Difference/Recurrence Equations Explained

  • Thread starter Thread starter Benny
  • Start date Start date
  • Tags Tags
    Difference
AI Thread Summary
Difference equations are mathematical expressions that relate a function's values at different points, similar to differential equations but focused on discrete variables. They often involve summations and can be solved using techniques analogous to those used for differential equations. The general solution of a first-order difference equation can be expressed in terms of an initial value and a constant, with solutions often taking the form of geometric series. Understanding the relationship between the terms in a difference equation is crucial for finding solutions, as demonstrated through various examples and transformations. Overall, mastering difference equations requires recognizing their structure and applying appropriate mathematical methods.
Benny
Messages
577
Reaction score
0
Hi I'm just wondering what difference/recurrence equations are about. From what I can recall from the bits and pieces I've readed, difference equations are somewhat similar or an analogue of differential equations. I've read through some things on first and second order difference equations but I don't really understand what they are about.

My book introduces difference equations as summations. They seem to be equations involving summations with different indices. I know what I've said is a little vague so here is an example question I have(not from the book I referred to).

Q. Find the general solution of the following difference equation.

y_{n + 1} - y_n = 2

Answer: y_n = A\left( {\frac{1}{2}} \right)^n + 2

Can someone explain to me what the equation actually means?. Any sort of explanation or references to useful websites would also be good. As I alluded to before I don't actually see the connection between difference equations and differential equations. Is it like when you have a real function f(x) and you the sequence f(n)? Any help would be appreciated.
 
Physics news on Phys.org
I have no idea what this could mean. Assuming that solution is correct:

y_{n+1} - y_{n} = A\left (\frac 12 \right )^{n+1} + 2 - A\left (\frac 12 \right )^n - 2 = -\frac 12A\left (\frac 12 \right )^n

so

2 = -\frac 12A\left (\frac 12 \right )^n

A = -2^{n+2}

y_n = -4 + 2 = -2
 
The two main ways difference and differential equations are related are
1. Analogous formula
2. differential equations can be seen as differential equations where the differences are taken to zero in a limit sense.
your equation
y_{n+1}-y_n=2
perform indefinite summation on both sides
\sum_{n=0}^{n-1}(y_{n+1}-y_n)=\sum_{n=0}^{n-1}2

y_{n}-y_0=2n

y_n=y_0+2n
where y0 is the constant to be found by initial conditions.
The equation means find the functions with the given difference.

http://mathworld.wolfram.com/RecurrenceEquation.html
 
Last edited:
What you have is an arithmetic series and your "answer" is incorrect. The correct solution is

y_n = y_1 + (n-1)\times 2

where y_1 is an initial value. The answer you show looks like the solution of a geometric series. Did you copy the problem correctly?
 
I left out a factor 2 for the first term in the question. It should be 2y_{n + 1} - y_n = 2 and the answer is the one that I included with my first post.
 
In that case, I would subtract y_n from both sides to make it look like a difference equation and see where it leads:

2 (y_{n+1} - y_n}) = 2 - y_n

or

y_{n+1} - y_n = \frac {2-y_n}{2}

The right side suggests defining a new variable z_n such that z_n = y_n - 2 or y_n = z_n + 2[/tex]. Substitute that into the equation:<br /> <br /> z_{n+1} + 2 - z_n - 2 = -\frac {1}{2} z_n<br /> <br /> or<br /> <br /> z_{n+1} - z_n = -\frac {1}{2} z_n<br /> <br /> Notice the additive constants are now gone! In fact, if we now just add z_n to both side we obtain the simple relation<br /> <br /> z_{n+1} = \frac {1}{2} z_n<br /> <br /> which is just a simple geometric series with the solution<br /> <br /> z_n = \left( \frac {1}{2} \right)^{n-1} z_1<br /> <br /> You should be able to go from there and find the general solution for y_n using the defintion of z_n that we used.
 
We can do other things, inspired by the methods of differential equations.

