PROJECT. DOCUMENT IDENTIFICATION D2.2 - Report on low cost filter deposition process DISSEMINATION STATUS PUBLIC DUE DATE 30/09/2011 ISSUE 2 PAGES 16

Transcription

1 GRANT AGREEMENT NO. ACRONYM TITLE CALL FUNDING SCHEME PROJECT 2WIDE_SENSE WIDE spectral band & WIDE dynamics multifunctional imaging SENSor ENABLING SAFER CAR TRANSPORTATION FP7-ICT STREP TITLE DOCUMENT IDENTIFICATION D2.2 - Report on low cost filter deposition process DISSEMINATION STATUS PUBLIC DUE DATE 30/09/2011 DELIVERABLE D2.2 ISSUE 2 PAGES 16 WP/TASK WP2, T2 WP TITLE STUDY ON LOW COST FILTER DEPOSITION PROCESS PARTNER IN CHARGE III-VLAB EXECUTIVE SUMMARY The deliverable D2.2 concerns the low cost deposition process. The deposition at pixel level has been replaced by deposition and process on sapphire substrate. The filter is afterward deposited on back-thinned VisSWIR sensor. The choice of spectral bands and designs has been presented in report D2.1. A design based on 57 layers is proposed and tested. The experimental results present efficient stop band without overlap. The flowchart for process is detailed from alignment procedures to dielectric etching. Finally, we briefly present how process could be adapted before dicing the readout circuit to be compatible to a complete backside filter deposition and process. Status: Public 1/16

2 AUTHORS III-VLab Jean-Luc Reverchon / T2.2. leader APPROVAL III-VLab ERIC COSTARD / PROGRAM LEADER ANNE ROUVIE / WP2.4 LEADER AUTHORIZATION EUROPEAN COMMISSION A. FERNANDEZ-RANADA SHAW / PROJECT OFFICER ISSUES DATE MODIFICATIONS AUTHOR 1 07/10/2011 Jean-Luc Reverchon 2 13/12/2011 CORRECTIONS APPROVED BY PROGRAM LEADER Jean-Luc Reverchon DISTRIBUTION LIST OPT, CRF, UNIPR, ADA, NIT;RPL WORK PACKAGE LEADERS PUBLIC DISSEMINATION Status: Public 2/16

3 CONTENT 1. Introduction Principe for filters deposition at pixel level Direct process on back thinned visible extended sensors Process on sapphire substrate before flip-chip Flip-chip report of filter on backside VisSWIR imager Flowchart for process on sapphire substrate Filters design adapted to the glued filter arrays Design of high-pass filters Fabrication Filter deposition Alignment marks SiO 2 / TiO 2 etching Reporting of backside Toward low cost deposition process Conclusion LIST OF ACRONYMS Status: Public 3/16

4 1. Introduction The filter deposition at pixel level could have been considered as a dielectric deposition on the backside of Focal Plane Arrays (FPA). It would require three dielectric depositions and process for delimitation of pixels on individual FPAs. It would be risky as long as readout and detection circuits are hybridised individually. Consequently we have made the choice to report an array of filters initially processed on sapphire by flip-chip procedure. We report here the process and the precautions required in terms of alignment, gluing, cross-talk and total transmission of the device. We will qualify the design of filters and deposition procedure in terms of rejection and stop band overlap. Two filters have been deposited for this aim in a preliminary work. The space between the InP contact layer and filters differs between the different kinds of pixels, is filled with a epoxy transparent from visible to 1,670 µm. The design is consequently adapted. Status: Public 4/16

5 2. Principe for filters deposition at pixel level 2.1. Direct process on back thinned visible extended sensors The filter deposition at pixel level could have been considered as a dielectric deposition on backside of FPAs (Readout Circuit & 2D array of InGaAs photodiodes) with processing for delimitation of pixels. It would require three dielectric depositions (1 clear pixel & 3 pixels with different high pass filters) plus processing on individual FPA as shown figure n 1. It would be risky as long as readout and detection circuits are hybridised individually. a) b) F1 InP InGaAs InP InGaAs ROIC ROIC c) d) InGaAs InGaAs ROIC ROIC Figure n 1: Flowchart for deposition and process on hybrid 2.2. Process on sapphire substrate before flip-chip As a consequence, we proposed in D2.1 to fabricate filters on sapphire substrate just before hybridising the array of filters on the backside of the individual FPA as shown figure n 2. It is the better solution to show the proof of concept with prototypes even though, we still think that collective process is possible and compatible to low cost imaging devices. As a good spatial resolution is required, we must take care of optical cross-talk when reporting array of dielectric filters by flip-chip. Hence, metallic mask is deposited and processed in order to provide alignment marks prior to all dielectric Status: Public 5/16

