Direct imaging of surface states hidden in the third layer of Si (111)-7 × 7 surface by pz-wave tip

Abstract

The direct visualization of the surface states of dimers in Si (111)-7 × 7 is still challenging although the DAS model has been proposed 33 years ago. In this letter, based on the partial density of states of the atoms in the reconstructed layers examined by the first principles calculations, scanning tunneling spectroscopy images with a pz-wave tip were carried out, which show a skull pattern at ∼0.7 eV, and 9 bright short lines at ∼1.00 eV in each unit cell, exhibiting obviously the electronic states from the third-layer atoms and dimers. The improved sensitivity (intensity and resolution) of scanning tunneling spectroscopy was ascribed to the enhanced tunneling matrix elements for pz-wave tip compared with s-wave tip. This discovery provides us a platform to explore the quantum states hidden in the deep of the materials surface.

Low-temperature insulating phase of the Si(111)–7×7 surface

“…This energy scale is critical for the Si(111)-7×7 surface and makes different datasets compatible with a metallic 16,18,20 or insulating 17 ground state. Low-temperature STS data 21,22,31,32,47 show the opening of a hard or soft gap below 20 K. However, the published STS spectra are integrated over the unit cell, or taken only on some kinds of the adatoms, or not suitably corrected for the systematic error caused by the non-equilibrium transport of carriers, which affects the energy scale at low temperatures 21,31 . Hence, these analyses do not allow us to identify the adatom-dependent effects of a gap opening at E F .…”

Evidence for “Unusual” Exchange–Correlation on Si(111) 7 × 7: Limitations of Density Functional Calculations for Charge Transfer Interactions on Semiconductor Surfaces

“…In 2018, Wang et al studied the 7 × 7 from heavily boron‐doped (0.01 Ω cm) sample at 78 K and found that Si‐decorated tips resolved more features in the 7 × 7 images than Pt–Ir tips. [ 37 ] This followed the observations in 1986 that W tips that impacted the surfaces produced higher corrugations and sharper images which was exploited to obtain stable high‐resolution tips to study Si surfaces. [ 106 ] Both work attributed the higher resolution to having a Si atom(s) on the tip with its p z orbital.…”

“…[13] Spin-restricted calculations for possible magnetic surface states were also performed. As in all DFT calculations, the DAS AA bands are partially filled and produce a metallic ground state [21][22][23][24][25][26][27][28][29][30][31][32][33][34][35][36][37] which is contrary to current experimental measurements showing an energy gap at E F [13,94,99,100] While most of the occupied AA DOS show relatively narrow features, characteristic of localized states, an AA band occurs on the corner AA, as shown in Figure 4b, as low as À0.2 or À0.25 eV for a spin-unrestricted ferromagnetic configuration. [13] Such a ferromagnetic band as calculated is too close to E F to account for the state at %0.5 eV shown in Figure 3a but becomes a possibility as X-C corrections are considered.…”

“…[18][19][20] Much of our current understanding of silicon surfaces is based on the agreement of the measured surface states arising in empirical, [21,22] approximate, [23] and more rigorous "ab-initio" DFT energy band calculations for the lowest energy structure. [13,[24][25][26][27][28][29][30][31][32][33][34][35][36][37] All these calculations over the past 35 years support the DAS model, but experiments tell us otherwise. [10][11][12][13][14][15] This leads to a consideration of the weakness in current theoretical constructs and the limitations of the physics now used to model such 2D surface systems.…”

Artificial-intelligence-driven scanning probe microscopy

“…Supervised learning has been applied to SPM in a recent study 22 where ML assists a human operator in detecting and repairing a specific type of probe defect in the particular case of hydrogen-terminated silicon. In this work, a trained CNN assessed the quality of acquired SPM images; if necessary, a wellestablished probe-conditioning protocol was executed 23 . In another recent study 24 , ML was successfully used to determine imaging quality directly from a small number of acquired scan lines, without requiring complete images.…”