Al2O3 Films on

Subscriber access provided by UIC Library

Article

Interface Properties of Atomic Layer Deposited TiO2/ Al2O3 Films on In0.53Ga0.47As/InP Substrates Chhandak Mukherjee, Tanmoy Das, Chandreswar Mahata, Chinmay Kumar Maiti, Ching K Chia, Sing Yang Chiam, D. Z. Chi, and Goutam K. Dalapati ACS Appl. Mater. Interfaces, Just Accepted Manuscript • DOI: 10.1021/am405019d • Publication Date (Web): 28 Jan 2014 Downloaded from http://pubs.acs.org on January 30, 2014

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

ACS Applied Materials & Interfaces is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 54

ACS Applied Materials & Interfaces

1 3

2

Interface Properties of Atomic Layer Deposited 5

4

7

6

TiO2/Al2O3 Films on In0.53Ga0.47As/InP Substrates 9

8 10 1 12

C. Mukherjee1*, T. Das1** , C. K. Maiti1, G. K. Dalapati2, H. Gao2, C. K. Chia2, M. K. Kumar2, S. 15

14

13

Y. Chiam2, and D. Z. Chi2 16 17 18 19 **

20

Author correspondence electronic mail: [email protected]

21 23

2 1

24

VLSI Engineering Lab, Department of Electronics and ECE, Indian Institute of Technology-

25

Kharagpur, India 721302 26 27 *

28

Currently with IMS Laboratory (Laboratoire de l'Intégration du Matériau au Système), University of Bordeaux 1, UMR CNRS 5218. 16, Avenue Pey Berland. 33607 Pessac Cedex, Bordeaux, France, 32

31

30

29

Institute of Materials Research and Engineering, A*STAR, (Agency for Science, Technology and

2

3 35

34

Research), 3 Research Link, Singapore 117602 36 38

37

Abstract 39 40 42

41

Electrical and interfacial properties of metal-oxide-semiconductor (MOS) capacitors fabricated 4

43

using atomic layer deposited bilayer TiO2/Al2O3 films on In0.53Ga0.47As/InP substrates are 46

45

reported. Vacuum annealing at 350oC is shown to improve the interface quality. Capacitance47 49

48

Voltage (C-V) characteristics with higher accumulation capacitance, negligible frequency 51

50

dispersion, small hysteresis and low interface state density (~1.5×1011 cm-2eV-1) have been 52 53

observed for MOS capacitors. Low frequency (1/f) noise characterization and inelastic electron 54 56

5

tunneling spectroscopy (IETS) studies have been performed to determine defects and interface 57 58

1 59 60 ACS Paragon Plus Environment


1 3

2

traps and explain the lattice dynamics and trap state generation mechanisms. Both the IETS and 5

4

1/f noise studies reveal the spatial locations of the traps near the interface and also the nature of 6 7

the traps. The IETS study further revealed the dynamic evolution of trap states related to low 10

9

8

frequency noise sources in the deposited TiO2/Al2O3 stacks. It is shown that deposition of an 12

1

ultrathin layer of TiO2 on Al2O3 can effectively control the diffusion of As in the dielectric and 13 14

the oxidation states of In and Ga at the In0.53Ga0.47As surface. 16

15 17 19

18

Keywords: Atomic Layer deposition, Vacuum Annealing, InGaAs, Low Frequency Noise, 20 21

Inelastic Electron Tunneling Spectroscopy, Lattice Dynamics, Traps. 23

2 24 25 26 27 28 29 30 31 32 3 34 35 36 37 38 39 40 41 42 43 4 45 46 47 48 49 50 51 52 53 54 5 56 57 58


Page 2 of 54

Page 3 of 54


1 3

2

1. Introduction 4 6

5

Recently, III-V compound semiconductors have received intense research interest as possible 8

7

channel material in MOSFETs owing to their high bulk electron mobility and breakdown field 9 10

than that of silicon 1. The ternary alloy, such as In0.53Ga0.47As with lattice matched to InP is one 13

12

1

of the most attractive material due to its high low-field electron mobility and high saturation 15

14

velocity. Thermodynamic instability and poor quality of the gate insulator (low active defect 16 18

17

density is essential to avoid Fermi level pinning at the interface) have so far impeded their use as 20

19

channel materials 2. It is thus important to find suitable gate dielectrics and develop optimum 21 2

passivation techniques for III-V semiconductor surfaces. Alternative high-k gate dielectrics 23 25

24

provide an opportunity for considering III-V compound semiconductors as channel materials in 27

26

MOSFET applications 3, 4. 28 30

29

Atomic layer deposition (ALD) Al2O3 is known to provide a relatively low interface defect 31 3

32

density during growth 35

34

5, 6

and an unpinned Fermi level on InGaAs (100) substrates. Al2O3 has a

moderate dielectric constant compared to other high-k oxides and is known to provide a 36 37

relatively low interface defect density and an unpinned Fermi level on InGaAs(100) substrates. 40

39

38

Low dielectric constant of Al2O3 hinders scaling of gate oxide capacitance density 7, and also the 42

41

hysteresis voltage is high 8. On the other hand, titanium dioxide (TiO2) has a wide range of 43 4

dielectric constant of 10–86, and is a potential candidate as gate oxide in InGaAs MOSFETs. 47

46

45

Surface passivation of III-V substrates using TiO2 has shown that the hysteresis voltage is low 49

48

compared to other dielectrics, but the presence of a thick interfacial layer is observed for directly 50 51

deposited TiO2 9. Thus, it is difficult to find a single dielectric which satisfies all the 54

53

52

requirements for end-of-roadmap MOS gate dielectrics (high dielectric constant, low interface 56

5

trap density, high thermal stability, etc.), making bilayer gate dielectrics an interesting option. 57 58



Page 4 of 54

1 3

2

Over the past decade, various efforts to improve MOS capacitor performance have focused on 5

4

understanding the device physics, optimizing the morphology through advanced processing 6 7

methods, and developing high performance dielectric materials. 9

8

1

10

The performance of MOS devices is critically dependent on the properties of both the 13

12

active materials and their interfaces. When the gate dielectric is only a few atomic layers thick, 14 15

conventional dielectric characterization tools, such as infrared spectroscopy and Raman 16 18

17

spectroscopy are ineffective in revealing structural and compositional information of the gate 20

19

oxide. In contrast, the inelastic electron tunneling spectroscopy (IETS) technique is more 21 2

sensitive for metal-oxide-semiconductor structures with ultrathin gate dielectrics 25

24

23

10

, and is

suitable in the study of phonon density of states, superconducting gaps, electrode phonons, 27

26

barrier phonons and electron density of states 28

11, 12

of high-k dielectrics

13

. A wide variety of

30

29

information about MOS capacitors can be obtained by IETS with excellent sensitivity and 32

31

resolution, including: phonon modes of the gate electrodes, dielectric, and substrate; various 34

3

vibrational modes of the bonding structures at interfaces; impurities in the gate dielectric and at 35 37

36

interfaces; trap features and other electronic defects. It has long been known that the low 39

38

frequency noise behavior of MOS structures is governed by the trapping/de-trapping 40 41

mechanisms in the gate dielectric. From the nature and the magnitude of the 1/f noise spectra, the 42 4

43

physical mechanisms involved as well as the nature of traps and interface qualities can be 46

45

assessed. From both IETS and 1/f noise data the nature of traps and interface properties can be 47 48

investigated. 50

49

52

51

In this paper, we report on the interfacial properties of ultrathin (~9 nm) amorphous ALD 54

53

TiO2/Al2O3 bilayer gate dielectric on In0.53Ga0.47As substrates. Lattice vibrations and phonon 5 56

modes of In0.53Ga0.47As lattice are studied in detail and are described by analytical expressions, 58

57


Page 5 of 54


1 3

2

identification of the trap-features including trap-assisted conduction and charge-trapping are 5

4

performed. The trap features are compared for both the vacuum annealed and as-deposited 6 7

devices. The low frequency noise study is used to extract trap information and a comparison of 10

9

8

1/f noise in vacuum annealed and as-deposited devices is made. Although reports on IETS 12

1

studies are available in the literature for TiO2-Al2O3/ GaAs systems 14, reports on phonon modes 13 14

and lattice dynamics of TiO2-Al2O3/InGaAs systems using IETS are not. In the present work, 17

16

15

IETS is used as a non-destructive characterization tools to probe phonons of the substrate, gate19

18

dielectrics, traps and microscopic bonding structures to investigate into the lattice dynamics and 20 2

21

interface properties of atomic layer deposited Al/TiO2-Al2O3/In0.53Ga0.47As/InP MOS capacitors. 23 25

24

2. Experimental Details 26 27 28

2.1 Fabrication of ALD Al/TiO2-Al2O3/In0.53Ga0.47As/InP MOS Capacitor 29 31

30

Metal-oxide-semiconductor (MOS) capacitors were fabricated on p-In0.53Ga0.47As/InP 3

32

substrates. The wafers were degreased using isopropanol, cleaned in HF solution (1%) for 3 mins 34 36

35

to remove the native oxide and then dipped in NH4OH solution for 10 mins. A thin layer of 38

37

Al2O3 was deposited on p-In0.53Ga0.47As using trimethyl aluminium (SAFC Hitech, Haverhill, 39 40

MA, USA; 99.9%) and H2O as the precursors in a viscous flow-type (0.6 Torr working pressure) 43

42

41

atomic layer deposition equipment (f·XALD ALD equipment, Azimuth Technologies Pte Ltd., 45

4

Singapore) with a N2 flow rate of 50 sccm at 170°C. After that, TiO2 films were deposited under 46 47

similar conditions. Vapors of TiCl4 (Merck & Co., Inc., Whitehouse Station, NJ, USA; 99%) and 50

49

48

H2O precursors were sequentially introduced into the chamber with an exposure time of 0.1 s and 52

51

purged by 50-sccm N2 flow for 10 s between the two exposures. For Al2O3 deposition with 53 5

54

TMA, 30 cycles were used for the deposition. Typical growth rate was 1Å/cycle. For the ALD of 56 57 58



1 3

2

TiO2 films, using TiCl4+2H2O > TiO2+4HCl reaction, 175 cycles of TiCl4 were used for the 5

4

deposition of 7 nm TiO2 with a typical growth rate of 0.4Å/cycle 15, 16 . Vacuum annealing (VA) 7

