Origin of traps and charge transport mechanism in hafnia
D. R. Islamov, V. A. Gritsenko, C. H. Cheng, and A. Chin
Citation: Applied Physics Letters 105, 222901 (2014); doi: 10.1063/1.4903169 View online: http://dx.doi.org/10.1063/1.4903169
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/105/22?ver=pdfcov Published by the AIP Publishing
Articles you may be interested in
Charge trapping defects in Si / SiO 2 / Hf ( 1 − x ) Si x O 2 film stacks characterized by spectroscopic second-harmonic generation
J. Vac. Sci. Technol. B 29, 04D101 (2011); 10.1116/1.3591433 Device characteristics of HfON charge-trap layer nonvolatile memory J. Vac. Sci. Technol. B 28, 1005 (2010); 10.1116/1.3481140
Characterization of fast charge trapping in bias temperature instability in metal-oxide-semiconductor field effect transistor with high dielectric constant
Appl. Phys. Lett. 96, 142110 (2010); 10.1063/1.3384999
Reduction in charged defects associated with oxygen vacancies in hafnia by magnesium incorporation: First-principles study
Appl. Phys. Lett. 93, 223104 (2008); 10.1063/1.3040306
Mechanism of improved channel carrier mobility for stacked Y 2 O 3 ∕ Hf O 2 gate dielectric Appl. Phys. Lett. 89, 173501 (2006); 10.1063/1.2363141
Origin of traps and charge transport mechanism in hafnia
D. R. Islamov,1,2,a)V. A. Gritsenko,1,2,b)C. H. Cheng,3and A. Chin4,c) 1
Rzhanov Institute of Semiconductor Physics, Siberian Branch of Russian Academy of Sciences, Novosibirsk 630090, Russian Federation
2
Novosibirsk State University, Novosibirsk 630090, Russian Federation
3
Department of Mechatronic Technology, National Taiwan Normal University, Taipei 106, Taiwan
4
National Chiao Tung University, Hsinchu 300, Taiwan
(Received 9 October 2014; accepted 19 November 2014; published online 1 December 2014) In this study, we demonstrated experimentally and theoretically that oxygen vacancies are responsible for the charge transport in HfO2. Basing on the model of phonon-assisted tunneling
between traps, and assuming that the electron traps are oxygen vacancies, good quantitative agreement between the experimental and theoretical data of current-voltage characteristics was achieved. The thermal trap energy of 1.25 eV in HfO2 was determined based on the charge
transport experiments.VC 2014 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4903169]
Knowledge about charge transport mechanisms hafnia (hafnium oxide, HfO2) is crucial for modern
microelec-tronics, because high-j HfO2is used as a gate dielectric in
high-speed MOSFETs1,2and FinFETs3,4and a blocking in-sulator in Si-oxide-nitride-oxide-silicon-type (SONOS) flash memory cells.5,6Hafnia is a promising candidate to used as active medium in resistive random access memory (RRAM), which would involve combining the most favorable proper-ties of both high-speed dynamic random access memory and non-volatile flash memory.7–9 Hereby, MOSFETs and SONOS need high-j dielectrics with low leakage currents, while RRAM requires dielectric medium with reversible resistive switching. Managing the process of hafnia films synthesis to control leakage currents can create high-quality devices for various purposes. However, an unresolved physics is the nature of defects and traps that are responsible for the charge transport in HfO2. The atomic structure of
defect that affects the localization and charge transport still remains unclear. Currently, the accepted hypothesis is that oxygen vacancies are responsible for charge transport in dielectric.9–11 Although many studies have investigated the theory of the atomic and electronic structure of oxygen vacancies in hafnia,11–18 direct experimental data regarding the presence of oxygen vacancies in hafnia were reported recently.19It was shown that oxygen vacancies in hafnia are responsible for blue luminescence band at 2.7 eV and a lumi-nescence excitation band at 5.2 eV, and a hypothesis that the oxygen vacancies in hafnia act as traps in charge transport through the dielectric was discussed.19 In this case, thermal energy traps in HfO2are equal to a half of the Stokes
lumi-nescence shiftWt¼ ð5:2 2:7Þ=2 ¼ 1:25 eV.
