Physics · Book 4 · Bachelor Year 2

University Physics — Year 2

University Physics — Year 2 · Bachelor Year 2

16Guided Waves and Cavities

A radar’s pulse travels from the transmitter to the dish inside a rectangular copper pipe; the light of an internet link travels ten thousand kilometres inside a glass thread thinner than a hair; the microwaves of an oven bounce between its walls and settle into a pattern of hot and cold spots. A wave confined by conducting or refracting walls is no longer a plane wave free to go anywhere: it must satisfy the boundary conditions on the walls, and that selects the shapes it may take — the modes — and forbids it below a cut-off frequency. This chapter builds the simplest guide, two parallel conducting plates, from the reflections of Chapter 15; then the rectangular waveguide, the closed cavity, and, in the ray picture, the optical fibre.

16.1 Waves between two conducting plates

Proposition 16.1 (Modes of the parallel-plate guide)

Two perfectly conducting planes y=0y = 0 and y=ay = a bound a vacuum. Seek a wave travelling along zz with E=E(y)ei(ωtkgz)ex\vect E = E(y)\,\eu^{\iu(\omega t - k_gz)} \vect e_x (the field parallel to the plates, perpendicular to the propagation). Maxwell’s equations in vacuum and the condition E=0E = 0 on the plates impose

E(y)=E0sinnπya,kg2=ω2c2(nπa)2,n=1,2,E(y) = E_0\sin\frac{n\pi y}a , \qquad k_g^2 = \frac{\omega^2}{c^2} - \Bigl(\frac{n\pi}a\Bigr)^2 , \qquad n = 1, 2, \dots

Mode nn propagates only above its cut-off frequency fn=nc/2af_n = nc/2a; below, kgk_g is imaginary and the field decays along the guide. Its phase and group velocities are

vφ=c1fn2/f2>c,vg=c1fn2/f2<c,vφvg=c2,v_\varphi = \frac c{\sqrt{1 - f_n^2/f^2}} > c , \qquad v_g = c\sqrt{1 - f_n^2/f^2} < c , \qquad v_\varphi v_g = c^2 ,

and its magnetic field has components along yy and along zz: a guided wave is not transverse in the direction of propagation.

Proof. Insert in ΔE=t2E/c2\Delta\vect E = \partial_t^2\vect E/c^2: Ekg2E=(ω2/c2)EE'' - k_g^2E = -(\omega^2/c^2)E, i.e. E+(ω2/c2kg2)E=0E'' + (\omega^2/c^2 - k_g^2)E = 0, whose solutions vanishing at y=0y = 0 and y=ay = a are the sines with ω2/c2kg2=(nπ/a)2\omega^2/c^2 - k_g^2 = (n\pi/a)^2; divE=0\operatorname{div} \vect E = 0 holds since ExE_x does not depend on xx. Faraday gives By=(kg/ω)E\underline B_y = (k_g/\omega)\underline E and Bz=(i/ω)yE\underline B_z = (\iu/\omega)\partial_y\underline E, the latter in quadrature. The velocities follow from the Klein–Gordon form of the dispersion relation (Chapter 8).

Remark 16.2 (The mode as two plane waves)

sin(nπy/a)eikgz\sin(n\pi y/a)\eu^{-\iu k_gz} is the sum of two plane waves with wavevectors (0,±nπ/a,kg)(0, \pm n\pi/a, k_g), of modulus ω/c\omega/c, bouncing between the plates at the angle θ\theta from the axis with cosθ=kgc/ω\cos\theta = k_gc/\omega: a guided mode is a plane wave zigzagging between the walls, reflected with r=1r = -1 at each one, the transverse standing wave being the condition that it interferes constructively with itself. The energy travels along the zigzag at cc, hence along the axis at ccosθ=vgc\cos\theta = v_g; the crests’ intersections with the axis run at c/cosθ=vφc/\cos\theta = v_\varphi. At cut-off the wave bounces back and forth across the guide and goes nowhere.

Left: a guided mode is a plane wave bouncing between the plates; the angle  opens as the frequency approaches cut-off. Right: the transverse profiles E(y) of the first two modes, vanishing on the conducting walls.
Left: a guided mode is a plane wave bouncing between the plates; the angle θ\theta opens as the frequency approaches cut-off. Right: the transverse profiles E(y)E(y) of the first two modes, vanishing on the conducting walls.
Left: the dispersion curves of the first three modes of a guide, each a hyperbola starting at its cut-off. Right: phase and group velocities of a mode against frequency; the energy crawls near cut-off and approaches c far above. Left: the dispersion curves of the first three modes of a guide, each a hyperbola starting at its cut-off. Right: phase and group velocities of a mode against frequency; the energy crawls near cut-off and approaches c far above.
Left: the dispersion curves of the first three modes of a guide, each a hyperbola starting at its cut-off. Right: phase and group velocities of a mode against frequency; the energy crawls near cut-off and approaches cc far above.

Example 16.3 (Numbers)

Plates 3cm3\,\mathrm{cm} apart: f1=5GHzf_1 = 5\,\mathrm{GHz}, f2=10GHzf_2 = 10\,\mathrm{GHz}; between those frequencies only one mode propagates (single-mode operation, which keeps a pulse from splitting into several with different vgv_g). At 8GHz8\,\mathrm{GHz} the mode n=1n = 1 has vg=0.78cv_g = 0.78c, vφ=1.28cv_\varphi = 1.28c, and the guide wavelength λg=2π/kg=4.8cm\lambda_g = 2\pi/k_g = 4.8\,\mathrm{cm} instead of 3.75cm3.75\,\mathrm{cm} in free space. At 4GHz4\,\mathrm{GHz} nothing passes: the field decays as ez/\eu^{-z/\ell} with =c/2πf12f2=1.6cm\ell = c/2\pi\sqrt{f_1^2 - f^2} = 1.6\,\mathrm{cm} — the principle of the mesh in an oven door, of the waveguide below cut-off used as a calibrated attenuator, and of the metal ducts of ventilation that let air but no radio through.

