On Infinitely Nested Radicals

Brian Thomas

1. Introduction

In On Infinitely Nested Radicals by Seth Zimmerman and Chungwu Ho, which appeared in the Feb. 2008 issue of Mathematics Magazine [1], the questions of convergence, density, and correspondence of rational numbers that can be written as infinitely nested radicals are explored. One example of a nested radical is the Golden Ratio, , which can be written as:

(1.1)

...

Many questions arise from the study of nested radicals, such as: 1. What can we say about the convergence of them? 2. What types of numbers (i.e. rationals, integers, transcendentals etc.) can we write as nested radicals? 3. For a number k, how many different sequences will converge to k? 4. If we take all possible limits of sequences, what can we say about their density? 2. An Example We begin by looking at a relatively simple example of a nested radical:

(2.1)

...

We need to have a formal definition of what a nested radical actually is, so let us define the sequence as follows:

(2.2)

Clearly when we extend Equation 2.1 out infinitely it is equal to

. Furthermore when we look at

the sequence we can use the Half-Angle identity to see that . Thus the sequence is bounded and monotonically increasing, and therefore converges. We can also see that as , , so and since the next section by taking an arbitrary value instead of 2. We will extend this idea in

3. Radicals of the form

As before we will define and for some rational number a and begin by looking at the limit of the sequence of s as n approaches infinity. We can very simply show that the sequence is increasing and furthermore inductively show it is bounded. Now, looking at , we use the equation . Squaring both sides and moving all terms to one side of the equation leaves us with the quadratic which has roots . But we also know that possibility of subtraction and conclude that: so , and since we can remove the

(3.1) This shows that the correspondence between a and r is one-to-one. Furthermore since a is rational and r must be the roots of a quadratic with rational coefficients, r cannot be transcendental. THEOREM 1: For each rational number . Proof: Fix h and let . Then we can substitute in for in Equation 3.1, and after simple algebraic manipulation we obtain the desired result. Uniqueness comes from the fact that Equation 3.1 is a one-to-one correspondence. is the unique rational number a such that

4. Radicals of the form

We will use the same definitions as in Section 3, , where that: is from Section 3. We will also define

and let . 3. Inductively we can show

(4.1) Thus we get . Using our results form the previous section we show that

, and furthermore we can apply to show that s must be a root of the equation and therefore again a transcendental will not be the limit of this type of nested radicals. Clearly, if we choose s, then there are infinitely many possible choices of a and b that will produce an equivalent nested radicals relation.

5. Radicals of the form Things begin to get considerably trickier when we start to deal with these types of nested radicals. For starters we will set and define the sequence similar to before, with Thus if then also is. To avoid imaginary numbers we restrict Inductively we can show the sequence of even terms is increasing and the sequence of odd terms is decreasing. They are both also bounded between 0 and and thus converge. We can furthermore define a function on

the interval , differentiate, and see that the minimum for this function must lie at an endpoint of the interval. We can use this to show the minimum is obtained at the left endpoint and that over the entire interval to the same limit: We will use this to show that the two sequences of odd and even terms converge

Since , both even and odd subsequences thus converge to the same limit and therefore From our earlier equations we can see that must be satisfied by u, and thus (5.1)

converges.

When we take sequence

, we define will only be defined when

and thus, . We will use this result to prove Theorem 2.

. The

THEOREM 2: Let h, a, and b be positive numbers. Then 1. 2. Proof: 2. Since and , , and

if and only if:

exists if and only if

From Equation 5.1 we get that Thus

Algebraic manipulation shows this is true if and only if 1. For to exist entirely in the reals we earlier specified that , leading to the inequality to an alternating series of 0s and bs). Thus

Thus condition 2 is proven. . However we have also established

(the inequality is strict because equality leads When we fix h, we notice that the

quadratic opens upward and thus the inequality is only satisfied when b lies between the two roots: But so and condition 1 is reached. . Furthermore, since

COROLlARY: For each positive integer k, there exist integers a and b such that a and b have to satisfy the conditions of Theorem 2, there are finitely many of them. 6. Alternating Series Radicals

We now turn to the most complex type of nested radicals that we will encounter in this paper, those of the

form:

or their alternate, and with

We will start by recursively defining a pair of sequences

their recursive definitions and . First we need to see that these converge. The xs have no problems converging if the ys are positive, but the ys bring up a more complicated case.

