# Polynomial bounded w.r.t supremum norm

1. Jan 25, 2010

### Somefantastik

1. The problem statement, all variables and given/known data

E1 = {pn(t) = nt(1-t)n:n in N};

E2 = {pn(t) = t + (1/2)t2 +...+(1/n)tn: n in N};

where N is set of natural numbers

is the polynomial bounded w.r.t the supremum norm on P[0,1]?

2. Relevant equations

supremum norm = ||*|| = sup{|pn(t)|: t in [0,1]}

3. The attempt at a solution

I know that the set is bounded on P[0,1] if I can show that the supremum norm is less than some constant for all n. Can someone give me some advice on how to show either polynomial is bounded or not?

I think E1 is unbounded, no matter your choice of t, since the polynomial's value will forever increase as n in increases. I feel the same can be said of E2 but I know this is wrong.

2. Jan 25, 2010

### ystael

You can answer this fairly directly, by computing the maximum and minimum values of each $$p_n$$; because polynomials are smooth functions, the extreme values of $$p_n$$ on $$[0,1]$$ must occur at a critical point of $$p_n$$ or at one of the endpoints.

The members of $$E_2$$ share a particular property which will save you even this computation.

3. Jan 26, 2010

### Somefantastik

I'm not seeing the special property of E2. Its min value must be at t = 0 for all n, and max when t = 1, but I can't pin down a value for sup. It looks like an expanded function but it doesn't look familiar.

factoring out a t makes p2(t) = t(1 + t + (1/2)t + (1/3)t2 + ... + (1/n)tn-1), but I'm still not seeing the trick.

Last edited: Jan 26, 2010
4. Jan 26, 2010

### Somefantastik

Also, at E1, the critical points are t = 0 and t = 1. At both of these points, no matter the value of N, the polynomial is zero. Does this mean that is my sup? That doesn't make sense to me b/c the shape of this function spikes in the middle between t = 0 and t = 1.

5. Jan 26, 2010

### Somefantastik

for E2, p'(t) gives the geometric series, which equals 1/(1-t) on the interval |t|<1. The second derivative is 1/(1-t)2 < 0

What can I do with this information? The series doesn't converge at either t = 0 or t = 1.

6. Jan 26, 2010

### ystael

If the function is zero at 0 and 1, and positive in the middle, then there must be another critical point between, no?

For $$E_2$$, put $$q_n(t) = t^n/n$$, so that $$p_n(t) = \sum_{1\leq j\leq n} q_n(t)$$. Where does each $$q_n$$ reach its maximum absolute value? What does this tell you about the $$p_n$$?

7. Jan 26, 2010

### Somefantastik

yes, and it'll be a function of n, but as n -> inf, isn't the max value of the function going to diverge to inf as well?

qn reaches its max value at t = 1 => $$p_n(1) = \sum_{1\leq j\leq n}q_n(1) = \sum_{1\leq j\leq n}\frac{1}{n}$$

so the value of this polynomial is the sum of this series and is finite and therefore this polynomial is bounded.

8. Jan 26, 2010

### ystael

That's precisely what the question asks you to determine.

Bounded? Uniformly in $$n$$? Look again.

9. Jan 26, 2010