Effective suppression of interface states in recessed-gate MIS-HEMTs by TMAH wet etching

The effect of tetramethylammonium hydroxide (TMAH) treatment prior to gate dielectric deposition on the performance of recessed-gate AlGaN/GaN metal-insulator-semiconductor high electron mobility transistors (MIS-HEMTs) was investigated. Through the use of TMAH wet etching, a low roughness etched surface of 0.173 nm was obtained. The capacitance–voltage characteristics of MIS heterostructures showed that the interface states reduced by one order of magnitude. When the temperature was increased to 473 K, the treated MIS-HEMTs delivered a small threshold voltage shift (ΔV TH) of ∼−0.53 V. From the dynamic measurement, the ΔV TH obtained without treatment was observed more severely (∼−1 V) when compared to the treated one (∼−0.01 V).


G
[3] Although p-GaN HEMTs have been commercially used, the small gate swing and high leakage of the devices limit further development.To solve this problem, the recess etching method, which removes part or all of the barrier layer under the gate, has attracted increasing attention. 4)][7] The Cl-based inductively coupled plasma (ICP) dry etching technology is the most common method for fabricating the thin barrier layer device, but this process leads to surface roughness and etching plasma damage. 8,9)ecently, substantial work has been done on surface optimization solving the damage problems associated with dry etching. 10)Joglekar et al. applied tetramethylammonium hydroxide (TMAH) treatment to improve the sidewall surface of AlGaN/GaN heterostructure, getting a more vertical profile and lower regrown ohmic contact resistance. 11)Yoon et al. minimized the surface current in AlGaN/GaN MIS-HEMTs by using TMAH-based wet solution prior to surface passivation. 12)n this paper, we investigate the effects of TMAH wet etching by improving the interface states of recessed-gate MIS-HEMTs.The ICP etching with higher RF sources is applied, and the recessed AlGaN areas display fewer protrusions and pits.X-ray photoelectron spectroscopy (XPS) analysis of the Ga 2p 3/2 core level confirmed that the surface oxide layer can be further removed by using TMAH wet etching, which contributes to a suppressed distribution of interface states. 13)The treated devices have lower interface trap density and obtain a more stable threshold voltage.
The AlGaN/GaN heterostructure epitaxial material was grown through metal-organic CVD (MOCVD) on a p-type Si (111) substrate.The structure included from bottom to top: a 3.5 μm thick buffer layer, a 200 nm GaN channel layer, a 1 nm AlN spacer layer, a 25 nm thick AlGaN barrier layer with 25% Al composition, and a 2.5 nm thick GaN cap layer, exhibiting an electron mobility > 1800 cm -2 V -1 •s -1 and a 2DEG concentration of 8 × 10 12 cm −2 .
For the AlGaN barrier recess of the photoresist (PR) masked areas, ICP etching was employed with RF 40 MHz for 12 min; RF power of 55 W and BCl 3 as a primary etching gas at a low chamber pressure of 6 mTorr.The as-grown AlGaN surface exhibited an RMS roughness of 0.137 nm.As shown in Figs.1(a) and 1(b), the recessed AlGaN areas displayed fewer protrusions and pits, resulting in an RMS of 0.230 nm.For the TMAH-treated sample, wet etching was performed as described in the device fabrication section.The treated surface achieved good step morphology and the RMS was only 0.173 nm.Due to the anisotropic etching characteristic, the TMAH solution preferred to etch the side slopes of the sharp protrusions in the gate recess area, ultimately leading to smoothing of the surface morphology. 14)igures 1(c) and 1(d) compare the measured Ga 2p 3/2 corelevel spectra for samples without TMAH and with TMAH respectively.To obtain the XPS spectra, the Al 2 O 3 /AlGaN film surface is subjected to low-energy Ar ion etching at a slow rate of less than 1 nm min −1 , effectively minimizing etchinginduced damage.In the Ga 2p 3/2 core-level spectra, the binding energies of Ga-O and Ga-N are located at 1117.50 ± 0.05 eV and 1116.25 ± 0.05 eV, respectively. 15,16)The percentage contribution of Ga-O is reduced from 45.5% to 36.6%.This is because TMAH wet etching can effectively reduce the O component on the AlGaN surface. 17)Characterizations by atomic force microscopy (AFM) and XPS show that the TMAH treatment reduces the roughness of the AlGaN surface through anisotropic etching, eliminating the surface oxides and improving the interface quality.
As shown in Fig. 1(e), the device fabrication started from etching the gate region.The region (L G /W G = 2/100 μm) was etched with the PR mask by partially removing the 11-nmthick AlGaN layer using ICP.With the PR mask, 2.38% TMAH wet etching was immediately followed at 85 °C for 5 min.Then, the device active region was isolated by nitrogen ion implantation.After that, the ohmic metals were formed by e-beam evaporation of Ti/Al/Ni/Au and annealed at 875 °C for 30 s in ambient N 2 without cap layer protection.The contact resistance of ∼0.6 Ω•mm was derived by measuring linear transmission line method patterns.Following this, the 10-nm-thick Al 2 O 3 layer was deposited as a gate insulator on both the TMAH-treated and untreated AlGaN surface through thermal-based atomic layer deposition.The Ni/Au gate metal was deposited by the e-beam evaporation and none of the samples were subjected post-gate annealing.Finally, a 200 nm Si 3 N 4 layer was deposited by plasma-enhanced CVD on top of the wafer as a passivation layer.The gate-source (L GS ), gate-drain spacing (L GD ), and gate length (L G ) were 5 μm, 8 μm, and 2 μm, respectively.The recessed-gate MIS-HEMTs without TMAH treatment were fabricated simultaneously on the same sample.
The AC capacitance-voltage (AC-CV) measurement is performed on the recessed-gate MIS diodes with different frequencies ( f ) varying from 10 kHz to 1 MHz at 473 K.As shown in Fig. 2(a), a significant frequency dispersion appears at the second slope, which is due to the higher trapping/ emission electron of the interface states between the dielectric layer and AlGaN, resulting in additional capacitance and impeding the presence of second rising slope. 18)As demonstrated in the inset of Fig. 2(a), with the frequency f increasing from 1 MHz to 4 MHz at RT, there is a significant frequency 011004-2 © 2023 The Author(s).Published on behalf of The Japan Society of Applied Physics by IOP Publishing Ltd attenuation and a pronounced capture of carriers by the AlGaN barrier layer, as a result of the higher interface trap density. 19)y using TMAH treatment as a post-gate-recess process [Fig.[22] By analyzing the second rising slope in the AC-CV characteristics, the interface trap mapping in III-N MIS heterostructures can be determined. 23)Based on this principle, the detectable energy range of the interface state energy (E T ) is represented by where D it is the interface trap density, E f ∆ ( ) are the interface trap energy levels responding to small AC-signals at frequencies f 1 and ( ) represent the voltage at the beginning of the second step at f 2 and f 1 , and C B and C OX represent the barrier and dielectric capacitances.In order to map the interface traps in MIS heterostructures, multi-temperature/ frequency (multi-T/f ) AC-CV techniques, focusing on the second slope, have been employed [Fig.2(c)].By virtue of AC-capacitance method, D it with E C -E T from 0.30 to 0.61 eV of the MIS heterostructures with and without TMAH treatment is determined to be ∼3.17× 10 13 -1.01 × 10 14 cm −2 eV −1 and ∼8.93 × 10 12 -5.96× 10 13 cm −2 eV −1 .From D it -E T mapping, D it in the treated device is reduced by an order of magnitude compared with the interface obtained by ICP etching, which can effectively suppress the frequency attenuation.
