You are on page 1of 8

Will Summing the Series Help Achilles Cross

the Stadium?

Matt Fig


Regarding the first
1
version of Zeno's dichotomy paradox, it is often alleged that the story only
has a paradoxical aspect in the absence of an understanding of modern mathematical analysis.
Since the calculus easily handles infinite summation, it is further explained, there is nothing
mysterious about Achilles completing his race. Here it will be argued that the mathematical
analysis of the dichotomy does not resolve the paradox. It will also be found that it is the
assumption upon which such analysis rests that leads to the paradox in the first place.

Zeno invites us to consider Achilles at the starting line of a racecourse, and asks if it is possible
for him to cross the finish line. It is plain that before Achilles can reach the finish line he must
arrive at the halfway mark. Once he makes it to this point, but before he can cross the remaining
distance, he must cover half the distance ahead. When he makes it to the halfway mark of the
remaining distance, he must once more cover half of the distance that now remains, and so on.
Every time Achilles reaches the halfway point between the last mark and the finish line, he has
the other half to cover. And before he can cover this, he again must cover half. So, argues Zeno,
Achilles will never reach the finish line.

A Computational Approach

The process Zeno envisions can be likened to an instructional loop for getting Achilles across the
stadium. If we take the length of the stadium to be 1 unit for simplicity, then the instructions
could look like this pseudo-code:

Final_Goal = 1;
Next_Goal = 1/2;
Distance_to_Travel = 1/2;
while Next_Goal Final_Goal
Distance_to_Travel = (Final_Goal - Next_Goal)/2;
Next_Goal = Next_Goal + Distance_to_Travel;
end

The values of the Next_Goal (NG) and Distance_to_Travel (DT) at the beginning of the first
few iterations N can be tabulated as in Table 1 below.

Table 1 Summary of variable results as a function of iteration number.
N 1 2 3 4 5
NG 1/2 3/4

7/8 15/16 31/32
DT 1/2 1/4 1/8 1/16 1/32

1
Classification according to Salmon, p. 8.
1
Notice that the nth value of NG is the sum of the first through the nth values of DT. The
highlighted values are demonstrative of the trend, which holds for all iteration values.

In effect, Zeno asks if the loop will terminate. That is, whether the set of instructions within the
loop is capable of producing a value of NG such that the termination condition,
Next_Goal=Final_Goal, is reached. If the loop terminates, then Achilles does reach the finish
line.

If the code above is written on a standard system
2
, the loop will terminate after 53 iterations.
This is because the standard is incapable of distinguishing values that differ from 1 by less than
2
-52
, a system-dependent value known as eps. Thus the above termination condition is, on such a
system, functionally equivalent to the following.

(Final_Goal-Next_Goal) eps/2

Computational systems are available
3
with a value for eps of 2
-112
. On these systems the loop
will run more iterations, but will terminate in a similar manner. In order to properly evaluate
Zeno's question, we need to be able to operate on a computational system capable of representing
and distinguishing between any two rational numbers such that the original termination condition
can be kept without limitations imposed by the computational system itself. A system without
rounding error or memory limitations will be called an ideal computational system (ICS). Since
an ICS does not currently exist, we turn to an examination of the values tabulated above.
Throughout, we keep the original termination condition and structure of the computational
algorithm in mind, the question being how the code would behave on an ICS.

The NG and DT values can be written in an alternative form in order to gain insight into the
general term. The above table becomes

Table 2 Results from Table 1 in symbolic form.
N 1 2 3 4 5 n
NG 1 - 2
-1
1 - 2
-2
1 - 2
-3
1 - 2
-4
1 - 2
-5
1 - 2
-n

DT 2
-1
2
-2
2
-3
2
-4
2
-5
2
-n


Given the closed form representation of each term, we can re-write the instructional loop so that
each value is calculated in the symbolic form shown in Table 2.

Final_Goal = 1;
N = 1;
Next_Goal = 1 - 1/2
N;

Distance_to_Travel = 1/2
N
;
while Next_Goal Final_Goal
N = N + 1;
Distance_to_Travel = 1/2
N
;
Next_Goal = 1 - 1/2
N
;
end


2
I have in mind here IEEE 754 standards for double precision floating point arithmetic.
3
Quadruple precision floating point.
2
With this alternative formulation, we can answer Zeno's question. Zeno wants to know if there is
an iteration number N, computed on an ICS, such that the termination condition is met. Since an
ICS is capable of representing and distinguishing between any two rational numbers, an answer
to this question will be a value of iteration number n e such that the following equation is
satisfied.


