The Dirac delta function The Fourier Transform Some FT examples exp(iw0 t) cos(w0 t) Some Fourier Transform theorems t w t w Complex conjugate: f*(t) Shift: f(t–a ) Scale: f(at ) Sum: f(t) + g(t) Derivative: f'(t) Modulation: f(t)cos(w0 t) The intensity and phase and the spectral intensity and phase The 2D Fourier Transform t w The relative importance of intensity and phase The Dirac delta function Unlike the Kronecker delta-function, which is a function of two integers, the Dirac delta function is a function of a real variable, t. if t 0 (t ) 0 if t 0 (t) t The Dirac delta function if t 0 (t ) 0 if t 0 It’s best to think of the delta function as the limit of a series of peaked continuous functions. fm(t) = m exp[-(mt)2]/√p (t) f3(t) f2(t) f1(t) t Dirac -function Properties t (t ) dt 1 - - - (t - a) f (t ) dt (t - a) f (a) dt f (a) exp(iwt ) dt 2p (w - exp[i(w - w ')t ] - (t) dt 2p (w - w ' The Fourier Transform and its Inverse F (w ) f (t ) exp( -iw t ) dt - 1 f (t ) 2p - F (w ) exp(iw t ) dw The Fourier Transform of (t) is 1. (t ) exp(-iwt ) dt exp(-iw [0]) 1 - (t) 1 w t And the Fourier Transform of 1 is 2p(w): 1exp(-iwt ) dt 2p (w - 2p(w) 1 t w The Fourier transform of exp(iw0 t) F exp(iw0 t ) exp(iw0 t ) exp( -i w t ) dt - exp(-i [w - w0 ] t ) dt 2p (w - w0 ) - exp(iw0t) Im 0 Re 0 F {exp(iw0t)} t t 0 w0 w The function exp(iw0t) is the essential component of Fourier analysis. It is a pure frequency. The Fourier transform of cos(w0 t) F cos(w0t ) 1 2 1 2 cos(w t ) exp(-i w t ) dt 0 - exp(i w0 t ) exp(-i w0 t ) exp(-i w t ) dt - exp(-i [w - w ]t) dt 0 - 1 2 exp(-i [w w ]t) dt 0 - p (w - w0 ) p (w w0 ) F {cos(w0t )} cos(w0t) 0 t -w0 0 w0 w Scale Theorem F { f (at )} F (w /a) / a The Fourier transform of a scaled function, f(at): Proof: F { f (at )} f (at ) exp(-iw t ) dt - Assuming a > 0, change variables: u = at F { f (at )} f (u) exp(-iw [ u /a]) du / a - f (u) exp(-i [w /a] u) du / a - F (w /a) / a If a < 0, the limits flip when we change variables, introducing a minus sign, hence the absolute value. F(w) f(t) The Scale Theorem in action The shorter the pulse, the broader the spectrum! This is the essence of the Uncertainty Principle! Short pulse t w t w t w Mediumlength pulse Long pulse The Fourier Transform of a sum of two functions F(w) f(t) G(w) g(t) F {af (t ) bg (t )} aF { f (t )} bF {g (t )} t f(t)+g(t) Also, constants factor out. w t t w F(w) + G(w) w Shift Theorem The Fourier transform of a shifted function, f (t - a ) : F f (t - a) exp(-iwa)F (w) Proof : F f t - a f (t - a) exp( -iwt ) dt - Change variables : u t - a f (u ) exp( -iw[u a]) du - exp(-iw a) f (u ) exp( -iwu )du - exp(-iw a) F (w ) Application of the Shift Theorem Suppose we’re measuring the spectrum of E(t), but a small fraction of its irradiance, say e, takes a longer path to the spectrometer. The extra light has the field, √e E(t–a), where a is the extra path. The measured spectrum is: S (w ) F {E (t ) e E (t - a)} 2 e E(t-a) Spectrometer E(t) Performing the Fourier Transform and using the Shift Theorem: E(t) E (w ) e exp(-iw a) E (w ) 2 2 2 E (w ) 1 e exp(-iw a) 2 E (w ) 1 e cos(w a) - i e sin(w a) E (w ) 2 2 1 e cos(w a) e sin(w a) 2 2 Application of the Shift Theorem (cont’d) E (w ) 2 1 2 e cos(w a) e cos2 (w a) e sin 2 (w a) E (w ) 2 1 2 e cos(w a) e Neglecting e compared to √e and 1: E (w ) 2 1 2 e cos(w a) The contaminated spectrum will have ripples with a period of 2p/a. And these ripples will have a surprisingly large amplitude. If e = 1% (a seemingly small amount), these ripples will have an amazingly large amplitude of 2√e = 20%! The Fourier Transform of the complex conjugate of a function F * * f ( t ) F (-w ) Proof: F * f (t ) f * (t ) exp( -iwt )dt - f (t ) exp( iwt )dt - * f (t ) exp( -i[ -w ]t ) dt - F * ( -w ) * Derivative Theorem The Fourier transform of a derivative of a function, f’(t): F { f '(t )} iw F (w ) Proof: F { f '(t )} f '(t ) exp(-iw t ) dt - Integrate by parts: - f (t ) [( -iw ) exp( -iw t )] dt - iw f (t ) exp( -iw t ) dt - iw F (w The Modulation Theorem: The Fourier Transform of E(t) cos(w0 t) F E (t ) cos(w0t ) F 1 2 1 2 E (t ) cos(w0t ) exp( -i w t ) dt - E (t ) exp(i w0t ) exp(-i w0t ) exp(-i w t ) dt - E (t ) exp(-i [w - w0 ] t ) dt - E (t ) cos(w0t ) 1 2 E(t) exp(-i [w w ]t ) dt 0 - 1 E (w - w0 ) 2 F 1 E (w w0 ) 2 E(t)cos(w0t ) If E(t) = D(t), then: -w0 