- Email: [email protected]
Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide
G Model
ARTICLE IN PRESS
APSUSC-34537; No. of Pages 7
Applied Surface Science xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Applied Surface Science journal homepage: www.elsevier.com/locate/apsusc
Full Length Article
Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide Tivadar Lohner a,∗ , Miklós Serényi a , Edit Szilágyi b , Zsolt Zolnai a , Zsolt Czigány a , Nguyen Quoc Khánh a , Péter Petrik a,c , Miklós Fried a,c a Institute of Technical Physics and Materials Science, Centre for Energy Research, Hungarian Academy of Sciences, H-1121 Budapest, Konkoly Thege Miklós út 29-33, Hungary b Institute for Particle and Nuclear Physics, Wigner Research Centre for Physics, Hungarian Academy of Sciences, H-1121 Budapest, Konkoly Thege Miklós út 29-33, Hungary c Doctoral School of Molecular and Nanotechnologies, Faculty of Information Technology, University of Pannonia, H-8200 Veszprém, Egyetem u. 10, Hungary
a r t i c l e
i n f o
Article history: Received 29 July 2016 Received in revised form 21 October 2016 Accepted 29 November 2016 Available online xxx Keywords: Sputtering Damage Ellipsometry Rutherford backscattering Electron microscopy
a b s t r a c t Substrate surface damage induced by deposition of metal atoms by radiofrequency (rf) sputtering or ion beam sputtering onto single-crystalline silicon (c-Si) surface has been characterized earlier by electrical measurements. The question arises whether it is possible to characterize surface damage using spectroscopic ellipsometry (SE). In our experiments niobium oxide layers were deposited by rf sputtering on c-Si substrates in gas mixture of oxygen and argon. Multiple angle of incidence spectroscopic ellipsometry measurements were performed, a four-layer optical model (surface roughness layer, niobium oxide layer, native silicon oxide layer and ion implantation-amorphized silicon [i-a-Si] layer on a c-Si substrate) was created in order to evaluate the spectra. The evaluations yielded thicknesses of several nm for the i-a-Si layer. Better agreement could be achieved between the measured and the generated spectra by inserting a mixed layer (with components of c-Si and i-a-Si applying the effective medium approximation) between the silicon oxide layer and the c-Si substrate. High depth resolution Rutherford backscattering (RBS) measurements were performed to investigate the interface disorder between the deposited niobium oxide layer and the c-Si substrate. Atomic resolution cross-sectional transmission electron microscopy investigation was applied to visualize the details of the damaged subsurface region of the substrate. © 2016 Elsevier B.V. All rights reserved.
1. Introduction The technique of sputtering involves ejecting material from the target to the substrate such as a silicon wafer. Resputtering is reemission of the deposited material during the deposition process by ion or atom bombardment. Sputtered atoms ejected from the target have a wide energy distribution, typically up to tens of eV. The sputtered ions and atoms can reach the substrate condensing after undergoing a random walk. High-energy neutrals sputtered from the target will still have enough energy reaching the substrate even after some collisions in the working gas to damage the surface of the crystalline substrate. The radio frequency (rf) voltage may be applied directly to the cathode for metal sputtering or via capacitor to sputter insulators.
∗ Corresponding author. E-mail address: [email protected] (T. Lohner).
The high frequency alternating potential may be used to neutralize the insulator surface periodically with plasma electrons, which must be switched in a period that is short compared with the time required by the positive ions to travel to the surface of target. Thus, both the amplitude and frequency of rf should be high to yield high rates. The insulator target may consist of different material: oxides, nitrides etc., from that not only atoms but molecules with considerable kinetic energy can be deposited. This is the reason why surface damage of the substrate can be significant as a result of rf sputtering deposition. Substrate surface damage induced by deposition of metal atoms by rf sputtering or ion beam sputtering onto the single-crystalline silicon (c-Si) surface has been characterized earlier by I-V, C-V and deep-level transient spectroscopy measurements [1–3]. In our experiments niobium oxide layers were deposited by rf sputtering on c-Si substrates at room temperature in gas mixture of oxygen and argon. Optical properties of thin film structures can be derived from SE measurement, which is known to be a high-
http://dx.doi.org/10.1016/j.apsusc.2016.11.232 0169-4332/© 2016 Elsevier B.V. All rights reserved.
Please cite this article in press as: T. Lohner, et al., Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide, Appl. Surf. Sci. (2016), http://dx.doi.org/10.1016/j.apsusc.2016.11.232
G Model
ARTICLE IN PRESS
APSUSC-34537; No. of Pages 7
T. Lohner et al. / Applied Surface Science xxx (2016) xxx–xxx
2
Table 1 Duration, pressure, oxygen flow, DC voltage (wall potential) of rf sputtering for different samples. Sample
A495 A496 A497 A505
A504
A503
A247
Duration Pressure [Pa] Oxygen flow[ml/min] DC voltage [kV]
100 s 0.3 6 2.0
30 min 0.3 6 1.5
30 min 0.3 6 2.0
120 min 0.3 6 2.0
150 s 0.3 6 1.5
300 s 0.3 6 1.0
30 min 0.3 6 1.0
precision optical characterization technique [4–7]. Multiple angle of incidence (MAI) spectroscopic ellipsometry measurements were performed on the samples deposited by rf sputtering. Multilayer optical models (surface roughness layer, niobium oxide layer, a native silicon oxide layer and ion implantation-amorphized silicon (i-a-Si) layer on a c-Si substrate) were created in order to evaluate the spectra. The dielectric function of the niobium oxide was described by the Tauc–Lorentz dispersion relation or a Lorentz oscillator. High depth resolution RBS measurements combined with channeling [8] were performed to investigate the interface disorder between the thin deposited niobium oxide layer and the c-Si substrate. Atomic resolution cross-sectional transmission electron microscopy (HRTEM) investigation was applied to visualize the details of the damaged subsurface region of the c-Si substrate. Calculations were performed using SRIM code (Stopping and Range of Ions in Matter) in order to interpret the details of the damage formation [9]. 2. Experimental Niobium oxide layers were deposited by rf sputtering on c-Si substrates at room temperature using a Leybold Z400 apparatus. Prior to the rf sputtering deposition the native oxide was removed from the silicon substrates using diluted HF. The deposition was performed using niobium oxide powder target (Kurt J. Lesker Co.) in an Ar–O2 atmosphere. During the sample preparation the pressure was adjusted to 0.3 Pa, the oxygen flow was chosen to be 6 ml/min, the DC sputtering voltage (wall potential) was selected from the range of 1.0–2.0 kV, and the duration of the sputtering was between 100 s and 120 min (the details are listed in Table 1). For the ellipsometric characterization of the rf-deposited samples M-2000DI (Institute of Technical Physics and Materials Science, Centre for Energy Research, Budapest, Hungary) and M2000F (Department of Optics and Quantum Electronics, University of Szeged, Hungary) rotating compensator spectroscopic ellipsometers manufactured by the J.A. Woollam Co., Inc. were used [10]. Multilayer optical models with the Tauc-Lorentz dispersion relation [11,12] or with a Lorentz oscillator and reference data from the literature [10,13] was applied for evaluation of ellipsometric spectra using the WVASE32 software [10]. RBS analysis was performed using the Hungarian Ion-beam Physics Platform 5 MV Van de Graaff accelerator in Budapest, at the Institute for Particle and Nuclear Physics, Wigner Research Centre for Physics of the Hungarian Academy of Sciences. The ion beam of 1000-keV 4 He+ was collimated with 2 sets of four-sector slits to the necessary dimensions of 0.5 mm × 0.5 mm. An ion current of typically 20 nA was measured by a transmission Faraday cap [14]. The dose of the measurements was 20 C. The spectra taken on the samples were simulated with the same layer structure using the RBX program [15,16]. The microstructure of the films was investigated by HRTEM using a JEOL JM3010 transmission electron microscope operated at an acceleration voltage of 300 kV with an 0.17 nm point resolution. HRTEM images were taken in cross-sectional view.
Cross-sectional samples were prepared by creating a “sandwich” from two 1.8 mm × 0.5 mm pieces of the sample (film to film). This “sandwich” was mounted and glued (with an araldite-based glue) into the window of a Ti grid followed by mechanical thinning, polishing, and dimpling to a thickness of ca. 20 m in the middle. Thinning to electron transparency was achieved by ion-beam milling [17] using a Technoorg Linda ion mill with 10-keV Ar+ ions at an incidence angle of 5◦ with respect to the surface. In the final period of the milling process, the ion energy was decreased gradually to 0.3 keV (using a Technoorg-Linda Gentle Mill) to minimize ion-induced structural changes in the surface layers. Specimens prepared by small angle cleavage technique (SACT) [18] were investigated as well, in this case the possible disturbing effect of the ion-beam milling was avoided. The image taken on the cleaved piece was very similar to the one prepared by ion-beam milling.
