sensors

Article Second Generation Small Technology Using Hybrid Bond Stacking †

Vincent C. Venezia *, Alan Chih-Wei Hsiung, Wu-Zang Yang, Yuying Zhang, Cheng Zhao, Zhiqiang Lin and Lindsay A. Grant OmniVision Technologies, Inc., Santa Clara, CA 95054, USA; [email protected] (A.C.-W.H.); [email protected] (W.-Z.Y.); [email protected] (Y.Z.); [email protected] (C.Z.); [email protected] (Z.L.); [email protected] (L.A.G.) * Correspondence: [email protected]; Tel.: +1-408-567-3000 † This paper is an extended version of our paper published in Venezia, V.C.; Shih, C.; Yang, W.-Z.; Zang, Y.; Lin, Z.; Grant, L.A.; Rhodes, H. 1.0 µm pixel improvements with hybrid bond stacking technology. In Proceedings of the International Workshop, Hiroshima, Japan, 30 May–2 June 2017; pp. 8–11.

Received: 2 November 2017; Accepted: 13 February 2018; Published: 24 February 2018

Abstract: In this work, OmniVision’s second generation (Gen2) of small-pixel BSI stacking technologies is reviewed. The key features of this technology are hybrid-bond stacking, deeper back-side, deep-trench isolation, new back-side composite metal-oxide grid, and improved gate oxide quality. This Gen2 technology achieves state-of-the-art low-light image-sensor performance for 1.1, 1.0, and 0.9 µm pixel products. Additional improvements on this technology include less than 100 ppm white-pixel process and a high near- (NIR) QE technology.

Keywords: CIS; BSI; stacked; hybrid-bond; 1.0 µm; 0.9 µm; NIR

1. Introduction The first generation of back-side illuminated (BSI) stacked products used oxide bonding to physically bond wafers, and through-silicon vias (TSV) to form electrical connections [1,2]. In addition to the chip size reduction offered by wafer stacking, improvements in photo-diode performance are possible with “sensor only”-type processing that would negatively impact ASIC performance on non-stack BSI or even front-side illuminated sensors, such as new materials or thermal conditions [1]. In our second generation of stacking technologies, wafers are connected by the top-metal layer and the inter-layer dielectrics of the sensor and ASIC wafer, forming a hybrid oxide-oxide and metal-metal bond that physically and electrically connects the two wafers simultaneously. Hybrid bonding offers significant improvements in design and layout flexibility over the TSV stacking which consumes chip space to create deep vias and are aspect-ratio pitch limited; for examples, see Figures1 and2. While TSVs can only be placed outside the array, hybrid-bond connections can be made both outside (Figure2a) and inside the array (Figure2b). In addition, hybrid bonding connections do not interfere with sensor or circuit performance, as is the case with the TSV Keep-Out Zone (KOZ) [3]. Using the flexibility hybrid bonding offers, the overall chip size of OmniVision’s 1 um, 16 MP stacked-chip product was reduced by ~10% [4]. Back-side, deep-trench isolation (BS-DTI), as well as a buried color filter array (CFA), were implemented in OmniVision’s first generation of stacking products to improve image quality [1]; BS-DTI to control electrical and optical crosstalk, and the buried CFA to reduce the optical stack height. Figure3 is a cross section diagram comparing the Gen1 and Gen2 structures. In the Gen2 technology, the silicon thickness and the BS-DTI depth are increased to improve sensitivity while controlling crosstalk. A new composite oxide-metal grid (CMG) is introduced, in which an additional oxide-grid structure is added above the metal grid, separating neighboring color filters. As will be shown below,

Sensors 2018, 18, 667; doi:10.3390/s18020667 www.mdpi.com/journal/sensors Sensors 2018, 18, x 2 of 8

Sensors 2018, 18, 667 2 of 8

the Gen 2 pixel structure and process improves sensitivity without degrading crosstalk to achieve

Sensorsexcellent 2018, low-light 18, x performance. 2 of 8 Sensors 2018, 18, x Figure 1. First-generation BSI-CIS stacking with oxide-oxide bonding and TSV. 2 of 8

FigureFigure 1. 1. First-generationFirst-generation BSI-CIS BSI-CIS stacking stacking with with oxide-oxide oxide-oxide bonding bonding and and TSV. TSV. Figure 2. Hybrid bond-stacking schematic, showing the Cu-Cu interconnects outside the array (a) and within the array (b).