6 deposition. These marks are reference for future process steps: dielectric etching to individualize pixels and flip chip reporting on backside VisSWIR FPA. Metallic masks of thickness between 0.5 and 4 µm will also prevent optical crass-talk between pixels in case of misalignment. Figure n 2: Flowchart for collective filter fabrication and reporting on individual hybrid 2.3. Flip-chip report of filter on backside VisSWIR imager After dicing, sapphire substrate supporting arrays of filters are reported on the backside of a VisSWIR sensor with a flip chip process identical to the one employed to report the array of photodiodes on the readout circuits (figure n 3). So, the filters are close to the photodiodes in order to prevent optical cross-talk. The assembly based on indium bumps in the photo-diode / readout circuit assembly is replaced by gluing the filter array with epoxy glue similar to the one used to fill the space between indium bumps, also named under-filing. The kind of epoxy used here is chosen regarding its transmission properties from 400 nm to 1600 nm. Transmission is expected larger than 98% with a maximal thickness of 4 µm. An interest is to fill the space between pixels filters of different thickness in a way similar to the filling of the space between indium bump. The general assembly presented in figure n 2 is taken into account for the simulation of the filter transmission. Status: Public 6/16

7 Sapphire Sapphire Epoxy Figure n 3: Report of the 2D array of filters on the backside of the VisSWIR array 3. Flowchart for process on sapphire substrate The solution based on the 2D array of filters requires a substrate transparent from 400 nm to 1700 nm with mechanical properties compatibles to lithography processes, dicing and cost. It is why sapphire is preferred to glass. 3.1 Filters design adapted to the glued filter arrays The choice of bands has been define in WP4 (D4.1) after the technological feasibility studies. It describes the sensitivity and spectral range of the sensor. Due to the different requirements of the selected preventive safety applications, a wide spectral sensitivity is needed. Respect to current InGaAs imager the imager sensitivity in the visible spectral range is mandatory. Status: Public 7/16

8 The multispectral information of the scene is made available by the use of a filter pattern at pixel level. The filter pattern is related to the number of spectral data required by the selected functions. In particular, the road condition monitoring (RCM) functions will require two wavelength bandwidths to be monitored. High Beam Assist (HBA) and Traffic Sign Recognition (TSR) will require one/two wavelength bandwidths. Taking into account this assumption, the proposed filter pattern (Deliverable D6.5) will be the following one: C F4 F1 F2 Figure n 4: Proposed high pass filters configuration The pattern is based on a configuration of 4 pixels in which one is panchromatic without any filter on it. The others ones are high-pass filters, combination of stop band based on Pelletier design as follows: - Clear: no filter - F1: SWIR 1300 to1700 nm - F2: SWIR 1000 to1700 nm - F4: RED+NIR+SWIR <600 nm Figure n 5: Proposed high pass filters transmission with 10 pairs of TiO 2 / SiO 2 bilayers Status: Public 8/16

9 Transmission WIDE spectral band & WIDE dynamics multifunctional imaging Due to this selection, it is possible to select the green and the red bandwidths and two SWIR spectral ranges. By selecting these four spectral information is possible to obtain the following spectral ranges: C-F4 Green band F4-F2 Red+NIR band F2-F1 SWIR band 1000 to 1300 nm F1 SWIR band 1300 to 1700 nm The selection of the most useful filter pattern configuration has to be made considering all the targeted preventive safety applications. In the first config, RED+NIR spectral band is useful for taillight recognition (HBA function). In particular by placing C clear pixel under F4 pixel taillight reflection on the road should be minimized. RED+NIR band allows also traffic sign detection based on colour segmentation. SWIR bands, 1000 to 1300 nm and 1300 to 1700 nm, show different water absorption and can be useful to detect road status. The proposed optical filters have been suggested due to the current better trade off between the optical performances in term of trasmittance and rejection regions and fabrication (at pixel level) reliability. 3.2 Design of high-pass filters Due to a sufficient high selectivity, we base our design on 7 pairs of TiO 2 / SiO 2 layers. The total filter consisting in 57 layers deposited in 3 steps is simulated here after taking into account the real configuration: sapphire substrate + filters + epoxy. The epoxy s refractive index is around 1.5. We notice that the transmission is close to the previous simulations. The periodic filters define the stop band independently of substrate and glue thickness that only induces modulation Pelletier 57 layers 15 layers + epoxy = 29 layers + epoxy = 43 layers + epoxy = 57 layers Status: Public 9/ Lambda (nm) Figure n 6: Proposed high pass filters transmission on a sapphire substrate, with 15 layers per filter and epoxy glue

