What Is Etch Iep Endpoint

An etch Interferometric Endpoint (IEP Endpoint)—also known as in situ laser reflectometry or interferometric endpoint detection—is an optical metrology technique that measures real-time etch depth, instantaneous etch rate, and remaining film thickness by analyzing the constructive and destructive interference fringes of monochromatic or broad-band light reflected from a wafer during dry etching. Unlike Optical Emission Spectroscopy (OES), which relies on changes in plasma chemical glow when a bulk layer clears into a stopping material, IEP endpoint operates directly on optical phase shifts between the receding top surface and underlying static film interfaces. This makes IEP indispensable for "blind" recess etching—such as shallow trench isolation (STI) silicon recess, nanosheet source/drain cavity formation, and FinFET fin height trims—where no chemical stop layer exists. At a high level, how an IEP endpoint system directs light, counts sinusoidal fringes, and triggers recipe transitions looks like this:

A few principles are essential for understanding how IEP endpoint works, why it overcomes the blind-etch limitation of OES, and how fabs extract single-nanometer depth precision from sinusoidal optical signals:

The optical path length difference governs periodic fringe oscillation. When a coherent laser beam of vacuum wavelength $\lambda_0$ strikes a partially transparent or etching film of refractive index $n$ at normal incidence ($\theta = 0^\circ$), the incident wave splits into two primary components: beam 1 reflected directly from the receding top surface ($R_1$), and beam 2 transmitted into the film and reflected from the underlying interface ($R_2$). As the target film thickness $d(t)$ decreases, the optical path length difference ($\Delta = 2nd$) continuously diminishes. The reflected intensity alternates between constructive interference (intensity maxima) and destructive interference (intensity minima):

For a typical silicon dioxide film ($n \approx 1.46$) etched with a 670 nm red diode laser, one complete sinusoidal fringe period (peak-to-peak) corresponds to precisely $\Delta d = 670 / (2 \times 1.46) \approx 229.5\text{ nm}$ of removed material, while the distance from a peak to the adjacent valley corresponds to a quarter-wave shift of $\approx 114.7\text{ nm}$.

IEP functions where Optical Emission Spectroscopy fails. Conventional OES detects the appearance or disappearance of chemical reaction byproducts in the plasma glow ($SiF^*, CN^*$). While highly effective for clearing a film off a chemically distinct stop layer (e.g., $SiO_2$ clearing onto a $Si_3N_4$ etch stop), OES is completely blind when etching a material into itself—such as etching 120 nm into a bulk crystalline silicon substrate for shallow trench isolation, or recessing silicon-germanium ($Si_{0.7}Ge_{0.3}$) by a fixed 35 nm. Because the chemical reaction products do not change as the trench deepens, the OES trace remains completely flat. IEP provides continuous, real-time depth metrology throughout the entire step, enabling programmatic termination at exact specified depths.

Sub-fringe interpolation achieves sub-2nm stopping accuracy. A fundamental limitation of simple integer fringe counting is quantization: if a process engineer relies solely on counting whole peaks or valleys, the stopping resolution is limited to $\lambda / (4n)$ (approximately 50 nm to 115 nm depending on wavelength and material). To stop at arbitrary depths—such as exactly 37.5 nm of silicon recess—advanced IEP algorithms model the continuous sinusoidal function:

By tracking the instantaneous first ($dI/dt$) and second ($d^2I/dt^2$) time derivatives, the digital signal processor calculates the exact instantaneous phase angle $\theta(t) = \arccos\left(\frac{I(t) - I_0}{A}\right)$. The chamber controller terminates or throttles the RF bias at the precise fractional phase angle corresponding to the target specification, routinely delivering sub-2nm depth repeatability across wafer production lots.

Spot size, pattern density, and film roughness dictate signal quality. In production wafer fabs, deploying IEP involves strict optical and layout design rules:
1. *Dedicated Scribe-Line Test Pads:* While broad plasma emission (OES) captures the entire wafer, IEP focuses a coherent laser spot (typically 50 to 200 $\mu$m in diameter) through a quartz chamber viewport onto a dedicated, uniform test macro located in the die scribe lane.
2. *Diffraction & Pattern Averaging:* If the beam illuminates a dense array of active transistors instead of a blanket macro, diffraction and differing vertical recession rates between dense and isolated lines produce overlapping, out-of-phase interference patterns that wash out signal amplitude. Modern multi-wavelength spectrometers address this by collecting full reflectance spectra ($400\text{ to }1000\text{ nm}$) and fitting rigorous coupled-wave analysis (RCWA) models in real time.
3. *Surface Roughening & Loss of Coherence:* Plasma-induced micro-roughness or polymer deposition scatters incident photons diffusely, attenuating the specular reflection $R_1$ and damping fringe amplitude. Digital Kalman filters track signal envelope decay and dynamically boost gain to prevent lost fringe lock.

Read an IEP endpoint system through a *real-time-optical-phase-ruler* lens: rather than waiting for an interface to announce itself chemically, IEP continuously tracks the physical descent of the etch front nanometer by nanometer, providing the deterministic depth control required to sculpt the foundational active dimensions of modern sub-3nm transistors.