Open Access
Add to BibSonomyAbstract
A path-folded infrared image spectrometer with five sub-gratings and five linear-array detectors was applied to a broadband optical monitoring (BOM) system for thin film deposition. Through in situ BOM, we can simultaneously acquire the thickness and refractive index of each layer in real time by fitting the measured spectra, and modify the deposition parameters during deposition process according to the fitting results. An effective data processing method was proposed and applied in the BOM process, and it shortened the data processing time and improved the monitoring efficiency greatly. For demonstration, a narrow band-pass filter (NBF) at 1540 nm with ~10 nm full width at half-maximum (FWHM) had been manufactured using the developed BOM system, and the results showed that this BOM method was satisfying for monitoring deposition of thin film devices.
©2011 Optical Society of America
1. Introduction
Great attention is being always paid on different kind of monitoring techniques for film deposition, especially for depositing optical thin film filters, which needs precise control of the thickness of every layer [1, 2]. Among the various monitoring techniques, the optical monitoring approaches usually play a leading role. However, the conventional optical monitoring methods, including single wavelength monitoring (SWM) and its derived methods, have their intrinsic defects because of the limitation of the monitoring wavelength range. For example, the defects of turning point monitoring (TPM) include overshooting of control, low efficiency on non-quarter-wave film deposition, etc [2, 3]. Therefore, many efforts have been made on the improvement of optical monitoring techniques [4-11]. To surpass the limitation of SWM method, BOM method was proposed and studied in the past 30 years [9-11]. The BOM method has many advantages, such as higher monitoring precision, broad adaptation to various designs, especially to non-quarter-wave film deposition. The other reason why BOM is important is that the in situ broadband spectrum contains a great deal of optical information of the film, which allows us to take real time measurement of the optical properties of film, such as refractive index and extinction coefficient of every layer, to deal with unstable and complicated deposition conditions occurred randomly and make some real time corrections in situ on the film being deposited.
Spectrometer is one of the key instruments in a BOM system. For conventional mechanical spectrometers, it is very hard to meet the demands of high acquisition speed, high spectral resolution and broad spectral range. In other words, it will take longer time to acquire a spectrum with both higher resolution and broader spectral band, and the time consuming is usually unacceptable in many cases, such as depositing optical film devices, where the spectral characteristics of optical film devices under deposition need to be measured in situ. In order to overcome the above-mentioned shortages of the conventional mechanical spectrometers, we have developed an image spectrometer characterized by using multiple gratings and linear-array detectors or charge-coupled device (CCD) with short data acquisition time, wide spectral range and high resolution [12-14]. The features of the spectrometers were highly evaluated and considered suitable for in situ BOM [15, 16]. In this work, we have developed an infrared BOM system for an electron-beam-evaporation (EBE) film deposition equipment based on the development of the infrared image spectrometer, and manufactured an infrared NBF using BOM with an optimized method of data processing. We acquired and analyzed the real time transmittance spectrum of the film sample during deposition process.
2. Experiment
In this work, samples were prepared in an EBE film deposition system (AFDS-1100), which has a vacuum chamber with its inside volume larger than 1 m3. The chamber was built by Dongfang Gaide Vacuum-Tech Co., Ltd., Beijing, China. Two turbo-molecular pumps (TMH2101P) together with one Roots pump (WKP 500AS) and one dry backing vacuum pump (UniDry TM 050-4) have been applied to achieve high vacuum for the chamber. The base pressure is below 1.5 × 10−3 Pa. Two E-Beam emissive sources are used for the evaporation of two types of target material (SiO2 and Ta2O5). Over each copper crucible for loading target material, a baffle is used to control the deposition time of each material. A third baffle is placed under the sample holder to prevent polluting the sample during pre-heating and degasification process on target material and block the signal light for the spectrometer to acquire background spectrum. Two optical monitoring windows with diameter of 10 cm, covered by fused silica are placed in the center of both the upside and downside of the vacuum chamber.
Figure 1 shows the basic configuration of the BOM system. Broadband infrared light covering 1200-2000 nm comes out of the light source (UWS-1000, Santec Corp., Japan). The infrared light emitting from the light source is conducted through a single mode fiber (SMF, core diameter of 9μm), and then expands to a collimated beam with diameter of ~3 mm via a-collimator. The beam is reflected by a mirror and enters the vacuum chamber through the fused silica window and then is incident vertically on the sample under deposition. The transmitted light goes upwards out of the vacuum chamber and is coupled into another SMF, and finally collected by the infrared spectrometer.