10 4. Fabrication 4.1 Filter deposition Before depositing filters on the final substrate, filters are deposited on sapphire substrate in order to tune the thickness and refine the process. The duration for deposition of F1 (64 nm x 14) and F2 (84 nm x 14), i.e. the thinnest filters, is respectively around 1h50 and 2h30. We present hereafters simulated and measured filters. Measurements are made by transmission and reflectivity on silicon substrate. Simulations are operated with McLoad software and the experimental values from TiO 2 and SiO 2. We obtain the following spectra. We obtain the following experimental spectra for F1 and F2: Status: Public 10/16

11 Figure n 7: Experimental transmissions for filters based on 15 layers We remark the cut-off very close to the simulated one: nm experimentally against 593 nm with simulation for F1-727 nm experimentally against 743 nm with simulation for F2 As a conclusion, the filters design is very close to design. The transmission is lower than 1% in stop bands excepted for some few narrow peaks. Status: Public 11/16

12 4.2 Alignment marks Specific marks are used for alignment to features required for backside flip-chip presented figure 3. Figure n 8: Alignment features on sapphire substrate These marks must recover the ohmic contact visible from backside as presented figure n 8 (left). The result is presented figure n 8 (right) with the ohmic contact from VisSWIR photodiode array centred inside the alignment marks from the 2D array of filters. It shows that the precision for the alignment of the 2D array of filters with the 2D array of photodiodes is in the µm range. Figure n 8: Left: backside of a VisSWIR imager after substrate removal. The ohmic contact is apparent. Right: Alignment of filters vs ohmic contact just after flip-chip. Metallisation is delimiting pixels in order to prevent backscattering and optical cross talks. A large metallisation of 4 µm will prevent cross-talk even several µm of misalignment but the aperture will be limited. If we manage to position filters with a precision better than 1 µm, we can use filters with metallisation of 0.5 µm. Metallisation of 0.5, 1, 1.5, 2, 2.5, 3 or 4 µm are available. All these values allow finding a trade-off between aperture limitation and cross-talk. Status: Public 12/16

13 Figure n 9: Metallic mask to avoid optical cross-talk 4.3 SiO 2 / TiO 2 etching SiO 2 is currently used as a hard mask or passivation layer. The plasma solutions to etch SiO 2 are well known. On the contrary, TiO 2 is rarely used and few references exist 1. We use a Chloride based etching process. Etching selectivity between SiO 2 / TiO 2 and spin coating resist is around 1. As the maximum dielectric filter s thickness is around 1.6 µm, a 2 µm spin coating is compatible with the etching process. 4.4 Reporting of backside The 2widesense VGA FPA is shown figure n 10 from backside. We observe some residual InP from substrate covering the ohmic contact in some places. The reason is the protection of edges for wet etching. This drawback could prevent a good contact between filter and photodiodes arrays. A particular attention will be paid in order to avoid it in deliverable D K. Baik et al. J. Appl. Phys. 108, (2010) Status: Public 13/16

14 Figure n 10: 2Widesense FPA after substrate thinning revealing InP residue 5. Toward low cost deposition process A low cost deposition process is not available as long as 2D arrays of photodiodes are individually hybridized to ROIC. The question of low cost filter is closely related to collective hybridization before ROIC wafer dicing. Hybridization schemes compatible to collective hybridization will be detailed in D2.6. It consists roughly in: - Low temperature indium bumps with pick and place before reflow - Microtubes compatible to smaller pitch but will a large number of steps - Molecular bonding - 3D hybridisation across ROIC substrate Another element required for low cost fabrication is the use of dielectric filters or doped / coloured polymers. Coloured polymers are extensively used with RGB filters. Nevertheless the question of polymer transparency in SWIR is still open, and specific developments are required. Dielectric filters are currently deposited on 8 inch wafers and can still be employed for low cost devices. Status: Public 14/16

15 6. Conclusion We have chosen to use a separate 2D array of filters fabrication before report by flip chip on individual VisSWIR FPA. A process on backside FPA would have been too much risky, and it would have altered the yield for the final deliverable D2.9. The simulated transmission filters is very close to the experimental ones. The report of filters arrays is challenging in terms of alignment. Nevertheless, the first tests have shown that our procedure is adapted to filter deposition. Given the accuracy of experimental transmission spectra versus simulation, we decide to deposit the last two filters F3 and F4 directly on the processed substrate. The first deposition tests show that we can be confident in the realization of deliverables D2.9, i.e. the wide dynamics full spectrum filtered FPA modules to be delivered March This version of deliverable D2.2 will be completed once filters fabrication would be achieved, i.e. just before D2.9 delivery in March Status: Public 15/16

16 LIST OF ACRONYMS 2D FPA HBA InGaAs NIR RCM ROIC SWIR VGA VisSWIR 2 dimension Focal Plane Array High Beam assistant Indium Galium Arsenide Near Infrared Road Condition Monitoring Readout Circuit Short Wave Infra-Red Video Graphics Array Visible Short Wave Infra-Red Status: Public 16/16