As shown in Figs.3(a) and 3(b), the TMAH-treated recessedgate MIS-HEMTs deliver more stable DC performances with the temperature measurement (T m ) elevated from 300 K to 473 K.The treated device exhibits improved V TH shift (ΔV TH ) of −0.53 V at V DS = 5 V, which is almost a 50% reduction compared to ΔV TH of −0.8 V for the untreated device.
Due to the strong polarization effect, the Fermi level position is deep into the bandgap at the dielectric/AlGaN interface. 24)Figures 3(d) and 3(e) show the band diagrams at pinch-off, E F relative to E C at Al 2 O 3 /AlGaN interface in the untreated device is deeper than the interface in the treated device, as a result of the higher D it and strong polarization field in the AlGaN barrier layer.By using Shockley-Read-Hall statistics, the time constants for electron emission (t e ) and electron capture (t c ) from the interface trap energy level to the conduction band can be calculated: where v , th s , n N , C and N s are the electron thermal velocity, capturing cross section, density of state at the conduction band, and electron concentration at the interface, respectively.Compared to the electron capture, electron emission is more important (t c <t e ). 25)At high temperature T 2 , interfacial traps with a wider energy range [ E T T 2 ∆ ( ) in Figs.3(d) and 3(e)] emit electrons and become empty during the DC scan of V G . 23,26)That is because the electron emission process of the interfacial trap increases exponentially at a higher temperature.Higher ΔV TH with a higher temperature towards negative bias direction in the case of without TMAH treatment, can be argued as follows.Since the density of the interface states is larger without TMAH treatment as analyzed by AC-CV measurement, more electrons from filled trap states are emitted, which are "frozen" at T 1 (RT) but can be emitted at higher temperature T 2 .The quantity of emitted electrons corresponds to an effective charge (ΔQ + ) comparing the high temperature characteristic to the RT characteristic and, hence, V TH is shifted towards negative bias.Without TMAH treatment, ΔV TH is larger because more electrons are emitted (larger ΔQ + ).In comparison, the treated device exhibits enhanced V TH -thermal stability, as the number of interfacial traps in the range E T T 2 ∆ ( ) above the Fermi energy level is lower. 27)nsidering that the excitation of trap states at the interface is a dynamic process, a double pulse I-V approach was also employed to characterize the transfer characteristics.Pulse transfer curves were obtained at the pulse period of 100 ms, with a quiescent (Q) state duration of 99 ms.A stress pulse ranging from −7 V to 1 V with the step of 1 V was applied to the gate, while a 0 V stress pulse was applied to the drain.In   011004-4 © 2023 The Author(s).Published on behalf of The Japan Society of Applied Physics by IOP Publishing Ltd the non-quiescent (NQ) state, the gate voltage (V GS, NQ ) was swept from −8 V to 2 V at V DS, NQ of 5 V to obtain the dynamic transfer characteristics, as shown in Fig. 4(a).
As depicted in Fig. 4(b), the untreated device exhibits pronounced variations in the measured I D -V GS, NQ curves at different V GS, Q .It is evident that at V GS, Q > −3 V, the obtained current values exceed the DC current with a gate voltage sweeping from −4 V to 2 V.When V GS, Q is less than −4 V, the I D measured under the pulse test is generally lower than the DC value.In contrast, the treated devices show minimal changes in I D value [Fig.4(c)].
The V TH of the device was extracted under the criterion of I D = 1 mA mm −1 through the pulse transfer curves, as shown in Fig. 4(d).For the different treatment, ΔV TH exhibits significant variations at different V GS, Q .Without the wet etching repairment, when the pulsed stress exceeds −3 V, ΔV TH shows an average shift of −1 V.When the pulsed stress is less than −4 V, ΔV TH shows an average shift of 0.6 V.However, with TMAH treatment, the maximum value of ΔV TH for the device is −0.16V, and the value fluctuates around ±0.01 V.
As shown in Fig. 5, ΔV TH in the pulse I-V measurement may be associated with the corresponding interface states.Under the condition of short stress time and the gate bias at Q state, the interface states are excited.Upon switching to the NQ state, the emitted and captured electrons are not released in time, leading to the shift of V TH .However, during the static testing, there is sufficient time for interface states to recover, and ΔV TH is slight.When V GS, Q is greater than V TH at the static testing, electrons are emitted from the Al 2 O 3 /AlGaN interface [Fig.5(a)], resulting in additional accumulation of negative charge.Conversely, electrons are captured at V GS, Q < V TH at the static testing [Fig.5(b)], leading to additional storage of positive charges. 28)For the sample without the treatment, significant ΔV TH is observed, which is probably due to the large amount of V GS, Q -induced electron emission or capture.But no obvious ΔV TH is observed in the treated device, suggesting that the reduced interface states are the dominant cause for improving threshold voltage stability.
In summary, the effect of TMAH treatment on the improved performance of recessed-gate MIS-HEMTs has been investigated.Compared to the untreated devices featuring manifest thermally induced V TH shift, the TMAHtreated devices exhibit thermally stable V TH of −0.53 V.By virtue of the AC-capacitance method, D it in the treated devices is determined to be ∼8.93 × 10 12 −5.96× 10 13 cm −2 eV −1 with E C -E T from 0.30 to 0.61 eV.The improved high temperature stability of treated recessed-gate MIS-HEMTs is attributed to the reduction of interface trap density.Pulse I−V measurements further demonstrate that the enhanced interface quality by TMAH process could contribute to threshold voltage stability in MIS-HEMTs, which benefits from the suppressed interface states.The results show that TMAH treatment can provide a simple and tractable method for optimizing recessed-gate MIS-HEMTs.

