Silicon Dopant Quantification in Atom Probe Tomography

K. DeRocher, M. McLean, F. Meisenkothen
National Institute of Standards and Technology,
United States

Keywords: atom probe tomography, dopant quantification


With the ever-increasing demand for semiconductor devices to shrink into the 3 nm regime without compromising (and in many cases, enhancing) computing power, there is a growing need to characterize these materials at device dimensions [1]. One important challenge is the accurate quantification and localization of dopant atoms, knowledge of which is vital to make sure the device functions properly. Atom probe tomography (APT) has become a valuable tool to perform this kind of analysis on semiconductor device features on the nano-scale [2, 3, 4]. APT utilizes a needle-shaped sample usually prepared by focused ion beam milling. In the atom probe, this sample is subjected to a high electric field and repeated voltage, or laser (thermal) pulses are used to evaporate atoms or clusters of atoms from the apex of the sample tip [5, 6]. These ions accelerate through the electric field and hit a 2-dimensional, position sensitive detector that measures the ion flight time (which is related to its identity), the arrival sequence, and the position of the impact on the detector. Together, this information can be combined to create a 3D reconstruction of the original sample. While this characterization technique can deliver chemical information down to concentrations of 10’s of µg/g (ppmat), it has also been shown that, for some materials, experimental conditions affecting the electric field in the atom probe can bias the quantitative analysis results [7,8]. Previously we’ve found that a standards-based approach can dramatically improve the accuracy of APT composition measurements in phosphorous-doped silicon. Using a reference material with a known dose of phosphorous (NIST SRM 2133, P implant in Si depth profile standard, [9]), we carried out APT measurements under a variety of analysis conditions and constructed a calibration curve we then applied to a sample doped with a known concentration of phosphorous. Comparing our composition measurement without correction to values obtained after application of the calibration curve showed an improvement from 26% relative error to less than 4%. We’ve now extended this analysis approach to arsenic-doped silicon. Using another reference material (NIST SRM 2134, As implant in Si depth profile standard, [9]) we performed APT analysis at varying field strengths to generate a calibration curve. We then analyzed an As-doped Si wafer sample after first establishing the arsenic concentration using secondary ion mass spectrometry (SIMS). Application of the calibration curve to data from the wafer sample improved both the accuracy and repeatability of our As concentration measurement. Looking forward, we plan to apply a similar measurement strategy to improve APT analysis accuracy in other materials systems [10].