6

was carried out for some sample at 350°C for 5 mins at process pressure of ~1×10-6mTorr. The 10

9

8

Al metal, deposited by evaporation, was used as the gate electrode (area: 2.5x10-4 cm2). Finally, 12

1

a low resistance ohmic back contact was formed by depositing Au on p-In0.53Ga0.47As substrate 13 14

by DC sputtering. 16

15

18

17

2.2 IETS Characterization Setup 19 20

The IETS measurements were performed at liquid nitrogen temperature (77K). The setup 21 23

2

included an E5263A 2-channel high speed source monitor unit, a SR830 lock-in amplifier and a 25

24

computer interface to control the bias sources to through GPIB connection and to capture the 26 27

measured spectra. E5263A 2-channel high speed source monitor unit provided the necessary gate 30

29

28

biases through the lock-in amplifier. The spectra were obtained by measuring the second 32

31

derivatives of the gate tunneling current (Ig) through the metal-insulator-semiconductor tunneling 3 34

junctions by the SR830 lock-in amplifier detection, with an AC modulation signal (generated 37

36

35

from oscillator output of the lock-in amplifier at f = 890 Hz with an amplitude (Vω) of 2 mV) 39

38

superimposed on a slowly varying dc gate voltage. The current, thus obtained, is amplified by an 40 42

41

operational amplifier. A time constant of 3 seconds was used in the experiment. A notch filter in 4

43

the lock-in amplifier was used to remove the unwanted harmonic frequencies. 45 46 48

47

3. Characterization and Analyses 49 51

50

3.1 Characterization Using XPS 53

52

In order to understand the interface quality between TiO2/Al2O3 and p-In0.53Ga0.47As, thin films 54 56

5

of TiO2/Al2O3 on p-InGaAs was prepared using the same process recipe and analyzed by x-ray 57 58


Page 6 of 54

Page 7 of 54


1 3

2

photoelectron spectroscopy high-resolution x-ray photoelectron spectroscopy (XPS) performed 5

4

under a UHV condition at room temperature with VG ESCALAB 220i-XL XPS system. The 7

6

take-off-angle was 90o using monochromatic AlKα (1486.7 eV) excitation sources. Figure 1(a)10

9

8

(d) shows the core-level spectra of Ga 3d and As 3d of TiO2/Al2O3/p- In0.53Ga0.47As for with and 12

1

without vacuum annealed sample. 13 14 15

In Figure 1 the indium and gallium bulk peaks are located at ~17.05 eV and 19.10 eV such that 16 18

17

the binding energy separation between the two is fixed at 2.05 eV. There are two peaks attributed 20

19

to indium oxide, consistent with those seen in the XPS spectra; at 0.57 eV and 1.15 eV from the 21 2

bulk peak, again tentatively assigned as In2O and a mixed phase, InxGayOz (with x>y initially) 25

24

23

respectively17. The remaining peaks are due to gallium oxides with a chemical shift at +0.55 eV 27

26

and +0.85 eV from the bulk Ga peak, attributed to Ga1+ (Ga2O) and Ga2+ (GaO) 18. From the Ga 28 30

29

3d spectrum it is evident that Ga-O feature has multiple oxidation states for different samples 32

31

including Ga1+ and Ga2+ probably due to Ga2O and GaO formation 19. It is important to note that 34

3

without annealed sample has resulted in a Ga-O peak that is centered at a lower binding energy 35 37

36

for bulk GaAs samples. But for vacuum annealed sample the resulting Ga–O feature with low 39

38

intensity presents in the higher energy position. According to Hinkle et al.20 it is possibly due to 40 41

charge redistribution from second nearest neighbor changes in the Ga-O bonding environment. 4

43

42

Thus vacuum annealing can suppress the formation of Ga2Ox native oxide at the interface with 46

45

higher oxidation state of Ga, which is more closely related to the interface defects formation, 47 48

causes to improve the interface quality 51

50

49

21

. The O 2s spectrum consists of the following two

peaks: Ga-O-Ti and Ga-O-Al bonds at 22.33 eV, and 23.9 eV attributed to formation of mix 53

52

bonding in the interface 54

22

. As the electron affinity of Al is quit higher that of Ti so we have

5 56 57 58



1 3

2

assumed that higher O 2s peak is related with Al-O-Ga. The O 2s spectra of these samples are 5

4

evident along with Ga3d spectra which are close to reported results 23. 6 8

7

Arsenic oxide is the well-known species that causes oxide-induced Fermi level pinning 9 10

through dissociation into As atoms on the GaAs surface.24 Figure 1(c)-(d) illustrates the As 3d 13

12

1

XPS spectra with fitted Gaussian curves of both the devices with and without vacuum annealing. 14 15

In order to obtain a valid fit for the As 3d spectra, we have considered doublets at ~40.03 and 16 18

17

~40.75 eV for metallic As bonding, where they have a peak with a separation of 0.7 eV and As20

19

O with a chemical shift of +9eV for without vacuum annealed sample indicates that the sample 21 2

is selectively oxidized during deposition. A model has been proposed by Robertson which 25

24

23

describes the defect density of the GaAs and InGaAs surface in terms of Ga and As dangling 27

26

bonds at the conduction and valence band edges 25 which would support the assignment of these 28 30

29

peaks. After vacuum annealing at 350°C all native oxides were completely removed, as shown in 32

31

Figure 1(d). This result suggests that VA is effective to eliminate native oxide and could 34

3

contribute to improvement of InGaAs MOS interface properties. Additionally strong Ti 3p signal 35 37

36

at 36.4-38.2 eV in the As3d region for TiO2/Al2O3 is a further evidence of TiO2 layer on Al2O3 39

38

which act as a barrier to As diffusion into the dielectric 26. Furthermore, after vacuum annealing, 40 41

the intensity of Ti 3p peaks increases, with a significant shift in the higher binding energy by 42 4

43

~0.7 eV being concurrently observed. This phenomenon may be correlated with the phenomenon 46

45

in which the thickness of the barrier layer (BL) region increases due to vacuum annealing 47 48

causing significantly suppression of arsenic oxides to and led to improvement of InGaAs MOS 51

50

49

interface properties. From Figure 1(a) and (b) it can be seen that the ratio of In-As: In2Ox has 53

52

been increased which is due to In related oxide increased after vacuum annealing. In terms of 54 56

5

Gibbs free energy of oxide formation, As oxide is less stable than the other oxides such as 57 58


Page 8 of 54

Page 9 of 54


1

5

4

3

2

Ga2O3, GaAsO4, In2O3, and InAsO4 (As2O3:−137.7 kcal/mol; Ga2O3:

-238.6 kcal/mol;

27

. Based on our XPS

GaAsO4:~-223 kcal/mol; In2O3:-198.6 kcal/mol; InAsO4:-209.4kcal/mol) 6 7

results, after ALD deposition the formation of the undesirable native oxide, especially for arsenic 10

9

8

oxide, appears to be significantly suppressed on In0.53Ga0.47As surface. 1 13

12

Figures 2(a) and (b) show the In 3d spectra for ALD TiO2/Al2O3 gate stack on p-In0.47Ga0.53As. 14 15

In-O bonds were evidenced by fitting G-L curves of the In3d spectra for the sample with and 16 18

17

without VA. Oxidation of In after ALD is proved by the In 3d5/2 line decomposition in a bulk 20

19

peak at 444.45 eV and an In2O3 (In3+ at 445.56) eV component. However, regardless of the exact 21 2

assignment of oxidation states, the spectra suggest that the formation of In-O is clearly increased 25

24

23

after vacuum annealing sample at 350°C. 26 28

27

Figure 2(c) illustrates the Ti 2p photoelectron spectra at the surface of the TiO2 films. Ti 2p 29 30

XPS spectra before sputtering displays mainly two peaks around 458.7 and 464.5 eV, 3

32

31

respectively, assigned to Ti 4+ 2 p3/ 2 and Ti 4 + 2 p1/ 2 in the TiO2. The two peaks are located with 34 35

energy separation of 5.8 eV and from G-L fitting an area ratio of Ti 4 + 2 p3 / 2 : Ti 4 + 2 p1/ 2 was found 38

37

36

as 3:1. These results show that the film at the surface is fully oxidized stoichiometric of Ti-O 28. 39 40

The shape of the doublet provides a little evidence for presence of non-stoichiometric oxide of 43

42

41

Ti 3+ 2 p3 / 2 and Ti 3+ 2 p1/ 2 which may come from Aluminum oxide layer. The extra peaks after 4 46

45

1min etching indicate the lower contents of O in Ti 3+ 2 p3 / 2 and Ti 3+ 2 p1/ 2 states in the Ti2O3 48

47

structure, at around 457.1 and 462.4 eV with the binding energies neighboring to Ti4+ states. 49 51

50

The energy band gap of gate dielectrics can be determined from the threshold energy for O1s 53

52

photoelectron energy loss caused by the interband transition across the energy band gap of the 5

54

high-k dielectrics 29. As discussed earlier, the broad peak at a higher binding energy of the zero57

56 58



Page 10 of 54

1 3

2

loss peak can be interpreted within the framework of inter band transition losses, and this can be 5

4

used to determine the band gap of high-k dielectrics. The band gap of the Al2O3 and TiO2 films 6 7

were determined by the O1s core level spectrum, as shown in Figure 2(d). By taking the baseline 10

9

8

of energy loss just after the zero-loss peak, the linear extrapolation of the initial slope onset 12

1

represents the band gap of the oxide. The measured band gaps are expected to be within an error 13 14

limit of ~0.05 eV. The band gaps were determined to be 6.65±0.5 eV for Al2O3, which are in 17

16

15

good agreement with reported values for amorphous phase Al2O3 30. 18 19 20

3.2 ToF SIMS depth chemical analysis 21 23

2

It is important to analyze inter-diffusion around the interface for characterization of high-k gate 24 26

25

dielectric film. The ToF-SIMS technique is highly surface sensitive and gives information from 28

27

the upper layers to the in depth interface structures of the gate stacks. 29 31

30

Figure 3 shows SIMS depth profile of the As, Ga, and In diffusion in TiO2/Al2O3 gate 32 3

stacks for the sample with and without vacuum annealing, analyzed with Ar+ ion beam. From the 36

35

34

Figure 3, it can be noticed that As diffusion significantly reduced by one order after vacuum 38