A lot of studies of the charge transport in dielectrics described experiment results by Poole-Frenkel (PF) mecha-nism in hafnia-based structures.20–23 The most part of HfO2
transport investigations did not get into account neither charge trap density, which depends on thin film fabrication technology, nor phonon influence on electron and hole
transport, which might be significant at high temperatures. Only few of investigations explained their results quantita-tively as well as qualitaquantita-tively, getting agreement of the phenomenological parameters such as dynamic permittivity, trap energy, frequency factor, etc. However, these studies involved complex mathematical calculations.24
In this letter, phonon-assisted tunneling between traps (PATT) conduction mechanism in HfO2 was developed.
Proposed model has simple mathematics and is in good quan-titative and qualitative agreement with experimental data. It was clearly shown that oxygen vacancies are responsible for the charge transport in HfO2and HfO2-based devices.
Transport measurements were performed for metal-insu-lator-semiconductor (MIS) and metal-insulator-metal (MIM) structures. For the MIS Si/HfOx/Ni structures, the
20-nm-thick amorphous hafnia was deposited on an-type Si (1 0 0) wafer by using the atomic layer deposition (ALD) system. Tetrakis dimethyl amino hafnium (TDMAHf) and water vapor were used as precursors at a chamber temperature of 250C for HfOxfilm deposition.
Another set of MIS samples with 8-nm-thick HfOxfilms
was fabricated using physical vapor deposition (PVD). A pure HfO2 target was spattered by an electron beam, and
HfO2was deposited on then-type Si (1 0 0) wafer. Low
tem-perature post-deposition annealing (PDA) during 15 min at 400C was applied to prevent the growth of interfacial SiOx.
25
Structural analysis shows that the resulting HfOx
films were amorphous. To fabricate Si/TaN/HfOx/Ni MIM
structures, we deposited the 8-nm-thick amorphous hafnia on 100-nm-thick TaN films on Si wafers, using PVD. We did not apply any post-deposition annealing to produce the most non-stoichiometric films. All samples for transport measure-ments were equipped with round 50-nm-thick Ni gates with a radius of 70 lm. Transport measurements were performed using a Hewlett Packard 4155B semiconductor parameter an-alyzer and an Agilent E4980A precision LCR meter.
The experimental current-voltage (I-V) characteristics in MIS(PVD) structures, measured at different temperatures T with a positive bias applied to the Ni contact, are shown in Fig. 1, graphed by different characters in PF ðlogðIÞ–pffiffiffiffiVÞ plot. The current grows exponentially with increasing of the a)Electronic mail: [email protected]
b)Electronic mail: [email protected] c)
Electronic mail: [email protected]
gate voltage and temperature. We attempted to explain the experimental data by using isolated trap ionization model
J¼ eN2=3
P; (1)
whereJ is the current density, e is the electron charge, N is the charge trap density,P is the probability of trap ionization per second and has different dependencies on electric fieldF and temperature. In term of PF model, the probability is
P¼ exp Wt bPF ffiffiffi F p kT ; (2)
is the frequency factor which was defined as ’ Wt=h, Wt
is thermal trap energy (the energy of thermal ionization of the trap),h is the Planck constant, bPFis Poole-Frenkel
coef-ficient, F is the electric field, and k is the Boltzmann constant.26 Experimental I-V characteristics and results of the fitting procedure are shown in Fig.1. As can be seen, Frenkel model (2) describes the experiment data qualita-tively very good. However, quantitative fitting procedure returns nonphysical fitting parameter values: the slopes of the fitting lines with PooleFrenkel coefficient bPF¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi e3=pe
0e1
p
give the dynamic permittivity e1ðTÞ ¼ 10–20,
which is much higher than e1ðHfO2Þ ¼ 4. e0 is vacuum
permittivity (dielectric constant). Further fittings return N 4 cm3 and Wt¼ 0:3–0:4 eV. Found value, the charge
trap density of N 4 cm3 at ’ Wt=h 10 14
s1, corre-sponds to one trap per2600 Ni contacts, thus this is unreal-istic value. Taking these into account, it was concluded that despite the fact that PF model describes the experiment data qualitatively, there is no quantitative agreement between experiments and theory. We tried to describe our experi-ments with other charge transport models in dielectrics, Hill model of overlapped traps ionization,27 and the model of multiphonon trap ionization.28 However, the fitting proce-dures involved in these models returned the nonphysical fitting parameter values as well as PF model.
