Interference metal/dielectric filters integrated on CMOS image sensors SEMICON Europa, 7-8 October 2014

Transcription

1 Interference metal/dielectric filters integrated on CMOS image sensors SEMICON Europa, 7-8 October 2014

2 Outline Spectral filtering applications Consumer Multispectral Prior art Organic resists and external filters Fully on-chip filters: Fabry-Perot, plasmonic Metal/Dielectric Fabry-Perot filters Advantages of MD filters Determination of layer optical constants Filter patterning process for staircase architecture Demonstrations on CMOS with Ag and Cu filters Performance optimization and robustness evaluation Nanostructured vs staircase MD filters

3 Outline Spectral filtering applications Consumer Multispectral Prior art Organic resists and external filters Fully on-chip filters: Fabry-Perot, plasmonic Metal/Dielectric Fabry-Perot filters Advantages of MD filters Determination of layer optical constants Filter patterning process for staircase architecture Demonstrations on CMOS with Ag and Cu filters Performance optimization and robustness evaluation Nanostructured vs staircase MD filters

4 Spectral filtering for consumer applications RGB and RGB+ Traditional color imaging Image dehazing Skin smoothing Shadow detection Object segmentation High sensitivity RGB C. Fredembach et al, Color Imaging Conference 2008 IR Ambient Light Sensing (ALS) and Active 3D Imaging Biometric authentification Wireless: screen on/off, gesture interfaces Cleaning robots: collision avoidance Domestic appliances: liquid monitoring Industrial machines: object detection Automotive: user interfaces, cabin occupancy 2-4 different filters per chip Courtesy of STMicroelectronics

5 Spectral filtering in multispectral applications Precision Agriculture & Food Aerial and ground based equipments for monitoring against drought, disease, nutrient stress, etc. Determination of quality attributes and ripeness stage in fruits and vegetables Biomedical Non invasive diagnostics based on fluorescence or label-free tissue imaging: erythema, wound or burn, vascular imaging and blood oxygenation analysis Clinical ICG applications: visualisation of blood flow in the retina Machine Vision, Process Manufacturing Real-time automated quality control Sorting of raw materials, finished goods, waste Defense & Security Discrimination between targets and decoys, defeating camouflage, forensic evidence, surveillance camera Requires 4 different filters per chip

6 Outline Spectral filtering applications Consumer Multispectral Prior art Organic resists and external filters Fully on-chip filters: Fabry-Perot, plasmonic Metal/Dielectric Fabry-Perot filters Advantages of MD filters Determination of layer optical constants Filter patterning process for staircase architecture Demonstrations on CMOS with Ag and Cu filters Performance optimization and robustness evaluation Nanostructured vs staircase MD filters

7 Organic resists and external filters Each filtering function requires both On-chip organic resists Interference filters on external glass substrate RGB or Ambient Light Sensing (ALS) IR-pass For multiple filters, patterned external filter is required

8 Fully on-chip filters IMEC XFab - All-dielectric Fabry-Perot: staircase - Multiple masks, partial etching - Thick stack > 1µm - Limited spectral range: R & NIR - Very high transmission and rejection - All-dielectric Fabry-Perot: nanostructured - One mask - Thick stack > 1µm - Limited spectral range: R & NIR - Very high transmission and rejection Glasgow Univ. - Metallic nanostructured filters (plasmonic) - One mask - Thin: typ 0,15µm - Limited spectral range: e.g. visible - Low transmission and rejection

9 Target of present study New technology of on-chip filters Thin: <1µm Extended spectral range: VIS + NIR Integrable on CMOS and robust to process errors Performance optimized on criteria defined by STMicroelectronics

10 Outline Spectral filtering applications Consumer Multispectral Prior art Organic resists and external filters Fully on-chip filters: Fabry-Perot, plasmonic Metal/Dielectric Fabry-Perot filters Advantages of MD filters Determination of layer optical constants Filter patterning process for staircase architecture Demonstrations on CMOS with Ag and Cu filters Performance optimization and robustness evaluation Nanostructured vs staircase MD filters

11 MD on-chip filters Thin filter stack (typ. 400nm) suitable for small pixels Extended spectral range with silver: nm High transmission, very high rejection Dielectric Metal B G R IR 2 Fabry-Perot cavities for enhanced rejection

12 Material choice Two developed technologies n(cu) k(ag) n(ag) k(cu) Ag filters Optical constants of bulk Ag and Cu (from Palik) Extended spectral range VIS + IR due to high k/n ratio of Ag Requires dedicated tools in clean room due to Ag contamination, possible in back-end with no subsequent critical step Cu filters Limited spectral range (G) + R + IR due to lower k/n ratio below 550nm Immediately accessible in clean room

13 Characterization of material optical constants Methodology Identify a set of appropriate filters Simultaneous fitting of RT spectral data Extract n&k optical constants of materials within deposited filters n(ag) k(ag) Fitted vs measured R and T of filters for n&k extraction. Fits were simultaneous for stacks with 17nm Ag (a-e), and for stacks with 40nm Ag layers (f-h), respectively

