Skip to content

Energy Budget Analyzer

Track where every photon's energy goes as it traverses the pixel stack. At each wavelength, the incident energy splits into reflection, per-layer absorption, and transmission.

Energy Budget Analyzer

Visualize where photon energy goes at each wavelength — how much is reflected, absorbed in each layer, or transmitted through the pixel stack.

27.4%21.5%45.3%5.8%
Reflection Microlens Planarization Color Filter BARL Silicon (QE) Transmission
At λ=550nm: QE=45.3%, R=27.4%, Other losses=27.3%

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

Photon Energy Accounting

Plain-English Intuition

Think of every incoming photon as a coin. Some coins bounce off the stack, some pass through the active silicon without being used, some are spent heating filters or coatings, and only the coins absorbed in the photodiode silicon can become signal. The budget tells you which bucket is stealing photons.

The energy-budget view decomposes incident optical power into reflection, escape transmission, useful silicon absorption, and parasitic absorption in non-silicon layers. It is a diagnostic layer on top of the TMM result: it explains why QE changes, not just that QE changes.

Assumptions

  • The stack is treated as passive, so incident energy must split into reflection, transmission, and layer absorption.
  • Layer absorption is an optical power bookkeeping term; it does not guarantee charge collection in the photodiode.
  • The browser model uses simplified layer spectra and therefore supports diagnosis rather than process sign-off.

Outputs

  • Per-wavelength fractions for reflected, transmitted, useful silicon-absorbed, and parasitic absorbed energy.
  • A diagnosis of whether QE loss is dominated by front-surface reflection, filter/coating absorption, or insufficient silicon absorption.

Validation Example

  • For every wavelength, verify that $R+T+\sum A_j$ stays close to 1 before interpreting layer tradeoffs.
  • Removing an absorbing filter should reduce parasitic absorption and move that energy into reflection, transmission, or silicon absorption.

Core Equations

Incident power normalization
$$P_{\text{inc}}(\lambda)=1$$
  • \(P_{\text{inc}}\): Incident optical power at a wavelength, normalized to unity
  • \(\lambda\): Wavelength at which the stack is evaluated

All bars in the budget are fractions of the same incident power, so channels and wavelengths can be compared directly.

Power conservation residual
$$\varepsilon = 1-R-T-\sum_jA_j$$
  • \(\varepsilon\): Numerical energy-balance residual
  • \(R,T\): Reflected and transmitted power fractions
  • \(A_j\): Absorption fraction in layer $j$

A large residual means the optical model, interpolation, or layer accounting should be inspected before interpreting QE.

Layer absorption from flux difference
$$A_j=P_{\text{top},j}-P_{\text{bot},j}, \quad P=\frac{\operatorname{Re}(EH^*)}{P_{\text{inc}}}$$
  • \(P_{\text{top},j},P_{\text{bot},j}\): Normalized Poynting flux at the top and bottom of layer $j$
  • \(E,H\): Electric and magnetic field amplitudes recovered from the transfer matrix
  • \(H^*\): Complex conjugate of magnetic field amplitude

This assigns absorption to physical layers instead of only reporting total stack loss.

Useful and parasitic buckets
$$A_{\text{use}}=A_{\text{Si}}, \quad A_{\text{parasitic}}=\sum_{j\notin\text{Si}}A_j$$
  • \(A_{\text{use}}\): Absorption that can potentially contribute to signal
  • \(A_{\text{parasitic}}\): Absorption in filters, coatings, metal, or other non-photodiode layers
  • \(A_{\text{Si}}\): Absorption in the silicon photodiode layer

A design that reduces reflection but increases parasitic absorption may not improve sensor signal.

Optical QE upper bound
$$QE_{\text{ext}}(\lambda)=\eta_{\text{cc}}(\lambda)A_{\text{Si}}(\lambda), \quad QE_{\text{ext}}\le A_{\text{Si}}$$
  • \(QE_{\text{ext}}\): External quantum efficiency after carrier collection effects
  • \(\eta_{\text{cc}}\): Carrier-collection probability for generated electron-hole pairs

The budget reports optical absorption; electrical QE can only be lower unless carrier collection is ideal.

Design sensitivity
$$S_{x_k}(\lambda)=\frac{\partial A_{\text{Si}}(\lambda)}{\partial x_k}$$
  • \(S_{x_k}\): Sensitivity of silicon absorption to design variable $x_k$
  • \(x_k\): Layer thickness, material index, angle, or wavelength-dependent filter parameter

Sensitivity helps identify whether a QE loss is controlled by BARL thickness, silicon depth, filter absorption, or angle.

Model Interpretation

  • Reflection-limited loss points to coating or BARL work; parasitic absorption points to material or thickness choices; escape transmission points to silicon thickness or light trapping.
  • Energy accounting is the bridge between a spectral curve and an engineering action.
  • The budget is most useful when compared across a before/after design change rather than read as a standalone number.

Diagnosis Workflow

  • First check $\varepsilon$ to confirm power conservation, then inspect whether loss sits in $R$, $T$, $A_{\text{parasitic}}$, or missing $A_{\text{Si}}$.
  • If $R$ dominates, tune BARL/coating; if $A_{\text{parasitic}}$ dominates, revisit color-filter or metal absorption; if $T$ dominates, adjust silicon thickness or backside trapping.
  • Compare the same budget for R/G/B because a fix for one channel can move loss into another channel.

Calibration Targets

  • Use measured reflectance/transmittance spectra to anchor $R$ and $T$ before trusting layer absorption allocation.
  • Replace generic $n,k$ with process-specific ellipsometry for color filters, BARL layers, silicon, and passivation.
  • Validate $A_{\text{Si}}$ against measured QE only after estimating carrier collection probability.

Known Missing Physics

  • The budget does not include lateral redistribution: photons absorbed in silicon outside the photodiode ROI may become crosstalk.
  • It does not model electrical collection, diffusion, recombination, or depletion-region geometry.
  • Structured pixels need RCWA/FDTD field maps to decide where within silicon absorption occurs.

References