Atomic imaging of atomic layer deposition oxide nucleation with trimethylaluminum on As-rich InGaAs(001) 2 × 4 vs Ga/In-rich InGaAs(001) 4 × 2

Abstract

Formation of a contaminant free, flat, electrically passive interface to a gate oxide such as a-Al(2)O(3) is the critical step in fabricating III-V metal oxide semiconductor field effect transistors; while the bulk oxide is amorphous, the interface may need to be ordered to prevent electrical defect formation. A two temperature in situ cleaning process is shown to produce a clean, flat group III or group V rich InGaAs surface. The dependence of initial surface reconstruction and dosing temperature of the seeding of aluminum with trimethylaluminum dosing is observed to produce an ordered unpinned passivation layer on InGaAs(001)-(4 × 2) surface at sample temperatures below 190 °C. Conversely, the InGaAs(001)-(2 × 4) surface is shown to generate an unpinned passivation layer with a seeding temperature up to 280 °C. For both reconstructions, the chemical drive force is consistent with formation of As-Al-As bonds. The optimal seed layer protects the surface from background contamination.

Arsenic decapping and pre-atomic layer deposition trimethylaluminum passivation of Al2O3/InGaAs(100) interfaces

“…dangling bond density over a wide temperature range from 25 C and 280 C. 13 These prior STM/STS results are consistent with our C-V result showing that the effect of TMA dosing on the interface defect density appears to be temperature independent.…”

“…Figure 1 plots multi-frequency C-V curves for the MOS capacitors fabricated on n-type and p-type InGaAs(100) surfaces which were decapped at 370 C (low-T decap) and 460 C (high-T decap) followed by the normal ALD-Al 2 O 3 procedure without any TMA pre-dosing. Based on previous studies of InGaAs(001) with As 2 decapping and annealing at similar temperatures to those used in this research, 13,14 it is believed that the InGaAs surfaces decapped at 370 C and 460 C in this research form predominantly As-rich 2 Â 4 surface reconstructions and In/Ga-rich 4 Â 2 surface reconstructions, respectively. From Figure 1, the high-T decapped n-InGaAs sample (In/Ga-rich surface) shows larger frequency dispersion over the gate bias range from À2 V to 0 V and larger C-V stretch-out than the low-T decapped n-InGaAs (As-rich surface) capacitor, which indicates a higher density of interface defects on the high-T decapped n-InGaAs.…”

“…This suggests that the Fermi level may be pinned between the midgap and the valence band (VB) edge of the InGaAs semiconductor, 15 which is consistent with a previous report that the Fermi level is pinned on a clean n-InGaAs(100) 4 Â 2 surface, and that TMA dosing at 270 C (the ALD-Al 2 O 3 temperature) does not move the position of the Fermi level. 13 We can also observe that the accumulation capacitance density of the high-T decapped capacitor is slightly smaller than that of the low-T decapped sample. There are several possible reasons for this, including more efficient initiation of ALD on this surface than on the low-T decapped substrate, or possible native oxide formation on the high-T decapped InGaAs surface because of the longer exposure of the decapped InGaAs surface to the background oxidant in the ALD chamber at high temperatures during heating and cooling of the substrate.…”

Influence of plasma-based in-situ surface cleaning procedures on HfO2/In0.53Ga0.47As gate stack properties

“…To reduce the large trap densities at interfaces between In 0.53 Ga 0.47 As and the gate dielectric (typically Al 2 O 3 or HfO 2 ), various surface preparation and passivation methods are currently being investigated. [1][2][3][4][5][6][7][8][9] Interface trap densities (D it ), at least around midgap, are relatively straight-forward to detect for In 0.53 Ga 0.47 As, because of its narrow band gap (0.75 eV (Ref. 10)); they cause, for example, a frequency-dependent hump in the depletion region of capacitance-voltage (CV) characteristics, 11,12 and peaks in normalized conductance maps.…”

The influence of surface preparation on low temperature HfO2 ALD on InGaAs (001) and (110) surfaces

“…Details of the (2 × 4) reconstruction are given elsewhere. 20,21 Samples were subsequently exposed to atomic H generated from an Oxford Applied Research TC-50 thermal gas cracker. The hydrogen doses are reported in Langmuir (1 L = 1 s at 1 × 10 −6 Torr) based on the H 2 pressure; however, the cracking efficiency is reported to be closer to 50% (Oxford Applied Research) but could not be verified; therefore, the reported atomic H doses are an upper limit.…”