16.2 The rectangular waveguide and the cavity

Proposition 16.4 (The rectangular guide; the TE10_{10} mode)

A hollow metal pipe of rectangular section a×ba \times b (a>ba > b) carries modes whose cut-off frequencies are

fmn=c2(ma)2+(nb)2;f_{mn} = \frac c2\sqrt{\Bigl(\frac ma\Bigr)^2 + \Bigl(\frac nb\Bigr)^2} ;

the lowest, the TE10_{10} mode (f10=c/2af_{10} = c/2a), has E=E0sin(πx/a)ei(ωtkgz)ey\vect E = E_0\sin (\pi x/a)\,\eu^{\iu(\omega t - k_gz)}\,\vect e_y: one half-sine across the wide side, uniform across the narrow side, vanishing on the two walls x=0x = 0, x=ax = a — exactly the parallel-plate mode, the side walls y=0y = 0, bb being perpendicular to E\vect E and imposing nothing on it. The guide is used between f10f_{10} and the next cut-off; its magnetic field lines close in the xzxz plane and the currents they induce in the walls are what a slot must not cut. The power carried is 14ε0cE02ab1f102/f2\tfrac14\varepsilon_0cE_0^2\,ab \sqrt{1 - f_{10}^2/f^2}.

Proof. The general mode is cos\cos/sin\sin products in xx and yy with the boundary conditions on the four walls, giving kg2=ω2/c2(mπ/a)2(nπ/b)2k_g^2 = \omega^2/c^2 - (m\pi/a)^2 - (n\pi/b)^2 (admitted in general; the TE10_{10} case is the proposition above). Power: the mean Poynting vector along zz is E02sin2(πx/a)kg/2μ0ωE_0^2\sin^2(\pi x/a) \,k_g/2\mu_0\omega integrated over the section.

The TE_10 mode of a rectangular guide: the electric field spans the narrow dimension with a half-sine profile across the wide one; the magnetic field closes in loops in the plane of the wide faces, a half guide wavelength long.
The TE10_{10} mode of a rectangular guide: the electric field spans the narrow dimension with a half-sine profile across the wide one; the magnetic field closes in loops in the plane of the wide faces, a half guide wavelength long.

Proposition 16.5 (Cavity resonator)

Closing a guide with two conducting walls a length dd apart turns it into a cavity, in which the guided wave must form a standing wave along zz too: kg=pπ/dk_g = p\pi/d, and the resonance frequencies of a rectangular box a×b×da \times b \times d are

fmnp=c2(ma)2+(nb)2+(pd)2.f_{mnp} = \frac c2\sqrt{\Bigl(\frac ma\Bigr)^2 + \Bigl(\frac nb\Bigr)^2 + \Bigl(\frac pd\Bigr)^2} .

The number of modes with frequency below ff grows as 8πVf3/3c38\pi V f^3/3c^3 for a box of volume VV (admitted: count the points of the lattice (m/2a,n/2b,p/2d)(m/2a, n/2b, p/2d) in an eighth of a sphere, with two polarizations) — the density of modes that the theory of thermal radiation will need (Chapter 26). Each mode has a quality factor QQ set by the wall losses (the surface resistance of Chapter 14), typically 10310^310410^4 for copper at microwave frequencies, 101010^{10} for a superconducting cavity.

Proof. Standing waves along the three directions with nodes on the six walls; the wavevector (mπ/a,nπ/b,pπ/d)(m\pi/a, n\pi/b, p\pi/d) has modulus ω/c\omega/c.

Example 16.6 (Ovens, radars, clocks)

A microwave oven of 30×30×2030 \times 30 \times 20 cm3^3 near 2.45GHz2.45\,\mathrm{GHz} has dozens of modes within a few percent of the magnetron’s frequency: the field is a superposition of standing waves, with maxima 6cm6\,\mathrm{cm} apart, which the turntable (and a "mode stirrer") smooth out. A radar’s magnetron is itself a cavity resonator; a particle accelerator is a chain of superconducting cavities whose TM mode pushes the bunches; the atoms of a caesium clock cross a microwave cavity tuned to 9.19GHz9.19\,\mathrm{GHz}.

16.3 The optical fibre in the ray picture

Proposition 16.7 (Step-index fibre)

A glass core of index n1n_1 surrounded by a cladding of slightly lower index n2n_2 guides light by total internal reflection at the core boundary. A ray entering the end face from air at the angle α\alpha from the axis is trapped if sinα<n12n22\sin\alpha < \sqrt{n_1^2 - n_2^2}, the numerical aperture; rays of different angles travel different lengths, and a pulse entering a fibre of length LL is spread by

Δt=n1Lc(n1n21),\Delta t = \frac{n_1L}c\Bigl(\frac{n_1}{n_2} - 1\Bigr) ,

the modal dispersion — some 50ns50\,\mathrm{ns} per kilometre for n1n2=0.01n_1 - n_2 = 0.01, which limits a multimode fibre to a few megabits per second over ten kilometres. A single-mode fibre has a core so thin (9µm9\,\text{µ}\mathrm{m}) that only one mode propagates, like the guide between its first two cut-offs; what remains is the chromatic dispersion of the glass (Chapter 8).

Proof. Total reflection at the core–cladding interface needs sinθ>n2/n1\sin\theta > n_2/n_1 for the angle θ\theta from the normal to the boundary, i.e. cosθ>n2/n1\cos \theta' > n_2/n_1 for the angle θ\theta' from the axis inside; refraction at the entrance gives sinα=n1sinθ\sin\alpha = n_1\sin\theta', whence sinα<n11n22/n12\sin\alpha < n_1\sqrt{1 - n_2^2/n_1^2}. The steepest trapped ray travels L/cosθ=Ln1/n2L/\cos\theta' = Ln_1/n_2 against LL for the axial one, at c/n1c/n_1.

