Desmond Jacelon
Posted on Saturday, 21 June, 2003 - 10:06 pm:

Is there any difference, in terms of definition, between ex and exp(x)?
Matthew Buckley
Posted on Sunday, 22 June, 2003 - 07:40 am:

Nope, there is no difference at all - just two ways of writing the same thing!
Ian Short
Posted on Sunday, 22 June, 2003 - 12:10 pm:

By convention they denote the same thing, although strictly the former is a multi-valued function and the latter is single-valued (for complex numbers).

For complex numbers define,

exp(z)= n=0 zn /n!

For w not equal to 0 define

wz =exp(zlog(w))

=exp[z(ln|w|+i(argw+2nπ))]

(any integer n).

So if w=e then ez =exp(z)exp(2nπiz) has many values. Take n=0 for the convention.

Ian

Desmond Jacelon
Posted on Sunday, 22 June, 2003 - 01:59 pm:

Thanks.

One question though: is the definition of wz purely arbitrary?

Desmond
Philip Ellison
Posted on Sunday, 22 June, 2003 - 02:08 pm:

eiθ can be seen to be equal to cosθ+isinθ by considering the power series expansions of ex , cosx and sinx. However, this assumes that the expansion of ex is valid for complex numbers, which at A-level we are simply told is acceptable, but which I vaguely recall reading somewhere is simply a matter of definition, i.e. we define eiθ as cosθ+isinθ.
Alex Miller Alex
Posted on Sunday, 22 June, 2003 - 09:12 pm:

as Philip said,
-

-Alex Miller
William Hall
Posted on Sunday, 22 June, 2003 - 11:08 pm:

The above is all true, but the definition of zw is altogether more subtle. We define zw by

zw = ew(logz)

where we are taking logarithms to the base e. There are several things we have to consider here:

1) What it means to take the natural logarithm of a complex number - if we write z=r eiθ (where r is real and θ is the argument of z), then we may write that

logz=log(r eiθ )=logr+log eiθ =logr+iθ,

2) Note how we have to be careful with this, because the above complex logarithm is multivalued (i.e. since e2iπ =1, then we could add any multiple of 2π to logz). Hence we restrict the argument of z to be within some range of width 2π (i.e. encompassing all of the complex plane), but such that there is no overlap, i.e. we may let (for example)

-π<θπ

Such a restriction of the argument of z is known as a branch cut (in fact, the above branch is known as the principal branch). Another possibility would be 0θ<2π.

The whole point of this is to prevent the function z zw being multivalued (i.e. the function is well defined). However, the range of θ which we choose is purely arbitrary; in fact, different branches define different functions.

However, it is relatively easy to show that this definition coincides with the (obviously single value) zn for integer n (in this case, the branch cut, since a change in the argument by 2π does not change zn , as e2iπ =1).

Bill