37

annealing at temperature of 350oC. This results confirms the As 3d results in previous section. 39 41

40

From the XPS analysis it was found that the native oxide consist of Ga2O3, In2O3, As2O3, and 43

42

elemental As (As–As bonds). After annealing in vacuum at 350°C, the As oxides partially 4 45

disappear and In2O3 are dramatically increased to a level near the detection limit as well as the 48

47

46

Ga2O3 feature intensity also significantly increases. The reaction paths for Ga oxide 50

49

decomposition upon the exposure of atomic hydrogen from GaAs native oxide have been 51 52

discussed by Yamada et al. 31 . By partial oxide removal at lower temperatures via the conversion 54

53 5 56 57 58


Page 11 of 54


1 3

2

of the stable Ga2O3 oxide to a volatile Ga2O oxide can be roughly explained with the following 5

4

reaction path: 6

9

8

7

2(InxGa1−xAs) +As2O3→x(In2O3)+(1–x)Ga2O3+4As

(1)

Ga2O3+4Ga→3Ga2O

(2)

10

13

12

1

15

14

Also from the XPS analysis it can be found that Ga substrate peak intensity increases whereas 17

16

Ga-O bond was reduced. From this finding it seems that Ga2O3 partially transform to its lower 18 19

oxidation state Ga2O which is also seen from SIMS study. Oxygen transfer from As to Ga and In 2

21

20

which has been further confirmed from Figure 1 and 2. Increase in In oxide related species can 24

23

be seen from Figure 3(a), (b). Due to artifacts it seems that, In and Al signal concentration slowly 25 26

builds up in the interface layer and peaks close to the surface. Now in our revised manuscript, we 29

28

27

have discounted the first 2 nm close to the surface in order to get the meaningful Al and In 31

30

profiles in our layers. Though after VA the artifacts becomes less prominent due to the lower 32 34

3

concentration oxygen content in the interfacial layer as well as a obstacle to oxygen absorption 36

35

as indicated by the XPS and SIMS analysis 32. This result relates our previous XPS experiments 37 38

(In3d) where the ratio of In-O: In has been significantly increased after vacuum annealing. From 39 41

40

Equations (1) and (2) the transformation of As2O3 to In2O3 and Ga2O has been shown. This 43

42

suggests that is it only arsenic oxide that is being removed with vacuum annealing as increased 4 45

oxygen bond transfer to gallium. So from the SIMS analysis it can be concluded that elementary 48

47

46

As can be reduced to improve the electrical property of the gate stack. 49 51

50

3.3 Electrical Characterization 52 54

53

Figures 4(a) and (b) show the frequency dependent room temperature C-V characteristics of 56

5

Al/TiO2/Al2O3/p- In0.53Ga0.47As MOS capacitors with and without vacuum annealing. Compared 57 58



Page 12 of 54

1 3

2

to the MOS capacitors without VA, the In0.53Ga0.47As capacitors with additional VA, exhibit 5

4

substantially reduced frequency dispersion in accumulation, reduced C-V stretch-out, decreased 6 7

capacitance response in depletion, and reduced intrinsic carrier response in inversion, indicating 10

9

8

that a thin layer of Al2O3 can effectively passivate electrically-active defects in Al/ 12

1

TiO2/Al2O3/In0.53Ga0.47As gate stacks. It is also evident that the MOS capacitors with additional 13 14

VA treatment show a higher maximum accumulation capacitance (Cox) compared to devices 17

16

15

without VA. Both the devices show the frequency dispersions in accumulation and depletion 19

18

regions, as shown in Figure 4. 20 21 2

The amounts of frequency dispersions in accumulation capacitance, evaluated as (Cox), are 25

24

23

22% and 10.7% for with and without VA, respectively. The frequency dispersion in 27

26

accumulation is attributed to the formation of an inhomogeneous layer at interface between gate 28 30

29

32

31

dielectric and substrate

33

. From Figure 4, it is noted that the vacuum annealing improved the

frequency dispersion by reduced interfacial layer formation. The amounts of frequency 34

3

dispersion in the depletion region, evaluated as the change of voltage (∆V) at flatband between 35 37

36

500 and 50 kHz, are 290 and 160 mV for with and without VA devices, respectively. As the 39

38

frequency dispersion in the depletion region is mainly due to the presence of interface traps, the 40 41

result indicates that the vacuum annealing reduces the interface trap density. 43

42

45

4

The hysteresis voltage was determined by sweeping the gate voltage from inversion to 47

46

accumulation and then sweeping back to obtain the difference of the flatband voltage. This 49

48

phenomenon is believed to be due to the presence of interfacial electrons and/or mobile charge in 50 52

51

the oxide. The hysteresis voltage was found reduce to ~150 mV after VA for TiO2/Al2O3/p54

53

In0.53Ga0.47Asstacks, while that for as deposited samples was ~330 mV, as shown in Figure 4 5 56

(inset).The C-V characteristics for VA devices show a sharper transition from the depletion 58

57


Page 13 of 54


1 3

2

region to the accumulation region than that obtained in without VA sample, with an excellent C5

4

V curve fitting agreement with a simulated C-V as shown (see Figure 4 inset). The equivalent 6 7

oxide thickness (EOT) was extracted by fitting the C-V curves in Figure 4 (inset) with the 10

9

8

simulated C-V curve that accounts for quantum mechanical effects. The values of EOT are 3.1 12

1

and 2.2 nm for with and without VA sample, respectively. The energy distribution of Dit was 13 14

evaluated for MOS interfaces with and without VA by the low temperature conductance method 17

16

15

which utilizes both measured capacitance and extracted parallel conductance data 34, 35. 18 20

19

A conductance method is considered to be the most sensitive method for determining 21 2

interface trap density (Dit). The peak conductance versus frequency [Gp/ω-log(f)] curves are 25

24

23

shown in Figure 5 in the gate voltage regime from +0.26 to +1 V at frequency range of 5027

26

500kHz. By measuring peak values of equivalent parallel conductance versus frequency required 28 30

29

by the high frequency conductance method, the dependence of Dit as a function of the applied 32

31

voltage is shown in Figure 4 Inset. The Dit near the midgap calculated to be about 2.9×1012 cm−2 34

3

eV−1 for ALD TiO2/Al2O3/p-In0.53Ga0.47As devices without. The Dit of the ALD TiO2/Al2O3/p35 37

36

In0.53Ga0.47As interfaces decreases by vacuum annealing. The MOS interfaces with VA exhibit 39

38

Dit of the order of ~9×1011 cm-2 eV-1 near the midgap. From the distribution of Dit around the 41

40

midgap, for vacuum annealed sample, all Dit values lie in the low 1011 cm−2 eV-1, implying good 4

43

42

interface quality achieved after vacuum annealing. 45 47

46

To extract the extremely slow traps near the midgap which have very low characteristic 49

48

trapping frequencies for both electrons and holes is very difficult for higher band gap materials 50 52

51

like GaAs. As a consequence, these states cannot be probed within the usual frequency range of 54

53

100 Hz - 1 MHz An alternative method to evaluate the density of midgap states in GaAs is to 5 56

thermally activate them at elevated temperature. Hot semiconductor will strongly increase the 58

57



Page 14 of 54

1 3

2

characteristic emission frequencies. In elevated temperature, it has been shown the possibility to 5

4

access interface defects over an increased portion of the semiconductor energy gap and assesses 7

6

their effect on the C-V characteristics 36. 9

8 10

Figure 6 shows the high temperature (125oC) C-V characteristics of Al/ TiO2/Al2O3/p13

12

1

In0.53Ga0.47AsMOS capacitors for both with and without VA devices. The accumulation 14 15

capacitance did not change and the CV curves did not shift horizontally with temperature. The 16 18

17

humplike features in Figure 6(a) may have resulted from the weak inversion due to the increased 20

19

intrinsic minority carrier generation (ni) from elevated temperatures (125°C). The trap response 21 2

time affect the increased ni, making the Gp/ω peak of the G-V curves shift towards higher 25

24

23

frequencies. In addition, border traps have responses at low measurement frequencies 27

26

37

, which

correspond to the trap levels near midgap. Thus, the increased contributions to measured Gp/ω 28 30

29

may lead to an overestimated Dit values obtained at elevated temperatures. 31 3

32

Figure 7 shows the current density-voltage (J-V) characteristics of the InGaAs MOS 35

34

capacitors measured at VFB-1V. The leakage current density for TiO2/Al2O3/p-In0.53Ga0.47As film 36 37

is 1×10-5A cm-2, whereas, with EOT of 2.4 nm, vacuum annealed MOS capacitors demonstrate a 40

39

38

low JG of 1.4×10-4A cm-2 at VFB-1V. The VA TiO2/Al2O3/p-In0.53Ga0.47As gate stacks show a 42

41

comparable JG as in recent reports 43

38, 39

. The large decrease, two orders of magnitude in leakage

45

4

current after VA, is due to suppressed surface states assisted tunneling. Figure 7 inset shows the 47

46

gate leakage current density at VFB ±1V against EOT. The result shows that, even though ALD 49

48

high-κ on p-In0.53Ga0.47As is leakier than high-κ on Si, it can achieve a comparable or lower 52

51

50

leakage current than SiO2 on Si. With EOT of 2.3 nm, the p-In0.53Ga0.47As MOS capacitor with 54

53

vacuum annealing, demonstrates a low JG of ~1×10−5 A/cm2 at VG=VFB-1 V. 5 56 57 58


Page 15 of 54


1 3

2

4. IETS study of lattice dynamics, traps and defects 4 6

5

In this section, we investigate the electron-phonon interaction via inelastic electron tunneling 8

7

spectroscopy (IETS) measurements in In1-xGaxAs samples. The optical-phonon behavior in In19 10 xGaxAs

13

12

1

was originally referred to as the ‘mixed mode’ or ‘one-two mode’ by Brodsky and

Luckovsky 15

14

40

. ‘InAs-like’ features were hardly detected in the spectra at low In contents, and

their intensities were found still to be low at higher In contents. In0.53Ga0.47As has two 16 18

17

longitudinal optic (LO) and two transverse optic (TO) branches, and both the LO branches 20

19

exhibit electron-phonon coupling. It is found that the dominant electron-phonon coupling is to 21 2

the higher-energy 'GaAs-like' LO modes. Due to the original lattice mismatch (∆r/r~7%) 23 25

