1. Introduction
A charge trap flash (CTF) memory is widely accepted as a candidate for the next generation nonvolatile memory devices due to its significant advantages such as localized charge storage and coupling-free structure[1, 2], which are particularly important for 3D applications. To achieve a large memory window under low program/erase (P/E) voltages on CTF memory devices, the typical silicon nitride trapping layer can be replaced by high-k materials such as HfO2 with small equivalent nitride thickness[3-7]. Although some work has been done for HfO2 material optimization[8], investigation of its reliability characteristics, such as the poor retention property, is still an important issue and becomes more important for future high-density applications. To improve the retention characteristics, it is founded that the post-deposition annealing (PDA) process is essential for the properties of HfO2[9]. However, the detailed retention degradation mechanism of HfO2 annealed in a different ambient has not yet been investigated[10]. To further understand the properties of the retention degradation mechanism as a guide for device performance optimization in CTF, it is of great importance to directly investigate the vertical and lateral charge loss characteristics for HfO2 annealed in a different ambient, and the properties of trap energy. However, conventional charge pumping and 1/f noise technology mainly focus on the Si/SiO2 interface properties, and cannot investigate the properties of the HfO2 trapping layer directly[11, 12]. Kelvin probe force microscopy (KFM) is a very powerful tool in investigating the retention characteristics by measuring contact potential difference (CPD) line profiles and calculating the charge density after charge injection[13]. Although the trapping properties and retention characteristics of electrons have been determined in some studies[9], only a limited number of reports have discussed the hole trapping properties in the HfO2 trapping layer. Furthermore, the trapping properties, retention characteristics, and trap energy have never been simultaneously determined on one set of samples using the same measurement method. Since variations in growth conditions, dielectric stack structures, and measurement techniques can result in a large variation of extracted charge trapping properties, a systematical study of all these trapping properties for both electrons and holes on one specific sample set using the same experimental method is highly desirable. In this work, we employ the technique of variable temperature KFM to study the trapping properties, retention characteristics, and trap energy (of both electrons and holes), in HfO2 annealed in a different ambient.
2. Experiment setup
The structure of HfO2-oxide-silicon was fabricated on an 8-12 Ω⋅cm p-type Si (100) substrate. After standard cleaning, a 4-nm SiO2 layer was thermally grown at 900 ℃ in dry O2 ambient. Then, a 10 nm HfO2 layer was subsequently deposited by atomic layer deposition at 150 ℃. Finally, all devices were treated by post-deposition annealing for 60 s in a N2 or O2 ambient at 1000 ℃. Data retention behavior was visualized by KFM technology, with Bruker's newest scanning probe microscopy system (Multimode 8). All samples before measurement had been prebaked at 105 ℃ for more than 120 min to desorb humidity.
Charge injection was done by scans (speed of 0.6 Hz and vertical electric field of 12 MV/cm) when the probe was under contact mode, where a 0.6-Hz scan speed guarantees injected charge saturation, as illustrated in Fig. 1(a). In the KFM mode, the cantilever measures and records the surface topography. Then, 2-D potential images of injected charges were visualized in the interleave scan of the KFM mode along the recorded topography, as illustrated in Fig. 1(b).
For MAHOS devices, based on the fabrication of the HfO2-oxide-silicon structure, a 15 nm Al2O3 was deposited on the HOS by ALD at 200 ℃ as a blocking layer, followed by post-deposition annealing for 60 s in a N2 or O2 ambient. Finally, 200 nm Al gate was deposited by an e-beam evaporator as a gate electrode to complete the MAHOS structure. The electrical measurements were performed by an Agilent B1500 semiconductor characterization system.
3. Results and discussion
Based on the C-V measurements, the dielectric constants of the HfO2 samples are firstly extracted, of which there were 20. KFM tests were performed to explore the data retention characteristics and HfO2 samples annealed in different ambients. Figure 2 shows the measured contact potential difference (CPD) line profiles of a sample with a HfO2 trapping layer after charge injection. The line profiles are obtained by extracting a horizontal scan-line across the injection center in the 2-D potential image. The inset regions denote the trapped electrons, which present the evolution of the surface potential image at different elapsed times. The downward arrow in Fig. 2(b) describes that the CPD peak value decreases due to vertical decay and lateral spreading and the left arrow describes that the widening of CPD profiles only results from lateral spreading. We calculated the initial injected charge density from the CPD peak value, which is based on the assumption of two extreme charge distributions: charges are trapped in the trapping bulk (Ⅰ) or at the interface (Ⅱ)[14]. The two assumptions of the charge distribution have the same trend in a charge density calculation. For simplicity, we use distribution (Ⅱ) as our calculation method. Meanwhile, compared to the N2 PDA sample, a larger CPD peak value, larger charge injection and less decay are observed for the O2 PDA sample in Fig. 2(c).