3. Results and discussion First, we evaluated SE spectra measured on the Nb-oxide sample deposited with 120 min duration and 2.0 kV wall potential. Various optical models (one-layer model, two-layer model, three-layer model and four-layer model) were created. We consider the Mean Square Error (MSE) together with the measured and generated ellipsometric spectra in order to determine the quality of a model. Usually we create models with increasing complexity (with increasing number of sublayers) based on the information achieved from independent experiments for thin layers as high resolution RBS and cross sectional electron microscopy. Table 2 shows the optical models together with the results yielded by the evaluations. The second, third, fourth and fifth columns represent the sublayers of the optical models. The c-Si substrate was not indicated in Table 2. The complex dielectric function of the Nb-oxide layer was modeled by the Tauc-Lorentz dispersion relation [11,12]. The optical model “A” contains only one layer, the Nb-oxide layer. The value of the Mean Square Error (MSE = 45.81) does not indicate a good correspondence between the measured and the generated spectra. After introducing a second layer, a surface roughness layer (model “B”), the evaluation yielded a considerably lower MSE value (36.87). In the first three-layer model (model “C”) a silicon dioxide layer was inserted between the Nb-oxide layer and the c-Si substrate. A part of this silicon dioxide layer may originate from the native oxide formation during the time interval of placing the substrate into the processing chamber of the Leybold Z400 apparatus. Another part of this silicon dioxide layer may originate from a possible plasma oxidation prior to the start of the Nb-oxide layer formation. The evaluation performed using this three-layer model yielded MSE = 32.23. The second three-layer model (model “D”) contains an amorphous silicon layer between the Nb-oxide layer and the c-Si substrate. For the description of the complex dielectric function of this amorphous silicon we selected the complex dielectric function of i-a-Si [13] because the damage was introduced by energetic ions and/or atoms. The evaluation has been performed in the wavelength range from 282 nm to 953 nm or from 1.3 eV to 4.4 eV because the complex dielectric function of ion-implantation amorphized silicon is available only in this range. It is important to note that the complex dielectric function of i-a-Si differs from that of amorphous silicon deposited by vacuum evaporation or by lowpressure chemical vapor deposition [13]. The evaluation performed with the second three-layer model (model “D”) gave a remarkable reduction of MSE, a value of 25.3 was obtained. The four-layer model “E” contains a silicon dioxide
Please cite this article in press as: T. Lohner, et al., Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide, Appl. Surf. Sci. (2016), http://dx.doi.org/10.1016/j.apsusc.2016.11.232
G Model APSUSC-34537; No. of Pages 7
ARTICLE IN PRESS T. Lohner et al. / Applied Surface Science xxx (2016) xxx–xxx
3
Table 2 The structure of the optical models together with the results and the Mean Square Error (MSE) values yielded by the evaluation of spectra taken on Nb-oxide sample deposited by rf sputtering with 120 min duration. Optical model
Surface roughness layer [nm]
Nb-oxide layer (Tauc-Lorentz) [nm]
SiO2 layer [nm]
Ion-implantation-amorphized silicon (ia-Si) layer [nm]
MSE
A One-layer model B Two-layer model CThree-layer model DThree-layer model E Four-layer model
– 2.11 ± 0.07 2.56 ± 0.07 1.98 ± 0.05 2.25 ± 0.04
369.7 ± 0.3 368.7 ± 0.3 362.4 ± 0.3 362.3 ± 0.2 359.3 ± 0.2
– – 4.8 ± 0.2 – 2.8 ± 0.1
– – – 0.86 ± 0.02 0.80 ± 0.02
45.81 36.87 32.23 25.3 21.26
Fig. 3. Data-model differences in psi values for Nb-oxide sample deposited for 120 min.
Fig. 1. Measured and generated values at different angles of incidence (shown in the legend) in function of photon energy for the sample deposited by rf sputtering with 120 min duration and at a wall potential of 2.0 kV. The evaluation has been performed with the four-layer model (model “E”).
Fig. 4. Data-model differences in delta values for Nb-oxide sample deposited for 120 min.
Fig. 2. Measured and generated values at different angles of incidence (shown in the legend) in function of photon energy for the sample deposited by rf sputtering with 120 min duration and at a wall potential of 2.0 kV. The evaluation has been performed with the four-layer model (model “E”).
layer between the Nb-oxide layer and the i-a-Si layer, for the MSE a value of 21.26 was obtained. Figs. 1 and 2 show the measured and generated SE spectra of sample deposited with 120 min duration, the agreement between the measured and generated SE spectra is good. The data-model differences in psi and delta values for Nb-oxide sample deposited for 120 min are shown in Figs. 3 and 4. Possible reasons of the deviation between data and model: probably graded structure of the Nb-oxide layer, the Tauc-Lorentz dispersion relation is not the optimal choice for modeling the optical properties of the sputter deposited Nb-oxide, the mixture of ion-implantation amorphized silicon and single crystalline silicon does not describe accurately the damaged region of the single crystalline silicon substrate. The second sample considered here was deposited using a wall potential of 2.0 kV and a duration of 30 min for deposition. Table 3 shows the optical models together with the results yielded by the
evaluation. The second, third, fourth and fifth columns represent the sublayers of the optical models. The c-Si substrate was not indicated in Table 3. The introduction of new sublayers into the optical model resulted in decreasing MSE values, especially the introduction of the i-a-Si sublayer (MSE = 28.17 and 27.99 for models “D” and “E”, respectively). In the second four-layer model (model “F”) for the damaged silicon layer a mixture of c-Si and i-a-Si was chosen. The optical properties of the mixed layer were calculated by the effective medium approximation. This approach was used by Fried et al., they described the depth distribution of damage by a simple box-type model using only two parameters: the effective thickness and effective damage of the implanted layer [19,20]. For the volume fraction of the i-a-Si component a value of 29.7 ± 0.5% was yielded by the evaluation. In this case ten unknown parameters (layer thicknesses, volume fraction of i-a-Si, and the parameters of the Tauc-Lorentz dispersion relation) were defined and fitted. Considering the correlation matrix after the evaluation we observed that the correlation between the thickness of the partially damaged silicon layer (it is a mixture of c-Si and i-a-Si) and the volume fraction of the i-a-Si component is equal to 0.936. We considered this value too high, that is why we repeated the evaluation of the spectra selecting only one free parameter (the
Please cite this article in press as: T. Lohner, et al., Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide, Appl. Surf. Sci. (2016), http://dx.doi.org/10.1016/j.apsusc.2016.11.232
G Model APSUSC-34537; No. of Pages 7
ARTICLE IN PRESS T. Lohner et al. / Applied Surface Science xxx (2016) xxx–xxx
4
Table 3 The structure of the optical models together with the results and the Mean Square Error (MSE) values yielded by the evaluation of spectra taken on Nb-oxide sample deposited by rf sputtering with 30 min duration and with 2.0 kV wall potential. Optical model
Surface roughness [nm]
Nb-oxide layer (Tauc-Lorentz) [nm]
SiO2 layer [nm]
Ion-implantation−amorphized silicon (i-a-Si) layer [nm]
MSE
A One-layer model B Two-layer model CThree-layer model DThree-layer model E Four-layer model
– 1.06 ± 0.05 1.12 ± 0.06 0.76 ± 0.02 0.80 ± 0.02
184.9 ± 0.1 184.0 ± 0.1 183.4 ± 0.2 178.58 ± 0.05 178.18 ± 0.08
– – 0.6 ± 0.2 – 0.39 ± 0.06
78.36 74.91 74.89 28.17 27.99
F Four-layer model
0.82 ± 0.02
177.58 ± 0.08
0.92 ± 0.05
– – – 1.65 ± 0.01 1.65 ± 0.01 EMA-layer 6.1 ± 0.1 nm (70.3 ± 0.5% c-Si + 29.7 ± 0.5% ia-Si)
26.27
Table 4 Percentage of the i-a-Si component in the mixed layer (volume fraction of i-a-Si, f), and the thickness values of the mixed layer, d, yielded by the evaluation, and the “Amount of damage”, f•d, together with MSE values for Nb-oxide sample deposited by rf sputtering with 30 min duration and with a wall potential of 2.0 kV. Volume fraction of i-a-Si, f
Thickness of mixed layer, d [nm]
“Amount of damage”, f·d [nm]
MSE
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
11.2 ± 0.2 8.18 ± 0.06 6.00 ± 0.04 4.52 ± 0.03 3.55 ± 0.02 2.89 ± 0.02 2.43 ± 0.02 2.13 ± 0.01 1.86 ± 0.01 1.65 ± 0.01
1.12 1.64 1.80 1.81 1.78 1.73 1.70 1.70 1.67 1.65
43.05 29.02 26.33 26.92 27.60 28.09 28.20 28.09 28.05 27.99
Fig. 6. Measured and generated values at different angles of incidence (shown in the legend) in function of photon energy for the sample deposited by rf sputtering with 30 min duration and at a wall potential of 2.0 kV.