A step-index fibre in the ray picture: rays within the acceptance cone are trapped by total internal reflection; the zigzag ray arrives after the axial one — modal dispersion.
A step-index fibre in the ray picture: rays within the acceptance cone are trapped by total internal reflection; the zigzag ray arrives after the axial one — modal dispersion.

Method 16.8 (Guided-wave bookkeeping)

(1) Identify the walls and their condition (tangential E=0\vect E = 0 on a conductor; total reflection for a dielectric). (2) Write the field as a transverse standing wave times a progressive factor along the guide; the boundary conditions quantize the transverse wavenumber. (3) The dispersion relation kg2=ω2/c2k2k_g^2 = \omega^2/c^2 - k_\perp^2 gives the cut-off, vφv_\varphi, vgv_g, λg\lambda_g. (4) Count the propagating modes at the working frequency; aim for one. (5) For a cavity add the longitudinal condition and find the resonance frequencies; for the losses use the surface resistance.

16.4 Exercises

Exercise 16.1

Plates 2.0cm2.0\,\mathrm{cm} apart: cut-off frequencies of the first three modes; at 12GHz12\,\mathrm{GHz}, which modes propagate, and for the first one kgk_g, λg\lambda_g, vφv_\varphi, vgv_g; at 6GHz6\,\mathrm{GHz}, the decay length of the field.

Solution

Solution of Exercise 16.1.

fn=nc/2a=7.5f_n = nc/2a = 7.5, 1515, 22.5GHz22.5\,\mathrm{GHz}. At 12GHz12\,\mathrm{GHz} only n=1n = 1: kg=(2π/c)f2f12=196rad/mk_g = (2\pi/c)\sqrt{f^2 - f_1^2} = 196\,\mathrm{rad}/\mathrm{m}, λg=3.2cm\lambda_g = 3.2\,\mathrm{cm}, vφ=c/0.78=3.8×108m/sv_\varphi = c/0.78 = 3.8 \times 10^{8}\,\mathrm{m}/\mathrm{s}, vg=2.3×108m/sv_g = 2.3 \times 10^{8}\,\mathrm{m}/\mathrm{s}. At 6GHz6\,\mathrm{GHz}: =c/2πf12f2=1.1cm\ell = c/2\pi\sqrt{f_1^2 - f^2} = 1.1\,\mathrm{cm}.

Exercise 16.2

A standard X-band waveguide has a=22.9mma = 22.9\,\mathrm{mm}, b=10.2mmb = 10.2\,\mathrm{mm}. Cut-off of TE10_{10}, of TE20_{20} and TE01_{01}; the single-mode band; at 10GHz10\,\mathrm{GHz}: λg\lambda_g, vgv_g, the angle of the zigzag. Why is ba/2b \approx a/2 a good choice?

Solution

Solution of Exercise 16.2.

TE10_{10} 6.55GHz6.55\,\mathrm{GHz}, TE20_{20} 13.1GHz13.1\,\mathrm{GHz}, TE01_{01} 14.7GHz14.7\,\mathrm{GHz}: single mode from 6.556.55 to 13.1GHz13.1\,\mathrm{GHz}. At 10GHz10\,\mathrm{GHz}: 1(6.55/10)2=0.756\sqrt{1 - (6.55/10)^2} = 0.756, λg=4.0cm\lambda_g = 4.0\,\mathrm{cm}, vg=0.76cv_g = 0.76c, cosθ=0.756\cos\theta = 0.756, θ=41\theta = 41{}^{\circ}. ba/2b \approx a/2 keeps TE01_{01} above TE20_{20} (the widest single-mode band) while leaving the largest gap for the field (power before breakdown).

Exercise 16.3

A cavity of 30×30×2030 \times 30 \times 20 cm3^3: frequencies of the modes (1,1,0)(1,1,0), (1,0,1)(1,0,1), (2,1,1)(2,1,1), (3,2,1)(3,2,1); number of modes below 2.5GHz2.5\,\mathrm{GHz} from the counting formula; the lowest mode of a 9.19GHz9.19\,\mathrm{GHz} cubic cavity — its side.

Solution

Solution of Exercise 16.3.

f=1.5×108(m/0.3)2+(n/0.3)2+(p/0.2)2f = 1.5 \times 10^8\sqrt{(m/0.3)^2 + (n/0.3)^2 + (p/0.2)^2}: 0.71GHz0.71\,\mathrm{GHz}, 0.90GHz0.90\,\mathrm{GHz}, 1.35GHz1.35\,\mathrm{GHz}, 1.95GHz1.95\,\mathrm{GHz}. N=8πVf3/3c3=8π×0.018×1.56×1028/8.1×1025260N = 8\pi Vf^3/3c^3 = 8\pi \times 0.018 \times 1.56 \times 10^{28}/ 8.1 \times 10^{25} \approx 260. Cube: f=c2/2Lf = c\sqrt2/2L, L=2.3cmL = 2.3\,\mathrm{cm}.

Exercise 16.4

A fibre has n1=1.48n_1 = 1.48, n2=1.46n_2 = 1.46. Numerical aperture and acceptance half-angle; modal spread per kilometre; maximum bit rate over 2km2\,\mathrm{km} if one bit must last longer than the spread; the same with n1n2=0.003n_1 - n_2 = 0.003.

Solution

Solution of Exercise 16.4.