1
1 1
2
n
= (1.1)

Unfortunately there is no value of n, reachable by incrementation, such that this equation is
satisfied. Thus the loop will not terminate, and Achilles does not reach the finish line. It is
important to note that this result is independent of the speed of the computational system used.
So long as the loop is run on an ICS, which it must if the computation is to model Achilles'
travels through a continuum, no finite iteration number can satisfy the loop's termination
condition. Though one could use (1.1) to calculate Achilles position before any iteration, it is
important to remember that the loop models a process, that of movement between points.

A Modern Analysis Approach

The previous result runs counter to the claims sometimes found in calculus textbooks
4
and taught
in analysis courses. The claim is that Zeno is mistaken if he thinks Achilles will not reach the
finish line, though his description is essentially correct. In order to understand the source of the
disagreement, we will need to look at a few of the ideas behind the analytical approach to Zeno's
question. This summary will be brief, the reader is directed to the references for a more in depth
and rigorous treatment
5
.

To begin, we examine the notion of a sequence. A formal definition of a sequence can be given.

If to each n corresponds a definite number x
n
, then the numbers e

x
1
, x
2
, x
3
, ..., x
n
, ...

form a sequence, which is symbolized { }
1
n
n
x

=
, or{ }
n
x for short .

This is another way of saying that a sequence is a function whose domain is the natural numbers.
Example expressions for the terms in several sequences are

{ }
{ } { }
{ }
1
1 1
1 1 1 1 1
1. i.e, 1, , , ,..., ,...
2 3 4
2. 2 i.e, 2, 4,8,16,...,2 ,...
1 3 11 25 137
3. 1, 1 ... i.e., 1, , , , ,...
2 6 12 60
n n
n
n n
n n
n n
x x
n n
x x
x x x
n

=
>

= =
`
)
= =
| |
= = + + =
` |
) \ .


4
See Ellis and Gulick, p. 522 for a typical example.
5
Present purposes only demand our terminology works for sequences with positive rational terms.
3
Note that the nth term in example sequence 3 is the sum of the first n terms given in example
sequence 1. Thus example sequence 3 is a sequence of the sums of the terms of another
sequence. The idea of a sequence of sums is integral to understanding series, defined later.

One more definition will be needed
6
in order to define a limit of a sequence, that of a null
sequence. A sequence is called a null sequence if it satisfies the following definition.

If for any positive c e there corresponds an n
0
such that

|x
n
|<, if n > n
0


then { } is a null sequence.
1
n
n
x

=

In plain English this definition means that a sequence is a null sequence if and only if, for any ,
there is at most a finite number of terms larger than . It may be helpful here to recall the
definition of a sequence. The first symbolic representation of a sequence given in the definition
has the symbols outside the braces to serve as a reminder that the sequence is a function of the
natural numbers. The set of natural numbers begins at 1 and has no largest element, hence the
use of the symbol. If every term after a given term is smaller than a fixed value, then there are
an infinite number of terms smaller than that fixed value.

Of the example sequences, only the first is a null sequence. For any , it is clear that

|x
n
|<, if n >1/

It is now easy to see that both the distances Achilles must travel at the beginning of each iteration
and his positions, as computed by the looping algorithm above, form sequences based on the
iteration number. The distance sequence, {D
n
}, has the formula for generating the nth term


1
2
n n
D = (1.2)

The position sequence will be given the symbol {P
n
}. The formula for generating the nth term
of this sequence is


1
1
2
n n
P = (1.3)

Some features of (1.2) and (1.3) are noted without proof. First, (1.2) is monotonically
decreasing, that is, if k>m, then D
k
< D
m
. Likewise, (1.3) is monotonically increasing. Second,
the sequence (1.2) is a null sequence. We also note that the terms in (1.3) are the sum of the
terms of (1.2), analogous to the relation between example sequence 3 and example sequence 1.

In mathematical analysis, the notion of a limit is central. Here we will define this idea in terms
of null sequences. The definition of a convergent sequence and its limit is as follows.

6
Limits may, of course, be defined without reference to null sequences.
4
If { } is a sequence that is related to a distinct number L such that
1
n
n
x

=

{ } | |
n
L x

is a null sequence, then {x
n
} is said to converge, and L is the limit of sequence.
This is read as, "the limit of { }
1
n
n
x

=
is L." It is also written, { }
1
n
n
x L

=
= .
7


With these definitions in place we note that a null sequence is simply a sequence whose limit is
0. We can easily see that sequence (1.3) has a limit 1. This is shown by using the definition.


