Let x be a real number so that $x + \frac{1}{x} = 3$. How find the sum of the last two digits of $(x^{2})^{2013} + \frac{1}{(x^{2})^{2013}}$?
-
Put $x=2\cos y$ – lab bhattacharjee Dec 29 '15 at 13:22
-
What makes you think that this number should have a finite decimal representation? – A.P. Dec 29 '15 at 13:31
-
nice...............+1 – Bhaskara-III Dec 29 '15 at 13:33
-
1Conjecture: we have that $x^{2n}+1/x^{2n} = \mathop{\rm Lucas}(4 n)$, see here. So we have the result that the last two digits are 22. The complete number is here. – Fabian Dec 29 '15 at 18:45
-
@Fabian That conjecture's easily proved: first note that $\operatorname{Lucas}(4n) = (\varphi_1^4)^n + (\varphi_2^4)^n$, where $\varphi_i = (1 \pm \sqrt{5})/2$. Then observe that $\varphi_1^4 = (7+3\sqrt{5})/2 = \lambda_1^2$. The rest follows from the theory of linear recurrence sequences (see my answer below). But how do you compute the last two digits of $\operatorname{Lucas}(4n)$ without using a CAS? – A.P. Jan 05 '16 at 01:14
-
@A.P.: one way is to follow the reasoning here. You can see by solving the recurrence relation of the Lucas numbers mod 4 that the period is 6. For mod 25, the period is 20. So the period mod 100 is 60. Now $\text{Lucas}(4n) \mod 100$ has a period of 15. So the last two digits of $\text{Lucas}(4\times 2013) $ are the same as the ones of $\text{Lucas}(4\times 3) $. – Fabian Jan 05 '16 at 04:34
3 Answers
I haven't solved this yet but my approach so far is along the lines of:
$$\begin{align} x+\frac{1}{x}&=3\\ \therefore \left(x+\frac{1}{x}\right)^2&=3^2\\ \therefore x^2+\frac{1}{x^2}+2&=9\implies x^2+\frac{1}{x^2}=7 \end{align}$$ Now we make use of this result and notice the following pattern: $$\begin{align} x^2+\frac{1}{x^2}&=7\\ x^4+\frac{1}{x^4}&=\left(x^2+\frac{1}{x^2}\right)\left(x^2+\frac{1}{x^2}\right)-2&&=7\left(x^2+\frac{1}{x^2}\right)-2\\ x^6+\frac{1}{x^6}&=\left(x^2+\frac{1}{x^2}\right)\left(x^4+\frac{1}{x^4}\right)-\left(x^2+\frac{1}{x^2}\right)&&=7\left(x^4+\frac{1}{x^4}\right)-\left(x^2+\frac{1}{x^2}\right)\\ x^8+\frac{1}{x^8}&=\left(x^2+\frac{1}{x^2}\right)\left(x^6+\frac{1}{x^6}\right)-\left(x^4+\frac{1}{x^4}\right)&&=7\left(x^6+\frac{1}{x^6}\right)-\left(x^4+\frac{1}{x^4}\right)\\ &\cdots\\ x^{2n}+\frac{1}{x^{2n}}&=\left(x^2+\frac{1}{x^2}\right)\left(x^{2(n-1)}+\frac{1}{x^{2(n-1)}}\right)-\left(x^{2(n-2)}+\frac{1}{x^{2(n-2)}}\right)&&=7\left(x^{2(n-1)}+\frac{1}{x^{2(n-1)}}\right)-\left(x^{2(n-2)}+\frac{1}{x^{2(n-2)}}\right)\\ \end{align}$$ If we now let:$$a_n=x^{2n}+\frac{1}{x^{2n}}$$then this can be written as the following recurrence relation:$$a_n=7a_{n-1}-a_{n-2}$$which can be solved to yield:$$a_n=x^{2n}+\frac{1}{x^{2n}}=\left(\frac{7+3\sqrt{5}}{2}\right)^{n}+\left(\frac{7-3\sqrt{5}}{2}\right)^{n}, n=0,1,2,3,\cdots$$ We now need to compute the sum of last two digits for $n=2013$