NA=1.4821.462=0.24\mathrm{NA} = \sqrt{1.48^2 - 1.46^2} = 0.24, 1414{}^{\circ}; Δt=(n1L/c)(n1/n21)=68ns/km\Delta t = (n_1L/c)(n_1/n_2 - 1) = 68\,\mathrm{ns}/\mathrm{km}; 135ns135\,\mathrm{ns} over 2km2\,\mathrm{km}: 7Mbit/s7\,\mathrm{Mbit}/\mathrm{s}; with 0.0030.003: 10ns/km10\,\mathrm{ns}/\mathrm{km}, 50Mbit/s50\,\mathrm{Mbit}/\mathrm{s}.

Exercise 16.5 ★★

For the parallel-plate mode n=1n = 1: (a) find B\vect B from Faraday’s law and show it has a zz component in quadrature with EE. (b) Surface current on each plate (boundary relation). (c) Mean Poynting vector along zz and the power per unit width; check that its transverse component averages to zero. (d) Show that the energy velocity, power over mean energy per unit length, equals vgv_g.

Solution

Solution of Exercise 16.5.

(a) By=(kg/ω)E\underline B_y = (k_g/\omega)\underline E, Bz=(i/ω)yE=(iπ/aω)E0cos(πy/a)ei(ωtkgz)\underline B_z = (\iu/\omega)\partial_y\underline E = (\iu\pi/a\omega) E_0\cos(\pi y/a)\eu^{\iu(\omega t - k_gz)}: in quadrature. (b) At y=0y = 0, By=0B_y = 0 and js=eyB/μ0=(Bz/μ0)ex\vect j_s = \vect e_y\wedge\vect B/\mu_0 = (B_z/\mu_0)\vect e_x: along the field. (c) Πz=E02kgsin2(πy/a)/2μ0ω\langle\Pi_z\rangle = E_0^2k_g\sin^2(\pi y/a)/2\mu_0\omega, power per unit width E02kga/4μ0ωE_0^2k_ga/4\mu_0\omega; ΠyRe(EiBz)=0\langle\Pi_y\rangle \propto \operatorname{Re}(\underline E\,\iu\underline B_z^*) = 0. (d) Mean energy per unit length and width 14ε0E02a\tfrac14\varepsilon_0E_0^2a (electric and magnetic halves each 18ε0E02a\tfrac18\varepsilon_0E_0^2a); the ratio is c2kg/ω=vgc^2k_g/\omega = v_g.

Exercise 16.6 ★★

Below cut-off. (a) A 1mm1\,\mathrm{mm} hole in the 0.5mm0.5\,\mathrm{mm} mesh of an oven door at 2.45GHz2.45\,\mathrm{GHz}: treat it as a guide of width a=1mma = 1\,\mathrm{mm}; decay length and attenuation across the sheet in dB (compare Chapter 14). (b) A ventilation duct of 20cm20\,\mathrm{cm} section in a shielded room: up to what frequency does it block radio, and by how much does it attenuate 100MHz100\,\mathrm{MHz} over 1m1\,\mathrm{m}? (c) A "waveguide-beyond-cutoff attenuator" uses a tube of 1cm1\,\mathrm{cm} diameter at 1GHz1\,\mathrm{GHz}: attenuation per centimetre (use the plate formula with a=1cma = 1\,\mathrm{cm} as an estimate). (d) Why is the attenuation independent of the wall conductivity?

Solution

Solution of Exercise 16.6.

(a) f1=150GHzff_1 = 150\,\mathrm{GHz} \gg f: a/π=0.32mm\ell \approx a/\pi = 0.32\,\mathrm{mm}; e0.5/0.32\eu^{-0.5/0.32}: 14dB-14\,\mathrm{dB} in amplitude, 27-27 in power. (b) f1=750MHzf_1 = 750\,\mathrm{MHz}; at 100MHz100\,\mathrm{MHz} =6.4cm\ell = 6.4\,\mathrm{cm}: 136dB-136\,\mathrm{dB} over a metre. (c) f1=15GHzf_1 = 15\,\mathrm{GHz}, =3.2mm\ell = 3.2\,\mathrm{mm}: 27dB/cm27\,\mathrm{dB}/\mathrm{cm}. (d) The wave does not fit; evanescence is geometry, not dissipation.

Exercise 16.7 ★★

Power and breakdown. The X-band guide of Exercise 16.2 carries TE10_{10} at 10GHz10\,\mathrm{GHz}. (a) Power for E0=1MV/mE_0 = 1\,\mathrm{MV}/\mathrm{m}. (b) Air breaks down at 3MV/m3\,\mathrm{MV}/\mathrm{m}: maximum power; how do radars carry megawatts (pressurization, gases)? (c) Wall losses: the surface current at the broad wall is of order E0/μ0cE_0/\mu_0c; with Rs=0.026ΩR_s = 0.026\,\Omega for copper at 10GHz10\,\mathrm{GHz}, estimate the loss per metre and the attenuation in dB/m. (d) Compare with a coaxial cable’s 0.5dB/m0.5\,\mathrm{dB}/\mathrm{m}: why are guides used at these frequencies?

Solution

Solution of Exercise 16.7.

(a) P=14ε0cE02ab1f102/f2=120kWP = \tfrac14\varepsilon_0cE_0^2ab\sqrt{1 - f_{10}^2/f^2} = 120\,\mathrm{kW}. (b) ×9\times9: 1MW1\,\mathrm{MW}; pressurized nitrogen or SF6_6 raise the breakdown field. (c) jsE0/μ0c=2.7kA/mj_s \approx E_0/\mu_0c = 2.7\,\mathrm{kA}/\mathrm{m}; 12Rsjs2=9×104W/m2\tfrac12R_sj_s^2 = 9 \times 10^{4}\,\mathrm{W}/\mathrm{m}^{2} over 2a=0.046m22a = 0.046\,\mathrm{m}^{2} per metre: 4kW/m4\,\mathrm{kW}/\mathrm{m}, 3.5%3.5\% per metre, about 0.15dB/m0.15\,\mathrm{dB}/\mathrm{m}. (d) Three times less than the cable, and megawatts instead of kilowatts.

