Propagation of a Gaussian wave packet
February 15, 2016
We study the time evolution of an initially Gaussian pulse.
1 Gaussian integrals
Before starting our example, we show how to find the integral of a Gaussian curve,
∞ 2 2 I = dx e−(x−x0) /2σ ˆ −∞ First, change the integration variable to x − x ξ = √ 0 2σ so that ∞ √ 2 I = 2σ2 dξ e−ξ ˆ −∞ Now square I,
∞ ∞ 2 2 −ξ2 −ξ2 I = 2σ dξ e 1 dξ e 2 ˆ 1 ˆ 2 −∞ −∞ ∞ ∞ − ξ2+ξ2 = 2σ2 dξ dξ e ( 1 2 ) ˆ 1 ˆ 2 −∞ −∞
2 2 2 and change to polar coordinates, where ρ = ξ1 + ξ2 and dξ1dξ2 = ρdρdϕ. Then
2π ∞ 2 I2 = 2σ2 dϕ ρdρ e−ρ ˆ ˆ 0 0 ∞ 2 = 4πσ2 ρdρ e−ρ ˆ 0
2 ∞ −λ and another change of variable to λ = ρ gives a simple exponential, with 0 dλ e = 1, ´ ∞ I2 = 2πσ2 dλ e−λ ˆ 0 = 2πσ2
1 Therefore, ∞ 2 2 √ dx e−(x−x0) /2σ = 2πσ2 ˆ −∞
2 Initial conditions and the mode amplitude
We begin with an initially Gaussian pulse
2 2 u (x, 0) = e−x /2L eik0x
with zero initial rate of change, ∂u (x, 0) = 0 ∂t We use these initial data to find A (k), then use A (k) to find u (x, t). The solution for A (k) is found by integrating over the initial conditions
∞ 1 i ∂u A (k) = √ dx u (x, 0) − (x, 0) e−ikx 2π ˆ ω (k) ∂t −∞
∂u Substituting for u (x, 0) and ∂t (x, 0) gives a Gaussian integral,
∞ 1 2 2 A (k) = √ dx e−x /2L eik0xe−ikx 2π ˆ −∞
For the integral of a Gaussian, see the final section. integral, we add and subtract a constant to complete the square in the exponent,
1 1 h 2 2i − x2 − 2iL2 (k − k ) x = − x2 − 2iL2 (k − k ) x + iL2 (k − k ) − iL2 (k − k ) 2L2 0 2L2 0 0 0 1 2 1 = − x − iL2 (k − k ) − L2 (k − k )2 2L2 0 2 0 Then, performing the Gaussian integral,
∞ ∞ 2 2 1 2 2 1 2 2 −x /2L −ikx ik0x − (x−iL (k−k0)) − L (k−k0) dxe e e = dx e 2L2 2 ˆ ˆ −∞ −∞ ∞ 1 2 2 1 2 2 − L (k−k0) − (x−iL (k−k0)) = e 2 dx e 2L2 ˆ −∞ √ − 1 L2(k−k )2 = 2πL2e 2 0
Therefore, the mode amplitudes follow a Gaussian as well,
1 √ 1 2 2 [iL (k−k0)] A (k) = √ 2πLe 2L2 2π 2 − L (k−k )2 = Le 2 0
2 3 Dispersion relation
The time dependence of the wave packet is now given by
∞ 1 1 h i u (x, t) = √ dk A (k) ei(kx−ωt) + A∗ (k) e−i(kx−ωt) 2 2π ˆ −∞ ∞ 2 1 L − L (k−k )2 i(kx−ωt) −i(kx−ωt) = √ dke 2 0 e + e 2 2π ˆ −∞ ∞ 2 L − L (k−k )2 i(kx−ωt) = √ Re dke 2 0 e 2π ˆ −∞ where ω = ω (k). Therefor, to continue, we need the form of of the dispersion relation, ω (k). Consider the case of high- frequency waves in plasma, for which we found q 2 2 2 ω = k c + ωp
Now we let ωP kc so that we may expand the square root,
2 c 2 ω ≈ ωP 1 + 2 k 2ωP 1 ≈ ω 1 + l2 k2 P 2 P
c where we define the plasma wavelength, lP = . This gives our dispersion relation. The group velocity is ωP dω v = = ω l2 k g dk P P so the group velocity evaluated at the center of the packet is
2 v0 ≡ ωP lP k0 and we may write 1 ω ≈ ω + kv P 2 g Notice that the approximation means that v 1 g = l2 ω k c 2c P P 1 = kc 2ωP 1
so the group velocity is much less than the speed of light. Notice that this form also works for a Klein-Gordon matter wave when the momentum due to mass is much greater than the wave momentum, mc ~k. Expanding, r m2c4 ω = + k2c2 ~2
3 r mc2 2k2c2 = 1 + ~ ~ m2c4 mc2 1 2 ≈ 1 + ~ k2 ~ 2 m2c2 1 = ω 1 + λ2k2 C 2 1 = ω + kv C 2 g
c where λC = ~ is the Compton wavelength and ωC = the corresponding frequency. We continue below mc λC with a wave packet in plasma.