The infrared spectrometer consists of five sub-gratings and five linear-array detectors (5 × 512 pixels). Each of the five grating-detector pairs covers a short range respectively: 1430-1470 nm, 1470-1510 nm, 1510-1550 nm, 1550-1590 nm, 1590-1630 nm, resulting in a continuous range of 1430-1630 nm with a high resolution better than 0.08 nm. Using linear-array detectors, the spectrometer acquires the whole spectrum with 2560 data points in less than 50 ms. The excellent characteristics of broad spectral range, high resolution and short data-acquisition time of the infrared spectrometer match the basic demand of in situ BOM. The acquired spectral data is sent to a computer for analysis, and converted to the growing information of the film under deposition, such as the thickness and refractive index of each deposited layer.
In the experiment, a multi-layer film system was manufactured to demonstrate the effect of the infrared BOM system. NBF is widely used in optical communication field. The successful manufacture of NBF needs good performance of deposition system and monitoring techniques. The well-known quartz crystal deposition monitoring has only a limited monitoring effect on NBF manufacturing. The TPM method widely used in optical films manufacturing is also only effective with strictly steady deposition parameters. In our experiment, the e-beam emitter in EBE deposition system may shut down occasionally due to the e-beam current overload protection feature. The sudden shut-down of e-beam induces film deposition pause, which may generate a wrong signal that the transmittance or reflectance has gotten its extreme point. Once the wrong signal is sent to the system controller, it will cause a wrong operation. Such a mistake does not exist in BOM system, because the finish signal of deposition in the BOM system is independent of the deposition time or deposition rate.
In the experiment, a NBF system of Si/9L(HL)52H(LH)5/Air was chosen with central wavelength of 1540 nm, where H and L represent Ta2O5 with high refractive index (pre-set as 1.955) and SiO2 with low refractive index (pre-set as 1.466) respectively. The first 9L layer was designed to get a high transmission at central wavelength 1540 nm and introduce a peak structure in the transmittance spectrum, which would help to improve the effect of BOM. In designing the NBF, the optical thickness of each H layer or L layer is of one quarter of 1540 nm (385 nm). There might be some changes on n and d of each layer in practical deposition process. As the transmittance spectrum is collected, we can get the precise effective parameters by fitting 2560 sets of spectral data of transmittance. In the experiment, double-side polished n-type single crystal silicon wafer with crystal orientation of (111), thickness of 350 ± 20 μm, resistivity of 3-7 Ω•cm, was used as substrate for its excellent transparency in the wavelength range of 1430-1630 nm and high-temperature resistance. The silicon substrate was cleaned in a modified RCA process [17], followed by a 30 s diluted HF solution dip, and then a de-ionized water rinse. The cleaned silicon substrate was then put on the sample holder in the vacuum chamber immediately. Before deposition, the background light spectrum Sbg(λN) (with light source turn-off) and the initial transmitted light spectrum S 0(λN) (with light passing through a naked substrate, i.e., without any film on the substrate) were collected, and then the film deposition process started. The e-beam parameters for depositing Ta2O5 and SiO2 were 180 mA × 9 kV and 40 mA × 7 kV, respectively.
3. Data processing
As mentioned above, the theoretical calculation of transmittance is based on an Air/Film/Substrate structure, without considering the effect by the reverse side of the substrate. In this work, we have reduced the influence of the reverse side (without AR coating) of the substrate by employing a substrate with two sides not strictly parallel. In this case, multiple beam separation of the transmitted light happened, and only the directly transmitted light beam was collected, i.e., the light beam transmitted after internal reflections inside the substrate was not collected. With the intensity of the light from the source and the transmittance of the reverse side of the substrate being kept constant, the intensity of the directly transmitted light is proportional to the transmittance of the front side of the substrate T 0 or the transmittance of the film deposited Trea.