Exercise 16.8 ★★

Cavity quality. A copper cubic cavity of side 10cm10\,\mathrm{cm} in its lowest mode. (a) Frequency. (b) The stored energy is 12ε0E02V/4\tfrac12\varepsilon_0E_0^2V/4 (admitted) and the wall loss 12RsH02×\tfrac12R_sH_0^2\times area with H0E0/μ0cH_0 \approx E_0/ \mu_0c and Rs=1/γδR_s = 1/\gamma\delta: quality factor Q=ωW/PQ = \omega W/P — compute it. (c) Bandwidth f/Qf/Q of the resonance; ring-down time Q/ωQ/\omega. (d) A superconducting cavity has Rs1×108ΩR_s \sim 1 \times 10^{-8}\,\Omega: QQ, and why accelerators use them.

Solution

Solution of Exercise 16.8.

(a) f=c2/2L=2.12GHzf = c\sqrt2/2L = 2.12\,\mathrm{GHz}. (b) δ=1.4µm\delta = 1.4\,\text{µ}\mathrm{m}, Rs=12mΩR_s = 12\,\mathrm{m}\Omega; W=ε0E02V/8=1.1×1015E02W = \varepsilon_0E_0^2V/8 = 1.1 \times 10^{-15}\,E_0^2; P=12Rs(E0/377)2×6L2=2.5×109E02P = \tfrac12R_s(E_0/377)^2 \times 6L^2 = 2.5 \times 10^{-9}\,E_0^2: Q=ωW/P6000Q = \omega W/P \approx 6000. (c) 360kHz360\,\mathrm{kHz}; 0.45µs0.45\,\text{µ}\mathrm{s}. (d) Q1010Q \sim 10^{10}: the field of tens of megavolts per metre is sustained by kilowatts instead of gigawatts.

Exercise 16.9 ★★

Mode counting. (a) Show that the number of modes of a box of volume VV with frequency below ff is N(f)=8πVf3/3c3N(f) = 8\pi Vf^3/3c^3 (points (m,n,p)(m, n, p) in the eighth of a sphere of radius 2fL/c2fL/c for a cube, two polarizations). (b) The number of modes per unit volume and frequency,  ⁣dN/V ⁣df=8πf2/c3\dd N/V\dd f = 8\pi f^2/c^3. (c) A 1m31\,\mathrm{m}^{3} room: modes per hertz at 1GHz1\,\mathrm{GHz}; at 500THz500\,\mathrm{THz} (visible light). (d) Why does a small cavity have sparse modes while a big room has a continuum — and what is the oven’s case at 2.45GHz2.45\,\mathrm{GHz}?

Solution

Solution of Exercise 16.9.

(a) Points (m,n,p)(m, n, p) with m2+n2+p2(2fL/c)2m^2 + n^2 + p^2 \le (2fL/c)^2 fill an eighth of a sphere: 1843π(2fL/c)3\tfrac18\cdot\tfrac43\pi(2fL/c)^3, times two polarizations: 8πL3f3/3c38\pi L^3f^3/3c^3. (b) 8πf2/c38\pi f^2/c^3. (c) 9×1079 \times 10^{-7} per hertz (one per megahertz) at 1GHz1\,\mathrm{GHz}; 2×1052 \times 10^5 per hertz at 500THz500\,\mathrm{THz}. (d) Modes are discrete when their spacing exceeds their width; in the oven, one mode per 10MHz10\,\mathrm{MHz} or so around 2.45GHz2.45\,\mathrm{GHz} — dozens within a few percent.

Exercise 16.10 ★★★

TE10_{10} in full. In the guide 0<x<a0 < x < a, 0<y<b0 < y < b: E=E0sin(πx/a)ei(ωtkgz)ey\vect E = E_0\sin(\pi x/a)\eu^{\iu(\omega t - k_gz)}\vect e_y. (a) From Faraday, find Bx\underline B_x and Bz\underline B_z; check divB=0\operatorname{div}\vect B = 0. (b) Check Maxwell–Ampère in vacuum, and recover kg2=ω2/c2(π/a)2k_g^2 = \omega^2/c^2 - (\pi/a)^2. (c) Surface currents on the four walls (the broad walls y=0y = 0, bb and the narrow walls x=0x = 0, aa); which currents flow along zz, which across? (d) A slot cut along zz in the centre of a broad wall does not radiate, a slot across it does: explain.

Solution

Solution of Exercise 16.10.

(a) From curlE=tB\operatorname{\vect{curl}}\vect E = -\partial_t\vect B:

Bx=kgωE,Bz=iπaωE0cosπxaei(ωtkgz),\underline B_x = -\frac{k_g}\omega\underline E , \qquad \underline B_z = \frac{\iu\pi}{a\omega}E_0\cos\frac{\pi x}a\,\eu^{\iu(\omega t - k_gz)} ,

and xBx+zBz=(kgπ/aω)E0cos(πx/a)+(kgπ/aω)E0cos(πx/a)=0\partial_xB_x + \partial_zB_z = -(k_g\pi/a\omega)E_0\cos(\pi x/a) + (k_g\pi/a\omega)E_0 \cos(\pi x/a) = 0. (b) (curlB)y=zBxxBz=(i/ω)(kg2+π2/a2)E=(iω/c2)E(\operatorname{\vect{curl}}\vect B)_y = \partial_zB_x - \partial_xB_z = (\iu/\omega)(k_g^2 + \pi^2/a^2)\underline E = (\iu\omega/c^2)\underline E: kg2=ω2/c2π2/a2k_g^2 = \omega^2/c^2 - \pi^2/a^2. (c) Broad walls: js=±(BzexBxez)/μ0\vect j_s = \pm(B_z\vect e_x - B_x\vect e_z)/\mu_0 — along zz as sin(πx/a)\sin(\pi x/a), across as cos(πx/a)\cos(\pi x/a) (zero at the centre); narrow walls: along yy only. (d) A central longitudinal slot cuts only transverse currents, which vanish there; a transverse slot cuts the longitudinal ones and radiates: the slotted-waveguide antenna.

