Chapter 12Ring Resonators
Loop a waveguide back on itself and you get a filter, a sensor, a modulator or a laser mirror, depending on what you do with it.
- derive the all-pass and add-drop transfer functions from a coupler and a round trip;
- define the free spectral range, finesse, loaded and intrinsic Q and extinction ratio;
- explain under-, critical and over-coupling;
- extract loss and coupling from a measured resonance, and see why one spectrum is ambiguous;
- explain split resonances and thermal drift, and list the applications of rings.
How do coupling and loss shape a ring’s resonance, and which gap gives the resonance you want?
12.1Resonance condition
A ring of radius \(R\) and circumference \(L = 2\pi R\) is resonant when light returns in phase after one round trip, \(\beta L = 2\pi m\), that is when
\[m\lambda_m = n_\mathrm{eff}L, \qquad m = 1, 2, 3, \dots\]Adjacent resonances are separated by the free spectral range, which, as for the MZI (Chapter 11), involves the group index:
A 10 µm radius silicon ring (\(L = 62.8\) µm, \(n_g = 4.18\)) has FSR = \(1.55^2/(4.18 \times 62.8) = 9.15\) nm. For the four channels of the reference link, spread over 9.6 nm, the FSR must exceed that span, which calls for rings of about 9 µm radius or less.
12.2The all-pass ring
Couple the ring to a single bus waveguide through a short directional coupler with field self-coupling \(t\) and cross-coupling \(\kappa\) (\(t^2 + \kappa^2 = 1\), Chapter 9), and let \(a\) be the field transmission of one round trip (\(a^2 = e^{-\alpha L}\)) and \(\phi = \beta L\) its phase. Summing the light that circulates any number of times gives the through-port transmission [35]:
Off resonance \(T \approx 1\); on resonance (\(\phi = 2\pi m\)) it dips to \(T_\mathrm{min} = (a - t)^2/(1 - at)^2\). The dip reaches zero when \(t = a\), critical coupling: the light coupled back out of the ring then exactly cancels the light that passed by, because the coupler’s loss of transmission equals the ring’s loss per round trip. For \(t > a\) the ring is under-coupled (it loses more than it couples, giving a narrow, shallow dip); for \(t < a\) it is over-coupled (a broad, shallow dip).
Try it. Change the radius, gap and loss; the top view shows the ring to scale, with a magnified inset of the gap.
12.3The add-drop ring
A second bus on the other side of the ring, with couplings \(t_2\) and \(\kappa_2\), drops the resonant light into a separate port:
\[T_\mathrm{through} = \frac{t_2^2a^2 - 2t_1t_2a\cos\phi + t_1^2}{1 - 2t_1t_2a\cos\phi + (t_1t_2a)^2}, \qquad T_\mathrm{drop} = \frac{(1 - t_1^2)(1 - t_2^2)\,a}{1 - 2t_1t_2a\cos\phi + (t_1t_2a)^2}.\]With equal couplers and no loss, the drop port takes all the power on resonance and nothing off resonance: a bandpass filter. Rings in series on a bus, each with its own resonance, make a multiplexer or demultiplexer, exactly the receiver of the reference link (Chapter 14).
12.4Quality factor and finesse
The width of a resonance is described by its full width at half maximum, its quality factor \(Q = \lambda/\Delta\lambda_\mathrm{FWHM}\), and its finesse \(F = \mathrm{FSR}/\Delta\lambda_\mathrm{FWHM}\). For the all-pass ring
\[\Delta\lambda_\mathrm{FWHM} = \frac{(1 - at)\,\lambda^2}{\pi n_gL\sqrt{at}}.\]\(Q\) is also \(2\pi\) times the number of optical cycles the energy survives in the ring, and the photon lifetime is \(\tau = Q\lambda/(2\pi c)\). The loaded Q combines two loss channels: the intrinsic loss of the ring and the coupling out,
\[\frac{1}{Q_L} = \frac{1}{Q_i} + \frac{1}{Q_c}, \qquad Q_i = \frac{2\pi n_g}{\lambda\alpha},\]and at critical coupling \(Q_c = Q_i\), so \(Q_L = Q_i/2\). Near resonance the circulating power builds up above the input by \(\kappa^2/(1 - at)^2\), which reaches several hundred in good silicon rings: a milliwatt in the bus becomes hundreds of milliwatts in the ring, enough for nonlinear effects (Chapter 7).
