Prev: What are some of the more advanced convergence / divergence verification tests for infinite series'?
Next: Musatov ATTACKS Wikipedia
From: Maury Barbato on 11 May 2010 01:56 Hello, let R be the field of real numbers and f:(c,+inf) -> R^n a twice-differentiable function, where c is a real number and n an integer greater than 1. Let M_0, M_1, M_2 be the least upper bounds of |f(x)|, |f'(x)| and |f''(x)|, respectively, on (c, +inf). Can we have (M_1)^2 > 4*(M_0)*(M_2) for some f? Thank you very much for your attention. My Best Regards, Maury Barbato PS The question is from Rudin, Principles of Mathematical Analysis. If n = 1, the answer is negative, that is we have (M_1)^2 <= 4*(M_0)*(M_2) for every f. This is the proof. If h>0, then Tayolr's theorem shows f'(x) = [f(x+2h) - f(x)]/(2h) - h*f''(y), for some y in (x, x+2h). Hence |f'(x)| <= h*M_2 + (M_0)/h, and since h is arbitrary, you obtain, choosing the minimum value of h*M_2 + (M_0)/h over (0,+inf), (M_1)^2 <= 4*(M_0)*(M_2).
From: David C. Ullrich on 11 May 2010 08:17
On Tue, 11 May 2010 05:56:57 EDT, Maury Barbato <mauriziobarbato(a)aruba.it> wrote: >Hello, >let R be the field of real numbers and f:(c,+inf) -> R^n >a twice-differentiable function, where c is a real number >and n an integer greater than 1. >Let M_0, M_1, M_2 be the least upper bounds of |f(x)|, >|f'(x)| and |f''(x)|, respectively, on (c, +inf). >Can we have (M_1)^2 > 4*(M_0)*(M_2) for some f? > >Thank you very much for your attention. >My Best Regards, >Maury Barbato > >PS The question is from Rudin, Principles of Mathematical >Analysis. If n = 1, the answer is negative, that is >we have (M_1)^2 <= 4*(M_0)*(M_2) for every f. There's a little trick that allows one to deduce lots of things about vector-valued functions from the corresponding result for scalar-valued functions. Fix a unit vector v (ie |v| = 1) and define g(t) = <f(t), v>, the standard scalar product or "dot product". Note that |g(t)| <= M_0 and |g''(t)| <= M__2, hence |g'(t)|^2 <= 4 M_0 M_2. But |f'(t)| = sup_v |g'(t)|. >This is >the proof. If h>0, then Tayolr's theorem shows > >f'(x) = [f(x+2h) - f(x)]/(2h) - h*f''(y), > >for some y in (x, x+2h). Hence > >|f'(x)| <= h*M_2 + (M_0)/h, > >and since h is arbitrary, you obtain, choosing >the minimum value of h*M_2 + (M_0)/h over (0,+inf), >(M_1)^2 <= 4*(M_0)*(M_2). |