edfas.org ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 24 NO . 4 4 EDFAAO (2022) 4:4-11 1537-0755/$19.00 ©ASM International® A GUIDE TO ACCURATE SYSTEM CALIBRATION AND DATA EXTRACTION TO INCREASE THE SIGNIFICANCE OF SPECTRAL PHOTON EMISSION MICROSCOPY MEASUREMENTS Norbert Herfurth1 and Christian Boit2 1IHP - Leibniz-Institut für innovative Mikroelektronik, Frankfurt (Oder), Germany 2Semiconductor Devices, Berlin University of Technology, Berlin, Germany herfurth@ihp-microelectronics.com INTRODUCTION Spectral analysis of photon emission signals measured in failure analysis of semiconductor devices has been discussed from time to time in several scientific publications.[1-4] The results presented by these publications have not always been fully conclusive. One reason is that no standardized calibration is available for spectral photon emission microscopy setups. This article shows that the development of a meticulous calibration procedure is an important prerequisite for the measurement and extraction of significant and meaningful spectral information from photon emission. Step-by-step instructions are given for performing a specific calibration and careful photon emission microscopy measurements. This includes the extraction of device parameters such as electron temperature from field-assisted photoemission, as well as extracting the material band gap by analyzing recombination-based emissions. Even though the calibration procedure presented here is performed on InGaAs detectors, it can easily be applied to all common photon emission detectors. Photon emission (PE) is a contactless fault isolation technique for integrated circuits (ICs) that uses the electroluminescence of active electronic devices. The intensity of this light emission is usually very faint and not easily extracted from background noise. This noise is not easily reduced as it is mainly caused by the thermal activity of the detector material itself. Therefore, dedicated cooling of the photon emission detector is always required to be able to detect the faint photon emission signals from semiconductor devices. In order to extract the spectral information from these emissions, the PE signal must be filtered for a certain wavelength or spatially spread, for example, with a prism. Both methods have the consequence that the radiant power density per detector pixel is reduced even further. This aggravates the problem of the already low signal-to-noise ratio in photon emission measurements. Therefore, it has not become common practice to isolate spectral information from PE measurements. But what information gain can be expected from a proper PE spectrum? The advantage is that device models can be associated with PE events. For example, a field-effect transistor (FET) in saturation emits a spectrum that decreases toward higher photon energy with an exponential function. From the slope of this function, the device parameter electron temperature (Te) can be derived. This parameter is correlated with the kinetic energy gained by channel electrons in the electric field.[5] This means the information that can be extracted from a single pixel of a spectrally resolved PE, e.g., from an FET, is not essential, but the number of pixels that contribute to a correct quantitative calculation of the slope is. Thus, it is more valuable to extract only a few datapoints with a high degree of confidence than to use all measured raw data from an improperly calibrated system. MOTIVATION Although the spectrum of an FET in saturation should be a pure exponential function of the photon energy, very divergent spectral distributions of photon emission obtained with an InGaAs detector have been published, often with sharply increasing or decreasing gradients at the spectral edges of the detector sensitivity or, evenmore drastically, with erratic valleys in the lower photon energy regime of InGaAs detectors.[1,6] These unrealistic results were a motivation to develop this issue in a methodically firm manner. The measurable PE intensity reduces
RkJQdWJsaXNoZXIy MTMyMzg5NA==