The key technique of BOM method is to acquire the monitoring spectrum carrying film-deposition information. During deposition, the real time infrared spectrum Srea(λN, d), where d was the thickness of the layer under deposition, was collected, and the transmittance spectrum was given by
Where T 0(λN) was determined by
Where n(λN) was the refractive index of silicon wafer at the wavelength corresponding to the N th pixel of the linear-array detectors in the infrared spectrometer. The theoretical spectrum of transmittance Tthe(λN, d 0) was calculated using conventional transfer matrix method (TMM), where d 0 was the pre-designed thickness of the layer under deposition. The thickness of substrate was considered as infinite in theoretical simulation. Once the real time spectrum Trea(λN, d) of transmittance and the theoretical one Tthe(λN, d 0) had both been determined, a conventional merit function RE(d), which measured the difference between Trea(λN, d) and Tthe(λN, d 0), was then obtained as follows:
Where, Npxls was the total number of pixels of the linear-array detectors in the infrared spectrometer. In this work, the linear-array detectors included 2560 pixels, so Npxls = 2560. If the ideal case of experiment were considered, once d = d 0, then RE(d) = 0, and the thickness of deposited layer would be exactly determined. However, in a practical experiment, there were some factors that causing RE(d) not equal to zero when d = d 0, such as imperfect deposition conditions, noise in the spectrum, etc. Instead, when d = d 0, RE(d) got to a minimum. It was difficult to foresee the exact time when RE(d) got to the minimum (the similar disadvantage also exists in TPM method), so there would be a slight time delay to stop depositing a certain layer. We may call this phenomenon “over-shoot”. To eliminate over-shoot, the Eq [3]. was revised following the method of data fitting used in spectroscopic ellipsometry[18]. The real time information of depositing a layer would be extracted by fitting the following modified merit function:
The fitting target was the real time spectrum of transmittance, not the pre-designed theoretical one. In the fitting process, the trial refractive index ntes varied slightly, the trial thickness dtes varied from 0 to d 0. Although RE(ntes, dtes) ≠ 0 in fitting process, a minimum would be achieved with unique ntes and dtes (d = dtes = dfit, n = ntes = nfit). According to the values of nfit and dfit, the theoretically designed film system using the preset n and d (n = 1.466 for SiO2, n = 1.955 for Ta2O5, and nd = 385 nm) would be modified in real time and applied to the subsequent deposition process. The fitting process would cost much time because of a huge amount of calculation in each course of data processing. The delayed shooting time due to the longtime fitting process became the main problem for BOM method. To solve the time-consuming problem of fitting calculation, we proposed a method to simplify the calculation by only fitting spectral data at the most sensitive wavelengths, which corresponded to the extreme points (peaks or valleys) of the spectrum of transmittance. It was similar to TPM method, but independent of the deposition time. If the deposition conditions were stable, the refractive index of film material would keep a constant, and the only parameter unknown would be the thickness of the deposited layer. Of course, the film system was designed in advance before deposition. The extreme point wavelengths of the simulated spectrum of transmittance at different thicknesses or optical thicknesses were compiled into a database as shown in Tab. 1 .
Table 1. The data compiled before deposition include all extreme-point wavelengths of the simulated spectrum of transmittance at different thicknesses.
Generally, the database was small, and easy to build with computer. As the database was built, we could modify the Eq [4]. and got the following merit function:
Where C denoted the sum of extreme point wavelengths in the spectra obtained for monitoring at different thickness dtes, and C was usually less than 3 in the considered spectral range. Comparing the sum C(<3) of Eq [5]. with the sum Npxls ( = 2560) of Eq [4], we found that the amount of calculation had been reduced greatly by applying Eq [5]. in place of Eq [4]. As the simulation information had been stored in database in advance, there was no computing time for the theoretical spectrum of transmittance. The main work of computer was to find the extreme point wavelengths λN,rea(d) of the real time spectrum of transmittance.
On the basis of the film system of NBF mentioned above, we simulated the extreme point wavelengths in the range of 1430-1630 nm at different optical thickness of the deposited layer. In Fig. 2 , we found that once the deposition of every layer, such as the first layer (9L), had been finished, there was always an extreme point appearing at 1540 nm. As the deposition of one layer had been finished, the deposition process paused for a while, and the fitting of the n and d of the last deposited layer was performed. According to the fitting results, the film system would be revised, and then the deposition of next layer began.
