Jpn. J. Appl. Phys. Vol. 41 (2002) pp. 805–809 Part 1, No. 2A, February 2002 #2002 The Japan Society of Applied Physics
Evaluation of SiO2 Films and SiO2 /Si Interfaces by Graded Etching Yuichi M URAJI*, Kazuhiro Y OSHIKAWA, Masakazu NAKAMURA1 and Yoshitsugu N AKAGAWA Toray Research Center, Inc., 3-3-7 Sonoyama, Otsu, Shiga 520-8567, Japan 1 Department of Electronics and Mechanical Engineering, Chiba University, 1-33 Yayoi-cho, Inage-ku, Chiba 263-8522, Japan (Received August 3, 2001; accepted for publication October 26, 2001)
We have established a new preparation method for the evaluation of SiO2 films. The method, graded etching, a good replacement of conventional step etching, enables us to analyze the SiO2 films and the SiO2 /Si interfaces in detail with much less effort. Three kinds of SiO2 films (HCl, dry and wet oxides) have been evaluated after graded etching. Atomic force microscopy (AFM) and X-ray photoelectron spectroscopy (XPS) results revealed the presence of a SiOx (0 < x < 2) rich transition layer near the SiO2 /Si interfaces, the amount of which differs among the three kinds of oxides. AFM images also indicated that the SiO2 films were etched spottily at the beginning of etching, which results in different roughnesses of the etched surfaces for the three kinds of oxides. [DOI: 10.1143/JJAP.41.805] KEYWORDS: silicon dioxide, graded etching, chemical etching, atomic force microscopy (AFM), roughness, ellipsometry, X-ray photoelectron spectroscopy (XPS), transition layer
1.
Introduction
1 - 3 cm
Further down-scaling of gate dielectric thickness is an inevitable trend in order to continuously increase packing density and enhance device performance. Gate oxide direct tunneling current and gate oxide reliability have become more important issues as the scaling of the oxide thickness continues, with nonlinear scaling of the power supply due to the higher electric field across the gate oxide.1) Rasras et al. demonstrated a percolation model where oxide breakdown is controlled by the bulk oxide in which hole-induced electron trap generation occurs.2) It has also been reported that the interface roughness and the structural defects within the transition layer near the SiO2 /Si interface have an effect on the thin oxide reliability.3,4) In addition, hydrogen-containing species may degrade oxide reliability due to the formation of weak Si-H bonds at the SiO2 /Si interface and enhance the generation of carrier trapping sites in the bulk oxide,5) although, historically, a hydrogen-containing ambient annealing has been reported to be beneficial in passivating interfacial traps.6,7) Accordingly, it is very important to evaluate the roughness of SiO2 /Si interfaces, the transition layer and bulk defects to evaluate the quality of gate oxide layers for metaloxide-semiconductor (MOS) transistors. A step etching method, the iteration of HF etching and some measurements such as atomic force microscopy (AFM), is frequently used for this purpose.8–11) Step etching is, however, practically inefficient due to the iteration of etching and measurement. Moreover, it is difficult to investigate the structural variation near the Si interfaces in detail by step etching. We have, therefore, developed a new preparation method, graded etching, as a replacement for step etching. In graded etching a long and narrow Si chip is immersed in HF solution and then drawn up. A sample, as illustrated in Fig. 1, can be prepared by graded etching. Fabricating this kind of sample, we can obtain more detailed information on the SiO2 /Si interfaces because it keeps a SiO2 film with any remaining thickness. We will show the surface roughness measured by AFM, the SiO2 thickness by ellipsometry, and the chemical bonding state by X-ray photoelectron spectroscopy (XPS), after graded etching. HCl-, dry-, and wet-oxidized SiO2
SiO2 Si Fig. 1. Schematic illustration of a Si chip (top) and its cross section (bottom) after graded etching. This chip contains areas of bare Si and nonetched SiO2 surfaces and any remaining thickness between them. The slope can be changed according to the etching time or the density of HF solution.
films have been measured and compared. 2.
Experimental
Three kinds of silicon oxide films (dry oxide film thermally grown in O2 , wet oxide film in O2 and H2 , and HCl oxide film in O2 with addition of HCl) were formed on Si wafers at 900 C. The thicknesses of the oxide films were 3–5 nm. Each oxidized Si wafer was cut into a 1 5 cm2 chip. After cleaning the chips with acetone, graded etching was carried out with a 0.5% HF solution. During the graded etching, the chip is gradually immersed in HF solution with constant velocity and then drawn up. The drawing up velocity is more than 100 times as fast as the immersion velocity. AFM images were obtained with a Digital Instruments NanoScope III system, using tapping mode. The points measured by AFM are shown in Fig. 2. Si cantilevers (OMCL-AC120TS, Olympus Optical Co., Ltd.) were used for the imaging. The scan area was set to 1 1 m2 and the scan rate to 0.8 Hz. Mean square roughness (Rms ) was estimated from each AFM image obtained. The thickness of the remaining oxide film after the graded etching was measured with an ellipsometer (L116A, Gaertner Inc.) at the same points as the AFM observations.
