Math 2280-001 Wed Apr 22 9.1-9.3 Fourier Series. (On Friday we'll revisit forced oscillations, section 9.4, and explain the results we obtained playing "the resonance game" with convolution integrals last Friday.) , Finish Monday's notes. The key points to recall from Monday are that Fourier series are a way of expressing piecewise continuous functions f on the interval Kp, p (or equivalently, a 2 pKperiodic functions f), as infinite sum of trigonometric functions. If f has Fourier series N N a0 fw C an cos n t C bn sin n t 2 n=1 n=1 Then the partial sums > fN t = a0 2 > N C N > a cos nt C > b sin nt n=1 n n=1 n with 1 a0 d p p bn d f, sin n t VN = span 2 Kp an d f, cos n t are the projections of f onto a0 f t dt 1 = p = 1 p =! f, 1 2 O 1 2 p f t cos n t dt, n 2 ; Kp p f t sin n t dt, n 2 ; Kp 1 , cos t , cos 2 t , ..., cos N t , sin t , sin 2 t , ... sin N t . 2 These partial sums converge to f as N/N in the ways described in Monday's notes. , After finishing with Monday's notes, continue with today's... Exercise 1 If a function ("vector") is already in a subspace, then projection onto that subspace leaves the function fixed. Use that fact to very quickly compute the 2 pKperiodic Fourier series for a) f t = sin 5 t K 8 cos 10 t b) g t = cos2 3 t Fourier series for 2 LKperiodic functions: Theorem: Consider the vector space of piecewise continuous, 2 LKperiodic functions. Then the inner product 1 g, h d L L g u h u du KL makes p 2p kp p 2p kp u , cos u ,... cos u ,... sin u , sin u , ...sin u .... L L L L L L 2 into an orthonormal collection of functions. p L proof: The substitution u = t, equivalently u = t converts between 2 LKperiodic functions L p L L g u , h u and 2 pKperiodic functions g t ,h t . Also (verify this!!!) p p 1 , cos 1 L L KL p 1 g u h u du = p g Kp L t h p L t dt . p A Thus the Fourier series for a 2 LKperiodic function f is defined by N N a0 p p fw C an cos n u C bn sin n u 2 L L n=1 n=1 with > > 1 a0 = L an d bn d f, cos n f, sin n p u L = p u L = 1 L 1 p L g u du KL L f u cos n KL p f u sin Kp p u du, n 2 ; L p u du, n 2 ; L (and then we usually use the dummy variable t rather than u). As a result, Fourier series for 2 LKperiodic functions along with convergence theorems, are "equivalent" to ones for 2 pKperiodic ones, via this isometry of the two vector spaces. Exercise 2) a) Use the Fourier series for tent t : 3 2 1 K3 p K2 p 0 Kp p 2p 3p t p 4 1 K cos n t 2 p n odd n2 to deduce the Fourier series for the related function f u with period 2 that has graph > tent w 1 0.6 0.2 K3 K2 K1 0 1 2 3 t solution: f w 1 4 K 2 2 p > n odd 1 n2 cos n p u We talked about differentiating Fourier series term by term in Monday's notes. There is also: Theorem If f is piece-wise continuous, 2 LKperiodic, with Fourier series N N a0 p p f t w C an cos n t C bn sin n t 2 L L n=1 n=1 then the antiderivative may be computed by term by term antidifferentiation, and the corresponding series will converge for each t: > t f s ds = 0 a0 2 N tC >a n=1 n L np > N p sin n t C bn L n=1 L np > Kcos n p t C1 . L Exercise 3) (This is the first part of your homework exercise 9.3.19) Start with N t = saw t w 2 K1 > nC1 sin n t n and integrate to get the 2 pKperiodic function that on Kp, p is given by n=1 N 2 t2 p K1 n C 1 = C2 cos n t . 2 6 n=1 n2 a0 Hint: The value of the constant term is easiest to compute as . If you compare to the definite integral 2 formula in the Theorem you will reproduce one of the "magic" series. > 4 1 K3 p K2 p Kp 0 p t 2p 3p In your homework you will antidifferentiate twice more to get a formula for the periodic extension of t4 g t = . 24