Atomic defect classification of the H–Si(100) surface through multi-mode scanning probe microscopy

Abstract

The combination of scanning tunnelling microscopy (STM) and non-contact atomic force microscopy (nc-AFM) allows enhanced extraction and correlation of properties not readily available via a single imaging mode. We demonstrate this through the characterization and classification of several commonly found defects of the hydrogen-terminated silicon (100)-2 × 1 surface (H–Si(100)-2 × 1) by using six unique imaging modes. The H–Si surface was chosen as it provides a promising platform for the development of atom scale devices, with recent work showing their creation through precise desorption or placement of surface hydrogen atoms. While samples with relatively large areas of the H–Si surface are routinely created using an in situ methodology, surface defects are inevitably formed reducing the area available for patterning. By probing the surface using the different interactivity afforded by either hydrogen- or silicon-terminated tips, we are able to extract new insights regarding the atomic and electronic structure of these defects. This allows for the confirmation of literature assignments of several commonly found defects, as well as proposed classifications of previously unreported and unassigned defects. By combining insights from multiple imaging modes, better understanding of their successes and shortcomings in identifying defect structures and origins is achieved. With this, we take the first steps toward enabling the creation of superior H–Si surfaces through an improved understanding of surface defects, ultimately leading to more consistent and reliable fabrication of atom scale devices.

Ionic charge distributions in silicon atomic surface wires

“…While we use a height reference defined by these STM setpoints, previous work has estimated this to correspond to an absolute tip-sample distance of ∼700 pm as established by Rashidi et al 18 where they defined z = 0 as a point where noticeable changes in tip-structure from surface contact occurred. It is also found that the absolute tip-sample distance can vary slightly (∼100 pm) due to different apex functionalizations, 54 tip geometries, and surface dopant profiles. Thus, we stick with the easily accessible STM set-point to enable easy reproduction of our results.…”

“…88 Further in situ conditioning was done via controlled tip contacts with hydrogen desorbed patches of silicon until it returned characteristics corresponding to a Si terminated tip. 25,54 DB structures were created via controlled bias pulses. 20,23,89 The bias ranges for each set of measurements refer to the sample bias (V S ) and were chosen to probe as great a bias window as possible while trying to prevent any unwanted tip changes due to high tunneling current through the valence and conduction band of the crystal.…”

“…S1. † By using a charge sensitive Si-functionalized probe, 25,54 we are able to clearly distinguish between the positive, neutral, and negative charge states of DBs within bare dimers and DB wires, yielding insights into the association of DB charging with lattice distortion. We identify the formation of an ionic charge distribution within bare dimers and DB wires, with the DB wires only exhibiting ionic character when accompanied by changes in their overall charge state.…”

Extended iron phthalocyanine islands self-assembled on a Ge(001):H surface

“…Similarly, it has been reported that the passivation of semiconducting materials, which removes surface dangling bonds and significantly reduces surface reactivity, may also provide a sufficiently insulating layer for an efficient decoupling of molecular structures from the substrate influence. Among such surfaces, hydrogen-passivated Si(001):H [ 22 – 23 ], Si(111):H [ 24 ], and Ge(001):H [ 25 – 28 ] surfaces are most commonly mentioned. Iron phthalocyanines (FePc) have been studied on Si(111):H [ 24 ] and it was concluded that the molecules are weakly coupled to the substrate.…”

Predicting hydration layers on surfaces using deep learning

“…Our methodological approach builds upon the general philosophy of using machine learning to handle data analysis challenges in Scanning Probe Microscopy (SPM) [46][47][48][49] and the specic use of deep learning Convolutional Neural Networks (CNN) 50 to recognize features in high-resolution SPM images. Recent examples include conditioning of SPM tips, 51 identication of defects with STM 52,53 and nanostructures with AFM, 54 and making molecular structure predictions from AFM images. 55 However, to the best of our knowledge, no earlier studies have applied machine learning to SPM at solid-liquid interfaces.…”