4 Time evolution of the waveform
1 With this dispersion relation, ω = ωP + 2 kvg, the full time evolution of the wave is given by
∞ 2 L − L (k−k )2 i kx−ω + 1 kv t u (x, t) = √ Re dke 2 0 e ( P 2 g ) 2π ˆ −∞ Substititing for ω we have another Gaussian integral,
∞ 2 L − L (k−k )2 i kx− ω 1+ 1 l2 k2 t u (x, t) = √ Re dk e 2 0 e ( ( P ( 2 P )) ) 2π ˆ −∞ ∞ 2 L L 2 i 2 2 = √ Re dk exp − (k − k0) + ikx − iωP t − ωP l k t 2π ˆ 2 2 P −∞ Expanding the exponent,
L2 i L2 2 2 i 2 − (k − k )2 + ikx − iω t − ω l2 k2t = − (k − k )2 − ikx + iω t + ω l2 k2t 2 0 P 2 P P 2 0 L2 L2 P 2 L2 P P L2 2ix 2 iω l2 t = − k2 − 2kk + k2 − k + iω t + P P k2 2 0 0 L2 L2 P L2 L2 iω l2 t ix L2 2 = − 1 + P P k2 − 2 k + k − k2 + iω t 2 L2 0 L2 2 0 L2 P we complete the square, 2 2 2 r 2 ix ix iωP lP t 2 ix iωP lP t k0 + L2 k0 + L2 1 + k − 2 k0 + k = k 1 + − − 2 2 2 q 2 q 2 L L L iωP lP t iωP lP t 1 + L2 1 + L2 2 2 ix ! ix 2 iωP lP t k0 + L2 k0 + L2 = 1 + 2 k − 2 − 2 L iωP lP t iωP lP t 1 + L2 1 + L2 Two constant exponentials,
" 2 2 ix 2 # " 2 2 # L 2 2 L k0 + L2 1 2 2 1 k0L + ix exp − k0 + iωP t + 2 = exp − k0L + 2iωP t + 2 iωP l t 2 2 2 L 2 P 2 2 L + iωP lP t 1 + L2
4 come out of the integral, while the remaining Gaussian integral gives
∞ 2 ix ! √ 1 2 2 k0 + L2 2π dk exp − L + iωP lP t k − 2 = iωP l t p 2 2 ˆ 2 1 + P L + iωP l t −∞ L2 P
Combining these,
" 2 # L 1 1 k L2 + ix u (x, t) = Re√ exp − k2L2 + 2iω t + 0 2 0 P 2 L + ilP ct 2 2 L + ilP ct where we have used lP ωP = c. We easily check that at t = 0 we recover the initial state,
" 2 # 1 1 k L2 + ix u (x, 0) = Re exp − k2L2 + 0 2 0 2 L2 " 2 # 1 1 k L2 + ix = Re exp − k2L2 + 0 2 0 2 L2
2 2 = e−x /2L eik0x
5 Finding the real part
Finally, we find the real part of the waveform. Separating the real and imaginary parts of the argument of the exponential,