Fig. 1 .
Fig. 1. 1 × 1 μm 2 AFM images of the surfaces with ICP etching for 12 min: (a) without TMAH treatment, and (b) with TMAH treatment.XPS images of the surfaces after ICP etching for 12 min: (c) without TMAH treatment, and (d) with TMAH treatment.(e) Schematic device structure of the recessed-gate MIS-HEMTs.

Fig. 2 . 3 ©
Fig. 2. Frequency-dependent C-V measurement of the recessed-gate MIS diodes (a) without TMAH treatment and (b) with TMAH treatment at 473 K. Inset: the AC-CV characteristics at RT. (c) D it -E T mapping in the recessedgate MIS diodes, using multi-f AC-CV techniques.Inset: structure cross section of Al 2 O 3 /AlGaN/GaN diode.

Fig. 3 .
Fig. 3. Multi-T transfer characteristics of the recessed-gate MIS-HEMTs (a) without TMAH treatment and (b) with TMAH treatment, with T m increasing from 300 K to 473 K. (c)-(d) Schematic band diagrams of trap state excitation at different temperatures (T 1 and T 2 ) for the recessed-gate MIS-HEMTs (c) without TMAH treatment and (d) with TMAH treatment.

Fig. 4 .
Fig. 4. (a) A schematic of the double pulse I-V approach to obtain the transfer characteristics.(b)-(c) The transfer characteristics with the V GS, Q varied from −7 to 1 V during the double pulse measurement for the recessedgate MIS-HEMTs (b) without TMAH treatment and (c) with TMAH treatment.(d) The variation of threshold voltage with the applied gate stressing voltage under the dynamic measurement.

Fig. 5 . 5 ©
Fig. 5. Schematic band diagrams for electron emitting or capturing of the Al 2 O 3 /AlGaN interface when measuring at V GS, NQ and (a) V GS, Q <V TH and (b) V GS, Q > V TH .