Exercise 16.11 ★★★

Dielectric guide. A slab of index n1n_1 and thickness aa in a medium n2<n1n_2 < n_1 guides light by total reflection. (a) For a ray at angle θ\theta from the normal to the faces, the guided condition is that the round-trip phase across the slab, 2k0n1acosθ2φr2k_0n_1a\cos\theta - 2\varphi_r (φr\varphi_r the phase shift of total reflection, taken 0\approx 0 here), is a multiple of 2π2\pi: number of modes for a=50µma = 50\,\text{µ}\mathrm{m}, λ=1.5µm\lambda = 1.5\,\text{µ}\mathrm{m}, n1=1.48n_1 = 1.48, n2=1.46n_2 = 1.46 (admit the mode count 2an12n22/λ\approx 2a\sqrt{n_1^2 - n_2^2}/\lambda per polarization). (b) Thickness for a single mode. (c) Compare with a single-mode fibre core of 9µm9\,\text{µ}\mathrm{m}. (d) Why can the dielectric guide, unlike the metal one, not have a cut-off for its lowest mode?

Solution

Solution of Exercise 16.11.

(a) 2an12n22/λ=2×50×0.242/1.5162a\sqrt{n_1^2 - n_2^2}/\lambda = 2 \times 50 \times 0.242/1.5 \approx 16 per polarization. (b) a<λ/2NA=3.1µma < \lambda/2\mathrm{NA} = 3.1\,\text{µ}\mathrm{m}. (c) A round core with NA=0.12\mathrm{NA} = 0.12 is single-mode below d=2.405λ/πNA=10µmd = 2.405\lambda/\pi\mathrm{NA} = 10\,\text{µ}\mathrm{m}: the 9µm9\,\text{µ}\mathrm{m} standard. (d) A grazing ray is always totally reflected: the lowest mode exists at every frequency, its evanescent tails merely spreading into the cladding.

Exercise 16.12 ★★★

Graded-index fibre. To reduce modal dispersion the core index decreases from the axis outward, n(r)=n112Δ(r/a)2n(r) = n_1\sqrt{1 - 2\Delta(r/a)^2} with Δ0.01\Delta \approx 0.01. (a) Explain qualitatively why an off-axis ray, though longer, can take the same time as the axial one. (b) Using the ray equation in a stratified medium, n(r)cosθ(r)=n(r)\cos\theta(r) = const, show that a ray launched from the axis at a small angle oscillates sinusoidally about it (expand nn to second order). (c) Period of the oscillation for a=25µma = 25\,\text{µ}\mathrm{m}. (d) The residual modal spread is of order n1LΔ2/2cn_1L\Delta^2/2c: value per kilometre, compared with the step-index n1LΔ/cn_1L\Delta/c.

Solution

Solution of Exercise 16.12.

(a) The off-axis ray spends its time in lower index, where light is faster: the longer path is compensated. (b) n(r)cosθ=n1cosθ0n(r)\cos\theta = n_1\cos\theta_0 with nn1(1Δr2/a2)n \approx n_1(1 - \Delta r^2/a^2) and cosθ1θ2/2\cos\theta \approx 1 - \theta^2/2: ( ⁣dr/ ⁣dz)2=θ022Δr2/a2(\dd r/\dd z)^2 = \theta_0^2 - 2\Delta r^2/a^2, a harmonic oscillation r=r0sin(2Δz/a)r = r_0\sin(\sqrt{2\Delta}\,z/a). (c) 2πa/2Δ=1.1mm2\pi a/\sqrt{2\Delta} = 1.1\,\mathrm{mm}. (d) n1LΔ2/2c=0.25ns/kmn_1L\Delta^2/2c = 0.25\,\mathrm{ns}/\mathrm{km} against 49ns/km49\,\mathrm{ns}/\mathrm{km}: two hundred times better.

Optical fibres lit from their far ends: light trapped by total internal reflection in a core thinner than a hair, and guided round every gentle bend.
Optical fibres lit from their far ends: light trapped by total internal reflection in a core thinner than a hair, and guided round every gentle bend.

16.5 Problem: The oven cavity and the optical fibre

Problem 16.1

Weekend problem — two guided-wave systems in every kitchen and every city: the oven that cooks with standing waves and the fibre that carries the internet

Part I — From the magnetron to the cavity. A magnetron feeds 800W800\,\mathrm{W} at 2.45GHz2.45\,\mathrm{GHz} into a rectangular waveguide (a=8.6cma = 8.6\,\mathrm{cm}, b=4.3cmb = 4.3\,\mathrm{cm}) that opens into the oven cavity (30×30×2030 \times 30 \times 20 cm3^3); walls steel, γ=1×106S/m\gamma = 1 \times 10^{6}\,\mathrm{S}/\mathrm{m}.