14 Non patterned filters: measurements vs design Ag diel Measurements vs designs with n&k from single layer characterization Measurements vs designs with n&k from multi-filter characterization Selected technique Applied Optics 53, 1663 (2014)

15 Filter patterning process SiN AlN Ag l 1 l 2 l 3 l 4 No partial etching: excellent control of filter peak wavelength throughout the whole wafer 2 dielectric materials: AlN for encapsulation of Ag and etching stop, SiN to form the staircase No patterning of metal 2 x (N-1) lithography/etching steps for 2 FP cavities and N filters Enables to choose any arbitrary set of filter wavelengths

16 Patterned Ag/diel filters on CMOS Leti / STMicroelectronics demonstration of RGB filters on FSI CMOS wafers with 1,75µm pixels No IR-blocking external filter SEM top view of Ag/diel patterned filters First color image with Ag/diel RGB filters without IR-blocking filter Applied Optics 19, (2011) Measured QE Absolute values are not provided due to absence of micro-lenses for this first demonstration

17 Cu/diel filters on CMOS Cu is suitable for filtering above 550nm Cu is commonly used in clean room First demonstration of ALS and NIR-bandpass Cu/SiN interference filters on separate wafers Low-H SiN developed for enhanced adherence with Cu ALS IR Absolute values are not provided due to absenced of micro-lensesfor this first demonstration Ongoing developments for patterned ALS + NIR filters on the same chip

18 ALS design optimization ALS Reference Integration scheme Métal 4 Alu Métal 4 Alu Métal 3 2 OH OH 2 Métal 3 Métal 2 Métal 1 1 H SiO2 H2O H2 PMDH H2O H Photodiode O2H H Photodiode 2H PMD H O2H Métal 2 Métal 1 1 H 2 SiO Specifications Minimize ALS Error (stability vs illuminant) Minimize ALS Dark (inverse of sensitivity) Includes STMicro customer black window An integration scheme Strategy Optimize nominal performances Evaluate dispersion under process errors + incidence angle Applied Optics 53, 4493 (2014) Performances of optimized stack designs out of spec in spec

19 ALS under process dispersion Reference (green resist + external filter) Cu filter Simulation of process errors on layer thickness and refractive index, with gaussian statistics (±3σ) Dispersion data on layer parameters s Cu =5%, s diel = 2,7%, s n = 1,6% s Cu =3,75%, s diel = 2%, s n = 0,16% s Cu =2,5%, s diel = 1,35%, s n = 0,13% out of spec 60 in spec 30 nominal performance of Cu filter under θ variations 0 15 nominal performance of reference ALS performances under process dispersion are in specifications for angles of incidence up to 30

20 IR band pass design optimization IR band pass Reference 2 Integration schemes [A] Métal 4 Alu Métal 4 Alu Métal 4 Alu Métal 4 Alu [C] [D] Métal 3 Métal 2 Métal 1 1 H2O H SiO2 2 H2 PMDH OH H2O H Photodiode 2 2 Métal 3 OH Métal 2 O2H Métal 1 1 H2O H H SiO2 Photodiode H2 2H PMDH PMD H OH Métal 3 H2O H Métal 2 O2H Métal 1 1 H Photodiode 2 SiO OH O2H H Photodiode 2 2H PMD H O2H Métal 3 Métal 2 Métal 1 1 H 2 SiO [B] Specifications 4 criteria to maximize transmission and rejection Includes STMicro black resist + black window 2 integration schemes Strategy Optimize nominal performances Evaluate dispersion under process errors Performances of optimized stack designs out of spec in spec out of spec in spec

21 IR band pass under process dispersion Reference (black resist + external filter) Cu filter Dispersion data on layer parameters s Cu =5%, s diel = 2,7%, s n = 1,6% s Cu =3,75%, s diel = 2%, s n = 0,16% s Cu =2,5%, s diel = 1,35%, s n = 0,13% In yellow: same dispersion as for red cloud, but s diel reduced to 0,5% for two CMOS passivation layers only Ref filter nominal performances Cu filter nominal performances in spec out of spec out of spec in spec Dispersion of performances is a critical issue for IR-BP Cu filter Lower dispersion with tight control of two layer only

22 Nanostructured vs staircase filters Simpler technology: 1 single lithography/etching step per FP cavity for N filters Flat surface: easier to form microlenses Best choice for multi- and hyperspectral applications with > 4 filters Staircase filters Nanostructured filters C u SiN SiO 2 First demonstration with Cu and SiN nanostructures embedded in SiO2 Design of Ag/nanostructured diel filters with extended spectral range

23 Summary FP filter advantages On-chip solution for RGB+ and multispectral Easily adjustable spectral responses High transmission and rejection MD FP filter advantages Low filter thickness, suitable for small pixels Extended spectral range with Ag Few technological steps with nanostructured version This work has been carried out in a collaboration with STMicroelectronics