- 5,434
-
-
-
The expression above for $x^4+\frac{1}{x^4}$ was bothering me a bit as it seemed to break the pattern, but then I noticed that it does indeed fit as it can be rewritten as:$$x^4+\frac{1}{x^4}=7(x^2+\frac{1}{x^2})-2=7(x^2+\frac{1}{x^2})-(x^{0}+\frac{1}{x^{0}})$$ – Mufasa Dec 29 '15 at 15:47
-
Here's a nice fact: if $u_n := x^n + 1/x^n$, then $u_{kn} = u_k u_{k(n-1)} - u_{k(n-2)}$. That's because $u_{kn} = (\lambda_1^k)^n + (\lambda_2^k)^n$ and $\lambda_1^k \lambda_2^k = (\lambda_1 \lambda_2)^k = 1$ (see my answer for the notation). Sadly, this isn't really helping me to compute $u_{2\cdot 2013} \pmod{100}$... maybe it can help you. :( – A.P. Dec 29 '15 at 16:32
Recall that a linear recurrence sequence (with constant coefficients) is a sequence $U = \{u_n\}_{n=0}^{\infty}$ with terms in $\Bbb{C}$ such that $$ u_n = c_1 u_{n-1} + \dotsc + c_m u_{n-m} \quad \text{for all } n \geq m \tag{1} \label{eq:1} $$ where $c_1,\dotsc,c_m$ are complex constants with $c_m \neq 0$. The numbers $u_0,\dotsc,u_{m-1}$ are called the initial values of $U$.
Now, while $U$ may satisfy various relations of type $\eqref{eq:1}$, it can be shown that there is a unique relation of minimal length $k$, the order of $U$. The polynomial $$ f_U(X) = X^k - c_1 X^{k-1} - \dotsc - c_k $$ is the companion polynomial of $U$, and its study leads to a very interesting result:
Theorem: Let $\lambda_1,\dotsc,\lambda_h$ be the distinct (complex) roots of $f_U$, with multiplicities $e_1,\dotsc,e_h$ respectively. Then there are polynomials $g_1,\dotsc,g_h \in \Bbb{C}[X]$ with $\deg g_i \leq e_i$ such that $$ u_n = g_1(n) \lambda_1^n + \dotsc + g_h(n) \lambda_h^n \quad \text{for all } h \geq 0. \tag{2} \label{eq:2} $$ Conversely, every sequence satisfying $\eqref{eq:2}$ is a linear recurrence sequence.
Proof: Cfr. Theorem 9.10 from these notes.
Why did I bring this up? Well... since $x + \frac{1}{x} = 3$, we know that $x$ is a root of the polynomial $X^2 - 3X + 1$, thus $x$ is either $$ \lambda_1 := \frac{3 + \sqrt{5}}{2} \quad \text{or} \quad \lambda_2 := \frac{3 - \sqrt{5}}{2}. $$ Interestingly, these are the inverse of each other, thus $$ x^n + \frac{1}{x^n} = \lambda_1^n + \lambda_2^n =: u_n. $$ Thus $\{u_n\}_{n=0}^\infty$ is the linear recurrence sequence defined by the relation $$ u_n = 3 u_{n-1} - u_{n-2} $$ with initial data $u_0 = \lambda_1^0 + \lambda_2^0 = 2$ and $u_1 = \lambda_1 + \lambda_2 = 3$.
This leads to a simple recursive algorithm to compute (exactly) the last two digits of any $u_n$, and doing this modulo $100$ is very fast even for large $n$. Here's a naive way to compute $u_{2 \cdot 2013} \pmod{100}$ in Mathematica:
rec[n_Integer] := LinearRecurrence[{3, -1}, {3, 7}, {n, n}] // First;
Mod[rec[2*2013], 100]
(* 22 *)
Also, if you don't trust what I wrote above, you can easily test that this recurrence does indeed give the numbers we want:
AllTrue[Range[100], rec[#] == N[((3 - Sqrt[5])/2)^# + ((3 + Sqrt[5])/2)^#] &]
(* True *)

- 9,728
Square your equation to get: $$x^2+\frac{1}{x^2}=3^2 - 2$$ And again: $$x^4+\frac{1}{x^4}=(3^2 - 2)^2-2$$ So: $$x^{2^n}+\frac{1}{x^{2^n}}=(\ldots((3^2 - 2)^2-2)^2 -2)\ldots)^2 -2$$
Define: $$a_{n+1} = a_n^2 - 2$$ Now let $a_n = 100 b_n + y_n$. So that:$$a_{n+1} = 100^2 +200 b_n y_n+ y_n^2 - 2$$ $$y_{n+1} = y_n^2 - 2\mod 100$$ Start with $y_1=7$. $y_2 = 47$ and $y_3 = 107 \equiv 7$.
Edit:
This works to get us a answer if we had $2048$ instead of $2013$. Working on an alternate solution.

- 33,018
-
1
-
1I don't think this is complete answer, so what are the last two digits of the number? – Bhaskara-III Dec 29 '15 at 13:34
-
-
Could you please address my comment, too? You say that $x^{2n} + 1/x^{2n} = a_n^2 - 2$, while what you (almost) proved is that $x^{2^n} + 1/x^{2^n} = a_n^2 - 2$. For example, you can easily check that $x^4 + 1/x^4 = 47$ and $x^8+1/x^8=2207$=47^2-2... – A.P. Dec 29 '15 at 14:11
-
1... , while $x^6+1/x^6=322$. In case you're wondering how I got those, $x^2 - 3x + 1$ has two roots, $(3 \pm \sqrt{5})/2$, which are one the inverse of the other. Thus $x^n + 1/x^n = \left((3-\sqrt{5})^n + (3+\sqrt{5})^n\right)/2^n$. – A.P. Dec 29 '15 at 14:13
-
-
@nbubis You're computing $$x^{2^n}+\frac{1}{x^{2^n}}$$, not $$x^{2n}+\frac{1}{x^{2n}}$$. – GNUSupporter 8964民主女神 地下教會 Dec 29 '15 at 15:28