THEOREM 3: Let a and b be positive real numbers such that n, Proof by Induction: Let Furthermore, . Then So we obtain . So since

Then for each positive integer

, . And since we defined

. So then the theorem is true for . Now we turn to the general case. Suppose Theorem 3 holds for some n. We will show that it holds for . Since we have . Since , . . , we can use this to get the desired result:

Since , We can now use this to show:

So now we ask the question of what numbers (6.1)

and

This proves Theorem 3 for the general case. satisfy the conditions of Theorem 3?

(6.2)

(6.3)

. Now we set

, where

(6.4) , We can also see that some exists such that Equation 6.4 is satisfied if and only if . Making the equation an equality rather than an inequality and solving the equation results in (Note: We can find an explicit formula for but for our purposes the approximation will suffice.). We now are prepared to investigate the convergence of the sequences 7. Theorem 4 and Its Implications We will begin by defining THEOREM 4: Let , , and following: and be defined as above. For any . , let and be real numbers and satisfy the and and what the sequences converge to.

1. 2. Then, and . .

, and

Proof: First a general outline of the proof: In Part I, we will use the conditions to show that the initial conditions of Theorem 3 are satisfied, which will lead us to the conclusion that our functions are well defined and furthermore we will be able to use the results of Theorem 3 in Part II. Part II will show that the sequences converge and what they converge to. Part I. Let Condition 2 if and only if , where be given. Now from our above analysis we know that . is the positive root of . By . This is true . Using the familiar quadratic

equation and taking the positive root we obtain or . We used solely if and only if statements so therefore Conditions 1 and 2 being satisfied is equivalent to the initial condition of Theorem 3, , and thus the results of Theorem 3 are applicable.

Part II. Now we will combine the results of Part I and Theorem 3. Since .

(7.1)

(7.2)

(because

(7.3)

(Condition 2)

(7.4)

(7.5)

(uses Condition 2 again)

(7.6)

(7.7)

(7.8)

Since

, and

. Furthermore since

, we can

look at the first part of Equation 7.8 as follows: . Thus, looking back at 7.8 we can conclude that . By definition . Thus we have proven Theorem 4. It is clearly apparent that . As such the limit of our alternating series depends on the initial sign; if we switch all signs around, somewhat unsurprisingly, we obtain a different result. Furthermore if we have a sequence where the first sign alternates, this sequence will never converge to a limit (assuming we choose identical constant values). This is easy to see because we can view the sequence as the union of two disjoint subsequences, x and y, which do not converge to the same limit so clearly the entire sequence can not converge to a limit. Furthermore we can analyze what numbers are possible for our limits of sequences x and y to be: , so

First off we note Condition 1,

(since we know c).

For the sequence

, we can choose any positive number

and construct an alternating series that and . We

approaches . The numbers just must satisfy our initial parameters, can use a very similar argument for the s. 8. Conclusion

There are many different routes of exploration available in this field. Some examples the authors give include questioning into the infinitely many possible nested radicals, where each individual sign can be either positive or negative. Clearly any such construction would be larger than that of all negatives and smaller than that of all positive signs. We could also explore into what would happen if we allowed the as and bs to be sequences rather than constants. Right off the bat we expect that these sequences will be convergent (or at least not go too either infinity) although it is definitely possible to create a sequence that the infinitely nested radical converges yet the sequence does not (for example, let and this turns into our alternating infinite radicals. Another interesting area would be that of complex numbers and what happens if you allow the constants to be complex (or even negative, which could lead to complex numbers as our limits). There are lots of interesting avenues into which research in this area could head and it is an extremely fascinating area of research. 9. References [1] C. Ho, S. Zimmerman, On Infinitely Nested Radicals, Mathematics Magazine, Feb. 2008.