Fig. 7. Data-model differences in psi values for Nb-oxide sample deposited for 30 min. Fig. 5. Measured and generated values at different angles of incidence (shown in the legend) in function of photon energy for the sample deposited by rf sputtering with 30 min duration and at a wall potential of 2.0 kV.
layer thickness) for the mixed layer containing c-Si and i-a-Si. The volume fraction of the i-a-Si was set to 10%, 20%, 30% and so on, the evaluations were performed for all these volume fraction values one-by-one. The results are shown in Table 4. The lowest MSE value (26.33) was obtained for the assumption of a mixture of 30 volume percent i-a-Si and 70 volume percent c-Si components in the mixed layer modeling the damage caused by rf sputtering. In the present case we may consider the 30 volume percent i-a-Si as effective damage and the 6 nm as effective thickness of the disordered silicon layer after [19,20]. Figs. 5 and 6 show the measured and generated SE spectra of sample deposited with 30 min duration and 2.0 kV wall potential, concerning the evaluation the details described earlier for the mixed layer (30 volume percent i-a-Si and 70 volume percent c-Si components) were applied. The correspondence between the measured and generated SE spectra is good. The data-model differences in psi and delta values for Nb-oxide sample deposited for 120 min are shown in
Fig. 8. Data-model differences in delta values for Nb-oxide sample deposited for 30 min.
Figs. 7 and 8. The complex dielectric function of the Nb-oxide layer of evaluation detailed above for sample deposited at 2.0 kV with 30 min duration has been saved as reference data. The SE spectra of samples deposited at 1.0 and 1.5 kV with 30 min duration were evaluated using a four-layer optical model, the above speci-
Please cite this article in press as: T. Lohner, et al., Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide, Appl. Surf. Sci. (2016), http://dx.doi.org/10.1016/j.apsusc.2016.11.232
G Model APSUSC-34537; No. of Pages 7
ARTICLE IN PRESS T. Lohner et al. / Applied Surface Science xxx (2016) xxx–xxx
5
Table 5 Thickness values of the surface roughness layer, the Nb-oxide layer, the SiO2 layer and thickness of a mixed layer consists of 70 volume fraction of c-Si and 30 volume fraction of i-a-Si yielded by the evaluation of SE spectra of Nb-oxide samples deposited using various wall potentials and with 30 min rf sputtering deposition. Wall potential [kV]
Duration of rf sputtering deposition [min]
Surface roughness layer [nm]
Nb-oxide layer [nm]
SiO2 layer [nm]
Thickness of the mixed layer (70% c-Si + 30% i-a-Si) [nm]
MSE
1.0 1.5 2.0
30 30 30
1.7 ± 0.1 1.2 ± 0.1 0.8 ± 0.1
27.3 ± 0.1 79.7 ± 0.1 177.6 ± 0.1
3.9 ± 0.1 3.1 ± 0.1 0.9 ± 0.1
1.0 ± 0.1 3.9 ± 0.1 6.00 ± 0.1
22.15 22.20 26.36
Nb-oxide layer 2.88 nm Silicon dioxide layer 2.80 nm EMA layer (crystalline Si + 26% ion-impl. amorphized Si) 4.75 nm Single crystalline silicon substrate
Fig. 11. The three-layer optical model constructed on basis of the results of the RBS and HRTEM measurement. Fig. 9. Cross sectional HRTEM image of a Nb-oxide sample rf deposited for 100 s (2.0-kV wall potential). The thicknesses of the amorphous layers are indicated in the image. The observed thicknesses were used to the construction of the optical model for the evaluations of ellipsometry measurements.
Fig. 12. The refractive index n in function of photon energy for Nb-oxide layers deposited for 120 min, 30 min and 100 s.
Fig. 10. Measured and simulated RBS spectra taken on Nb oxide sample rf deposited for 100 s (2.0 kV wall potential).
fied optical reference data of Nb-oxide layer was built in into this optical model. The thicknesses of the four layers (surface roughness layer, Nb-oxide layer, SiO2 layer and the mixed layer with 30 volume percent i-a-Si and 70 volume percent c-Si components) were considered free parameters. The results of the evaluation are displayed in Table 5. The thickness of the mixed layer increases with increasing wall potential. A third part of our investigation involved studying of rf sputtering-deposited samples with 100, 150 and 300 s deposition durations. Fig. 9 displays the cross sectional HRTEM image and gives the layer structure for sample deposited with 100 s duration and 2.0 kV wall potential. The thickness of the rf sputtering deposited amorphous Nb-oxide film is 2.9 nm. A layer of 4.5 nm thickness sandwiched between the amorphous Nb-oxide film and the c-Si substrate can be interpreted by an amorphous SiOx and an i-a-Si component. A 0.6 nm thick surface roughness was assessed, too. Fig. 10 shows the measured and simulated RBS spectra, the thickness value of the Nb-oxide film (2.88 nm) is practically the
same as determined by HRTEM. Beneath the 2.8 nm thick SiO2 layer a partially damaged silicon layer was identified with thickness of 4.75 nm. Using the compositional and structural data obtained from RBS and HRTEM studies an optical model was created for the Nb-oxide sample deposited for 100 s and 2.0 kV wall potential (Fig. 11). The complex refractive index of the 2.88 nm thick Nb-oxide film was evaluated using a Lorentz oscillator [10]. The complex refractive index of the 2.88 nm thick Nb-oxide film was built into a four-layer optical model and the thickness values of the Nb-oxide film, the SiO2 layer and the thickness of a mixed layer consists of 74 volume fraction of c-Si and 26 volume fraction of i-a-Si were defined as free parameters. The results of the evaluations are compiled in Table 6. The thickness of the mixed layer increases with increase of wall potential. The refractive index n and the extinction coefficient k of the Nboxide layer displayed in function of photon energy for Nb-oxide samples deposited for 120 min, 30 min and 100 s in Figs. 12 and 13, respectively. The values of the refractive index and the extinction coefficient in function of photon energy for samples deposited with duration of 30 m and 120 m are similar to values published by Hála et al. [21]. The values of the refractive index and the extinction coefficient in function of photon energy for samples deposited with duration of 100 s show strange shape, certainly the initial growth stage and nucleation may play a role.
Please cite this article in press as: T. Lohner, et al., Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide, Appl. Surf. Sci. (2016), http://dx.doi.org/10.1016/j.apsusc.2016.11.232
G Model
ARTICLE IN PRESS
APSUSC-34537; No. of Pages 7
T. Lohner et al. / Applied Surface Science xxx (2016) xxx–xxx
6
Table 6 Thickness values of the Nb-oxide layer, the SiO2 layer and thickness of a mixed layer consists of 74 volume fraction of c-Si and 26 volume fraction of i-a-Si yielded by the evaluation of SE spectra of Nb-oxide samples deposited using various wall potentials. Wall potential [kV]
Duration of rf sputtering deposition [sec]
Surface roughness layer [nm]
Nb-oxide layer [nm]
SiO2 layer [nm]
Thickness of the mixed layer (74% c-Si + 26% i-a-Si) [nm]
MSE
1.0 1.5 2.0
300 150 100
0.6 (fixed) 0.6 (fixed) 0.6 (fixed)
2.4 ± 0.1 2.2 ± 0.1 2.7 ± 0.1
3.0 ± 0.1 3.4 ± 0.1 3.2 ± 0.1
2.5 ± 0.1 3.3 ± 0.1 4.5 ± 0.1
16.75 17.3 19.07
thickest sample (360 nm in 120 min) the growth rate (3 nm/min) is significantly lower than for samples with deposition duration of 10 min and 30 min. The reason may originate that the target (sintered ceramic body) is not 100% dense, it contains pores. The sputtering by Ar ions causes erosion of the surface and the change of the surface morphology explores the pores. A significant increase of the effective area of target can increase the sputtering rate. The change of the morphology in the nanometer scale is accompanied by graying of the target (low-reflectance surface nanostructures created with Ar ion bombardment). The thickest sample (360 nm in 120 min) has been deposited in 2007, the other samples have been deposited in the first half of 2016.