  1. Cut-off frequencies of the guide’s TE10_{10}, TE20_{20}, TE01_{01} modes; check that only TE10_{10} propagates at 2.45GHz2.45\,\mathrm{GHz}.
  2. Guide wavelength, phase and group velocities at 2.45GHz2.45\,\mathrm{GHz}.
  3. Field amplitude E0E_0 in the guide for 800W800\,\mathrm{W} (TE10_{10} power formula); compare with the breakdown field of air.
  4. Surface current amplitude on the broad wall (of order E0/μ0cE_0/\mu_0c); skin depth and surface resistance of the steel; loss per metre of guide.
  5. Modes of the cavity: list those with frequencies within 3%3\% of 2.45GHz2.45\,\mathrm{GHz} (try (m,n,p)(m, n, p) up to 44); how many?
  6. Distance between the hot spots of one such mode along each axis; why does the plate turn, and what is a "mode stirrer"?
  7. The cavity’s quality factor when loaded with food is about 200200: bandwidth of its response; why a mismatch between the magnetron’s frequency and the modes does not matter much.
  8. The empty cavity has Q3000Q \approx 3000: stored energy at 800W800\,\mathrm{W} input, mean energy density, field amplitude; compare with question 3 and with breakdown. What protects the magnetron when the oven runs empty?
  9. Time for the energy to travel the 20cm20\,\mathrm{cm} of guide from the magnetron to the cavity.
  10. The magnetron is fed by a half-wave rectified supply and emits only during half of each mains cycle: peak power during the bursts for a mean of 800W800\,\mathrm{W}; what does the cavity do between bursts (use Q200Q \approx 200)?

Part II — The door and the cut-off.

  1. The door mesh has 1.5mm1.5\,\mathrm{mm} holes in a 1mm1\,\mathrm{mm} sheet: decay length in a hole and attenuation in dB across the sheet.
  2. The door’s edge is sealed by a quarter-wave "choke": a slot λ/4\lambda/4 deep around the door frame, which presents an open circuit at the gap (recall Chapter 15): depth of the slot at 2.45GHz2.45\,\mathrm{GHz}.
  3. Legal leakage is 5mW/cm25\,\mathrm{mW}/\mathrm{cm}^{2} at 5cm5\,\mathrm{cm}: to what fraction of 800W800\,\mathrm{W} through a 100cm2100\,\mathrm{cm}^{2} door does that correspond?
  4. A 30cm30\,\mathrm{cm} duct of 4cm4\,\mathrm{cm} diameter vents the cavity: is it below cut-off at 2.45GHz2.45\,\mathrm{GHz}? Attenuation over its length.
  5. Why must the mesh be electrically bonded to the door frame all round (what would a gap in the bond behave like)?

Part III — The fibre. A step-index fibre: core n1=1.465n_1 = 1.465, cladding n2=1.460n_2 = 1.460, core diameter 50µm50\,\text{µ}\mathrm{m} (multimode) or 9µm9\,\text{µ}\mathrm{m} (single-mode); λ=1.55µm\lambda = 1.55\,\text{µ}\mathrm{m}; attenuation 0.2dB/km0.2\,\mathrm{dB}/\mathrm{km}.

  1. Numerical aperture and acceptance angle; is it easy to couple light in?
  2. Modal spread per kilometre for the multimode fibre; maximum bit rate over 10km10\,\mathrm{km} (one bit per spread time).
  3. Number of modes of the multimode fibre (admit N12(πdNA/λ)2N \approx \tfrac12(\pi d\,\mathrm{NA}/\lambda)^2); of the single-mode one (d=9µmd = 9\,\text{µ}\mathrm{m}: show the same formula gives about 11).
  4. In the single-mode fibre the remaining spreading is chromatic: with ω0.2m2/s\omega'' \approx -0.2\,\mathrm{m}^{2}/\mathrm{s} (Chapter 8) and 50ps50\,\mathrm{ps} pulses, distance over which a pulse doubles; bit rate over 100km100\,\mathrm{km}.
  5. Attenuation over 100km100\,\mathrm{km} in dB and as a power ratio; input 1mW1\,\mathrm{mW}: output power; with amplifiers every 80km80\,\mathrm{km}, how many for a 6000km6000\,\mathrm{km} transatlantic link?
  6. Frequency and photon energy at 1.55µm1.55\,\text{µ}\mathrm{m}; photons per second in 1mW1\,\mathrm{mW}, and per bit at 10Gbit/s10\,\mathrm{Gbit}/\mathrm{s}, at the input and after 100km100\,\mathrm{km}.
  7. Why 1.55µm1.55\,\text{µ}\mathrm{m} (think of the two loss mechanisms of silica: Rayleigh scattering, falling as 1/λ41/\lambda^4, and infrared absorption rising beyond 1.6µm1.6\,\text{µ}\mathrm{m})?
  8. Fresnel reflection at a cleaved fibre end facing air: loss in dB; why connectors use index-matching gel or polished physical contact.
  9. Why does a fibre not radiate at a gentle bend, and why does it lose light at a sharp one (think of the angle of the ray on the core boundary)?
  10. Compare the two guides of this problem: what confines the wave, what limits the bandwidth, what the losses.
Solution

Solution of Problem 16.1.

1. c/2a=1.74GHzc/2a = 1.74\,\mathrm{GHz}; TE20_{20} and TE01_{01}: 3.49GHz3.49\,\mathrm{GHz}: only TE10_{10} at 2.45GHz2.45\,\mathrm{GHz}.

2. 1(1.74/2.45)2=0.70\sqrt{1 - (1.74/2.45)^2} = 0.70: λg=17.4cm\lambda_g = 17.4\,\mathrm{cm}, vφ=1.42cv_\varphi = 1.42c, vg=0.70c=2.1×108m/sv_g = 0.70c = 2.1 \times 10^{8}\,\mathrm{m}/\mathrm{s}.

3. E02=4P/ε0cab=3200/6.9×106E_0^2 = 4P/\varepsilon_0c\,ab\sqrt{\cdot} = 3200/6.9 \times 10^{-6}: E0=21kV/mE_0 = 21\,\mathrm{kV}/\mathrm{m}, a hundredth of breakdown.

4. jsE0/μ0c=57A/mj_s \approx E_0/\mu_0c = 57\,\mathrm{A}/\mathrm{m}; δ=10µm\delta = 10\,\text{µ}\mathrm{m}, Rs=0.1ΩR_s = 0.1\,\Omega; 12Rsjs2×2a=28W/m\tfrac12R_sj_s^2 \times 2a = 28\,\mathrm{W}/\mathrm{m}: 3.5%3.5\% per metre — steel is lossy, the guide is short.

