Semiconductor devices are some of the most complex engineered devices in today’s world. Their manufacturing often requires an even more complex sequence of steps, first to understand and tailor the material properties so that desired electrical, optical or mechanical properties are possible, followed by very intricate and controlled fabrication processes so that these devices can be made reliably and reproducibly and at a reasonable cost. The above sequence of events demands that the practitioner is able to measure the material and device properties at every step of the design and production process to ensure that yields are optimal and to remove defective material or components as early as possible in the production process as every step add cost and complexity. A vast number of techniques have been developed for these types of material, product and process characterization. Amongst these, optical or spectroscopic techniques are usually preferred. The preference for optical spectroscopy in semiconductor material and process characterization derives from several factors. Optical spectroscopic (OS) techniques are often non-contact, hence reducing the chances of contamination. Secondly,manyOS techniques are comparatively fast and require little to no sample preparation.
Given the large number of material and process parameters that can affect the behavior of the final semiconductor product, it follows that control of the material design and fabrication process is necessarily a multi-modal effort. Meaning that many measurement techniques are required to measure and control all the parameters that lead to desired device behavior. Instrument vendors have accordingly developed a wide variety of specialized optical spectroscopy instruments often specialized on each of the required techniques. In addition, as device features go from micro to nano, many of these instruments are based on a microscope or other submicron to nanometer measurement platforms. For example, it is commonin a semiconductor research or fabrication facility to have one micro-Raman spectroscopy instrument used to characterize crystallinity of epitaxial deposition or stress and a separate Photoluminescence instrument to measure wafer homogeneity, etc. Beyond the cost burden of having multiple instruments to perform these necessary measurements, the task itself, in going from instrument to instrument has become quite challenging in recent time as the features of interest become smaller and approaching the micro to nanoscale.
In this article, we describe a novel approach to achieve multi-modality on one microspectroscopy platform (fig 1), enabling the practitioner to characterize various semiconductor samples using different spectroscopies with the benefit of sample co-location. In this approach, various complementary measurements can be performed at the same micro location and in so doing obtain deeper insights into sample or process. In addition, when there is a necessity to perform micro or nano measurements across different instruments, we present a new coordinate transformation technology (nano-GPS) that enables fast and accurate location of nanostructures across different measurement instruments.
Example 1: LEDs
One of the most important considerations of LEDs is the emission wavelength. For example, most common Lidar sources have semiconductor materials designed to havelight emission at 905nm or 1550nm. The primary material property that controls this parameter is the optical bandgap, which is readily measured using micro photoluminescence (fig 1a). After the material properties are understood, a fabrication process engineer could be interested in the uniformity of the epitaxial deposition on a large wafer, so that a Lidar laser die cut from one part of the wafer can be expected to perform and behave as one from any other part of the wafer. Once again, this property is readily characterized using a photoluminescence (PL) map across the wafer (figure 1b). If for some reason some of the dies from this wafer do not show optimal luminescence compared to others, the process engineer might be interested in understanding the nature of the defects causing this suboptimal luminescence efficiency and could use time-resolved PL to undertake these studies (figure1c). Alternatively, and as part of a QA process the engineer might want to measurement the performance of the device under conditions similar to actual use by measuring the electroluminescence of the device before it is packaged (figure 1d).