*
E-mail: Yuichi
[email protected] 805
806
Jpn. J. Appl. Phys. Vol. 41 (2002) Pt. 1, No. 2A
Y. MURAJI et al.
SiO2 Si Fig. 2. Schematic illustration of the points measured by AFM. AFM images were obtained at more than 20 points between the SiO2 surface and the SiO2 /Si interfaces.
The measurement area was around 3 1 mm2 . The longitudinal direction of the area was placed perpendicular to the slope direction of the sample. In thickness calculations, the refractive index of the film was fixed at 1.46 to eliminate the difficulty of determining both the refractive index and the thickness of very thin SiO2 films. The refractive index and the extinction coefficient of the substrate were calibrated by measuring a bare Si area for each Si chip. The XPS study was carried out with ESCALAB220iXL (VG) using monochromatic AlK as an excitation source (h ¼ 1486:6 eV). A hemispherical electron analyser was operated in fixed analyser transmission (FAT) mode by selecting a constant pass energy of 30 eV throughout the measurements. Under these conditions, the full-width at half-maximum (FWHM) of the Ag3d5=2 line was confirmed to be 0.8 eV. The diameter of the analysis area was around 1 mm. All of the measurements were performed at pressures of 1 109 Torr. 3.
Results and Discussion
3.1 AFM and XPS results obtained with graded etching Figure 3 shows the Rms (left) and the remaining SiO2 thicknesses (right) against etching time for the HCl oxide film. More than 20 measurement points are obtained even for the SiO2 film with an initial thickness of 3.5 nm. This is the
(1)
(2)
most advantageous point of the graded etching. Considering the areas marked (1)–(3) in Fig. 3, we can discuss the characteristics of the sample as follows: (1) The Rms increased immediately after etching started. The increase indicates the existence of certain structural nonuniformities, such as density fluctuation or structural defects. The etching rate of a microvolume is expected to depend on the density of the film, which is affected by the voids or fragile structure. When we assume that the fragile part is less than half the volume of the film (this assumption is reasonable because AFM images indicate the slight increase in the number of hollows), the Rms increases as the amount of fragile microvolume increases. (2) The Rms increased abruptly and the etching rate decreased at the remaining thickness of approximately 0.5 nm. This implies the existence of a transition layer, where the chemical structures having a slower etching rate are distributed nonuniformly. (3) We can evaluate the roughness of the SiO2 /Si interfaces after necessary and sufficient etching. This is also an advantage of graded etching. If we use onestep etching to measure the interface roughness, the possibility of overestimating the roughness due to the transition layer or roughening by the excessive etching of Si by HF exists. In this case, the Rms of the interface was estimated to be 0.145 nm. XPS measurements were carried out at the remaining SiO2 thicknesses of 3.8, 3.0, 0.7, 0.3 and 0 nm. The Si2p spectra obtained are shown in Fig. 4. Each spectrum is normalized by the peak intensity. The intensity of the peak corresponding to SiO2 (the binding energy is 103.6 eV) changes according to the thickness of the remaining SiO2 films. A trace of suboxide, SiOx (0 < x < 2), is noted between the SiO2 and Si peaks. Based on the spectra in Fig. 4, difference spectra were obtained between neighboring spectra and were decomposed into Si, SiOx and SiO2 components to investigate the etched
(3)
0.22
6
0.20
5
0.18
4
0.16
3
0.14
2
Thickness (nm)
Rms (nm)
Rms (left)
Thickness (right) 0.12
1
0.10
0 0
50
100
150
Etching Time (s) Fig. 3. Rms and remaining SiO2 thicknesses against etching time for a HCl oxide film measured after graded etching.
Fig. 4. XPS Si2p spectra at the remaining SiO2 thicknesses of 3.8, 3.0, 0.7, 0.3 and 0 nm. Each spectrum is normalized by the peak intensity.
Jpn. J. Appl. Phys. Vol. 41 (2002) Pt. 1, No. 2A Table I.
Y. MURAJI et al.
Results of difference spectrum obtained by XPS.