5. f=1.5×10811.1(m2+n2)+25p2f = 1.5 \times 10^8\sqrt{11.1(m^2 + n^2) + 25p^2} within ±3%\pm3\% of 2.45GHz2.45\,\mathrm{GHz}: (1,2,3)(1, 2, 3), (2,1,3)(2, 1, 3) at 2.51GHz2.51\,\mathrm{GHz}, (4,0,2)(4, 0, 2), (0,4,2)(0, 4, 2), (4,3,0)(4, 3, 0), (3,4,0)(3, 4, 0) at 2.50GHz2.50\,\mathrm{GHz} — about six.

6. Half-wavelengths a/ma/m, b/nb/n, d/pd/p: 77 to 10cm10\,\mathrm{cm} apart. The turntable drags the food through maxima and minima; a stirrer (a rotating metal fan) reshuffles the modes.

7. f/Q=12MHzf/Q = 12\,\mathrm{MHz}: the loaded resonances overlap into a continuum; the magnetron always finds a mode to feed.

8. W=QP/ω=3000×800/1.54×1010=0.16JW = QP/\omega = 3000 \times 800/1.54 \times 10^{10} = 0.16\,\mathrm{J}; u=8.7J/m3u = 8.7\,\mathrm{J}/\mathrm{m}^{3}; E04u/ε0=2MV/mE_0 \approx \sqrt{4u/\varepsilon_0} = 2\,\mathrm{MV}/\mathrm{m} — near breakdown: arcs; the reflected power goes back to the magnetron, protected (a little) by a dummy load or circulator — hence "never run it empty".

9. 0.2/2.1×108=1ns0.2/2.1 \times 10^8 = 1\,\mathrm{ns}.

10. 1.6kW1.6\,\mathrm{kW} peak; between bursts the field rings down in Q/ω=13nsQ/\omega = 13\,\mathrm{ns}: the cavity is empty most of the time.

11. a/π=0.48mm\ell \approx a/\pi = 0.48\,\mathrm{mm}; e1/0.48\eu^{-1/0.48}: 18dB-18\,\mathrm{dB} in amplitude, 36dB-36\,\mathrm{dB} in power.

12. λ/4=3.1cm\lambda/4 = 3.1\,\mathrm{cm}.

13. 5mW/cm25\,\mathrm{mW}/\mathrm{cm}^{2} ×\times 100cm2100\,\mathrm{cm}^{2} = 0.5W0.5\,\mathrm{W}: 0.06%0.06\%.

14. f1c/2d=3.75GHz>2.45GHzf_1 \approx c/2d = 3.75\,\mathrm{GHz} > 2.45\,\mathrm{GHz}: below cut-off; =1.7cm\ell = 1.7\,\mathrm{cm}, e30/1.7\eu^{-30/1.7}: 150dB-150\,\mathrm{dB}.

15. The wall currents must flow continuously into the mesh; a gap in the bond is a slot that cuts them — a slot antenna radiating outward.

16. NA=1.46521.4602=0.12\mathrm{NA} = \sqrt{1.465^2 - 1.460^2} = 0.12, 77{}^{\circ}: a narrow cone, needing a laser and a lens.

17. (n1L/c)(n1/n21)=17ns/km(n_1L/c)(n_1/n_2 - 1) = 17\,\mathrm{ns}/\mathrm{km}; 170ns170\,\mathrm{ns} over 10km10\,\mathrm{km}: 6Mbit/s6\,\mathrm{Mbit}/\mathrm{s}.

18. 12(πdNA/λ)2=75\tfrac12(\pi d\,\mathrm{NA}/\lambda)^2 = 75; for 9µm9\,\text{µ}\mathrm{m}: 2.42.4 — the single-mode regime (V=πdNA/λ=2.2<2.4V = \pi d\,\mathrm{NA}/\lambda = 2.2 < 2.4).

19. Δx0=(c/n)τ=1cm\Delta x_0 = (c/n)\tau = 1\,\mathrm{cm}: tsp=104/0.2=0.5mst_{\text{sp}} = 10^{-4}/0.2 = 0.5\,\mathrm{ms}, 100km100\,\mathrm{km}: 10Gbit/s10\,\mathrm{Gbit}/\mathrm{s} over that distance before compensation.

20. 20dB20\,\mathrm{dB}, a factor 100100: 10µW10\,\text{µ}\mathrm{W}; 6000/80=756000/80 = 75 amplifiers.

21. f=1.9×1014Hzf = 1.9 \times 10^{14}\,\mathrm{Hz}, hf=0.8eVhf = 0.8\,\mathrm{eV}; 8×10158 \times 10^{15} photons per second; 8×1058 \times 10^5 per bit at the input, 8×1038 \times 10^3 after 100km100\,\mathrm{km}.

22. The sum of the two losses is minimal near 1.55µm1.55\,\text{µ}\mathrm{m} (0.2dB/km0.2\,\mathrm{dB}/\mathrm{km}).

23. R=((1.4651)/2.465)2=3.6%R = ((1.465 - 1)/2.465)^2 = 3.6\%: 0.16dB0.16\,\mathrm{dB} per face, twice at a connector plus the gap’s interference; gel or physical contact removes the air.

24. On a gentle bend the ray still meets the boundary beyond the critical angle; on a sharp one the incidence on the outer wall falls below it and light leaks out.

25. Oven: metal walls, reflection with r=1r = -1; bandwidth set by the cavity’s modes; losses in the steel skin. Fibre: total internal reflection at a glass–glass boundary; bandwidth set by modal and chromatic dispersion; losses by scattering and absorption, 0.2dB/km0.2\,\mathrm{dB}/\mathrm{km}.