4. Conclusions
Fig. 13. The extinction coefficient k in function of photon energy for Nb-oxide layers deposited for 120 min, 30 min and 100 s.
Table 7 The values of projected range and straggling for bombarding silicon substrate with oxygen and argon ions based on SRIM-2013 calculations [9]. Longitudinal Straggling [nm]
Lateral Straggling [nm]
Oxygen ions into silicon substrate 1.0 4.3 5.8 1.5 2.0 7.1
3.2 4.1 4.9
2.3 3.0 3.6
Argon ions into silicon substrate 3.4 1.0 4.2 1.5 2.0 5.0
1.9 2.3 2.6
1.3 1.7 1.9
Ion energy [keV]
Projected Range [nm]
The formation of partially damaged layer on the surface of the silicon substrate during rf sputtering deposition can be explained by the bombardment of the surface of the single crystalline silicon substrate by energetic oxygen and argon ions and neutral atoms. The neutral atoms can be formed due to charge exchange mechanism. The energetic oxygen and argon ions and neutral atoms may penetrate into the c-Si substrate to a depth of several nm as one can see in Table 7 and they may remove silicon atoms from the crystalline lattice producing interstitial atoms and vacancies. In our interpretation the damaging effect originating from energetic particles bombarding the single crystalline silicon substrate will start at the beginning of the rf sputtering process and as the rf sputtering process proceeds, the thin damaged subsurface layer of the single crystalline silicon substrate will be buried by the newly deposited Nb-oxide material. The low growth rate of the thinnest sample (only 2.9 nm in 100 s) certainly a consequence of the initial stage of Nb-oxide layer growth including nucleation. Another sample (not included in the manuscript) deposited for 10 min exhibits a Nb oxide layer of 55.3 nm thickness. Its growth rate value (5.5 nm/min) is similar to the one obtained for the 30 min deposition (5.9 nm/min). For the
The experiments show that it is possible to characterize substrate surface damage induced by radiofrequency sputtering deposition of niobium oxide on c-Si substrate using spectroscopic ellipsometry. For the modeling of the partially damaged silicon layer a mixed layer with i-a-Si and c-Si components proved to be appropriate. For the case of a few nm thick rf sputtering deposited Nb-oxide layer independent methods such as cross HRTEM or high resolution RBS can deliver additional thickness values of the sublayers in the interface region. Sputtering at higher wall potential results in higher thickness of the mixed layer with i-a-Si and c-Si components in the damaged subsurface region. During the rf sputtering the energetic oxygen and argon ions and neutral atoms may penetrate into the c-Si substrate to a depth of several nm causing displacement of host silicon atoms and consequently producing damaged regions.
Acknowledgements Support from the National Development Agency grant OTKA K115852, and projects H2020-SMEINST-2-2014-683541 “OptiLight” FP7 “E450EDL” and “SEA4KET” and TÉT 12 DE-1-2013-0002 is greatly acknowledged.
References [1] S. Berg, L.P. Andersson, H. Norström, E. Grusell, Substrate surface damages by rf-sputtering, Vacuum 27 (1977) 189–191. [2] Stephen J. Fonash, S. Ashok, Ranbir Singh, Effect of ion-beam sputter damage on Schottky barrier formation in silicon, Appl. Phys. Lett. 39 (1981) 423–425. [3] F.D. Auret, O. Paz, N.A. Bojarczuk, Characterization of process-induced defects and device properties of ion beam sputter-deposited Mo contacts on Si, J. Appl. Phys. 55 (1984) 1581–1589. [4] H. Fujiwara, Spectroscopic Ellipsometry: Principles and Applications, John Wiley and Sons, Chichester, England, 2007. [5] J. Budai, B. Farkas, Z.L. Horváth, Zs. Geretovszky, On determining the optical properties and layer structure from spectroscopic ellipsometric data using automated artifact minimization method, Thin Solid Films 567 (2014) 14–19. [6] Z. Tóth, I. Hanyecz, A. Gárdián, J. Budai, J. Csontos, Z. Pápa, M. Füle, Ellipsometric analysis of silicon surfaces textured by ns and sub-ps KrF laser pulses, Thin Solid Films 571 (2014) 631–636. [7] E.R. Shaaban, M. El-Hagary, M. Emam-Ismail, A.M. Abd Elnaeim, S.H. Moustafa, A. Adel, Optical characterization of polycrystalline ZnSe1-xTex thin films using variable angle spectroscopic ellipsometry and spectrophotmetery techniques, Mater. Sci. Semicond. Process. 39 (2015) 735–741.
Please cite this article in press as: T. Lohner, et al., Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide, Appl. Surf. Sci. (2016), http://dx.doi.org/10.1016/j.apsusc.2016.11.232
G Model APSUSC-34537; No. of Pages 7
ARTICLE IN PRESS T. Lohner et al. / Applied Surface Science xxx (2016) xxx–xxx
[8] G. Mezey, E. Kótai, T. Lohner, T. Nagy, J. Gyulai, A. Manuaba, Improved depth resolution of channeling measurements in Rutherford backscattering by a detector tilt, Nucl. Instrum. Methods 149 (1978) 235–237. [9] J.F. Ziegler, M.D. Ziegler, J.P. Biersack, SRIM – the stopping and range of ions in matter (2010), Nucl. Instrum. Methods Phys. Res. Sect. B 268 (2010) 1818–1823. [10] J. A. Woollam Co., Inc. (www.jawoollam.com) (2016). [11] G.E. Jellison Jr., F.A. Modine, Parameterization of the optical functions of amorphous materials in the interband region, Appl. Phys. Lett. 69 (1996) 371–373. [12] G.E. Jellison Jr., F.A. Modine, Parameterization of the optical functions of amorphous materials in the interband region (erratum), Appl. Phys. Lett. 69 (1996) 2137. [13] M. Fried, T. Lohner, W.A.M. Aarnink, L.J. Hanekamp, A. van Silfhout, Determination of complex dielectric functions of ion implanted and implanted-annealed amorphous silicon by spectroscopic ellipsometry, J. Appl. Phys. 71 (1992) 5260–5262. [14] F. Pászti, A. Manuaba, C. Hajdu, A.A. Melo, M.F. da Silva, Current measurement on MeV energy ion beams, Nucl. Instrum. Methods B 47 (1990) 187–192. [15] E. Kótai, Computer methods for analysis and simulation of RBS and ERDA spectra, Nucl. Instrum. Methods B 85 (1994) 588–596.
[16] E. Kótai, Computer methods for analysis and simulation of RBS and ERDA spectra, Proc. of Fourteenth Int. Conf. on the Application of Accelerators in Research and Industry, Nov 6–9, 1996, Denton, Texas USA, CP392, edited by J.L. Duggan, I.L. Morgan, AIP Press, New York 1997, p. 631–634. [17] Á. Barna, Topographic kinetics and practice of low angle ion beam thinning, Mater. Res. Soc. Symp. Proc. 254 (1992) 3–22. [18] J.P. McCaffrey, Small-angle cleavage of semiconductors for transmission electron microscopy, Ultramicroscopy 38 (1991) 149–157. [19] M. Fried, G. Juhász, C. Major, P. Petrik, G. Battistig, Homogeneity check of ion implantation in silicon by wide-angle ellipsometry, 17th IEEE International Conference on Advanced Thermal Processing on Semiconductors—RTP 2009, New York, IEEE, 2009. Paper 5373448. 4 p. [20] C. Major, G. Juhasz, Z. Herczeg, Z. Labadi, P. Petrik, Z. Horvath, M. Fried, High speed spectroscopic ellipsometry technique for in-line monitoring in large area thin layer production, Collection entitled ICSE-7 Ms. Ref. No.: APSUSC-D-16-06772. [21] M. Hála, J. Capek, O. Zabeida, J.E. Klemberg-Sapieha, L. Martinu, Hysteresis-free deposition of niobium oxide films by HiPIMS using different pulse management strategies, J. Phys. D Appl. Phys. 45 (2012) 055204.