Fig. 2 Extreme point wavelengths of the spectrum of transmittance in 1430-1630 nm at different optical thickness in Si/L9(HL)5H2(LH)5/Air film system.
4. Results and discussion
According to the methods mentioned above, we have fulfilled an experiment of depositing an optical filter. During deposition, the spectra of transmittance, as shown in Fig. 3 , were obtained in situ for monitoring the deposition process.
Once the spectrum of transmittance was obtained, it would be analyzed immediately. The analyzed results would be applied to determining whether the deposition of a layer should stop or not. Figure 4 showed the results of comparison between the pre-designed spectra and the measured spectra, which was collected just after the deposition of a certain layer had been finished. Figure 4(a) displayed the results for the case that the 5th layer countdown of the filter had been deposited, and Fig. 4(b) to 4(d) showed the similar results for the cases corresponding to the 4th, 3rd, 2nd layer countdown of the filter had been deposited respectively. We found that the deposition effect differed from what we expected during real deposition process. We also observed the similar phenomenon from the change of transmittance versus deposition time at the central wavelength 1540 nm during deposition process as shown in Fig. 5 .
Fig. 4 Comparison between the measured spectra and the pre-designed spectra. The black lines show the pre-designed spectra, and the red lines show the experimental spectra.
From Fig. 5, we found that the transmittance curve did not show a symmetrical shape as it should be for an ideal thin film system during real deposition. It implied that the film structure and the material parameters for each deposited layer were different from those of theoretically designed ones. Therefore, it was necessary to amend the depositing film during deposition process. Using the fitting method mentioned above, we could obtain the thickness and optical constants of each deposited layer, and then the fitted parameters had afterwards been applied to theoretical design of the film with considering the deviation of spectra of the deposited film with those of the designed film. These measures would amend the thin film stacks on time and prevent the further deviation of performance of the deposited film with that of the designed film. By taking the sample used here as an example, during the deposition, at 6300 s from the beginning of deposition, the current of electron beam was wrongly turned off, and the transmittance of the film became invariable, usually this indicated the deposition of one layer had finished and it was time to deposit another layer according to conventional TPM method. However, at present case, the invariability of the transmittance of the film was ascribed to the experimental breakdown rather than the normal optical character of well deposited film. If the TPM was adopted here for monitoring the deposition process, this breakdown of the experimental system would result in failure of the deposition of the thin film stack. Fortunately, the BOM was applied here for monitoring the film deposition, and the breakdown of the electron beam had been noticed on time and the compensation had been done carefully in situ. Finally, a filter device in accordance with the designed one was obtained. In order to show the amending effects, we compared the spectra of transmittance for the deposited film during deposition with the theoretically calculated ones which had been amended, as shown in Fig. 6 . One may find, the theoretically calculated spectra for the amended films were in good agreement with the measured spectra. The spectrum of transmittance recorded in vacuum for the finally prepared narrow band filter was shown in Fig. 7 , and the difference between the measured spectrum and the calculated one was small enough, indicating that the amending measure was effective and successful.
Fig. 6 Comparison between the experimental transmittance spectra and the revised design transmittance spectra. The black lines show the revised design spectra, and the red lines show the experimental spectra. (a)-(d) show the results for the cases that the 5th to 2nd layers countdown of the filter had been deposited, respectively.
5. Conclusion
On the basis of an image spectrometer with multiple sub-gratings, a BOM system for thin-film deposition was developed. An optimized data processing method for broadband monitoring was proposed. During deposition, the spectra of transmittance of the thin film under deposition had been obtained and analyzed. The analyzed results were used to modify the deposition conditions. If any deviation of the optical properties of the film under deposition occurs, the deposition conditions will be amended in situ. In order to testify this method, it has been applied to an actual deposition process for preparing optical thin film filter devices, and a filter device with performance in accordance with the theoretical design has finally been obtained. The results show the validity of this broad band monitoring method on thin film deposition.
Acknowledgements
This work was supported by the National Science Foundation (NSF) project of China with the contract numbers 60778028, 60938004, by the STCSM project of China (Grant No. 08DZ1204600) and by the No.2 National Science and Technology Major Project of China (No. 2011ZX02109-004).
