Hyperbolic functions and complex exponential expressions
Exponentials encode growth and decay, trigonometric functions encode oscillation, and hyperbolic functions combine opposite exponential rates. Complex exponentials place these families in a common algebra without eliminating the need to choose a complex logarithm value or branch. This chapter develops the identities used in differential equations and coordinate models rather than assuming a symbolic substitution is always safe.
1 The complex exponential from an absolutely convergent series
For , define . Absolute convergence follows from the ratio , and convergence is uniform on every bounded disk by the same scalar exponential majorant. The absolutely convergent Cauchy product and binomial theorem give
Also as , with the remainder bounded by . Thus the difference quotient of proves derivative even for complex increments. The product identity gives , so the exponential never vanishes.
Separate even and odd terms for real to obtain Euler's identity , using the power-series definitions of the real sine and cosine. Hence
Real and imaginary parts of are the familiar damped oscillations and . A complex notation can therefore simplify a real linear calculation without changing the physical variable into a complex measured quantity.
2 Hyperbolic functions and their identities
For real , define
The first is even, the second odd. Multiplying the expressions gives
Their derivatives are and . The exponential addition law gives
with the corresponding plus-sign formula for . The plus sign differs from the circular cosine addition identity because the defining quadratic relation has a minus sign.
Since on the reals, , with equality only at zero by the arithmetic-geometric mean inequality or . Therefore is defined for all real , and
The curve lies on , explaining the term hyperbolic. No claim is made that its parameter is ordinary circular angle.
3 One complex variable links the two families
For complex , define
and use the same exponential formulas for . These agree with their real power series on the real axis. Substitution gives
For , expanding by the addition identities gives
Thus complex sine is not bounded in magnitude by 1: at its magnitude is , unbounded as . A real-variable range statement cannot be exported to complex arguments.
4 Inverse hyperbolic functions and logarithm domains
The real function is strictly increasing because its derivative is positive, and its limits are opposite infinities. Solving with gives
The logarithm argument is positive for every real . Differentiation gives derivative . Similarly, restricting to gives
with derivative only for . Solving for gives on .
For complex numbers, the exponential is periodic: . If has magnitude and angle , every value , , exponentiates to . Since the exponential is not injective, it has no two-sided inverse on its full domain.
A single-valued right inverse can nevertheless be selected on all of . Choose the principal argument and define . Then , but is discontinuous along the negative real axis. In particular, a single-valued selection is not necessarily a continuous logarithm branch.
There is no global continuous logarithm branch on , and hence no global holomorphic one. Here is the elementary circle obstruction. If a continuous satisfied everywhere, then would be a continuous function from into , so the intermediate value theorem forces it to be constant. But gives , a contradiction.
Finally, writing for a selected logarithm can fail by multiples of . Real positive arguments have their familiar unambiguous real logarithm.
The complex completeness and product proof behind the exponential
A complex Cauchy sequence converges because its real and imaginary parts are real Cauchy sequences; recombining their limits gives a complex limit in absolute value. Thus if , its complex partial sums converge, since the absolute value of every tail is bounded by the corresponding real nonnegative tail. The same argument is uniform over a domain when one common scalar majorant controls those tails.
For , the real series converges by an explicit geometric-tail test: after an index with , each following term is at most half its predecessor. This proves absolute convergence of the complex exponential and uniform convergence on without applying a real-valued theorem directly to complex terms.
Now let , be finite and set . Finite triangular sums give , so the nonnegative partial sums converge and converges absolutely. The square sum is the product of the two partial sums, hence tends to . Compare it with the triangular sum . If , at least one of exceeds . Consequently
Thus the Cauchy product equals the product of the sums. For , , its th coefficient is by the finite binomial theorem, proving the exponential addition law.
Finally , because . This proves the local quadratic remainder used in the complex derivative. All rearrangements in Euler's even/odd split are justified by absolute tails; the real sine and cosine series supply the two resulting sums. The proof concerns finite complex scalars, not arbitrary operators or conditional series.
5 Graded practice
E1 Foundation. Evaluate in rectangular complex form.
E2 Hyperbolic inversion. Solve and for real .
E3 Differentiation. Differentiate and verify satisfies .
E4 Complex identity. Derive the displayed formula for from the exponential definitions and use it to evaluate .
6 Full solutions
S1. Euler's identity gives . Its modulus is 2, consistent with .
S2. The first has unique solution . The second has two solutions because is even and strictly increasing only after restriction to the nonnegative half-line.
S3. The chain rule gives . For the second, and .
S4. Insert and the corresponding negative exponential into the sine definition; collecting real/imaginary parts yields the formula. At , , it gives .
Summary
Absolute power-series convergence justifies the complex exponential and its product law. Hyperbolic and trigonometric functions are related exponential combinations, but their real ranges and inverse domains differ. Complex logarithms need branch choices, and identities valid for positive real logs or bounded real trigonometric values cannot be transferred without checking those domains.