For a 10 µm radius silicon ring with 2 dB/cm of loss (\(a = 0.99855\), \(Q_i = 3.7\times10^5\)) and the coupler model of Chapter 9, Table 12.1 shows three gaps. A 200 nm gap over-couples (κ² = 1.2 %, \(t = 0.9938 < a\)): a broad 22 pm resonance only 4 dB deep. At 250 nm (κ² = 0.52 %) the ring is still slightly over-coupled, with an 11 dB dip and \(Q_L = 1.3\times10^5\). At 300 nm (κ² = 0.22 %, \(t = 0.99892 > a\)) it is slightly under-coupled: a 7.4 pm resonance, 17 dB deep, with \(Q_L = 2.1\times10^5\). Critical coupling at 250 nm would need 3.6 dB/cm of loss. Because the loss is not known until the chip is measured, test chips sweep the gap.
Table 12.1. A 10 µm silicon ring at 1550 nm, 2 dB/cm loss. Coupling from the femwell-fitted coupler model of Chapter 9.
| Gap | κ² | Regime | FWHM | Loaded Q | Extinction |
|---|---|---|---|---|---|
| 200 nm | 1.23 % | over-coupled | 22.3 pm | 70 000 | 4.1 dB |
| 250 nm | 0.52 % | slightly over-coupled | 11.7 pm | 132 000 | 11.0 dB |
| 300 nm | 0.22 % | slightly under-coupled | 7.4 pm | 211 000 | 16.7 dB |
12.5Extracting parameters from measurements
Fitting a measured through-port resonance with the all-pass formula gives \(a\) and \(t\), and so the propagation loss and the coupling, in a device a fraction of a millimetre long [48]. There is a catch: \(T(\phi)\) is symmetric in \(a\) and \(t\), so an under-coupled ring with \((a, t)\) and an over-coupled one with \((t, a)\) give identical spectra. The ambiguity is resolved by measuring more than one spectrum: a series of rings with different gaps (the coupling changes, the loss does not), an add-drop ring (whose drop port breaks the symmetry), or the phase response. Chapter 32 fits real data this way.
12.6Non-idealities
- Split resonances. Sidewall roughness backscatters a little light into the counter-propagating mode of the ring. In high-Q rings the clockwise and counter-clockwise modes then couple and split each resonance into a doublet.
- Thermal drift. Silicon’s thermo-optic coefficient moves a ring’s resonances by about 70–80 pm per kelvin, several times a typical resonance width (Chapter 18).
- Fabrication. A 1 nm change of width or thickness moves the resonance by about 0.6 or 1.4 nm (Chapter 6), so rings never land exactly where designed.
- Dispersion and bending. The FSR changes slowly with wavelength because \(n_g\) does, and in small rings the bend changes \(n_\mathrm{eff}\) and adds loss (Chapter 8).
12.7Applications
Rings are the most versatile passive device in silicon photonics: WDM filters and multiplexers (Chapter 14), compact modulators that shift their resonance with carriers (Chapter 20), evanescent-field sensors that track a resonance as molecules bind (Chapter 34), the wavelength-selective mirrors of tunable lasers, where two rings of slightly different radius form a Vernier filter (Chapter 26), and high-Q resonators for frequency combs and narrow-linewidth lasers.
Include all-pass rings of 10 µm radius with gaps of 200, 250 and 300 nm, and an add-drop ring. Their spectra give the propagation loss independently of the cutback, the coupling against gap, and the group index from the FSR.
Reading
- [35] Wim Bogaerts et al. “Silicon microring resonators.” Laser & Photonics Reviews, 6(1):47–73, 2012.
- [1] Lukas Chrostowski and Michael Hochberg. Silicon Photonics Design: From Devices to Systems. Cambridge University Press, 2015. https://www.cambridge.org/core/books/silicon-photonics-design/BF3CF13E8542BCE67FD2BBC7104ECEAB
- [48] Wei Shi, Ting Kai Chang, Han Yun, Wen Zhang, Yun Wang, Nicolas A. F. Jaeger and Lukas Chrostowski. “Differential measurement of transmission losses of integrated optical components using waveguide ring resonators.” Proc. SPIE 8412, Photonics North, 2012.
- [6] Amnon Yariv and Pochi Yeh. Photonics: Optical Electronics in Modern Communications, 6th edition. Oxford University Press, 2007.
- Resonances at \(m\lambda = n_\mathrm{eff}L\), spaced by FSR = \(\lambda^2/(n_gL)\): 9.15 nm for a 10 µm silicon ring.
- All-pass ring: \(T = (a^2 - 2at\cos\phi + t^2)/(1 - 2at\cos\phi + a^2t^2)\); critical coupling at \(t = a\) gives a full null.
- \(Q_i = 2\pi n_g/(\lambda\alpha)\) depends only on loss; \(Q_L = Q_i/2\) at critical coupling.
- For a 10 µm ring at 2 dB/cm, gaps from 200 to 300 nm move from over- to under-coupling; sweep the gap on a test chip.
- One through-port spectrum cannot tell under- from over-coupling; use several gaps or an add-drop ring.