24

between GaAs and InAs, structural disorder is superimposed to the chemical disorder in the 27

26

alloy. In0.53Ga0.47As is near lattice matched to InP substrate; although the substrate-induced strain 28 29

can cause shift in optical phonon frequencies. There have been several efforts to understand 32

31

30

lattice properties of these materials that depend on the effects of biaxial strain and confinement. 34

3

Disorder-activated TO (DATO) features and effects of microscopic strains on phonon frequency 35 36

shifts are also needed to be considered. 38

37 39 40 41

4.1 Effect of substrate strain on phonon frequencies and lattice dynamics 42 4

43

The empirical valence force field (VFF) approach 46

45

more accurate ab-initio models 47

42

41

offers an attractive alternative over the

due to its reduced computational cost for the analysis of

48

physical systems. The simplest and most widely used version of the VFF model was developed 49 51

50

by Keating 53

52

43

which gives a good description of LO, TO, and LA phonons modes in unstrained

covalent semiconductors. For modeling of the electronic states in strained nanostructures 54 5

modified Keating model including anharmonic corrections 57

56

45

58


44

,a

provides reasonable agreement


Page 16 of 54

1 3

2

with the experimental data. The original VFF model for diamond was extended to zinc-blende 5

4

structures by Martin 7

6

46

. Zunger et al. have included the stretch-bend interaction

47, 48

and the

cross-stretch interaction in successful descriptions of GaP quantum dots 49, InAs quantum dots 47, 10

9

8

InGaAs alloys and InAs/GaAs super-lattices 48. Sui et al. 50 described a modified Keating model 12

1

combined with the six parameter valence force field (VFF) 13

51, 52

model which describes phonon

14

dispersions in unstrained Si and Ge reasonably well. Steiger et al. 53 unified the VFF descriptions 17

16

15

for zinc blend crystals by including all nearest-neighbor as well as the coplanar second-nearest19

18

neighbor interactions, incorporating the Coulomb interaction within the rigid-ion approximation. 20 2

21

In the present work, the modified Keating/VFF model described by Sui et al. 23

50

is

24

extended for zinc blend structures to describe phonon frequency shifts of ‘InAs’ and ‘GaAs-like’ 27

26

25

modes due to strain in In0.53Ga0.47As/InP MOS structures and the substrate strain is calculated. 29

28

Additionally effects of microscopic strain are included and mixed mode behavior of In1-xGaxAs 30 32

31

is studied. The phonons of the TiO2/Al2O3 dielectrics are also observed from IETS. Optical 34

3

phonon modes are studied in both in vacuum-annealed and without vacuum anneal devices. A 36

35

comparative study of the trap-features is presented from IETS results of the two devices. Also 1/f 37 39

38

noise spectra are compared in the next section to assess the interface qualities of the two devices. 41

40

The inelastic peaks are identified from the second derivative I-V spectra. The spectra 42 43

obtained is an average of eight scans going up to 300 meV after removing the elastic tunneling 46

45

4

background. 48

47

Figures

8

(a)

and

(b)

show

the

IETS

spectra

of

the

two

Al2O3/TiO2/In0.53Ga0.47As/InP devices depicting the ‘InAs’ and ‘GaAs-like’ phonon modes. The 49 50

most intense peak in Fig. 8 (a) is found at 37.6 mV (with wave number 303.2 cm-1). This peak is 53

52

51

identified as the ‘GaAs-like’ longitudinal optical phonon (LO) at Г point. ‘GaAs-like’ transverse 5

54

optical (TO) phonon at Г point and transverse acoustic (TA) phonon at X point are observed at 56 57 58


Page 17 of 54


1 2

33 mV (wave number 266.1 cm-1) and 9.5 mV (wave number 76.6 cm-1), respectively. The less 5

4

3

prominent ‘InAs-like’ phonons are also observed, including the TO (Г) mode at 27 mV (wave 7

6

number 217.7 cm-1), the LO (X) mode at 22.5 mV (wave number 181.5 cm-1) and TA (X) mode 10

9

8

at 6.1 mV (wave number 49.2 cm-1); the longitudinal acoustic (LA) ‘InAs-like’ phonon at X 12

1

point is found at 17.8 mV (with wave number 143.55 cm-1). Figure 8 (b) also demonstrates the 13 14

same peaks at almost same voltage locations. All these phonon frequencies demonstrate shift 17

16

15

from the phonon frequencies of GaAs and InAs crystals. The strain due to small lattice mismatch 19

18

as well as internal microscopic strain and mass disorder 20

54

are accounted for these shifts in

2

21

phonon frequencies. As a first approach, the phonon frequencies are calculated using dispersion 24

23

and mode Gruneisen parameters (γi) 26

25

50

. Table I lists the analytic expressions of phonon

frequencies at the Γ and X points for different modes (which includes quasi-harmonic force 27 29

28

constants α, β, κ and τ which depend on strain to account for the anharmonic interactions) 31

30

The phonon frequencies of different modes in InAs 32

55

and GaAs

53

Table II lists the mode Gruneisen parameters used in the modified Keating model 36

35

.

are also listed in Table I.

3 34

42

50

to fit the

phonon data. 37 39

38

Under (001) biaxial strain, the elements of strain tensor in the strained-In0.47Ga0.53As film on 41

40

InP substrate are given as: 42 43

aInP − aIn 0.53 Ga0.47 As  , aIn0.53Ga0.47 As   2C ezz = − 12 exx ,  C11   exy = exz = eyz = 0,   4

exx = eyy =

52

51

50

49

48

47

46

45

(3)

53 54 5 56 57 58



Page 18 of 54

1 3

2

where the lattice constants for InP and In(1-x)GaxAs with Ga fraction x are given as 4 5

0

6 7

a InP = 5.8687 Α and aIn

1− x Gax As

0

56

= (6.0583 − 0.405 x) Α , respectively. For In0.53Ga0.47As films, we

9

8

have, aIn 10 12

1

0.53Ga0.47 As

0

= 5.86795 Α , which yields exx = -0.013% (compressive stress). C11 and C12 are the

second order elastic constants, respectively. Under biaxial strain (001), the expressions for 14

13

phonon frequencies of TO and LO phonons are expressed in first order expansion of strain and in 15 16

terms of Gruneisen parameters (γi) developed from the analytic expressions in 50: 18

17 19 20 21

nα α 0 + mβ β 0 + mκ κ 0 − nτ τ 0   ωTO ,0 (Γ ) = (1 − 3γ TO (Γ)exx ) ωTO ,0 (Γ ), 2(α 0 + β 0 + κ 0 − τ 0 )   n α + mβ β 0 + mκ κ 0 − nτ τ 0   ω LO (Γ ) = 1 + exx α 0  ω LO ,0 (Γ) = (1 − 3γ LO (Γ )exx ) ωLO ,0 (Γ ) 2(α 0 + β 0 + κ 0 − τ 0 )   

2

ωTO (Γ ) = 1 + exx

27

26

25

24

23

(4)

28

where the Gruneisen parameters for unstrained InAs and GaAS are listed in Table II. ωTO,0 (Г) 31

30

29

and ωLO,0 (Г) are listed in Table I for InAs and GaAs. These frequencies are calculated for both 3

32

InAs and GaAs and the theoretically calculated values of the ‘InAs’ and ‘GaAs-like’ modes are 34 36

35

listed in Table III. At the X point the LO and LA modes are calculated from the following 38

37

expressions developed from the original expressions described in 50. 39 40 41 43

42

nα α 0 + 2mβ β0 + mκ κ 0 + nττ 0   ωLO,0( X ) = (1 − 3γ LO ( X )exx ) ωLO,0( X ), 2(α 0 + 2β 0 + κ 0 + τ 0 )   n α + 2mβ β0 + mκ κ 0 + nττ 0   ωLA( X ) = 1 + exx α 0  ωLA,0( X ) = (1 − 3γ LA ( X )exx ) ωLA,0( X ), 2(α 0 + 2β0 + κ 0 + τ 0 )   

ωLO ( X ) = 1 + exx

50

49

48

47

46

45

4

(5)

The calculated LO and LA frequencies at X point are also listed in Table III. Similarly the 51 52

expressions for TO and TA at X point, are also developed using the modified Keating model and 5

54

53

the following expressions are obtained: 56 57 58


Page 19 of 54


1 3

2

 nα α 0 − nττ 0   ωTO0 ( X ) = (1 − 3γ TO ( X )exx ) ωTO ,0 ( X ),  2(α 0 − τ 0 )     mβ    ωTA( X ) = 1 + exx  ωTA0 ( X ) = (1 − 3γ TA ( X )exx ) ωTA,0 ( X ),  2    

ωTO ( X ) = 1 + exx

10

9

8

7

6

5

4

(6)

Where ω T O ( X ) and ω TA ( X ) are the TO and TA phonon frequencies, respectively, at the X point 1

0

0

13

12

for both InAs and GaAs as listed in Table I. From Eq. (6), the calculated TO and TA are also 15

14

listed in Table III. The most significant phonon frequencies obtained from our IETS 17

16

measurement in the two devices are also listed in Table III and they are compared with the 18 20

19

theoretically calculated values, both of which are in good agreement. The following empirical 2

21

expressions are developed for TO and LO phonon frequencies of In1-xGaxAs structures using Eq. 23 24

(3) and (4) and are given as a function of fraction x for ‘InAs’ and ‘GaAs-like’ modes: 26

25 27 29

28

ωTO (Γ ) InAs = 245.3 − 51.5 x − 3.54 x 2 ( cm −1 ) ,  30



31

ω LO (Γ) InAs = 265.1 − 50.6 x − 3.37 x 2 ( cm −1 ) ,  32



3

ωTO (Γ ) GaAs = 302.1 − 72.2 x − 5.08 x 2 ( cm −1 ) , 

34



35

ω LO (Γ) GaAs = 319.2 − 68.12 x − 4.56 x 2 ( cm −1 ) ,  37

36 38 39 40 41 42 43 4 45 46 47 48 49 50 51 52 53 54 5 56 57 58


(7)


Page 20 of 54

1 2 3

Table I. Expressions of Phonon frequencies in terms of quasi-harmonic force constants and the phonon frequencies of GaAs and InAs 4 5 6 7 8 9 10

