Optical metrology for efficient manufacturing of high-quality VCSELs

FBH research: 04.03.2025

Fig. 1: Post growth wafer maps, showing the uniformity of the cavity dip position.

The image has three graphs: Left (a): A reflectance spectrum with wavelength (x-axis) and reflectance (y-axis). Dotted and solid lines show measured and fitted data, with a dip between ~759.7 nm and 835.8 nm. Middle (b): Another reflectance spectrum for a fine-tuned VCSEL, with a slightly shifted dip (~760.2 nm to 825.6 nm). Right (c): A time-dependent reflectance graph (x: process time, y: reflectance at 488 nm). Measured (symbols) and fitted (line) data track growth, with pink and red curves showing n- and p-DBR deposition and temperature changes. — Fig. 2: (a) Measured (dotted) and fitted (solid) reflectance spectrum of a reduced 795 nm VCSEL layer sequence with 10 DBR periods for p- and n-DBR. (b) Measured (dotted) and fitted (solid) reflectance spectrum of the fine-tuned full VCSEL structure. (c) Measured (symbols) and fitted (line) in-situ 488 nm growth transient for the same VCSEL layer sequence as shown in (a) and (b).

Graph showing measured and simulated etch transients for a VCSEL layer sequence. The x-axis represents process time (s), and the y-axis shows reflectance. Additional curves in green and pink indicate the modeled etch rate and aluminum mole fraction. Labels mark p-DBR and n-DBR regions. — Fig. 3: Measured (symbols) and simulated etch transients (blue – from in-situ growth transient fit; red – from WLR fit) for the VCSEL layer sequence shown in Fig. 2.

The very narrow cavity of vertical cavity surface emitting lasers (VCSELs), which defines the emission wavelength, requires precise control of layer thicknesses and material composition. While state-of-the-art MOVPE machinery usually achieves ±1% growth rate accuracy, VCSELs require sub-percent precision. Only the use of in-situ metrology allows for corrections from run to run without time-consuming and inefficient post-growth characterization. The obtained data can then be fed forward to allow for precise etching during processing, exemplified with VCSELs emitting at 795 nm for a space mission.

To tune the Fabry-Perót dip (FP-Dip) defining emission wavelength and the stop band center (SBC) of the mirror reflectivity, white light reflectance (WLR) with a LayTec EpiX mapping system is used to optimize on-wafer uniformity (Fig. 1). However, this information relates to the whole layer stack and cannot identify root causes for deviation. Our calibration starts with growing a reduced VCSEL structure (Fig. 2a). The WLR spectrum is modelled and fitted to determine the adjustments for the full structure, including all DBR periods. This is assisted by high-resolution X-ray (HR-XRD) diffraction for Al mole fraction determination and in-situ growth rate determination to deduce the thicknesses of the two quarter wave layers of a DBR period. Fig. 2b shows the corresponding WLR spectrum of the full VCSEL structure with the FP-Dip position now very close to the 795 nm target wavelength as a result of reducing the quarter wave layer thicknesses of n- and p-DBR by 0.5 nm each.

In-situ VCSEL growth analysis is conducted using a spectroscopic in-situ metrology system (LayTec EpiCurveTT VCSEL) that provides multi-wavelength reflectance, emissivity-corrected temperature, and wafer curvature. Automatic analysis is triggered from the growth recipe after each DBR period. With the knowledge of the SBC wavelength target at growth temperature, we can spot deviations during the first half of the n-DBR growth, which can be counteracted for each individual wafer by adjusting process parameters for the subsequent layers. This way, cavity resonance and p-DBR stop band wavelengths will shift accordingly, so that the VCSEL as a whole will have a good alignment between n-DBR, cavity and p-DBR resulting in a cavity resonance wavelength inside the targeted window.

The plasma etching of the VCSEL mesa is monitored and controlled using in-situ optical metrology with multi-wavelength reflectance (LayTec TRIton). A modelled layer stack simulates the etch transient, using either ex-situ WLR data (Fig. 2b) or in-situ reflectance data (Fig. 2c). For the latter, we perform a full simulation of the growth transient yielding excellent agreement with the measurement. The simulation is then repeated by removing layers from the full structure yielding the transient expected in the etching process.

Fig. 3 shows WLR-based (red) and in-situ-based (blue) simulated transients, nicely aligned with the measured etch transient (symbols) after adjustment of the etch rates (green lines). The simulated etch transient shows all the features that are present in the measurement. This way it is possible to use the simulated trace to reliably hit the desired stopping point of the plasma etching process. Thus, connecting in-situ and ex-situ measurements of growth and device process combined with simulation of the expected results allows for efficient and “first time right” manufacturing of complex devices.

This work was funded by the German Federal Ministry of Education and Research within the QYRO project [1].

Publication

[1] BMBF, “QYRO: Development of a miniaturized quantum-based rotation sensor,” Quantentechnologien.de.
https://www.quantentechnologien.de/forschung/foerderung/leuchtturmprojekte-der-quantenbasierten-messtechnik/qyro.html