Please cite this article in press as: T. Lohner, et al., Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide, Appl. Surf. Sci. (2016), http://dx.doi.org/10.1016/j.apsusc.2016.11.232
7
ARTICLE IN PRESS
APSUSC-34537; No. of Pages 7
Applied Surface Science xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Applied Surface Science journal homepage: www.elsevier.com/locate/apsusc
Full Length Article
Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide Tivadar Lohner a,∗ , Miklós Serényi a , Edit Szilágyi b , Zsolt Zolnai a , Zsolt Czigány a , Nguyen Quoc Khánh a , Péter Petrik a,c , Miklós Fried a,c a Institute of Technical Physics and Materials Science, Centre for Energy Research, Hungarian Academy of Sciences, H-1121 Budapest, Konkoly Thege Miklós út 29-33, Hungary b Institute for Particle and Nuclear Physics, Wigner Research Centre for Physics, Hungarian Academy of Sciences, H-1121 Budapest, Konkoly Thege Miklós út 29-33, Hungary c Doctoral School of Molecular and Nanotechnologies, Faculty of Information Technology, University of Pannonia, H-8200 Veszprém, Egyetem u. 10, Hungary
a r t i c l e
i n f o
Article history: Received 29 July 2016 Received in revised form 21 October 2016 Accepted 29 November 2016 Available online xxx Keywords: Sputtering Damage Ellipsometry Rutherford backscattering Electron microscopy
a b s t r a c t Substrate surface damage induced by deposition of metal atoms by radiofrequency (rf) sputtering or ion beam sputtering onto single-crystalline silicon (c-Si) surface has been characterized earlier by electrical measurements. The question arises whether it is possible to characterize surface damage using spectroscopic ellipsometry (SE). In our experiments niobium oxide layers were deposited by rf sputtering on c-Si substrates in gas mixture of oxygen and argon. Multiple angle of incidence spectroscopic ellipsometry measurements were performed, a four-layer optical model (surface roughness layer, niobium oxide layer, native silicon oxide layer and ion implantation-amorphized silicon [i-a-Si] layer on a c-Si substrate) was created in order to evaluate the spectra. The evaluations yielded thicknesses of several nm for the i-a-Si layer. Better agreement could be achieved between the measured and the generated spectra by inserting a mixed layer (with components of c-Si and i-a-Si applying the effective medium approximation) between the silicon oxide layer and the c-Si substrate. High depth resolution Rutherford backscattering (RBS) measurements were performed to investigate the interface disorder between the deposited niobium oxide layer and the c-Si substrate. Atomic resolution cross-sectional transmission electron microscopy investigation was applied to visualize the details of the damaged subsurface region of the substrate. © 2016 Elsevier B.V. All rights reserved.
1. Introduction The technique of sputtering involves ejecting material from the target to the substrate such as a silicon wafer. Resputtering is reemission of the deposited material during the deposition process by ion or atom bombardment. Sputtered atoms ejected from the target have a wide energy distribution, typically up to tens of eV. The sputtered ions and atoms can reach the substrate condensing after undergoing a random walk. High-energy neutrals sputtered from the target will still have enough energy reaching the substrate even after some collisions in the working gas to damage the surface of the crystalline substrate. The radio frequency (rf) voltage may be applied directly to the cathode for metal sputtering or via capacitor to sputter insulators.
∗ Corresponding author. E-mail address: [email protected] (T. Lohner).
The high frequency alternating potential may be used to neutralize the insulator surface periodically with plasma electrons, which must be switched in a period that is short compared with the time required by the positive ions to travel to the surface of target. Thus, both the amplitude and frequency of rf should be high to yield high rates. The insulator target may consist of different material: oxides, nitrides etc., from that not only atoms but molecules with considerable kinetic energy can be deposited. This is the reason why surface damage of the substrate can be significant as a result of rf sputtering deposition. Substrate surface damage induced by deposition of metal atoms by rf sputtering or ion beam sputtering onto the single-crystalline silicon (c-Si) surface has been characterized earlier by I-V, C-V and deep-level transient spectroscopy measurements [1–3]. In our experiments niobium oxide layers were deposited by rf sputtering on c-Si substrates at room temperature in gas mixture of oxygen and argon. Optical properties of thin film structures can be derived from SE measurement, which is known to be a high-
http://dx.doi.org/10.1016/j.apsusc.2016.11.232 0169-4332/© 2016 Elsevier B.V. All rights reserved.
Please cite this article in press as: T. Lohner, et al., Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide, Appl. Surf. Sci. (2016), http://dx.doi.org/10.1016/j.apsusc.2016.11.232
G Model
ARTICLE IN PRESS
APSUSC-34537; No. of Pages 7
T. Lohner et al. / Applied Surface Science xxx (2016) xxx–xxx
2
Table 1 Duration, pressure, oxygen flow, DC voltage (wall potential) of rf sputtering for different samples. Sample
A495 A496 A497 A505
A504
A503
A247
Duration Pressure [Pa] Oxygen flow[ml/min] DC voltage [kV]
100 s 0.3 6 2.0
30 min 0.3 6 1.5
30 min 0.3 6 2.0
120 min 0.3 6 2.0
150 s 0.3 6 1.5
300 s 0.3 6 1.0
30 min 0.3 6 1.0
precision optical characterization technique [4–7]. Multiple angle of incidence (MAI) spectroscopic ellipsometry measurements were performed on the samples deposited by rf sputtering. Multilayer optical models (surface roughness layer, niobium oxide layer, a native silicon oxide layer and ion implantation-amorphized silicon (i-a-Si) layer on a c-Si substrate) were created in order to evaluate the spectra. The dielectric function of the niobium oxide was described by the Tauc–Lorentz dispersion relation or a Lorentz oscillator. High depth resolution RBS measurements combined with channeling [8] were performed to investigate the interface disorder between the thin deposited niobium oxide layer and the c-Si substrate. Atomic resolution cross-sectional transmission electron microscopy (HRTEM) investigation was applied to visualize the details of the damaged subsurface region of the c-Si substrate. Calculations were performed using SRIM code (Stopping and Range of Ions in Matter) in order to interpret the details of the damage formation [9]. 2. Experimental Niobium oxide layers were deposited by rf sputtering on c-Si substrates at room temperature using a Leybold Z400 apparatus. Prior to the rf sputtering deposition the native oxide was removed from the silicon substrates using diluted HF. The deposition was performed using niobium oxide powder target (Kurt J. Lesker Co.) in an Ar–O2 atmosphere. During the sample preparation the pressure was adjusted to 0.3 Pa, the oxygen flow was chosen to be 6 ml/min, the DC sputtering voltage (wall potential) was selected from the range of 1.0–2.0 kV, and the duration of the sputtering was between 100 s and 120 min (the details are listed in Table 1). For the ellipsometric characterization of the rf-deposited samples M-2000DI (Institute of Technical Physics and Materials Science, Centre for Energy Research, Budapest, Hungary) and M2000F (Department of Optics and Quantum Electronics, University of Szeged, Hungary) rotating compensator spectroscopic ellipsometers manufactured by the J.A. Woollam Co., Inc. were used [10]. Multilayer optical models with the Tauc-Lorentz dispersion relation [11,12] or with a Lorentz oscillator and reference data from the literature [10,13] was applied for evaluation of ellipsometric spectra using the WVASE32 software [10]. RBS analysis was performed using the Hungarian Ion-beam Physics Platform 5 MV Van de Graaff accelerator in Budapest, at the Institute for Particle and Nuclear Physics, Wigner Research Centre for Physics of the Hungarian Academy of Sciences. The ion beam of 1000-keV 4 He+ was collimated with 2 sets of four-sector slits to the necessary dimensions of 0.5 mm × 0.5 mm. An ion current of typically 20 nA was measured by a transmission Faraday cap [14]. The dose of the measurements was 20 C. The spectra taken on the samples were simulated with the same layer structure using the RBX program [15,16]. The microstructure of the films was investigated by HRTEM using a JEOL JM3010 transmission electron microscope operated at an acceleration voltage of 300 kV with an 0.17 nm point resolution. HRTEM images were taken in cross-sectional view.
Cross-sectional samples were prepared by creating a “sandwich” from two 1.8 mm × 0.5 mm pieces of the sample (film to film). This “sandwich” was mounted and glued (with an araldite-based glue) into the window of a Ti grid followed by mechanical thinning, polishing, and dimpling to a thickness of ca. 20 m in the middle. Thinning to electron transparency was achieved by ion-beam milling [17] using a Technoorg Linda ion mill with 10-keV Ar+ ions at an incidence angle of 5◦ with respect to the surface. In the final period of the milling process, the ion energy was decreased gradually to 0.3 keV (using a Technoorg-Linda Gentle Mill) to minimize ion-induced structural changes in the surface layers. Specimens prepared by small angle cleavage technique (SACT) [18] were investigated as well, in this case the possible disturbing effect of the ion-beam milling was avoided. The image taken on the cleaved piece was very similar to the one prepared by ion-beam milling.