Phonon Modes 1

Model 42

GaAs Phonon frequencies (cm-1)

12

InAs Phonon frequencies ( cm-1) 55

53

13 14 15

ωLO,TO (Γ) 16

 8(α + β + κ − τ )    M  

17 18

1 2

TO: 271.13

TO: 221

LO: 293.15

LO: 241

LO: 240.3

LO: 199.2

LA: 225

LA: 145.2

256.5

210

81.7

53.23

19 20 21

1

ωLO,LA (X)

 4(α + 2β + κ + τ )  2   M  

ωTO (X)

 8(α − τ )  2    M 

ωTA (X)

 8β  2   M 

2 23 24 25 26 27

1

28 29 30 31 32 3

1

34 35 36 37 38

Table II. Grüneisen parameters for GaAs and InAs computed including anharmonicity corrections 39 40 41 42 43 4 45

Material

γLO (Г)

γTO (Г)

γLO (X)

γLA (X)

γTO (X)

γTA (X)

InAs

1.06a

1.2a

1.7527 b

1.7569 b

1.956 b

0.3182b

GaAsc

1.23

1.39

1.01

1.22

1.73

-1.62

46 47 48 49 50 51 52 53 54

a

5

b

57

56

Reference 57 Reference 45 c Reference 53 58


Page 21 of 54


1 3

2

Table III. Theoretical and experimental values of the phonon frequencies of ‘InAs’ and ‘GaAslike’ modes 5

4 6 7

Phonon Frequencies

Calculated phonon frequency of ‘GaAs-like’ mode (mV)

Calculated phonon frequency of ‘InAs-like’ mode (mV)

Experimental values of phonon frequencies from IETS: ‘GaAslike’ mode (mV)

Experimental values of phonon frequencies from IETS: ‘InAs-like’ mode (mV)

1

10

9

ωLO (Γ)

36.4

27.41

37.5

29.4

ωTO (Γ)

33.64

29.9

34.2

26.8

ωLO (X)

29.8

24.72

-

23

ωLA (X)

27.92

18.01

-

18

ωTO (X)

31.82

26.06

-

-

ωTA (X)

10.12

6.6

10.04

6

8

12

15

14

13

18

17

16

19 21

20 2

25

24

23

28

27

26

31

30

29

32 3 34

In a binary material such as (InAs)1-x(GaAs)x the collective vibrations of all the harmonic 37

36

35

oscillators are described in terms of phonons. We will describe the optical phonons in the 39

38

(InAs)1-x(GaAs)x alloy by a set of effective InAs- and GaAs-like oscillators. TO phonons in In140 42

41 xGaxAs

4

43

display a two-mode behavior. The TO frequencies in pure GaAs and InAs differ

significantly: 271.13 and 221 cm-1, respectively, which is mainly due to the difference between 46

45

their reduced masses. Hence it is expected that the GaAs- and InAs-like characters to be 47 49

48

preserved in the alloy. Indeed, when the optical frequencies of the InAs and GaAs binary 51

50

compounds differ significantly, the Ga atoms are not expected to participate in InAs-like 52 53

vibrations, and vice versa. The almost linear dependencies of TO frequencies on x fraction can 56

5

54

be explained by two effects: mass disorder and microscopic strains 54. Mass disorder refers to the 57 58



Page 22 of 54

1 3

2

fact that when one is concerned with the response of one oscillator, the other is at rest, and vice 5

4

versa. Mass disorder lowers the phonon frequencies 6

54

. Compressive Microscopic strain

7

increases the phonon frequencies. In our current scope we will focus on the microscopic strain 10

9

8

effects on phonon frequencies. A localized mode of In in GaAs lies always in the optical 12

1

frequency range of the pure compound InAs, whereas, except for some rare "band modes", it is 13 14

located outside the In1-xGaxAs optical range ("local" or "gap" modes) 17

16

15

58

. The effect of the host

lattice is summed up by an equal stretching of the four bonds of the InAs4 unit, as would be 19

18

realized by an hydrostatic pressure. A shift ∆ωop between the mean value, , of the InAs 20 2

21

optical frequencies to the particular value ωop of the localized mode of InAs4 in In1-xGaxAs is 24

23

induced by this "hydrostatic pressure". If we define an averaged Grüneisen parameter for the 26

25

optical modes by , this shift will satisfy the following equation 58. 29

28

27

∆ωop

 ∆r  = −3 < γ op >  , < ωop >  r  32

31

30

(8)

ωop =< ωop > +∆ωop 3 35

34

where, (∆r/r) is the relative change in the bond lengths and the corresponding Grüneisen 37

36

parameter. For the determination of in either InAs or GaAs, since the Grüneisen 38 39

parameters are not known for optical modes belonging to any point of the Brillouin zone, the 42

41

40

following approximation can be made 58. 4

43

2 1 < γ op >= γ [TO (Γ )] + γ [ LO (Γ )], 3 3 2 1 < ωop >= ωop [TO (Γ )] + ωop [ LO (Γ )] 3 3 50

49

48

47

46

45

(9)

Table IV summarizes the average Grüneisen parameters as well the average optical phonon 51 52

frequencies (calculated using Tables I and II) for InAs and GaAs oscillators due to microscopic 5

54

53

strain. The phonon frequency shifts are calculated using (∆r/r)GaAs=+1.6x% and (∆r/r)InAs=56 57 58


Page 23 of 54


1 3

2

1.3(1-x)% according to reference 59. The calculated optical phonon frequencies as well as the 5

4

corresponding shifts are also summarized in Table IV for both ‘InAs’ and ‘GaAs-like’ modes. 6 7

Eq. (7) is modified for incorporating the microscopic strain effects and the empirical expressions 10

9

8

are given as a function of x fraction as: 1 12

ωTO (Γ ) InAs = 235.05 − 41.25 x − 3.54 x 2 ( cm −1 ) ,  13



14

ω LO (Γ ) InAs = 254.86 − 40.36 x − 3.37 x 2 ( cm −1 ) , 

15

(10)

 ωTO (Γ ) GaAs = 302.1 − 54.3 x − 5.08 x ( cm ) ,   ω LO (Γ ) GaAs = 319.2 − 50.22 x − 4.56 x 2 ( cm −1 ) ,  16

2

17 18 19

−1

21

20

The LO and TO mode phonon frequencies calculated in table IV for the ‘InAs-‘ and ‘GaAs-like’ 23

2

modes are in excellent agreement with the optical phonon modes observed from the experimental 25

24

IETS data as shown in Table III. 26 27 28 30

29

Table IV. Microscopic strain parameters and phonon frequencies in In0.53Ga0.47As 31 32 3

Phonon modes

< γ op >

GaAs-like

1.34

InAs-like

1.153

34 35

< ωop > (mV)

∆ωop (mV)

ωTO (Γ ) (mV)

ωLO (Γ) (mV)

+2.22x

28.23

1.043

34.6

37.4

-1.27(1-x)

34.53

-0.673

26.7

29.23

Expression of ∆ωop

38

37

36

39 40 41 42 4

43

In Figures 9 (a) and (b), the phonons of TiO2/Al2O3 gate stack are shown from IETS results of 46

45

the two devices (as-deposited and with vacuum anneal devices). The identified peaks of TiO2 are 47 49

48

consistent in the two devices. These peaks include TO phonon at 46 mV (with wave number 371 51

50

cm-1) and two LO phonon peaks at 54 and 98 mV (with wave numbers 435.5 cm-1 and 791.5 cm53

52

1

56

5

54

, respectively). The phonon modes observed for TiO2 agree well with the reported localized

vibration of TiO2 (110) 57

60

. The Al2O3 phonons vary slightly in the two devices. In the vacuum

58



Page 24 of 54

1 3

2

annealed sample, the peaks include LO phonons at 62.3, 86 and 103 mV (with wave numbers 5

4

502.4, 693.5 and 826.6 cm-1, respectively). The TO phonons are observed at 73.5 and 79 mV 7

6

(with wave numbers 592.7 and 637.1 cm-1, respectively). In the other sample, the Al2O3 LO 10

9

8

phonons are observed at 66, 89, 102.1 and 113.8 mV (with wave numbers 532.3, 717.7, 825 and 12

1

917.7 cm-1, respectively). The TO phonons are observed at 70, 81 and 85 mV (with wave 13 14

numbers 564.5, 653.2 and 685.5 cm-1, respectively). The Al2O3 phonon modes also agree well 17

16

15

with the surface phonons reported for thin aluminium oxide films 19

18

61

. The minor phonon shifts

may be caused due to the vacuum annealing of the bi-layer TiO2/Al2O3 gate dielectric reducing 20 2

21

the leakage as well as improves the interface quality. This is also reflected in the reduced peak 24

23

intensities in the vacuum annealed devices as observed from Figs. 9 (a) and (b). 25 26 27 29

28

Table V: The phonons of the TiO2/Al2O3 gate dielectric stack. 30 31 32

Phonons

Vacuumannealed sample (cm-1)

Sample without anneal (cm-1)

TO (Г5+)

395.2

371.8

LO (Г5+)

437.1

435.5

LO (Г1-)

796

791.5

TO (ε||)

592.7

564.5

TO (ε┴)

637.1

653.2

TO (ε┴)

-

685.5

LO (ε||)

502.4

532.3

LO (ε||)

826.6

825

LO (ε┴)

693.5

717.7

LO (ε┴)

-

917.7

3 34 36

35

TiO2 37 38 39 40 42

41

Al2O3

5

54

53

52

51

50

49

48

47

46

45

4

43

56 57 58


Page 25 of 54


1 3

2

4.2 Trap Study and Evaluation of Interface Quality 5

4

Figure 10 shows the IETS spectra of the two Al/TiO2/Al2O3/In0.53Ga0.47As/InP devices (both 6 8

7

without anneal and vacuum annealed devices) in the 130-350 mV range. It shows many trap 10

9

related features. Two distinct trap features are observed; trap-assisted conduction (increased 1 12

leakage current) and charge trapping (causes shifts in threshold voltage) where the former feature 15

14

13

includes a peak followed by a valley and for the later a valley preceding the peak 13. It is clearly 17

16

observed that the vacuum annealed sample have less trap features depicting a better quality of 18 19

the interface compared with the sample without anneal. Also, the leakage is evidently much 2

21

20

reduced due to annealing as the peak intensities of the traps in the vacuum-annealed sample is 24