{ }
1 1
1 1 1
2 2
n n n
P

| |
= =
` `
|
\ . ) )
(1.4)

It is straightforward to show that for any positive rational , the terms of sequence (1.4) are less
than if

( ) 2
1
log n
c
>

For example, if = 1/10, then P
4
is within of 1. If = 1/1000, then P
10
is within of 1. No
matter how small we make , we can use the above expression to find the term n above which the
sequence will give values between (1- ) and 1. Thus (1.4) is a null sequence and the limit of
(1.3) is 1.

The sequence (1.3) is given our attention because it is related to (1.2) in a special way. For each
term P
n
is the sum of the distances Achilles must travel from the first iteration of the
computational loop through the nth iteration. Thus we would expect that if we add up all the
distances traveled by Achilles through the nth iteration, we would get the total distance he has
run and this value would be equal to P
n
. Indeed this is the case.

Now the question arises, can we use our knowledge of sequences and limits to answer Zeno's
question? Is it meaningful to ask whether the sequence of sums given by {P
n
} will ever give the
total distance Achilles needs to travel in order to cross the finish line? When we ask such
questions, we are asking about infinite series. An infinite series, or series for short, is a special
kind of sequence. Namely, it is the sequence of sums of the terms of another sequence.

If { } is a sequence that is related to another sequence { }
1
n
n
x

= 1
n
n
y

=
by the relation

1 1 1 1
, ...
n n
x y x y y
>
= = + +
then { } is called an infinite series, and is symbolized as { }
1
n
n
x

= 1
1
n n
n
k
x y

=
=
=

.


7
The limit of a sequence defines its limit as uniquely as a Dedekind cut. See Knopp p. 68 for discussion.
5
Since the nth term of a sequence given as a series is the sum of the first n terms of another
sequence, we speak of a series as a sequence of the partial sums of another sequence. The sums
being partial because, for any term n, there are terms not included in the sum, namely all terms
m>n.

Infinite series allow us to understand the analytical approach Zeno's question. Much like any
other sequence, a series may have a limit. In this case the limit will be the limit of the partial
sums of another sequence. The limit of a series is known as the value, or the sum of the series.
Thus we say that the sum of the series is L, its limit.

With this in mind we interpret Zeno's question as follows. Will the sum of the individual
distances travelled by Achilles during each step equal the distance to the finish line? Since we
want to know the sum of an infinite number of distances, each given by the terms of (1.2), we
want to know the sum of the infinite series given by


1 1
1
2
n n
n n
D

= =
=

(1.5)

We have seen already that the partial sums of the sequence (1.2) are given by the terms of
sequence (1.3). Thus we already know the answer. The limit of the sequence (1.3) was shown
previously to be 1. Thus the sum of series (1.5) is 1. Since 1 is the distance Achilles must travel
if he is to cross the stadium, we answer Zeno in the affirmative: Achilles reaches the finish line.

Why the different results?

We have seen two different answers to the question Zeno poses. When modeled as a
computational loop on an ICS, the journey Achilles takes is seen to never end. When examined
as the sum of an infinite number of finite distances, Achilles is seen to reach the finish line. How
can we reconcile these different conclusions?

I believe the difference lies in a misinterpretation of the terminology used in the analytical
approach. The sum of a series is, strictly, not a sum at all. When speaking of the sum of an
infinite series, one is not really speaking of the arithmetic addition
8
of an infinite number of
things. The sum of a series is nothing other than its limit, if it has one, as previously defined.
Careful examination of the definition of a limit shows that the sum of a series is none other than
the number defined by the limit of the underlying sequence of partial sums. As such, the sum of
a series is the number L for which, given any , one can find a term of the underlying sequence
such that there are an infinite number of terms between (L- ) and L
9
.

When stated in terms of limits rather than sums, the discovery that the limit of the infinite series
(1.5) is the finite value 1 does not lend the same confidence that Zeno has been answered at all.
Even a cursory observation of the terms produced by the computational loop indicate that
Achilles approaches the finish line with each successive iteration! In fact, the results achieved

8
Knopp p. 100 says, "... the addition of an infinite number of terms is something quite undefined, something
perfectly meaningless." Emphasis original.
9
For a monotonic increasing sequence, as here. For a decreasing sequence the terms will be between L and L+ .
6
on a non-ICS can be put in terms given by analysis, so that some agreement can be found after
all.