Under the same estimating strategy, charge retention characteristics are extracted from different samples with a normalized electron and hole density as shown in Fig. 3. The overall charge decay of the N2-PDA HfO2 trapping layer is larger than O2-PDA ones. The overall charge decay of holes is larger than the electrons.
Considering that the retention property has a strong relation with the vertical decay and lateral diffusion as illustrated in Fig. 2(b), we integrate charge densities from the extracted CPDs to obtain the remaining proportion after 7560 s, which remain at 0.9216, 0.8889, 0.7948, and 0.7491, corresponding to the electron density of the samples annealed in an O2 or N2 ambient, and the hole density of the samples annealed in an O2 or N2 ambient, respectively. We can conclude that the O2 PDA condition is effective in reducing the total charge loss.
To further understand the charge loss characteristics in HfO2, the lateral charge spreading is another concern to be evaluated. As illustrated in Fig. 4, the electron density profile can be fitted by a Gaussian distribution. Selected potential line profiles are fitted by the following equation to qualify the diffusion coefficient[15].
p=U2√πDp(t+t0)exp[−x24Dp(t+t0)], |
(1) |
where p, U, Dp, t and x are the charge density, the constant related to the initial charge density, the diffusion coefficient, elapsed time, and lateral dimension, respectively. The fitting results are illustrated in Table 1.
![]() |
The results show that the lateral diffusion of HfO2 material annealed in O2 ambient is smaller than the N2-PDA ones. The lateral diffusion coefficient of holes is larger than electrons as well. The lateral charge spreading characteristics evaluated by the diffusion coefficient are in good agreement with the aforementioned total charge loss properties.
To determine the effective trap energies of charges trapped in the HfO2 trapping layer, we studied their retention behavior at elevated temperatures. We assume that the thermal emission, which is the thermal excitation of trapped electrons (holes) from the trap centers in HfO2 to the conduction (valence) band, followed by the oxide tunneling process, is the main decay process.
Firstly, using the calculated charge trap density in each sample, we can obtain the retention times τ at different temperatures, then trap energy Et can also be determined from the linear relationship between log(τT2) and 1/T.
log(τT2)=−log(aebb)+Et2.3kB1T. |
(2) |
Figure 5(a) shows the CPD decay profiles of the sample with a N2-PDA HfO2 trapping layer at 25 ℃; the 90 ℃ results show that CPD decay becomes more severe at high temperatures. Figure 5(b) shows the retention behavior at high temperatures, and we extracted the effective retention time τ from the slope, which is closely related to the thermal emission and band-to-band tunneling. Using τ values at different T, we obtain the effective trap energies by linear fitting as illustrated in Fig. 5(c) when we employ the Shockley-Read-Hall (SRH) model. The results are illustrated in Table 2.
![]() |
Considering all the above, it is concluded that, compared to N2-PDA, the O2-PDA HfO2 trapping layer shows the better retention characteristics in both the vertical and lateral directions. Vertical charge leakage and lateral charge spreading both play important roles in the charge loss mechanisms. The retention improvement is attributed to the deeper effective trap energies.
Finally, memory devices annealed at different ambients are fabricated and measured[16]. Figure 6 presents the high-frequency (100 kHz) C-V hysteresis of the two samples under ±15 V sweep voltage to observe the memory effect. The results show that the device annealed in the O2 ambient has a larger memory window. The memory effect of the MAHOS structure originates from the intrinsic HfO2 traps and the trap sites at the interface. Larger memory windows indicate that the PDA processes have an impact on the trap density of the AHOS structure, which agrees with the KFM result in Fig. 2(c).

Data retention properties with an identical initial flat-band voltage are shown in Fig. 7. We can see that the retention characteristics of the O2 PDA device at 25 ℃, 85 ℃ is better than the device annealed in the N2 ambient. Comparing these results with Fig. 3, the results are in good agreement.
With Al2O3 as the blocking oxide layer, the gate leakage current properties at 25 ℃ and 85 ℃ of samples annealed in N2 and O2 ambients are shown in Fig. 8. It is shown that the gate leakage current of the O2 PDA device is smaller than the device annealed in the N2 ambient at 25 ℃, 85 ℃. Comparing these results with Fig. 7, we can know that the post-deposition anneal in the O2 ambient improved the retention characteristics in MAHOS memory devices by reducing the leakage current and then improving the retention characteristics.
4. Conclusion
In summary, the HfO2 material annealed in O2-PDA shows a larger injected charge density, and a better retention characteristic in both the vertical and lateral directions than N2-PDA ones. Vertical charge leakage and lateral charge spreading both play important roles in the charge loss mechanisms. The retention improvement is attributed to the deeper effective trap energies. Finally, the electrical characteristics of memory devices are demonstrated from the experiment, which agreed with our characterization results.