To describe the experiments quantitatively and qualita-tively, we performed simulations based on the PATT model.29In this model, the probability of electron tunneling between traps per second is defined as following:
P¼ ffiffiffiffiffiffi 2p p hWt ms2 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiW opt Wt p exp Wopt Wt 2kT exp 2s ffiffiffiffiffiffiffiffiffiffiffi mW t p h sinh eFs 2kT ; kT Wt; eFs Wt; (3)
h¼ h=2p, Woptis the energy of optical excitation of the trap,
m* is the effective mass, and s¼ N1=3 is mean distance between traps. The results of this multi-parameter fitting procedure are shown in Fig. 2, graphed in solid lines. This procedure yielded the values of different transport parame-ters, N¼ 6.8 1019cm3, Wt¼ 1.25 eV, Wopt¼ 2Wt
¼ 2.5 eV, and m¼ ð0:3–0:4Þm
e(meis a free electron mass).
Quantitatively, there is full agreement between the PATT model and the experimental data. The trap thermal energy value of 1.25 eV that was obtained is close to that of 1.2 eV (Ref. 30) and Wt¼ 1:36 eV (Ref. 24) observed earlier, and
equal to a half of the Stokes luminescence shift.19
Furthermore, the trap optical energy value of Wopt¼
2:5 eV is close to the calculated value of 2.35 eV for the neg-atively charged oxygen vacancy in hafnia reported earlier.15
Fig.3shows the configuration diagram of a negatively charged oxygen vacancy (electron trap) in hafnia. A vertical transition with a value of 2.5 eV corresponds to the optical trap excitation, and transitions of 1.25 eV correspond to ther-mal trap energy.
The inset in Fig. 2 show experimental capacitance-voltage characteristics of the MIS(PVD) structure at different voltage limits. Positions of the right hysteresis branch ofC-V characteristics depend on the voltage limits, while the left
FIG. 1. Experimental current-voltage characteristics (characters) ofn-Si/ HfOx/Ni MIS(PVD) structure and simulation (lines) by Frenkel model of the
trap ionization(1)and(2)at different temperatures on PF plot.
FIG. 2. Experimental current-voltage characteristics (characters) of n-Si/ HfOx/Ni MIS(PVD) structure and simulation (lines) by PATT model (3) at
different temperatures. Inset: experimental capacitance-voltage characteris-tics of the MIS(PVD) structure at different voltage limits; arrows show sweep voltage directions.
branch is fixed. Sweep voltage directions show that the bulk charge is accumulating on the traps with the voltage grow-ing. These phenomena show that the electron traps are filled only, while the holes traps (if any) are empty. Voltage loop allows us to valuate the density of the filled electron trapsne
ⱗ 3.7 1018
cm3, which is less than 20 times than found value ofN.
The same procedure was applied to experiment data of the charge transport measurements in MIS(ALD) and MIM struc-tures. Experiment current-voltage characteristics compared with simulations in terms of the PATT model in MIS(ALD) are shown in Fig. 4. Fitting procedure returns the following parameters values: N¼ 2.5 1020cm3, Wt¼ 1.25 eV,
Wopt¼ 2Wt¼ 2.5 eV, and m* ¼ 0.8 me. Different values of
fit-ting parameters of MIS(ALD) and MIS(ALD) structures have only the trap densityN and effective mass m*. The difference of effective mass values is due to bulk space charge (due to captured in traps electrons and holes), which is adequately addressed in Ref.31. Neither thermal trap energy nor optical trap energy depends on film fabrication technology.
The inset in Fig. 4 shows experimental capacitance-voltage characteristics of the MIS(ALD) structure at differ-ent voltage limits. With increasing of the voltage limits, the right hysteresis branch of C-V characteristics moves right,
and the left branch moves left. Using the voltage shifts, one can valuate the density of the filled holes and electron traps: 5nh neⱗ1:1 1019cm3, respectively. It should be noted
thatne N/20 as well as for MIS(PVD) structures.
The bulk charge due to trapped charge carriers blocks the carriers injections from the semiconductor and metal electrodes at low voltages and leads to deviation of simulated I(V, T) from experiment data. To simplify Fig. 4, the mis-matched data at V < 3 V are not shown. To get full agree-ment of simulations with experiagree-mental data, the bulk charge must be taken into account in solving of Poisson’s equation with Shockley-Read-Hall equations for electrons and holes.31
Fig. 5 shows the experimental data of current-voltage measurements in MIM structures at different temperatures. The solid lines present the results of simulations in terms of PATT model(3). MIM structures have the following param-eter values: N¼ 5.5 1020cm3, Wt¼ 1.25 eV, Wopt¼ 2Wt
¼ 2.5 eV, and m* ¼ 0.9 me.