3. Results and discussion First, we evaluated SE spectra measured on the Nb-oxide sample deposited with 120 min duration and 2.0 kV wall potential. Various optical models (one-layer model, two-layer model, three-layer model and four-layer model) were created. We consider the Mean Square Error (MSE) together with the measured and generated ellipsometric spectra in order to determine the quality of a model. Usually we create models with increasing complexity (with increasing number of sublayers) based on the information achieved from independent experiments for thin layers as high resolution RBS and cross sectional electron microscopy. Table 2 shows the optical models together with the results yielded by the evaluations. The second, third, fourth and fifth columns represent the sublayers of the optical models. The c-Si substrate was not indicated in Table 2. The complex dielectric function of the Nb-oxide layer was modeled by the Tauc-Lorentz dispersion relation [11,12]. The optical model “A” contains only one layer, the Nb-oxide layer. The value of the Mean Square Error (MSE = 45.81) does not indicate a good correspondence between the measured and the generated spectra. After introducing a second layer, a surface roughness layer (model “B”), the evaluation yielded a considerably lower MSE value (36.87). In the first three-layer model (model “C”) a silicon dioxide layer was inserted between the Nb-oxide layer and the c-Si substrate. A part of this silicon dioxide layer may originate from the native oxide formation during the time interval of placing the substrate into the processing chamber of the Leybold Z400 apparatus. Another part of this silicon dioxide layer may originate from a possible plasma oxidation prior to the start of the Nb-oxide layer formation. The evaluation performed using this three-layer model yielded MSE = 32.23. The second three-layer model (model “D”) contains an amorphous silicon layer between the Nb-oxide layer and the c-Si substrate. For the description of the complex dielectric function of this amorphous silicon we selected the complex dielectric function of i-a-Si [13] because the damage was introduced by energetic ions and/or atoms. The evaluation has been performed in the wavelength range from 282 nm to 953 nm or from 1.3 eV to 4.4 eV because the complex dielectric function of ion-implantation amorphized silicon is available only in this range. It is important to note that the complex dielectric function of i-a-Si differs from that of amorphous silicon deposited by vacuum evaporation or by lowpressure chemical vapor deposition [13]. The evaluation performed with the second three-layer model (model “D”) gave a remarkable reduction of MSE, a value of 25.3 was obtained. The four-layer model “E” contains a silicon dioxide
Please cite this article in press as: T. Lohner, et al., Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide, Appl. Surf. Sci. (2016), http://dx.doi.org/10.1016/j.apsusc.2016.11.232
G Model APSUSC-34537; No. of Pages 7
ARTICLE IN PRESS T. Lohner et al. / Applied Surface Science xxx (2016) xxx–xxx
3
Table 2 The structure of the optical models together with the results and the Mean Square Error (MSE) values yielded by the evaluation of spectra taken on Nb-oxide sample deposited by rf sputtering with 120 min duration. Optical model
Surface roughness layer [nm]
Nb-oxide layer (Tauc-Lorentz) [nm]
SiO2 layer [nm]
Ion-implantation-amorphized silicon (ia-Si) layer [nm]
MSE
A One-layer model B Two-layer model CThree-layer model DThree-layer model E Four-layer model
– 2.11 ± 0.07 2.56 ± 0.07 1.98 ± 0.05 2.25 ± 0.04
369.7 ± 0.3 368.7 ± 0.3 362.4 ± 0.3 362.3 ± 0.2 359.3 ± 0.2
– – 4.8 ± 0.2 – 2.8 ± 0.1
– – – 0.86 ± 0.02 0.80 ± 0.02
45.81 36.87 32.23 25.3 21.26
Fig. 3. Data-model differences in psi values for Nb-oxide sample deposited for 120 min.
Fig. 1. Measured and generated values at different angles of incidence (shown in the legend) in function of photon energy for the sample deposited by rf sputtering with 120 min duration and at a wall potential of 2.0 kV. The evaluation has been performed with the four-layer model (model “E”).
Fig. 4. Data-model differences in delta values for Nb-oxide sample deposited for 120 min.
Fig. 2. Measured and generated values at different angles of incidence (shown in the legend) in function of photon energy for the sample deposited by rf sputtering with 120 min duration and at a wall potential of 2.0 kV. The evaluation has been performed with the four-layer model (model “E”).
layer between the Nb-oxide layer and the i-a-Si layer, for the MSE a value of 21.26 was obtained. Figs. 1 and 2 show the measured and generated SE spectra of sample deposited with 120 min duration, the agreement between the measured and generated SE spectra is good. The data-model differences in psi and delta values for Nb-oxide sample deposited for 120 min are shown in Figs. 3 and 4. Possible reasons of the deviation between data and model: probably graded structure of the Nb-oxide layer, the Tauc-Lorentz dispersion relation is not the optimal choice for modeling the optical properties of the sputter deposited Nb-oxide, the mixture of ion-implantation amorphized silicon and single crystalline silicon does not describe accurately the damaged region of the single crystalline silicon substrate. The second sample considered here was deposited using a wall potential of 2.0 kV and a duration of 30 min for deposition. Table 3 shows the optical models together with the results yielded by the
evaluation. The second, third, fourth and fifth columns represent the sublayers of the optical models. The c-Si substrate was not indicated in Table 3. The introduction of new sublayers into the optical model resulted in decreasing MSE values, especially the introduction of the i-a-Si sublayer (MSE = 28.17 and 27.99 for models “D” and “E”, respectively). In the second four-layer model (model “F”) for the damaged silicon layer a mixture of c-Si and i-a-Si was chosen. The optical properties of the mixed layer were calculated by the effective medium approximation. This approach was used by Fried et al., they described the depth distribution of damage by a simple box-type model using only two parameters: the effective thickness and effective damage of the implanted layer [19,20]. For the volume fraction of the i-a-Si component a value of 29.7 ± 0.5% was yielded by the evaluation. In this case ten unknown parameters (layer thicknesses, volume fraction of i-a-Si, and the parameters of the Tauc-Lorentz dispersion relation) were defined and fitted. Considering the correlation matrix after the evaluation we observed that the correlation between the thickness of the partially damaged silicon layer (it is a mixture of c-Si and i-a-Si) and the volume fraction of the i-a-Si component is equal to 0.936. We considered this value too high, that is why we repeated the evaluation of the spectra selecting only one free parameter (the
Please cite this article in press as: T. Lohner, et al., Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide, Appl. Surf. Sci. (2016), http://dx.doi.org/10.1016/j.apsusc.2016.11.232
G Model APSUSC-34537; No. of Pages 7
ARTICLE IN PRESS T. Lohner et al. / Applied Surface Science xxx (2016) xxx–xxx
4
Table 3 The structure of the optical models together with the results and the Mean Square Error (MSE) values yielded by the evaluation of spectra taken on Nb-oxide sample deposited by rf sputtering with 30 min duration and with 2.0 kV wall potential. Optical model
Surface roughness [nm]
Nb-oxide layer (Tauc-Lorentz) [nm]
SiO2 layer [nm]
Ion-implantation−amorphized silicon (i-a-Si) layer [nm]
MSE
A One-layer model B Two-layer model CThree-layer model DThree-layer model E Four-layer model
– 1.06 ± 0.05 1.12 ± 0.06 0.76 ± 0.02 0.80 ± 0.02
184.9 ± 0.1 184.0 ± 0.1 183.4 ± 0.2 178.58 ± 0.05 178.18 ± 0.08
– – 0.6 ± 0.2 – 0.39 ± 0.06
78.36 74.91 74.89 28.17 27.99
F Four-layer model
0.82 ± 0.02
177.58 ± 0.08
0.92 ± 0.05
– – – 1.65 ± 0.01 1.65 ± 0.01 EMA-layer 6.1 ± 0.1 nm (70.3 ± 0.5% c-Si + 29.7 ± 0.5% ia-Si)
26.27
Table 4 Percentage of the i-a-Si component in the mixed layer (volume fraction of i-a-Si, f), and the thickness values of the mixed layer, d, yielded by the evaluation, and the “Amount of damage”, f•d, together with MSE values for Nb-oxide sample deposited by rf sputtering with 30 min duration and with a wall potential of 2.0 kV. Volume fraction of i-a-Si, f
Thickness of mixed layer, d [nm]
“Amount of damage”, f·d [nm]
MSE
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
11.2 ± 0.2 8.18 ± 0.06 6.00 ± 0.04 4.52 ± 0.03 3.55 ± 0.02 2.89 ± 0.02 2.43 ± 0.02 2.13 ± 0.01 1.86 ± 0.01 1.65 ± 0.01
1.12 1.64 1.80 1.81 1.78 1.73 1.70 1.70 1.67 1.65
43.05 29.02 26.33 26.92 27.60 28.09 28.20 28.09 28.05 27.99
Fig. 6. Measured and generated values at different angles of incidence (shown in the legend) in function of photon energy for the sample deposited by rf sputtering with 30 min duration and at a wall potential of 2.0 kV.