23

much less. The trap features marked as 1 and 2 in the forward bias are associated with trap25 27

26

assisted conduction while 3 is associated with charge trapping in the IETS of the vacuum 29

28

annealed sample. In the sample without anneal, the trap features 1, 2 and 4 are trap-assisted 31

30

conduction while the feature 3 is charge-trapping. The width of the trap feature is given 32 34

3

by ∆V = qNt / Cox , with Nt being the areal trap density and Cox being the gate dielectric 35 36

capacitance 13. Traps that respond in the voltage range 130-350 mV are mainly oxide traps. The 39

38

37

major trap features are mainly visible in the 150-200 mV, and in the 250-300 mV ranges. The 41

40

common interpretation for the origin of these traps is that they may arise due to some 42 4

43

characteristics bonding-defects or they may be electrically activated. These can be excited by the 46

45

energy of the applied voltage and appear in the 2nd derivative of the I-V characteristics as peaks. 47 48

Proximity of both trap-assisted conduction and charge trapping features (as also observed in Fig. 51

50

49

10) may indicate similar origins of the trap states but with slight difference in surroundings 13. 52 53 54 5 56 57 58



Page 26 of 54

1 3

2

5. 1/f Noise Characteristics 4 6

5

The Comparing the trap-features of the vacuum-annealed and without anneal devices from the 8

7

IETS spectra, it is clear that the leakage and interface quality improves due to vacuum anneal. 9 10

This is further verified from the 1/f noise characteristics of the two devices as depicted in Fig. 11. 13

12

1

The 1/f noise measurement setup includes an Agilent E5263A two-channel high-speed source 15

14

monitor unit (SMU), a SR 570 LNA, and an Agilent 35670A dynamic signal analyzer. The SMU 16 18

17

provides the necessary bias, the minute fluctuations in the drain source voltage are amplified 20

19

using the LNA, and the output of the LNA is fed to the dynamic signal analyzer that performs 21 2

the fast Fourier transform on the time-domain signal to yield the noise power spectral density. 23 25

24

The minute fluctuations due to noise were amplified to the measurable range using the low-noise 27

26

SR570 preamplifier. 28 30

29

In Fig. 11, the current power spectral densities of two different devices, both without anneal, 31 3

32

show a significant higher noise level compared to the current spectral density of the vacuum 35

34

annealed sample. Also, another interesting thing to note is that the devices without anneal, show 36 37

a 1/f dependency on frequency signifying a large number of active traps. While the vacuum 40

39

38

annealed sample show a 1/f 2 dependency signifying a less number of traps and a better interface 42

41

quality. A random telegraph noise (RTN) fluctuation is caused due to a single trap and has a 4

43

Lorentzian (1/f 2) power spectral density 47

46

45

62

, when multiple traps are active the superposition of

multiple RTNs (Lorentzians) result into a 1/f noise spectrum 63. 48 49 51

50

6. Conclusions 52 53

In conclusion, TiO2/Al2O3 films on In0.53Ga0.47As substrates by atomic layer deposition and 56

5

54

fabrication of Al/TiO2/Al2O3/In0.53Ga0.47As/InP MOS capacitors are reported. The interfacial 57 58


Page 27 of 54


1 3

2

properties and band alignment have been investigated using XPS. The band gap was determined 5

4

to be ~6.7±0.05 eV for TiO2/Al2O3.It is shown that the addition of an ultrathin TiO2 layer on 6 7

Al2O3 helps to reduce of In-O and As-O formation at the interface. Vacuum annealing used for 10

9

8

the decomposition and desorption of native oxide (EOT ~2.2 nm) resulted in superior MOSCAP 12

1

characteristics in terms of low hysteresis and frequency dispersion, moderate Dit (~1.5×1011 cm-2 13 14

eV-1) and low leakage current density of 9.1×10-6A cm-2 at VFB-1V. 17

16

15

Inelastic electron tunneling spectroscopy technique was used to study the lattice structure 19

18

and strain-induced shifts in electron-phonon interactions. Phonon frequencies of the ‘InAs-’ and 20 2

21

‘GaAs-like’ modes are identified from the IETS spectra. Theoretical models for phonon 24

23

frequencies under strain are validated with experimental IETS data. Empirical relations are 26

25

developed for phonon frequencies of the TO and LO vibrations of InAs- and GaAs-like modes. 27 29

28

Effects of microscopic strain are also incorporated for computation of theoretical TO and LO 31

30

phonon modes. Trap-features (such as trap-assisted conduction and charge-trapping) and 32 3

electrically active bonding defects are identified from the IETS in the vacuum annealed (and 36

35

34

without) devices. Comparison between the two devices reveals the vacuum annealed devices 38

37

show a superior interface quality and reduced leakage due to annealing. The phonons of the 39 40

TiO2/Al2O3 gate stacks are also identified from the IETS. Lastly, the interface qualities are 43

42

41

compared with the 1/f noise spectra of the two devices which revealed a lower noise level and 45

4

less active traps in the vacuum annealed devices. Based on the experimental data reported in the 46 48

47

present study, ALD high-k TiO2/Al2O3 bilayer may be used as the gate dielectric in future III-V 50

49

CMOS technology. 51 52 53 54 5 56 57 58



Page 28 of 54

1 3

2

References 4 6

5

[1] 7 8

Spicer, W. E.; Chye, P. W.; Skeath, P. R.; Su, C. Y.; Lindau, I. J. Vac. Sci. Technol. 1979, 16, 1422-1424.

9

12

1

10

[2]

Hasegawa, H.; Akazawa, M. Appl. Surf. Sci. 2008, 255, 628-632.

15

14

[3]

Dalapati, G. K.; Chia, C. K.; Mahata, C.; Krishnamoorthy, S.; Tan, C. C.; Tan, H. R.;

13

16 17

Maiti, C. K.; Chi, D. IEEE Trans. Electron Devices 2013, 60, 192-199. 18 20

19

[4] 21 2

Dalapati, G. K.; Tong, Y.; Loh, W. Y.; Mun, H. K.; Cho, B. J. IEEE Trans. Electron Devices 2007, 54, 1831-1837.

23 24 25

[5] 27

26 28

Frank, M. M.; Wilk, G. D.; Starodub, D.; Gustafsson, T.; Garfunkel, E.; Chabal, Y. J.; Grazul, J.; Muller, D. A. Appl. Phys. Lett. 2005, 86, 152904-152906.

29 31

30

[6] 32

Chang, C. H.; Chiou, Y. K.; Chang, Y. C.; Lee, K. Y.; Lin, T. D.; Wu, T. B.; Hong, M.;

3

Kwo, J. Appl. Phys. Lett. 2006, 89, 242911-242913. 34 35

38

37

36

[7]

Peacock, P. W.; Robertson, J. J. Appl. Phys. (Melville, NY, U. S.) 2002, 92, 4712-4721.

41

40

[8]

Goel, N.; Majhi, P.; Chui, C. O.; Choi, D.; Harris, J. S. Appl. Phys. Lett. 2006, 89,

39

42

163517-163519. 43 4 45

[9] 48

47

46

Dalapati, G. K.; Sridhara, A.; Wong, A. S. W.; Chia, C. K.; Lee, S. J.; Chi, D. Z. J. Appl. Phys. (Melville, NY, U. S.) 2008, 103, 034508-034512.

49

52

51

50

[10]

Petit, C.; Salace, G. Rev. Sci. Instrum. 2003, 74, 4462-4467.

5

54

[11]

Giaever, I.; Megerle, K. Phys. Rev. 1961, 122, 1101-1111.

53

56 57 58


Page 29 of 54


1

4

3

2

[12]

Chynoweth, A. G.; Logan, R. A.; Thomas, D. E. Phys. Rev. 1962, 125, 877-881.

7

6

[13]

He, W.; Ma, T. –P. Appl. Phys. Lett.2003, 83, 5461-5463.

[14]

Ma, T. P. Sci. China Ser. F, 2011, 54, 980-989.

[15]

Jevasuwan ,W.; Urabe, Y.; Maeda, T.; Miyata, N.; Yasuda, T.; Yamada, H.; Hata, M.;

5

10

9

8

1 13

12

15

14

Taoka, N.; M. Takenaka, M.; and S. Takagi, S. Materials, 2012, 5, 404-414. 16 18

17

[16] 20

19

Cameron, M.A.; Gartland, I.P.; Smith, J.A.; Diaz, S.F.; and S. M. George, S.M.

Langmuir, 2000, 16, 7435-7444. 21 2 23

[17] 26

25

24

Sun, Y.; Liu, Z.; Machuca, F.; Pianetta, P.; Spicer, W.E.J. Vac. Sci. Technol.,A, 2003, 21,

219. 27

30

29

28

[18]

Zhu, M.; Tung, C.H.; and Yeo, Y. C. Appl. Phys. Lett. 2006, 89, 202903.

3

32

[19]

Dalapati, G. K.; Kumar, M. K.; Chia, C. K.; Gao, H.; Wang, B. Z.; Wong, A. S. W.;

31

35

34

Kumar, A.; Chiam, S. Y.; Pan, J. S.; Chi, D. Z. J. J. Electrochem. Soc., 2010, 157, H82537

36

H831. 38 40

39

[20] 41

Hinkle, C. L.; Sonnet, A. M.; Vogel, E. M.; McDonnell, S.;Hughes, G. J.; Milojevic, M.;

43

42

Lee, B.; Aguirre-Tostado, F. S.; Choi, K.j.; Kim, H. C.; Kim, J.; Wallace, R. M. Appl. Phys. 45

4

Lett. 2008, 92, 071901. 46 48

47

[21] 49

Hinkle, C. L.; Milojevic, M.; Brennan, B.; Sonnet, A. M.; Aguirre-Tostado, F. S.;

51

50

Hughes, G. J.; Vogel, E. M; Wallace, R. M. Appl. Phys. Lett. 2009, 94, 162101-1162101-3. 52 53 54 5 56 57 58



Page 30 of 54

1 3

2

[22] 5

4

Mahata, C.; Mallik, S.; Das, T.; Maiti, C. K.; Dalapati, G. K.; Tan, C. C.; Chia, C. K.;

Gao, H.; Kumar, M. K.; Chiam, S. Y.; Tan, H. R.; Seng, H. L.; Chi, D. Z.; Miranda, E. Appl. 6 7

Phys. Lett., 2012, 100, 062905-1- 062905-4. 9

8

1

10

[23] 13

12

O’Connor, E.; Monaghan, S.; Long, R. D.; O’Mahony, A.; Povey, I. M.; Cherkaoui, K.;

Pemble, M. E.; Brammertz, G.; Heyns, M.; Newcomb, S. B.; Afanas’ev, V. V.; Hurley, P. K. 14 15

Appl. Phys. Lett. 2009, 94, 102902-102904. 16 17 19

18

[24] 21

20

Hinkle, C. L.; Sonnet, A. M.; Vogel, E. M.; McDonnell, S.; Hughes, G. J.; Milojevic, M.;

Lee, B.; Aguirre-Tostado, F. S.; Choi, K. J.; Kim, H. C.; Kim, J.; Wallace, R. M. Appl. Phys. 23

2

Lett., 2008, 92, 071901-071903. 24 25

28

27

26

[25]

Robertson, J. Appl. Phys. Lett. 2009, 94, 152104.

