Skip to content

Silicon Absorption Depth

Visualize how photon absorption depth in silicon varies with wavelength using the Beer-Lambert law. Understand why pixel thickness is a critical design parameter for quantum efficiency.

Silicon Absorption Depth Visualizer

Explore how light penetration depth in silicon depends on wavelength using Beer-Lambert law.

Light (550 nm)SiliconPD depth0 um1 um2 um3 umdelta = 1.56 um
101102103104105106alpha (cm-1)1000 um100 um10.0 um1.0 um100 nm10 nmdelta (um)4005006007008009001000Wavelength (nm)
Absorption coeff. alpha
6,397 cm-1
Penetration depth delta
1.563 um
Absorbed in PD
72.2%
Absorbed in full Si
85.3%

Model scope

Use this browser tool for intuition, relative trends, and design-space exploration. Its local simplified model is not a substitute for RCWA/FDTD sign-off, silicon calibration, or vendor process data.

Physics Notes

Beer-Lambert Absorption in Silicon

Plain-English Intuition

Silicon swallows different colours at different depths. Blue light is absorbed within the first few hundred nanometres near the surface, while red and near-infrared photons sneak much deeper before being absorbed — sometimes deeper than the photodiode reaches. That is why thin BSI pixels often lose near-IR sensitivity.

Silicon absorption depends strongly on wavelength. Blue light is absorbed near the surface, while red and NIR photons penetrate much deeper.

Assumptions

  • Silicon absorption follows Beer-Lambert attenuation using wavelength-dependent absorption coefficient data.
  • The model treats silicon as a uniform slab and does not include surface texturing, interference, or carrier collection probability.
  • Absorbed photons are optical absorption events, not automatically collected electrons.

Outputs

  • Absorption coefficient, penetration depth, and absorbed fraction versus wavelength and silicon thickness.
  • A wavelength-regime map showing why blue absorbs near the surface and red/NIR requires thicker silicon.

Validation Example

  • At fixed thickness, blue absorption should be higher than red/NIR absorption for typical visible silicon data.
  • Increasing silicon thickness should help long wavelengths more than short wavelengths once blue absorption is already saturated.

Core Equations

Absorption coefficient
$$\alpha(\lambda) = \frac{4\pi k(\lambda)}{\lambda}$$
  • \(\alpha\): Absorption coefficient ($1/\mu\text{m}$)
  • \(k\): Extinction coefficient
  • \(\lambda\): Incident wavelength

The extinction coefficient k is converted to an absorption coefficient.

Intensity decay
$$I(z, \lambda) = I_0 e^{-\alpha(\lambda)z}$$
  • \(I(z)\): Intensity at depth $z$
  • \(I_0\): Intensity at the surface ($z=0$)

This is the Beer-Lambert law for a uniform absorbing medium.

Absorbed fraction
$$A(d, \lambda) = 1 - e^{-\alpha(\lambda)d}$$
  • \(A(d)\): Absorbed fraction within thickness $d$
  • \(d\): Silicon physical thickness

This ignores front-surface reflection and interference.

Model Interpretation

  • Thin BSI pixels may collect visible light well but lose NIR photons.
  • DTI, backside texture, and reflectors can increase the effective path length.
  • Carrier collection depth is a separate electrical problem not included here.

Wavelength Regimes

  • Blue light has large $\alpha$, so it is absorbed near the surface and is sensitive to surface passivation and shallow collection.
  • Red and NIR light have smaller $\alpha$, so a thin silicon layer can transmit photons without useful absorption.
  • The absorption length $1/\alpha$ is a useful scale, but 63% absorption at one length is not the same as high QE.

Device Implications

  • Thicker silicon improves red/NIR absorption but may worsen crosstalk if carriers are generated far from the intended photodiode.
  • Backside reflectors or light trapping can convert escape transmission into another absorption opportunity.
  • Collection probability $\eta_{\text{cc}}(z)$ should be multiplied with the absorption profile for electrical QE.

Known Missing Physics

  • Beer-Lambert absorption ignores interference from upper films and coherent standing waves inside the stack.
  • The model does not include doping, temperature, strain, or free-carrier absorption changes in silicon.
  • It predicts where photons are absorbed, not whether generated carriers are collected.