Fig. 7. Data-model differences in psi values for Nb-oxide sample deposited for 30 min. Fig. 5. Measured and generated values at different angles of incidence (shown in the legend) in function of photon energy for the sample deposited by rf sputtering with 30 min duration and at a wall potential of 2.0 kV.
layer thickness) for the mixed layer containing c-Si and i-a-Si. The volume fraction of the i-a-Si was set to 10%, 20%, 30% and so on, the evaluations were performed for all these volume fraction values one-by-one. The results are shown in Table 4. The lowest MSE value (26.33) was obtained for the assumption of a mixture of 30 volume percent i-a-Si and 70 volume percent c-Si components in the mixed layer modeling the damage caused by rf sputtering. In the present case we may consider the 30 volume percent i-a-Si as effective damage and the 6 nm as effective thickness of the disordered silicon layer after [19,20]. Figs. 5 and 6 show the measured and generated SE spectra of sample deposited with 30 min duration and 2.0 kV wall potential, concerning the evaluation the details described earlier for the mixed layer (30 volume percent i-a-Si and 70 volume percent c-Si components) were applied. The correspondence between the measured and generated SE spectra is good. The data-model differences in psi and delta values for Nb-oxide sample deposited for 120 min are shown in
Fig. 8. Data-model differences in delta values for Nb-oxide sample deposited for 30 min.
Figs. 7 and 8. The complex dielectric function of the Nb-oxide layer of evaluation detailed above for sample deposited at 2.0 kV with 30 min duration has been saved as reference data. The SE spectra of samples deposited at 1.0 and 1.5 kV with 30 min duration were evaluated using a four-layer optical model, the above speci-
Please cite this article in press as: T. Lohner, et al., Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide, Appl. Surf. Sci. (2016), http://dx.doi.org/10.1016/j.apsusc.2016.11.232
G Model APSUSC-34537; No. of Pages 7
ARTICLE IN PRESS T. Lohner et al. / Applied Surface Science xxx (2016) xxx–xxx
5
Table 5 Thickness values of the surface roughness layer, the Nb-oxide layer, the SiO2 layer and thickness of a mixed layer consists of 70 volume fraction of c-Si and 30 volume fraction of i-a-Si yielded by the evaluation of SE spectra of Nb-oxide samples deposited using various wall potentials and with 30 min rf sputtering deposition. Wall potential [kV]
Duration of rf sputtering deposition [min]
Surface roughness layer [nm]
Nb-oxide layer [nm]
SiO2 layer [nm]
Thickness of the mixed layer (70% c-Si + 30% i-a-Si) [nm]
MSE
1.0 1.5 2.0
30 30 30
1.7 ± 0.1 1.2 ± 0.1 0.8 ± 0.1
27.3 ± 0.1 79.7 ± 0.1 177.6 ± 0.1
3.9 ± 0.1 3.1 ± 0.1 0.9 ± 0.1
1.0 ± 0.1 3.9 ± 0.1 6.00 ± 0.1
22.15 22.20 26.36
Nb-oxide layer 2.88 nm Silicon dioxide layer 2.80 nm EMA layer (crystalline Si + 26% ion-impl. amorphized Si) 4.75 nm Single crystalline silicon substrate
Fig. 11. The three-layer optical model constructed on basis of the results of the RBS and HRTEM measurement. Fig. 9. Cross sectional HRTEM image of a Nb-oxide sample rf deposited for 100 s (2.0-kV wall potential). The thicknesses of the amorphous layers are indicated in the image. The observed thicknesses were used to the construction of the optical model for the evaluations of ellipsometry measurements.
Fig. 12. The refractive index n in function of photon energy for Nb-oxide layers deposited for 120 min, 30 min and 100 s.
Fig. 10. Measured and simulated RBS spectra taken on Nb oxide sample rf deposited for 100 s (2.0 kV wall potential).
fied optical reference data of Nb-oxide layer was built in into this optical model. The thicknesses of the four layers (surface roughness layer, Nb-oxide layer, SiO2 layer and the mixed layer with 30 volume percent i-a-Si and 70 volume percent c-Si components) were considered free parameters. The results of the evaluation are displayed in Table 5. The thickness of the mixed layer increases with increasing wall potential. A third part of our investigation involved studying of rf sputtering-deposited samples with 100, 150 and 300 s deposition durations. Fig. 9 displays the cross sectional HRTEM image and gives the layer structure for sample deposited with 100 s duration and 2.0 kV wall potential. The thickness of the rf sputtering deposited amorphous Nb-oxide film is 2.9 nm. A layer of 4.5 nm thickness sandwiched between the amorphous Nb-oxide film and the c-Si substrate can be interpreted by an amorphous SiOx and an i-a-Si component. A 0.6 nm thick surface roughness was assessed, too. Fig. 10 shows the measured and simulated RBS spectra, the thickness value of the Nb-oxide film (2.88 nm) is practically the
same as determined by HRTEM. Beneath the 2.8 nm thick SiO2 layer a partially damaged silicon layer was identified with thickness of 4.75 nm. Using the compositional and structural data obtained from RBS and HRTEM studies an optical model was created for the Nb-oxide sample deposited for 100 s and 2.0 kV wall potential (Fig. 11). The complex refractive index of the 2.88 nm thick Nb-oxide film was evaluated using a Lorentz oscillator [10]. The complex refractive index of the 2.88 nm thick Nb-oxide film was built into a four-layer optical model and the thickness values of the Nb-oxide film, the SiO2 layer and the thickness of a mixed layer consists of 74 volume fraction of c-Si and 26 volume fraction of i-a-Si were defined as free parameters. The results of the evaluations are compiled in Table 6. The thickness of the mixed layer increases with increase of wall potential. The refractive index n and the extinction coefficient k of the Nboxide layer displayed in function of photon energy for Nb-oxide samples deposited for 120 min, 30 min and 100 s in Figs. 12 and 13, respectively. The values of the refractive index and the extinction coefficient in function of photon energy for samples deposited with duration of 30 m and 120 m are similar to values published by Hála et al. [21]. The values of the refractive index and the extinction coefficient in function of photon energy for samples deposited with duration of 100 s show strange shape, certainly the initial growth stage and nucleation may play a role.
Please cite this article in press as: T. Lohner, et al., Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide, Appl. Surf. Sci. (2016), http://dx.doi.org/10.1016/j.apsusc.2016.11.232
G Model
ARTICLE IN PRESS
APSUSC-34537; No. of Pages 7
T. Lohner et al. / Applied Surface Science xxx (2016) xxx–xxx
6
Table 6 Thickness values of the Nb-oxide layer, the SiO2 layer and thickness of a mixed layer consists of 74 volume fraction of c-Si and 26 volume fraction of i-a-Si yielded by the evaluation of SE spectra of Nb-oxide samples deposited using various wall potentials. Wall potential [kV]
Duration of rf sputtering deposition [sec]
Surface roughness layer [nm]
Nb-oxide layer [nm]
SiO2 layer [nm]
Thickness of the mixed layer (74% c-Si + 26% i-a-Si) [nm]
MSE
1.0 1.5 2.0
300 150 100
0.6 (fixed) 0.6 (fixed) 0.6 (fixed)
2.4 ± 0.1 2.2 ± 0.1 2.7 ± 0.1
3.0 ± 0.1 3.4 ± 0.1 3.2 ± 0.1
2.5 ± 0.1 3.3 ± 0.1 4.5 ± 0.1
16.75 17.3 19.07
thickest sample (360 nm in 120 min) the growth rate (3 nm/min) is significantly lower than for samples with deposition duration of 10 min and 30 min. The reason may originate that the target (sintered ceramic body) is not 100% dense, it contains pores. The sputtering by Ar ions causes erosion of the surface and the change of the surface morphology explores the pores. A significant increase of the effective area of target can increase the sputtering rate. The change of the morphology in the nanometer scale is accompanied by graying of the target (low-reflectance surface nanostructures created with Ar ion bombardment). The thickest sample (360 nm in 120 min) has been deposited in 2007, the other samples have been deposited in the first half of 2016.
