Acceptance bandwidth

Acceptance bandwidth quantifies the range of input conditions over which an optical process (nonlinear conversion, grating coupling, filter operation) maintains useful efficiency. It is typically specified as a full-width at half-maximum (FWHM) of the efficiency versus the relevant input parameter.

Three principal acceptance bandwidths in optical-system design:

1. Spectral acceptance bandwidth ($\mathrm{\Delta }\lambda$): wavelength range over which the device is efficient
2. Angular acceptance bandwidth ($\mathrm{\Delta }\theta$): incidence angle range over which the device is efficient
3. Temperature acceptance bandwidth ($\mathrm{\Delta }T$): temperature range over which the device stays in spec (for nonlinear crystals where index varies with temperature)

Spectral acceptance bandwidth for phase-matched nonlinear processes.

For a phase-matched second-harmonic generation in a crystal of length $L$, the SHG efficiency falls off with input wavelength shift as ${\text{sinc}}^2(\mathrm{\Delta }kL\mathrm{/}2)$. The FWHM in wavelength:

$$\mathrm{\Delta }{\lambda }_{\text{FWHM}}\approx \frac{0.886{\lambda }^2}{2L\mathrm{\mid }{n}_g(2\omega )-n_g(\omega )\mathrm{\mid }},$$

where $n_g$ is group index. For 10-mm PPLN at 1064 nm pump: $\mathrm{\Delta }\lambda \approx 0.2$ nm. For 1-mm PPLN: $\mathrm{\Delta }\lambda \approx 2$ nm. Shorter crystal → broader bandwidth but lower peak efficiency.

Spectral acceptance bandwidth for gratings.

For a uniform-period Bragg grating of length $L$ in fiber or waveguide, the width of the reflection band, measured between the first zeros on either side of the peak, is

$$\mathrm{\Delta }{\lambda }_0=\frac{{\lambda }_B^2}{\pi n_{g}L}\sqrt{(\kappa L{)}^2+{\pi }^2},$$

where $\kappa$ is the coupling coefficient. In a strong grating ($\kappa L\gg \pi$) the light is reflected within a length of order $1\mathrm{/}\kappa$ and never reaches the far end, so the width tends to ${\lambda }_B^2\kappa \mathrm{/}(\pi n_g)$ and no longer depends on $L$; in a weak grating ($\kappa L\ll \pi$) the light passes through the whole length and the width tends to ${\lambda }_B^2\mathrm{/}(n_{g}L)$. For typical telecom fiber Bragg gratings, $\mathrm{\Delta }\lambda =0.05-0.5$ nm. A larger $\kappa$ widens the band and also raises the peak reflectivity reached in a given length, $R={\tanh }^2(\kappa L)$, so strong gratings are short and broad, while a weak grating has to be long to reflect well, which makes its band narrow.

Angular acceptance bandwidth.

For grating couplers, the angular acceptance is set by diffraction from the illuminated length of the grating, $N\mathrm{\Lambda }$. A plane wave couples efficiently over a range of angles of about

$$\mathrm{\Delta }\theta \approx \frac{\lambda }{N\mathrm{\Lambda }\cos {\theta }_0},$$

where $N$ is the number of grating periods illuminated and ${\theta }_0$ the angle of best coupling. For uniform illumination this is a little more than the full width at half maximum of the coupling efficiency versus angle, which is $0.89\lambda \mathrm{/}(N\mathrm{\Lambda }\cos {\theta }_0)$. For a silicon photonic grating coupler with $N=20$ periods and $\mathrm{\Lambda }=600$ nm at ${\theta }_0=10\mathrm{°}$ and $\lambda =1550$ nm, $\mathrm{\Delta }\theta \approx 7.5\mathrm{°}$. A longer grating accepts a narrower range of angles.

Temperature acceptance bandwidth in nonlinear crystals.

Lithium niobate's refractive indices depend on temperature; thermal expansion changes the QPM period:

$$\mathrm{\Delta }{T}_{\text{FWHM}}\approx \frac{5.57}{L\mathrm{\mid }\mathrm{\partial }\mathrm{\Delta }k\mathrm{/}\mathrm{\partial }T\mathrm{\mid }},$$

For 10-mm PPLN at 1064 nm: $\mathrm{\Delta }T\approx 3$°C. Temperature stabilization to 0.1 – 1°C is therefore required for stable operation.

Trade-offs.

Bandwidth engineering techniques.

• Crystal length / grating length: shorter → wider bandwidth (and lower efficiency)
• Apodization: smooth the structure's envelope for cleaner spectral response
• Chirped designs: periodically-poled crystals or gratings with spatially-varying period give intentional broadband response
• Multi-section designs: cascaded sections with different periods give multi-band response
• Group-velocity dispersion compensation: matched dispersion in subsequent elements compensates initial dispersion for broadband ops

Comparison: temporal bandwidth vs spectral acceptance bandwidth. A nonlinear conversion of an ultrafast pulse requires bandwidth matching: the pulse spectrum must fit within the crystal's spectral acceptance bandwidth. For a transform-limited 100 fs pulse at 1064 nm the spectral width is about 16.6 nm. In PPLN, where a 10 mm crystal accepts 0.21 nm, accepting 16.6 nm takes a crystal only about 0.13 mm long, and the single-pass efficiency falls with it.

Efficiency–bandwidth trade. For a fixed crystal material and wavelength in the low-conversion limit, peak SHG efficiency scales as $L^2$ while the acceptance bandwidth scales as $1\mathrm{/}L$, so efficiency × bandwidth² stays roughly fixed. Halving the crystal length doubles the bandwidth and cuts the peak efficiency by a factor of four.

References: Boyd, Nonlinear Optics, Ch. 2; Agrawal, Nonlinear Fiber Optics, Ch. 10; Othonos & Kalli, Fiber Bragg Gratings, Ch. 4 for grating bandwidth treatment.