An artifact feature of experimentalI-V-T curves shows that the zero current is observed at nonzero but negative voltages as shown in Fig. 5. This phenomenon is due to dis-placement current
ID¼ C dV=dt; (4)
C is capacity of the sample and dV=dt¼ þ0.3 V/s is voltage sweep rate. Taking into account (4) in simulation of I-V-T characteristics (1)and(3)with found fitting parameters, the artifact feature is described with good agreement as shown by dashed lines in Fig.5.
The difference between different MIS ad MIM struc-tures in effective mass is due to bulk space charge. However, it is important to notice that the trap’s energy parameters are invariants of grown structures and film fabrication tech-niques. Consequently, we found that the nature of charge carrier transport in hafnia and hafnia-based structures is
FIG. 3. Configuration coordination energy diagram of trap ionization pro-cess on negative charged oxygen vacancy in hafnia (lower term is filled ground state, upper term is excited empty state).
FIG. 4. Experimental current-voltage characteristics (characters) ofn-Si/ HfOx/Ni MIS(ALD) structure and simulation (lines) by PATT model(3)at
different temperatures. Inset: experimental capacitance-voltage characteris-tics of the MIS(ALD) structure at different voltage limits; arrows show sweep voltage directions.
FIG. 5. Experimental current-voltage characteristics (characters) in Si/TaN/ HfOx/Ni MIM structures at different temperatures. Black, red, and blue solid
lines representI-V simulations by PATT model at positive bias. Dashed lines showI-V simulations taking into account displacement current(4).
PATT. This charge transport model is more simple than based in quasi-continuous spectra of charge trap energy 1.4–2.4 eV, proposed by Vandelliet al.32
These results combined with spectra measurements and quantum chemical simulations19 show that, namely, oxygen vacancies are responsible for charge transport in HfOx, and
the oxygen vacancies play the role of charge traps.
Previous experiments in charge transfer have demon-strated that hafnia conductivity is bipolar (or two-band):32–34 electrons are injected from a negatively shifted contact in the dielectric, and holes are injected from a positively shifted electrode in the dielectric. For the reason of simplicity, the current study took into account electron conductivity only. This assumption is supported by evaluated values of the filled traps densities, when the holes traps are mostly empty. This means that the hole current is much lower than the electron current caused by blocking holes injection from the metal gate due to high barrier for holes on Ni/HfO2interface.
33,35
To summarize, we examined the transport mechanisms of HfO2, demonstrating that transport in hafnium oxide is
described by the PATT model. Simulating the current-voltage characteristics of this model and comparing experimental data with calculations revealed the energy parameters of the traps in hafnia: the thermal trap energy of 1.25 eV and the optical trap energy of 2.5 eV. PATT charge transport model describes experiment data results with excellent qualitative and quantita-tive agreement, while standard PF model has qualitaquantita-tive agree-ment only with unrealistic values for material parameters. These results jointly with earlier ones19facilitated in determin-ing that oxygen vacancies act as charge carrier traps.
Our results can be used to predict the leakage currents in HfO2-based devices and applications. High-quality
MOSFET and FinFET transistors and SONOS flash memory require low leakage currents through the gate dielectrics and blocking insulator, while different states in resistive memory cells must be distinguishable over a wide range of tempera-tures. Temperature dependence of memory window (resist-ance ratio in different states) in resistive memory might be predicted as well.
This work was particularly supported by National Science Council, Taiwan (Grant No. NSC103-2923-E-009-002-MY3) (growing test structures, preparing samples, and performing transport measurements) and by the Russian Science Foundation (Grant No. 14-19-00192) (calculations and modeling).
1T. P. Ma, H. M. Bu, X. W. Wang, L. Y. Song, W. He, M. Wang, H.-H.
Tseng, and P. J. Tobin,IEEE Trans. Device Mater. Reliab.5, 36 (2005).
2J. Robertson,Rep. Prog. Phys.
69, 327 (2006).
3
M. A. Pavanello, J. A. Martino, E. Simoen, R. Rooyackers, N. Collaert, and C. Claeys,Solid-State Electron.51, 285 (2007).
4A. Tsormpatzoglou, D. H. Tassis, C. A. Dimitriadis, M. Mouis, G.
Ghibaudo, and N. Collaet,Semicond. Sci. Technol.24, 125001 (2009).
5
L.-J. Chen, Y.-C. Wu, J.-H. Chiang, M.-F. Hung, C.-W. Chang, and P.-W. Su,IEEE Trans. Nanotechnol.10, 260 (2011).
6P.-H. Tsai, K.-S. Chang-Liao, T.-C. Liu, T.-K. Wang, P.-J. Tzeng, C.-H.
