), (c). log(−1 +√ 3i).
9.5. Show that
(a). Log(1 +i)2= 2Log(1 +i) but (b). Log(−1 +i)2= 2Log(−1 +i).
9.6. Find the limit limy→0+[Log(a+iy)−Log(a−iy)] when a > 0, and whena <0.
9.7. Evaluate the following and find their principal values:
(a). (1 +i)i, (b). (−1)π, (c). (1−i)4i, (d). (−1 +i√ 3)3/2. 9.8. Establish (9.9).
9.9. Evaluate the following:
(a). sin−1√
5, (b). sinh−1i.
9.10.Prove the following inequalities:
e−2y
1 +e−2y < |tan(x+iy)−i| < e−2y
1−e−2y, y >0 e−2y
1 +e−2y < |cot(x+iy) +i| < e−2y
1−e−2y, y >0 e2y
1 +e2y < |tan(x+iy) +i| < e2y
1−e2y, y <0 e2y
1 +e2y < |cot(x+iy)−i| < e2y
1−e2y, y >0
Answers or Hints
9.1. (a). ez =−2 iffez=eln(2)eiπ iff z= ln 2 +iπ+i(2kπ), k∈Z, (b).
ez= 1 +√
3i= 2eiπ/3 iffz= ln 2 +iπ3 +i(2kπ), k∈Z,(c). e2z−1=ei0iff
Elementary Functions II 63 2z−1 =i(2kπ) iffz=12 +i(kπ), k∈Z, (d). sinz= 2 iffeiz−e−iz= 4i iff e2iz −4ieiz−1 = 0 iff eiz = (4i±√
−16 + 4)/2 = (4i±√
12i)/2 = (2±√
3)i= (2 +√
3)eiπ/2, (2−√
3)eiπ/2so z= ln(2 +√
3) +iπ
2 + 2kπ , or ln(2−√
3) +iπ
2 + 2kπ
, k∈Z.
9.2. tanz1−tanz2= 0 if and only if sin(z1−z2) = 0.
9.3. If Logz= Log|z|+iArgz= Logr+iΘ,thenu= Logr, v= Θ.Thus, ur= 1/r, uΘ= 0, vr= 0, vΘ= 1.
9.4. (a). Log(−ei) = lne+i(−π/2) = 1−iπ/2, (b). Log(1−i) = ln√
2−iπ/4,(c). log(−1 +√
3i) = ln 2 +i2π3 + 2kπi, k∈Z.
9.5. (a). Log(1 +i)2= Log(2i) = ln 2 +iπ2 = 2 ln√
2 +iπ4
= 2Log(1 + i), (b). Log(−1 +i)2 = Log(−2i) = ln 2−iπ/2 and 2Log(−1 +i) = 2
ln√
2 +i3π4
= ln 2 +i3π2.
9.6. 0 whena >0,and 2πiwhena <0.
9.7. (a). (1 +i)i =eilog(1+i) =ei(ln√2+i(π/4+2kπ)) =e−(π/4+2kπ)eiln√2, k∈Z,(b). (−1)π=eπlog(−1)=eπ(ln 1+i(π+2kπ))=eiπ2(1+2k), k ∈Z, (c). eπei(2 ln 2),(d). e3/2(ln 2+i2π/3)= 2√
2eiπ =−2√ 2.
9.8. 1 +itanw= 2eiw/(eiw+e−iw), 1−itanw= 2e−iw/(eiw+e−iw).
9.9. (a). (4k+ 1)π/2±iLog(√
5 + 2),(b). (4k+ 1)πi/2.
9.10.Follow the same method as in Example 8.2.