0.26
SiO2 thickness (nm)
SiO2 (%)
SiOx (%)
3.8–3.0 0.7–0.3
89 89
11 11
0.3–0
67
33
0.24 0.22
substances. The results are shown in Table I. These indicate that the SiOx component is larger within the thickness range of 0.3–0 nm, i.e., near the SiO2 /Si interface. It is, therefore, concluded that the increase in the Rms in the transition layer is due to the nonuniform distribution of the SiOx structure in the transition layer. Strictly speaking, the difference in the decay length of photoelectrons between Si and SiO2 must be taken into account for the differential analyses. The estimated error of the values in Table I is about 5%, which includes both the normalization error and an error in the curve fitting. The error is, however, not sufficiently large to prevent us from concluding that a large amount of SiOx exists near the interface because the difference spectra were obtained between those which have small differences in SiO2 thickness. 3.2 Comparison of the three kinds of SiO2 films Figures 5(a) and 5(b) show plots of the Rms and the SiO2 thicknesses against the etching time for dry and wet oxide films, respectively. The Rms of both films also increases near the Si interfaces as seen in the HCl oxide film. The wet oxide film exhibits the largest Rms near the SiO2 surface among the films evaluated. Here, we will compare the three kinds of oxide films. Figure 6 illustrates the relationship between the remaining thickness and roughness of the three kinds of oxide films. From this figure, we can conclude the following: (1) Comparing Rms at 0 nm, the roughnesses of the SiO2 /Si interfaces are in the order of dry > wet > HCl. 0.26
HCl Dry Wet
0.12 0
1
2
3
4
5
Thickness (nm) Fig. 6. Relationship between the remaining thickness and roughness of the three kinds of oxide films.
(2) Rms changes abruptly near 0.5 nm, suggesting that every oxide film has a transition layer near the SiO2 /Si interface. The thicknesses of the transition layers are almost the same for the three oxide films. However, because the maximum Rms values at around 0.3 nm are not the same among the three oxides, the density of microvolume which is not easily etched differs among them. The amounts of the microvolumes are in the order of wet ; dry > HCl. (3) The Rms values near the SiO2 surfaces (1–2 nm from the surface) differ among the three films. Since the Rms is supposed to increase as the nonuniformity becomes 0.26
7 Rms (left)
0.24
6
0.22
0.18 3 0.16 Thickness (right)
0.14 0.12 0.10 100
Etching Time (s)
150
2
5
0.20 4 0.18 3 0.16 0.14
1
0.12
0
0.10
Thickness (right)
2 1 0
0
50
100
150
Etching Time (s)
Fig. 5. Rms and remaining SiO2 thicknesses against etching time of (a) dry and (b) wet oxide film after graded etching.
Thickness (nm)
4
Thickness (nm)
0.20
6
0.22
5
50
0.18
0.14
Rms (left)
0
0.20
0.16
7
0.24
Rms (nm)
Rms (nm)
0 HCl. Conclusions (1), (2) and (4) are the same as those obtained by step etching.10) However, graded etching has made the experiment far more efficient and precise. Conclusion (3) is newly obtained using graded etching. From the point of view that the evaluation of SiO2 /Si interfaces is important for a discussion of the quality of SiO2 films, we put emphasis on the analysis of the interfaces after graded etching. However, graded etching also revealed an interesting phenomenon near the SiO2 surfaces. It was shown that there were spotty regions with faster etching rates near the surfaces of SiO2 films. Their structures are to be studied in detail. In this work, we applied graded etching for AFM, XPS and ellipsometry. Graded etching can be applied for other measurement techniques, for example, Fourier transform infrared spectroscopy (FT-IR) and secondary ion mass spectroscopy (SIMS). Moreover, graded etching is applicable for SiON and SiN films. It can be said that graded etching is a useful pretreatment for many kinds of analyses of thin films. 1) Y. Taur and E. J. Nowak: IEDM Tech. Dig. 1997, p. 215. 2) M. Rasras, I. De Wolf, G. Groeseneken, R. Degraeve and H. E. Maes:
Jpn. J. Appl. Phys. Vol. 41 (2002) Pt. 1, No. 2A IEDM Tech. Dig. 2000, 22.3.1 3) E. Hasegawa, A. Ishitani, M. Tsukiji, K. Akimoto and N. Ohta: J. Electrochem. Soc. 142 (1995) 273. 4) M. Nagamine, H. Itoh, H. Satake and A. Toriumi: IEDM Tech. Dig. 1998, p. 593. 5) D. J. DiMaria, E. Cartier and D. Arnold: J. Appl. Phys. 73 (1993) 3367. 6) E. H. Nicollian, C. N. Bergland, P. F. Schmidt and J. M. Andrews: J. Appl. Phys. 42 (1971) 5654. 7) F. C. Hsu, J. Hui and K. Y. Chiu: IEEE Electron Device Lett. 6 (1985)
Y. MURAJI et al.
809
369. 8) S. J. Fang, H. C. Lin, J. P. Snyder, C. R. Helms and T. Yamanaka: J. Electrochem. Soc. 143 (1996) 329. 9) S. J. Fang, W. Chen, C. R. Helms and T. Yamanaka: J. Electrochem. Soc. 143 (1996) 338. 10) K. Yamabe, K. Kiao and M. Murata: 109–114, JSAP Catalog Number: AP002201. 11) H. Hashimoto, N. Nagai, H. Oyanagi, S. Ichimura and T. Maeda: 209– 214, JSAP Catalog Number: AP002201.