As previously noted, any non ICS will have a value known as eps. This value effectively sets the
'fineness' of the resolution between rational numbers, and thus the total number of rational
numbers representable in any given computational environment. Each system may have a
different value for eps, with some being smaller and others being larger, but the functional
definition remains the same. For any computational system CS, eps is the smallest number
which, when added to 1 on CS, produces a number detectably different from 1 by CS. Thus the
discovery that the limit of (1.5) is 1 is simply another way of saying that the computational loop
will terminate, for some iteration number, when run on any non ICS no matter how small the
system's value for eps. This connects the two approaches, and points the way to understanding
why Achilles cannot reach the finish line if he runs according to Zeno's description. If Achilles
does run as Zeno describes, he will not reach the finish line both because the instructions within
the loop are incapable of achieving the termination condition and because the numerical value of
the finish line is not a member of the sequence of partial sums.

Again it might be objected that the computational model does not take into account the effect of
the diminishing time taken to traverse the distance between points as the iteration number
increases. One could certainly remedy the situation as follows. Say for convenience that all of
the arithmetical and comparison operations within the loop occur instantaneously on an ICS.
Then one could simply insert a line into the instructional loop that pauses the system for the
appropriate amount of time, the length of the pause being a function of velocity and iteration
number, such that if the computation were initiated as Achilles began his run the loop would
track Achilles' position in real time. Were this done, the sequence of pause times would be a null
sequence, much like sequence (1.2). Still, the termination condition depends only on the spatial
location of the runner, not on how little time elapses between comparisons! As we have seen,
the instructions that calculate the spatial location of Achilles at all times are arithmetically
incapable of producing a value that will satisfy the termination condition for any iteration
number reachable by incrementation.

The Source of the Paradox

Finally we are in a position to understand the source of the paradox, for Zeno's conclusion is
paradoxical precisely because every school child knows that a runner can reach the finish line
provided a finite distance to run, no matter how Zeno argues! It would help to expand Zeno's
contention in the form of an argument so that the premises are open to inspection.

1. In order to travel from 0 to 1, Achilles must sequentially occupy each point in the ordered set
, given by the union of {0}, {P
n
}
10
and {1}.
2. To sequentially occupy each point in an ordered set is to move from a previous point to the
next point until all points have been occupied.
3. There is no point immediately previous to 1 in .
Therefore it is not possible to travel from 0 to 1.

( )
10
The partition of [0,1] Zeno gives is not critical. For example, either
2
1
6
n
n t

or
1
1
2
3
n
n
n

=

could substitute for


the series (1.5).
7

Since universal experience tells us that the conclusion is false, we must reject one of the
premises. The third premise has been argued for at length above. {P
n
} does not include the
number 1, but merely approaches 1 as a limit, thus between any element of {P
n
} and 1 there are
an infinite number of other elements. Premise two, which is really more of a definition, also
seems completely obvious. For it is difficult to imagine a more basic exposition of the term
'sequential' than that given in the premise, or to conceive of another way to describe moving
from a starting point to an ending point when there are other points between. This premise
amounts to little more than the statement that there will be a definite time t
n
at which Achilles'
location will coincide with each point in the set, and that for all times n and m,

m n
m n
P P
V
t t



where V is the constant velocity at which Achilles travels. Since the final two premises are
sound, the only premise left is the first premise. This one also seems obviously true, and hence
the difficulty of resolving the paradox. But it is the first premise that is most easily rejected.

The reason the first premise may be rejected is that it is based on an unjustified assumption about
the connection between physical reality and the mathematical continuum. To clarify, consider
the following definitions.

D1. The actual line is the physical space between starting line and finish line, inclusive.
D2. The real line is the mathematical representation of the actual line as a continuum.

Whatever infinite partition is used in the first premise, it is clear that at bottom the assumption
common to them all is that the real line provides an appropriate model of the actual line. That is,
that there is no physically smallest distance. But if reality itself has something of a smallest
distance, an eps
R
, then Zeno's first premise is false. In that case it could be said that just as there
is no ICS in reality, there is no spatial continuum in reality. If the actual line is not a continuum
then Achilles will cross the finish line when he gets to within eps
R
of the finish line.

In addition to finding no compelling reason to accept this underlying assumption, a powerful
argument against it has been given in the form of an argument against the existence of an actual
set with an infinite number of members, namely, the famous Hilbert's Hotel. If the actual line
were partitioned into an infinite number of finite distances, all of the absurdities discovered in
that argument would attend this partitioning. Thus we are adequately justified in rejecting the
first premise, rather than either of the other two, of Zeno's argument to resolve the paradox.

References

Ellis, Robert and Denny Gulick. Calculus with Analytic Geometry, Harcourt and Brace, 1986
Knopp, Konrad. Theory and Applications of Infinite Series, Dover, 1990
Salmon, Wesley, ed. Zeno's Paradoxes, Hackett Publishing, 2001

See also: Great Ideas of Modern Mathematics by Jagjit Singh for a different, but interesting,
interpretation of the meaning of this paradox.

8

You might also like