0 w0 w The Fourier transform and its inverse are symmetrical: f(t) F(w) and F(t) f(w) (almost). If f(t) Fourier transforms to F(w), then F(t) Fourier transforms to: F (t ) exp( -i w t ) dt - Rearranging: 1 2p 2p F (t ) exp(i[ - w ] t ) dt - Relabeling the integration variable from t to w’, we can see that we have an inverse Fourier transform: 1 2p 2p - F (w ) exp(i[ - w ]w ) dw 2p f (-w ) This is why it is often said that f and F are a “Fourier Transform Pair.” Fourier Transform Magnitude and Phase As with any complex quantity, we can decompose f(t) and F(w) into their magnitude and phase. f(t) can be written: f(t) = Mag{f(t)} exp[ -i Phase{f(t)}] where Mag{f(t)}2 is often called the intensity, I(t), and Phase{f(t)} is called the phase, f(t). They’re the same quantities we’re used to for light waves. Analogously, F(w) = Mag{F(w)} exp[ -i Phase{F(w)}] The Mag{F(w)}2 is called the spectrum, S(w), and the Phase{F(w)} is called the spectral phase, j(w). Just as both the intensity and phase are required to specify f(t), both the spectrum and spectral phase are required to specify F(w), and hence f(t). Calculating the Intensity and the Phase It’s easy to go back and forth between the electric field and the intensity and phase. The intensity: I(t) = |E(t)|2 The phase: Im[ E (t )] f (t ) - arctan Re[ E (t )] Equivalently, f(t) = -Im{ ln[E(t) ] } Im E(ti) -f(ti) Re Intensity and Phase of a Gaussian The Gaussian is real, so its phase is zero. Time domain: A Gaussian transforms to a Gaussian Frequency domain: So the spectral phase is zero, too. The spectral phase of a time-shifted pulse Recall the Shift Theorem: F Time-shifted Gaussian pulse (with a flat phase): So a time-shift simply adds some linear spectral phase to the pulse! f (t - a) exp(-iw a)F (w) What is the spectral phase anyway? The spectral phase is the abs phase of each frequency in the wave-form. All of these frequencies have zero phase. So this pulse has: w1 w2 j(w) = 0 w3 Note that this wave-form sees constructive interference, and hence peaks, at t = 0. w4 w5 w6 0 t And it has cancellation everywhere else. Now try a linear spectral phase: j(w) = aw. By the Shift Theorem, a linear spectral phase is just a delay in time. And this is what occurs! w1 j(w1) = 0 w2 j(w2) = 0.2 p w3 j(w3) = 0.4 p w4 j(w4) = 0.6 p w5 j(w5) = 0.8 p w6 j(w6) = p t The spectral phase distinguishes a light bulb from an ultrashort pulse. Complex Lorentzian and its Intensity and Phase a Real component 0 0 Imaginary component w Real part Imag part 1 1 a - iw a -w L(w ) 2 2 i 2 a iw a iw a - iw a w a w2 1 Intensity(w ) 2 a w2 Im[ L(w )] Phase(w ) - arctan arctan(w / a) Re[ L(w )] Intensity and Phase of a decaying exponential and its Fourier transform Time domain: (solid) Frequency domain: Light has intensity and phase also. We usually extract out the carrier frequency. A light wave has the time-domain electric field: E (t ) Re I (t ) exp i(w0t - f (t )) Intensity Equivalently, vs. frequency: The minus signs are just conventions. Phase (neglecting the negative-frequency component) E (w ) S (w ) exp -ij (w ) Spectrum Spectral Phase Knowledge of the intensity and phase or the spectrum and spectral phase is sufficient to determine the light wave. Fourier Transform with respect to space If f(x) is a function of position, F (k ) f ( x) exp(-ikx) dx - x F {f(x)} = F(k) We refer to k as the “spatial frequency.” k Everything we’ve said about Fourier transforms between the t and w domains also applies to the x and k domains. The 2D Fourier Transform F = (2){f(x,y)} f(x,y) = F(kx,ky) f(x,y) exp[-i(kxx+kyy)] dx dy If f(x,y) = fx(x) fy(y), then the 2D FT splits into two 1D FT's. But this doesn’t always happen. x F y (2){f(x,y)} A 2D Fourier Transform: a square function f(x,y) Consider a square function in the xy plane: f(x,y) = rect(x) rect(y) The 2D Fourier Transform splits into the product of two 1D Fourier Transforms: F {f(x,y)} = sinc(kx/2) sinc(ky/2) This picture is an optical determination of the Fourier Transform of the square function! x F y (2){f(x,y)} Fourier Transform Magnitude and Phase Rick Linda Pictures reconstructed using the spectral phase of the other picture Mag{F [Linda]} Phase{F [Rick]} Mag{F [Rick]} Phase{F {Linda]} The phase of the Fourier transform (spectral phase) is much more important than the magnitude in reconstructing an image.