4. Conclusions
Fig. 13. The extinction coefficient k in function of photon energy for Nb-oxide layers deposited for 120 min, 30 min and 100 s.
Table 7 The values of projected range and straggling for bombarding silicon substrate with oxygen and argon ions based on SRIM-2013 calculations [9]. Longitudinal Straggling [nm]
Lateral Straggling [nm]
Oxygen ions into silicon substrate 1.0 4.3 5.8 1.5 2.0 7.1
3.2 4.1 4.9
2.3 3.0 3.6
Argon ions into silicon substrate 3.4 1.0 4.2 1.5 2.0 5.0
1.9 2.3 2.6
1.3 1.7 1.9
Ion energy [keV]
Projected Range [nm]
The formation of partially damaged layer on the surface of the silicon substrate during rf sputtering deposition can be explained by the bombardment of the surface of the single crystalline silicon substrate by energetic oxygen and argon ions and neutral atoms. The neutral atoms can be formed due to charge exchange mechanism. The energetic oxygen and argon ions and neutral atoms may penetrate into the c-Si substrate to a depth of several nm as one can see in Table 7 and they may remove silicon atoms from the crystalline lattice producing interstitial atoms and vacancies. In our interpretation the damaging effect originating from energetic particles bombarding the single crystalline silicon substrate will start at the beginning of the rf sputtering process and as the rf sputtering process proceeds, the thin damaged subsurface layer of the single crystalline silicon substrate will be buried by the newly deposited Nb-oxide material. The low growth rate of the thinnest sample (only 2.9 nm in 100 s) certainly a consequence of the initial stage of Nb-oxide layer growth including nucleation. Another sample (not included in the manuscript) deposited for 10 min exhibits a Nb oxide layer of 55.3 nm thickness. Its growth rate value (5.5 nm/min) is similar to the one obtained for the 30 min deposition (5.9 nm/min). For the
The experiments show that it is possible to characterize substrate surface damage induced by radiofrequency sputtering deposition of niobium oxide on c-Si substrate using spectroscopic ellipsometry. For the modeling of the partially damaged silicon layer a mixed layer with i-a-Si and c-Si components proved to be appropriate. For the case of a few nm thick rf sputtering deposited Nb-oxide layer independent methods such as cross HRTEM or high resolution RBS can deliver additional thickness values of the sublayers in the interface region. Sputtering at higher wall potential results in higher thickness of the mixed layer with i-a-Si and c-Si components in the damaged subsurface region. During the rf sputtering the energetic oxygen and argon ions and neutral atoms may penetrate into the c-Si substrate to a depth of several nm causing displacement of host silicon atoms and consequently producing damaged regions.
Acknowledgements Support from the National Development Agency grant OTKA K115852, and projects H2020-SMEINST-2-2014-683541 “OptiLight” FP7 “E450EDL” and “SEA4KET” and TÉT 12 DE-1-2013-0002 is greatly acknowledged.
References [1] S. Berg, L.P. Andersson, H. Norström, E. Grusell, Substrate surface damages by rf-sputtering, Vacuum 27 (1977) 189–191. [2] Stephen J. Fonash, S. Ashok, Ranbir Singh, Effect of ion-beam sputter damage on Schottky barrier formation in silicon, Appl. Phys. Lett. 39 (1981) 423–425. [3] F.D. Auret, O. Paz, N.A. Bojarczuk, Characterization of process-induced defects and device properties of ion beam sputter-deposited Mo contacts on Si, J. Appl. Phys. 55 (1984) 1581–1589. [4] H. Fujiwara, Spectroscopic Ellipsometry: Principles and Applications, John Wiley and Sons, Chichester, England, 2007. [5] J. Budai, B. Farkas, Z.L. Horváth, Zs. Geretovszky, On determining the optical properties and layer structure from spectroscopic ellipsometric data using automated artifact minimization method, Thin Solid Films 567 (2014) 14–19. [6] Z. Tóth, I. Hanyecz, A. Gárdián, J. Budai, J. Csontos, Z. Pápa, M. Füle, Ellipsometric analysis of silicon surfaces textured by ns and sub-ps KrF laser pulses, Thin Solid Films 571 (2014) 631–636. [7] E.R. Shaaban, M. El-Hagary, M. Emam-Ismail, A.M. Abd Elnaeim, S.H. Moustafa, A. Adel, Optical characterization of polycrystalline ZnSe1-xTex thin films using variable angle spectroscopic ellipsometry and spectrophotmetery techniques, Mater. Sci. Semicond. Process. 39 (2015) 735–741.
Please cite this article in press as: T. Lohner, et al., Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide, Appl. Surf. Sci. (2016), http://dx.doi.org/10.1016/j.apsusc.2016.11.232
G Model APSUSC-34537; No. of Pages 7
ARTICLE IN PRESS T. Lohner et al. / Applied Surface Science xxx (2016) xxx–xxx
[8] G. Mezey, E. Kótai, T. Lohner, T. Nagy, J. Gyulai, A. Manuaba, Improved depth resolution of channeling measurements in Rutherford backscattering by a detector tilt, Nucl. Instrum. Methods 149 (1978) 235–237. [9] J.F. Ziegler, M.D. Ziegler, J.P. Biersack, SRIM – the stopping and range of ions in matter (2010), Nucl. Instrum. Methods Phys. Res. Sect. B 268 (2010) 1818–1823. [10] J. A. Woollam Co., Inc. (www.jawoollam.com) (2016). [11] G.E. Jellison Jr., F.A. Modine, Parameterization of the optical functions of amorphous materials in the interband region, Appl. Phys. Lett. 69 (1996) 371–373. [12] G.E. Jellison Jr., F.A. Modine, Parameterization of the optical functions of amorphous materials in the interband region (erratum), Appl. Phys. Lett. 69 (1996) 2137. [13] M. Fried, T. Lohner, W.A.M. Aarnink, L.J. Hanekamp, A. van Silfhout, Determination of complex dielectric functions of ion implanted and implanted-annealed amorphous silicon by spectroscopic ellipsometry, J. Appl. Phys. 71 (1992) 5260–5262. [14] F. Pászti, A. Manuaba, C. Hajdu, A.A. Melo, M.F. da Silva, Current measurement on MeV energy ion beams, Nucl. Instrum. Methods B 47 (1990) 187–192. [15] E. Kótai, Computer methods for analysis and simulation of RBS and ERDA spectra, Nucl. Instrum. Methods B 85 (1994) 588–596.
[16] E. Kótai, Computer methods for analysis and simulation of RBS and ERDA spectra, Proc. of Fourteenth Int. Conf. on the Application of Accelerators in Research and Industry, Nov 6–9, 1996, Denton, Texas USA, CP392, edited by J.L. Duggan, I.L. Morgan, AIP Press, New York 1997, p. 631–634. [17] Á. Barna, Topographic kinetics and practice of low angle ion beam thinning, Mater. Res. Soc. Symp. Proc. 254 (1992) 3–22. [18] J.P. McCaffrey, Small-angle cleavage of semiconductors for transmission electron microscopy, Ultramicroscopy 38 (1991) 149–157. [19] M. Fried, G. Juhász, C. Major, P. Petrik, G. Battistig, Homogeneity check of ion implantation in silicon by wide-angle ellipsometry, 17th IEEE International Conference on Advanced Thermal Processing on Semiconductors—RTP 2009, New York, IEEE, 2009. Paper 5373448. 4 p. [20] C. Major, G. Juhasz, Z. Herczeg, Z. Labadi, P. Petrik, Z. Horvath, M. Fried, High speed spectroscopic ellipsometry technique for in-line monitoring in large area thin layer production, Collection entitled ICSE-7 Ms. Ref. No.: APSUSC-D-16-06772. [21] M. Hála, J. Capek, O. Zabeida, J.E. Klemberg-Sapieha, L. Martinu, Hysteresis-free deposition of niobium oxide films by HiPIMS using different pulse management strategies, J. Phys. D Appl. Phys. 45 (2012) 055204.
Please cite this article in press as: T. Lohner, et al., Spectroellipsometric detection of silicon substrate damage caused by radiofrequency sputtering of niobium oxide, Appl. Surf. Sci. (2016), http://dx.doi.org/10.1016/j.apsusc.2016.11.232
7