Lin, L. S. Lee, and M.-J. Tsai, IEEE Electron Device Lett. 30, 775 (2009).
7
J. J. Yang, M. D. Pickett, X. Li, D. A. A. Ohlberg, D. R. Stewart, and R. S. Williams,Nat. Nanotechnol.3, 429 (2008).
8L. Goux, P. Czarnecki, Y. Y. Chen, L. Pantisano, X. P. Wang, R.
Degraeve, B. Govoreanu, M. Jurczak, D. J. Wouters, and L. Altimime,
Appl. Phys. Lett.97, 243509 (2010).
9Z. Wang, H. Y. Yu, X. A. Tran, Z. Fang, J. Wang, and H. Su,Phys. Rev. B85, 195322 (2012).
10
S. Yu, X. Guan, and H.-S. P. Wong, Appl. Phys. Lett. 99, 063507 (2011).
11S. Guha and V. Narayanan,Phys. Rev. Lett.98, 196101 (2007).
12D. Mu~noz Ramo, A. L. Shluger, J. L. Gavartin, and G. Bersuker,Phys. Rev. Lett.99, 155504 (2007).
13
A. S. Foster, F. Lopez Gejo, A. L. Shluger, and R. M. Nieminen,Phys. Rev. B65, 174117 (2002).
14J. X. Zheng, G. Ceder, T. Maxisch, W. K. Chim, and W. K. Choi,Phys. Rev. B75, 104112 (2007).
15
D. Mu~noz Ramo, J. L. Gavartin, A. L. Shluger, and G. Bersuker,Phys. Rev. B75, 205336 (2007).
16H. Jiang, R. I. Gomez-Abal, P. Rinke, and M. Scheffler,Phys. Rev. B
81, 085119 (2010).
17
K. Xiong, J. Robertson, M. C. Gibson, and S. J. Clark,Appl. Phys. Lett.
87, 183505 (2005).
18P. Broqvist and A. Pasquarello,Appl. Phys. Lett.
89, 262904 (2006).
19
T. V. Perevalov, V. S. Aliev, V. A. Gritsenko, A. A. Saraev, V. V. Kaichev, E. V. Ivanova, and M. V. Zamoryanskaya,Appl. Phys. Lett.104, 071904 (2014).
20
S. K. Sahoo and D. Misra,J. Appl. Phys.110, 084104 (2011).
21
S. Pan, S.-J. Ding, Y. Huang, Y.-J. Huang, D. W. Zhang, L.-K. Wang, and R. Liu,J. Appl. Phys.102, 073706 (2007).
22W. J. Zhu, T.-P. Ma, T. Tamagawa, J. Kim, and Y. Di, IEEE Electron Device Lett.23, 97 (2002).
23
S. D. Ganichev, E. Ziemann, W. Prettl, I. N. Yassievich, A. A. Istratov, and E. R. Weber,Phys. Rev. B61, 10361 (2000).
24D. S. Jeong and C. S. Hwang,Phys. Rev. B71, 165327 (2005).
25S. H. Lin, C. H. Cheng, W. B. Chen, F. S. Yeh, and A. Chin, IEEE Electron Device Lett.30, 999 (2009).
26
J. Frenkel,Phys. Rev.54, 647 (1938).
27R. M. Hill,Philos. Mag.23, 59 (1971). 28S. Makram-Ebeid and M. Lannoo,Phys. Rev. B
25, 6406 (1982).
29
K. A. Nasyrov and V. A. Gritsenko, J. Appl. Phys. 109, 093705 (2011).
30H. Takeuchi, D. Ha, and T.-J. King,J. Vac. Sci. Technol., A 22, 1337
(2004).
31
K. A. Nasyrov, V. A. Gritsenko, Y. N. Novikov, E.-H. Lee, S. Y. Yoon, and C. W. Kim,J. Appl. Phys.96, 4293 (2004).
32L. Vandelli, A. Padovani, L. Larcher, R. G. Southwick, W. B. Knowlton,
and G. Bersuker,IEEE Trans. Electron Devices58, 2878 (2011).
33
D. R. Islamov, V. A. Gritsenko, C. H. Cheng, and A. Chin,Appl. Phys. Lett.99, 072109 (2011).
34T. Ando, N. D. Sathaye, K. V. R. M. Murali, and E. A. Cartier,IEEE Electron Device Lett.32, 865 (2011).
35
V. V. Afanas’ev and A. Stesmans,J. Appl. Phys.102, 081301 (2007).