31

30

[26]

Miyazaki, S. Appl. Surf. Sci. 2002, 190, 66-74.

[27]

Hollinger, G.; Skheyta-Kabbani, R.; Gendry, M. Phys. Rev. [sect.] B 1994, 49, 11159-

29

34

3

32

35

11167. 36 37

40

39

38

[28]

Dhayal, M.; Junb, J.; Guc, H. B.; Park, K. H. J. Solid State Chem. 2007, 180, 2696-2701.

43

42

[29]

Miyazaki, S. J. Vac. Sci. Technol. B 2001, 19, 2212-2216.

[30]

Gao, K. Y.; Seyller, Th.; Ley, L.; Ciobanu, F.; Pensl, G.; Tadich, A.; Riley, J. D.; Leckey,

41

46

45

4

47

R. G. C. Appl. Phys. Lett. 2003, 83, 1830-1832. 49

48

52

51

50

[31]

Yamada,M. Jpn. J. Appl. Phys., Part 2, 1996, 35, 651.

56

5

54

[32]

Hartmann, J.M.; Benevent, V.; Barnes, J.P.; Veillerot, M.; Deguet, C. Semiconductor

53

Science and Technology. 2013, 28, 025017. 57 58


Page 31 of 54


1 3

2

[33] 5

4

Harris, H.; Biswas, N.; Temkin, H.; Gangopadhyay, S.; Strathman, M. J. Appl. Phys.

(Melville, NY, U. S.) 2001, 90, 5825-5831. 6 8

7

[34] 9

Nicollian, E. H.; Brews, J. R. MOS (Metal Oxide Semiconductor) Physics and

1

10

Technology; Wiley: New York, 2003. 12 14

13

[35] 16

15

Schroder, D. K. Semiconductor Material and Device Characterization, Wiley: New

York, USA, 1998. 17 18 19

[36] 2

21

20

Brammertz, G.; Martens, K.; Sioncke, S.; Delabie, A.; Caymax, M.; Meuris, M.; Heyns,

M. Appl. Phys. Lett. 2007, 91, 133510-133512. 23 25

24

[37] 26

Fleetwood, D. M.; Winokur, P. S.; Reber, R. A. Jr.; Meisenheimer, T. L.; Schwank, J. R.;

27

Shaneyfelt, M. R.; Riewe, L. C. J. Appl. Phys. (Melville, NY, U. S.) 1993, 73, 5058-5074. 29

28

31

30

[38] 3

32

Oh, H. J.; Lin, J. Q.; Lee, S. J.; Dalapati, G. K.; Sridhara, A.; Chi, D. Z.; Chua, S. J.; Lo,

G. Q.; Kwong, D. L. Appl. Phys. Lett. 2008, 93, 062107-062109. 34 36

35

[39] 37

Koveshnikov, S.; Tsai, W.; Ok, I.; Lee, J. C.; Torkanov, V.; Yakimov, M.; Oktyabrsky,

39

38

S. Appl. Phys. Lett. 2006, 88, 022106-022108. 40

43

42

41

[40]

Brodsky, M. H.; Luckovsky, G. Phys. Rev. Lett. 1968, 21, 990-993.

46

45

[41]

Musgrave, M.; People, J. Proc. R. Soc. London, Ser. A 1962, 268, 474-484.

[42]

Giannozzi, P.; de Gironcoli, S.; Pavone, P.; Baroni, S. Phys. Rev. [sect.] B 1991, 43,

4

49

48

47

51

50

7231-7242. 52 54

53

[43] 5

Keating, P. N. Phys. Rev. 1966, 145, 637-645.

56 57 58



Page 32 of 54

1 3

2

[44] 5

4

Lee, S.; Lazarenkova, O. L.; Von Allmen, P.; Oyafuso, F.; Klimeck, G. Phys. Rev. [sect.]

B 2004, 70, 125307-125313. 6 8

7

[45] 9

Lazarenkova, O.; Von Allmen, P.; Oyafuso, F.; Lee, S.; Klimeck, G. Superlattices and

1

10

Microstruct. 2003, 34, 553-556. 12

15

14

13

[46]

Martin, R. Phys. Rev. [sect.] B 1970, 1, 4005-4011.

18

17

[47]

Williamson, A. J.; Wang, L. W.; Zunger, A. Phys. Rev. [sect.] B 2000, 62, 12963-12977.

[48]

Kim, K.; Kent, P. R. C.; Zunger, A.; Geller, C. B. Phys. Rev. [sect.] B 2002, 66, 045208-

16

21

20

19

23

2

045222. 24

27

26

25

[49]

Fu, H.; OzoliĦš, V.; Zunger, A. Phys. Rev. [sect.] B 1999, 59, 2881-2887.

30

29

[50]

Sui, Z.; Herman, I. P. Phys. Rev. [sect.] B 1993, 48, 17938-17953.

[51]

Tubino, R.; Piseri, L.; Zerbi, G. J. Chem. Phys. 1972, 56, 1022-1039.

[52]

McMurry, H. L.; Solbrig, A. W. Jr.; Boyter, J. K.; Noble, C. J. Phys. Chem. Solids 1967,

28

31 3

32 34

38

37

36

35

28, 2359-2368. 39 41

40

[53] 42

Steiger, S.; Jelodar, M. S.; Areshkin, D.; Paul, A.; Kubis, T.; Povolotskyi, M.; Park, H. -

43

H.; Klimeck, G. Phys. Rev. [sect.] B 2011, 84, 155204-01-155204-11. 45

4

48

47

46

[54]

Rücker, H.; Methfessel, M. Phys. Rev. [sect.] B 1995, 52, 11059-11072.

51

50

[55]

Talwar, D. N.; Agrawal, B. K. Phys. Status Solidi B 1974, 63, 441- 452.

[56]

Adachi, S. Physical Properties of III-V Semiconductor compounds, John Wiley and Sons:

49

54

53

52

56

5

New York, USA, 1992. 57 58


Page 33 of 54


1 3

2

[57] 5

4

Groenen, J.; Carles, R.; Landa, G.; Guerret-Pie´court, C.; Fontaine, C.; Gendry, M. Phys.

Rev. [sect.] B 1998, 58, 10452-10462. 6

9

8

7

[58]

Carles, R.; Landa, G.; Renucci, J. B. Solid State Commun. 1985, 53, 179-182.

[59]

Mikkelsen, J. C.; Boyce, J. B. Phys. Rev. [sect.] B 1983, 28, 7130-7140.

[60]

Rocker, G.; Schaefer, J. A.; Gopel, W. Phys. Rev. [sect.] B 1984, 30, 3704-3708.

[61]

Frederick, B. G.; Apai, G.; Rhodin, T. N.; Phys. Rev. [sect.] B 1991, 44, 1880-1890.

[62]

Kirton, M. J.; Uren, M. J. Adv. Phys. 1989, 38, pp. 367-468.

[63]

Haartman, M. V. PhD dissertation: Royal Institute of Technology (KTH), Sweden; 2006.

10

13

12

1

16

15

14

19

18

17

2

21

20

23

26

25

24

27 28 29 30 31 32 3 34 35 36 37 38 39 40 41 42 43 4 45 46 47 48 49 50 51 52 53 54 5 56 57 58



Page 34 of 54

1 3

2

Figures 4 5 6 7 8

TiO2/Al2O3/In0.53Ga0.47As without VA

Ga3d 9 10

InxGayOz

In2Ox 1 12

As3d5/2

Ga-As

In-As 13

As3d

14

As3d3/2

25

24

23

2

21

20

19

18

17

16

Intensity (a.u.)

15

1+

Ga Ga2+

0

O2s

Ga3d

As

(a) (c)

As3d

TiO2/Al2O3/In0.53Ga0.47As with VA

26 27 28 29 30 31

(b) (d)

32

O2s 3 35

34

14 37

36

16

18

20

22

24

26

28 30

Binding Energy (eV) 38

32

34

36

38

40

42

44

46

Binding Energy (eV)

40

39

Figure 1: (a) & (b) Ga 3d and (c) & (d) As 3d core-level spectra from TiO2/Al2O3 /In0.53Ga0.47As 42

41

gate stack with and without vacuum annealing. 43 4 45 46 47 48 49 50 51 52 53 54 5 56 57 58


Page 35 of 54


1 2 3 4

In 3d 5

without VA

Ti2p Ti2p3/2

6 7 8 9

(4+)

In3d3/2

In3d5/2 10 1

Ti2p1/2

Ti2p1/2

23

2

21

20

19

18

17

16

15

14

13

Intensity (a.u.)