First, some notation:

Ey_n := y_{n+1}

Δ = E - 1
or, equivalently,

\Delta y_n := y_{n+1} - y_n


So, you have the equation 2Ey - y = 2. We'd like to rewrite it as Tide suggested. The "easy" way is to recognize you can subtract y from both sides, but if you didn't notice that, you could use E = 1 + Δ to get:

2 (1 + Δ) y - y = 2
2y + 2Δy - y = 2
2Δy + y = 2
Δy + (1/2) y = 1


First, the "plug and chug" method: this will be uninspired, but it does give you a solution to any difference equation with summations and products. You can safely skip down to the next section if you like.

In an ordinary differential equation we look for an integrating factor μ... we can do the same here. Notice that:

Δ(μy) = Eμ Δy + Δμ y

So, what we want is (Δμ) / Eμ = (1/2). Using Δ = E - 1, we can rearrange this to form:

Eμ = 2μ

By picking an initial value for \mu_0, in the general case we can write μ as a product. Here, however, we can find the closed form solution:

\mu_n = 2^n

So now, we have:

Δy + (1/2) y = 1
Δy + (Δμ)/(Eμ) y = 1
Eμ Δy + Δμ y = Eμ
Δ(μy) = Eμ

And now, we can take an antidifference (i.e. a sum). Notice that:

<br /> \sum_{k = 0}^{n-1} \Delta z_k = z_n - z_0<br />

We can apply this here to get:

<br /> \begin{equation*}<br /> \begin{split}<br /> \mu_n y_n - \mu_0 y_0 &amp;= \sum_{k = 0}^{n-1} E \mu_k \\<br /> 2^n y_n - y_0 &amp;= \sum_{k = 0}^{n - 1} 2^{k+1} \\<br /> &amp;= 2^{n+1} - 2 \\<br /> y_n &amp;= 2^{-n} y_0 + 2 - 2^{1-n} \\<br /> &amp;= A 2^{-n} + 2<br /> \end{split}<br /> \end{equation*}<br />

Where the last step was just collecting the coefficients.

#################################################
#################################################

Just like with differential equations, we recognize special forms that come up frequently. Recall that we are trying to solve:

Δy + (1/2) y = 1

The homogenous part is Δh + (1/2) h = 0. This has constant coefficients, so we can immediately posit the solution h_n = r^n. Plugging this in gives:

(r - 1) r^n + (1/2) r^n = 0

Which is easily solved to give r = (1/2). So, the homogenous solution is h_n = A (1/2)^n.

Now to find a particular solution of the original equation. Since the R.H.S. is also a constant, we can immediately posit a constant solution for y. (If the R.H.S. was a polynomial of degree n, we could posit a polynomial of degree n as a solution for y) This is easily seen to be y = 2.

So, the general solution is:

y_n = A (1/2)^n + 2
 
Thanks for the help guys but I still don't really understand well enough the main thing which characterises difference equations. I just read my book again and it says that first order recurrence relations for which n = 1 are exemplified by u_{n + 1} = au_n + k with u_0 specified. The solution of the homogeneous relation (where k = 0) is immediate, u_n = Ca^n.

I don't understand why the solution would be so 'obvious.' Can someone please show me how the given expression for u_n can be verified to be a solution of the homogeneous relation? Perhaps it will make it easier for me to understand it. Any help would be good.
 
If u_n = C a^n then u_{n+1} = C a^{n+1} and substituting into u_{n+1} = a u_n gives

C a^{n+1} = a \times C a^n

which is clearly true for any value of a. Therefore, C a^n is the general solution.

To obtain the solution of the homogeneous equation from scratch simply step down with the index like this:

u_{n+1} = a u_n = a \times a u_{n-1} = a \times a \times a u_{n-2} ...

and so on until you get to the first index on the right side. You should see the pattern that develops.
 
  • #10
Thanks again for your help Tide.
 
Back
Top