References and links
1. H. A. Macleod, “Monitoring of optical coatings,” Appl. Opt. 20(1), 82–89 (1981). [CrossRef] [PubMed]
2. H. A. Macleod, Thin-Film Optical Filters, 3rd ed.(Institute of Physics, 2001).
3. B. Vidal, A. Fornier, and E. Pelletier, “Optical monitoring of nonquarterwave multilayer filters,” Appl. Opt. 17(7), 1038–1047 (1978). [CrossRef] [PubMed]
4. C. C. Lee, K. Wu, C. C. Kuo, and S. H. Chen, “Improvement of the optical coating process by cutting layers with sensitive monitoring wavelengths,” Opt. Express 13(13), 4854–4861 (2005). [CrossRef] [PubMed]
5. C. C. Lee and K. Wu, “In situ sensitive optical monitoring with proper error compensation,” Opt. Lett. 32(15), 2118–2120 (2007). [CrossRef] [PubMed]
6. C. C. Lee and Y. J. Chen, “Multilayer coatings monitoring using admittance diagram,” Opt. Express 16(9), 6119–6124 (2008). [CrossRef] [PubMed]
7. F. Lai, X. Wu, B. Zhuang, Q. Yan, and Z. Huang, “Dual wavelengths monitoring for optical coatings,” Opt. Express 16(13), 9436–9442 (2008). [CrossRef] [PubMed]
8. A. V. Tikhonravov, M. K. Trubetskov, and T. V. Amotchkina, “Statistical approach to choosing a strategy of monochromatic monitoring of optical coating production,” Appl. Opt. 45(30), 7863–7870 (2006). [CrossRef] [PubMed]
9. B. Vidal, A. Fornier, and E. Pelletier, “Wideband optical monitoring of nonquarterwave multilayer filters,” Appl. Opt. 18(22), 3851–3856 (1979). [PubMed]
10. L. Li and Y. H. Yen, “Wideband monitoring and measuring system for optical coatings,” Appl. Opt. 28(14), 2889–2894 (1989). [CrossRef] [PubMed]
11. B. Badoil, F. Lemarchand, M. Cathelinaud, and M. Lequime, “Interest of broadband optical monitoring for thin-film filter manufacturing,” Appl. Opt. 46(20), 4294–4303 (2007). [CrossRef] [PubMed]
12. T. Han, Y. H. Wu, J. K. Chen, Y. F. Kong, Y. R. Chen, B. Sun, C. H. Xu, P. Zhou, J. H. Qiu, Y. X. Zheng, J. Miao, and L. Y. Chen, “Study of the high resolution infrared spectrometer by using an integrated multi-grating structure,” Rev. Sci. Instrum. 76(8), 083118 (2005). [CrossRef]
13. Y. R. Chen, B. Sun, T. Han, Y. F. Kong, C. H. Xu, P. Zhou, X. F. Li, S. Y. Wang, Y. X. Zheng, and L. Y. Chen, “Densely folded spectral images of the CCD spectrometer working in the full 200-1000nm wavelength range with high resolution,” Opt. Express 13(25), 10049–10054 (2005). [CrossRef] [PubMed]
14. M. H. Liu, S. X. Pan, Y. R. Chen, Y. F. Wu, Q. Y. Cai, P. H. Mao, Y. X. Zheng, and L. Y. Chen, “Path-folded infrared spectrometer consisting of 10 sub-gratings and a two-dimensional InGaAs detector,” Opt. Express 17(17), 14956–14966 (2009). [CrossRef] [PubMed]
15. G. Overton, “Spectrometry: CCD spectrometer folds spectral images from 200 to 1000 nm”, Laser Focus World, World News, 42–43, Feb. 2006 edition, www.lfw-digital.com.
16. R. Gaughan, “Folded Spectrometer offers Range, Resolution and Speed,” Photonics Spectra, Photonics Technology World, 20–22, Mar. 2006 edition, www.photonics.com.
17. W. Kern and D. A. Puotinen, “Cleaning solution based on hydrogen peroxide for use in semiconductor technology,” RCA Rev. 31, 187–206 (1970).
18. R. M. A. Azzam and N. M. Bashara, Ellipsometry and Polarized Light (North-Holland, Amsterdam, 1977).