12

Ti2p3/2

In-O

(a)

(4+)

(3+)

(c)

(3+) 456

458

460

462

464

466

468

with VA

In 3d

O-Al 24 25

Eg=6.65 eV 26 27

(b) 28 29

(d)

30 32

31

442 34

3

444

446

448

450

452

454

456 530

Binding Energy (eV) 35

535

540

545

550

Binding Energy (eV)

37

36

Figure 2: (a) & (b) In 3d core-level spectra from TiO2/Al2O3 /In0.53Ga0.47As gate stack, with and 39

38

without vacuum annealing (c)Ti 2p core-level spectra from TiO2/Al2O3/In0.53Ga0.47As, and (d) O 40 42

41

1s energy loss spectra for the TiO2/Al2O3 /In0.53Ga0.47As samples. 43 4 45 46 47 48 49 50 51 52 53 54 5 56 57 58



Page 36 of 54

1 2 3 4 5 6 7 8 9 10 1 12 13 14 15 16 17 18 19 20 21 2

Figure 3: ToF-SIMS profiles of TiO2/Al2O3/InGaAs gate stack for (a) without vacuum annealed, 25

24

23

and (b) with vacuum annealed sample. 26 27 28 29 30 31

Al/~7nm TiO2/~3nm Al2O3/In0.53Ga0.47As 32 3

48

47

46

45

4

43

42

41

40

39

38

37

36

2

Capacitance (fF/µm )

35

34

10

50k

8

8

4 2

4

50

20

16

500 k

12 ∆VFb=320 mV

2 f=100 kHz -1 0

1

Volatge (V)

2

(a)

15

8

10

4 f=100 kHz -1 0

∆VFb=190 mV

1

2

Volatge (V)

(b) 5 % dispersion= 10.7%

% dispersion= 22% Without Vaccum annealing

-1 0 1 Applied Volatge (V) 49

CV sweep forward CV sweep reverse Simulation

20

6 500 k

6

50k

CV sweep forward CV sweep reverse Simulation

10

With Vaccum annealing

2

51

-1 0 1 Applied Volatge (V)

52

Figure 4: Multi-frequency (50 kHz-500 kHz) C-V characteristics of Al/~7nm TiO2/~2nm 5

54

53

Al2O3/n-In0.53Ga0.47As MOS capacitors for (a) without vacuum annealing (VA) and (b) with VA. 56 57 58


Page 37 of 54


1 2 3 4 5 6


1.8

7 8

0.26 V

1.6

9

0.23 V 4.8

Dit

10


5.2

3.18

17

4.4

3.6 0.96

3.2

18

0.2

0.4

2

5

0.4

0.6 Bias (V)

0.8

6

10

3.06 3.04

2.98 0.2

0.4

0.6

0.8

1.0

Bias (V)

2.4 5

10

6

10

7

10

Log Frequency (Hz) 24

3.08

1.0

7

10

23

3.10

3.02

2.8

0.6

21

3.12

3.00

0.94

20

3.14

12

0.98

0.8

19

-2

4.0

-1

1.0

1V

1.00

-2

0.97 V

-1

1.2

Dit

3.16

12

1.4

Dit (eV cm x 10 )

Dit (eV cm x 10 )

16

15

14

11

13

12

1

2

Gp/ω (10 / eV.cm )

1.02

10

Log Frequency (Hz)

25 26

Figure 5: Gp/q vs log (f) characteristics of Al/TiAlO/GaAs MOS capacitors for (a) with and (b) 27 29

28

without vacuum aneealed sample. 30 31 32 3

Al/~7nm TiO2/~3nm Al2O3/In0.53Ga0.47As 34 35

Without Vaccum annealing

1.0 36

With Vaccum annealing

37 38

0.8

4

43

42

41

40

C/Cox

39

50kHz

50kHz

0.6 0.4

45

500kHz

47

46

0.2 49

48

o

-1

0

1

2

-2

-1

0

1

2

Applied Volatge (V)

Applied Volatge (V) 52

(b)

o

T=125 C

T=125 C -2

51

50

500kHz

(a)

54

53

Figure 6: C-V characteristics of Al/~7 nm TiO2/~2nm Al2O3/n-In0.53Ga0.47As MOS capacitors 56

5

for (a) without VA (b) with VA devices measured at 125°C. 57 58



Page 38 of 54

1 2 3 4 5 6

-1 Al/TiO2/Al2O3/p-In0.53Ga0.47As Without VA

9

Al/TiO2/Al2O3/p-In0.53Ga0.47As With VA

28

27

26

25

24

23

2

21

20

19

18

17

16

15

14

13

12

1

2

Cuurent density (A/cm )

10

-2

10

-3

10

2

8

1


10 7

10 0 10 -1 10 -2 10 -3 10 -4 10 -5 10 -6 10 -7 10 -8 10

-4

10

High-k/Si SiO2/Si This Work

1.0

1.5

2.0 2.5 EOT (nm)

3.0

3.5

-5

10

-6

10

30

29

-7

10 32

31

34

3

-3 36

35

-2

-1

0

1

2

3

Applied Voltage (V) 37 38 39

Figure 7: Gate leakage current density vs. gate voltage characteristics of Al/~7 nm TiO2/~2nm 42

41

40

Al2O3/n-In0.53Ga0.47As MOS capacitors with and without vacuum annealing. 43 4 45 46 47 48 49 50 51 52 53 54 5 56 57 58


Page 39 of 54


1 2 3 4 6

LO(Γ )

(a) Without Anneal Sample 5 7

14 15 16 17 18

TO(Γ )

13

'InAs'-like

LO(X)

12

LA(X)

1

TA(X)

10

TA(X)

9

TO(Γ )

'GaAs'-like

IETS (arb. unit)

8

19 20 21 2 23

0 24

10

20

30

40

26 27

(b) Vacuum Annealed Sample 28 29

36 37 38 39 40 41

'InAs'-like

TO(Γ )

35

LA(X)

34

TA(X)

3

TA(X)

32

TO(Γ )

IETS (arb. unit)

31

LO(Γ )

'GaAs'-like 30

LO(Γ )

VG (mV) 25

42 43 4 45 46

0 47

10

20

30

40

VG (mV) 48 49 51

50

Figure 53

52

8:

Phonons

of

‘InAs-’

and

‘GaAs-like’

modes

observed

in

ALD

Al/TiO2/Al2O3/In0.53Ga0.47As/InP MOS capacitors from IETS: (a) without vacuum anneal, (b) 54 56

5

vacuum-annealed samples. 57 58



Page 40 of 54

1 2 3 4

(a) Without Anneal Sample 5 6 7

TiO2 8

TO(Γ5 )

18

17

16

15

14

13

12

1

10

9

IETS (arb. unit)

+

Al2O3

5

TiO2

Al2O3

TiO2 LO(ε| |) + Al2O3 LO(Γ )

TO(ε_ |_)

−

LO(Γ1 )

TO(ε_ |_) Al2O3 Al2O3 TO(ε| |) LO(ε_ |_)

Al2O3 Al2O3 LO(ε_ |_) LO(ε| |)

19 20 21 2

40 23

60

80

24

100

120

VG (mV) 25 26 27

(b) Vacuum Annealed Sample 28 29

TiO2

41

40

39

38

37

36

35

34

3

32

+

IETS (arb. unit)

31

30

TO(Γ5 ) TiO2 TiO2

TO(ε_ |_)

+

LO(Γ5 ) Al O 2 3

Al2O3

LO(ε_ |_)

LO(ε| |) Al2O3 TO(ε| |)

42

−

LO(Γ1 )

Al2O3

Al2O3 LO(ε| |)

43 4 46

45

40 47

60

80

100

VG (mV) 50

49

48

Figure 9: Phonons of TiO2/Al2O3 bi-layer gate stack in ALD Al/TiO2/Al2O3/In0.53Ga0.47As/InP 51 53

52

MOS capacitors observed from IETS: (a) without vacuum anneal, (b) vacuum annealed samples. 54 5 56 57 58


Page 41 of 54


1 2 3 4 5 6 7 8

Without Anneal 9

23

2

21

20

19

18

17

16

15

14

13

12

1

IETS (arb. unit)

10

1

3

2 1

4 3

2

24

Vaccum Annealed 25 26 28

27

150 29

200

250

300

350

31

30

VG (mV) 3

32

35

34

Figure 10: IETS spectra showing trap-features (trap-assisted conduction and charge trapping) in 37

36

the two ALD Al/TiO2/Al2O3/In0.53Ga0.47As/InP MOS capacitor devices; the trap features and 38 39

leakage is more prominent in devices without annealing. 41

40 42 43 4 45 46 47 48 49 50 51 52 53 54 5 56 57 58



Page 42 of 54

1 2 3 4 5 6

Without Anneal

-17

10 7 8 9

-19

10

10 13

SI(A /Hz)

12

1

1/f

-21

10

16

2

15

14

-23

10

20

19

18

17

1/f

2

-25

10 21 23

2

Vacuum Annealed

-27

10 25

24

27

26 2

3

10 29

28

10

4

10

Frequency (Hz) 30 31 32 3

Figure 11: 1/f noise current spectra of the ALD Al/TiO2/Al2O3/In0.53Ga0.47As/InP MOS 36

35

34

capacitors at a gate bias of -1V; the vacuum-annealed samples show a lower noise level and a 38

37

Lorentzian nature of the noise spectral density. 39 40 41 42 43 4 45 46 47 48 49 50 51 52 53 54 5 56 57 58


Page 43 of 54


1 3

2

Table of Content Graphics 4 5 6 7 8 9 10 1 12 13 14 15 16 17 18 19 20 21 2 23 24 25 26 27 28 29 30 31 32 3 34 35 36 37 38 39 40 41 42 43 4 45 46 47 48 49 50 51 52 53 54 5 56 57 58



1 2 3 4 5 6 7 8 9 10 1 12 13 14 15 16 17 18 19 20 21 2 23 24 25 26 27 28 29 30 31 32 3 34 35 36 37 38 39 40 41 42 43 4 45 46 47 48 49 50 51 52 53 54 5 56 57 58 59 60 ACS Paragon Plus Environment

Page 44 of 54

Page 45 of 54





Page 46 of 54

Page 47 of 54





Page 48 of 54

Page 49 of 54



-1

1

2

2

3

0.53

0.47

2


Al/TiO2/Al2O3/p-In0.53Ga0.47As With VA

2

10

-2

10

-3

10

-4

10

-5

10

-6

10

-7

3 4

10

/Al O /p-In Ga Materials As Without VA ACSAl/TiO Applied & 10Interfaces Page 50 of 54 2

1


10

High-k/Si SiO2/Si This Work

0

-1

10 -2 10 -3 10 -4 10 -5 10 -6 10 -7 10 -8 10

1.0

1.5

2.0 2.5 EOT (nm)

3.0

3.5

6

5

13

12

1

10

9

8

7

-3

ACS Paragon Plus Environment -2 -1 0 1 2

Applied Voltage (V)

3

Page 51 of 54





Page 52 of 54

Page 53 of 54





Page 54 of 54

Al2O3 Films on

Recommend Documents