Back-side, deep-trench isolation (BS-DTI), as well as a buried (CFA), were implemented in OmniVision’s first generation of stacking products to improve image quality [1]; BS-DTI to control electrical and optical crosstalk, and the buried CFA to reduce the optical stack height. Figure 3 is a cross section diagram comparing the Gen1 and Gen2 structures. In the Gen2 technology, the silicon thickness and the BS-DTI depth are increased to improve sensitivity while controlling crosstalk. A new composite oxide-metal grid (CMG) is introduced, in which an

additional oxide-grid structure is added above the metal grid, separating neighboring color filters. Figure 2. Hybrid bond-stacking schematic, showing the Cu-Cu interconnects outside the array (a) As willFigure be 2.shownHybrid below, bond-stacking the Gen schematic, 2 pixel showing structure the Cu-Cu and interconnectsprocess improves outside thesensitivity array (a) andwithout and within the array (b). degradingwithin crosstalk the array (tob). achieve excellent low-light performance. Back-side, deep-trench isolation (BS-DTI), as well as a buried color filter array (CFA), were implemented in OmniVision’s first generation of stacking products to improve image quality [1]; BS-DTI to control electrical and optical crosstalk, and the buried CFA to reduce the optical stack height. Figure 3 is a cross section diagram comparing the Gen1 and Gen2 structures. In the Gen2 technology, the silicon thickness and the BS-DTI depth are increased to improve sensitivity while controlling crosstalk. A new composite oxide-metal grid (CMG) is introduced, in which an additional oxide-grid structure is added above the metal grid, separating neighboring color filters. As will be shown below, the Gen 2 pixel structure and process improves sensitivity without degrading crosstalk to achieve excellent low-light performance.

Figure 3. Gen1 (a) and Gen2 (b) schematic; Gen2 BSI stack has thicker silicon, narrower and deeper Figure 3. Gen1 (a) and Gen2 (b) schematic; Gen2 BSI stack has thicker silicon, narrower and deeper BS-DTI, and a composite metal-oxide, back-side grid. BS-DTI, and a composite metal-oxide, back-side grid.

2. Results and Discussion 2. Results and Discussion The quantum efficiency (QE) from 1 um, 16 MP Gen1 and Gen2 products is compared in The quantum efficiency (QE) from 1 um, 16 MP Gen1 and Gen2 products is compared in Figure4. Figure 4. The peak green and red QE are increased 10% and 15%, respectively, while the NIR QE The peak green and red QE are increased 10% and 15%, respectively, while the NIR QE (850 nm) is (850 nm) is increased by ~50%. The Gen2 angular response improvement was reported in [4]. increased by ~50%. The Gen2 angular response improvement was reported in [4]. Figure5a compares Figure 5a compares the Gen1 and Gen2 average crosstalk along the chief ray angle (CRA) of the the Gen1 and Gen2 average crosstalk along the chief ray angle (CRA) of the sensor; crosstalk is sensor; crosstalk is calculated from QE measurements along a diagonal line from array center to calculated from QE measurements along a diagonal line from array center to edge, where the angle Figure 3. Gen1 (a) and Gen2 (b) schematic; Gen2 BSI stack has thicker silicon, narrower and deeper is determinedFigure 3. Gen1 by the (a) and CRA Gen2 of the (b) modules schematic; lens Gen2 (Figure BSI stack5b). has The thicker crosstalk silicon, increases narrower at largeand deeper angles on BS-DTI, and a composite metal-oxide, back-side grid. a Gen1 sensor, while it remains relatively constant on a Gen2 stacked sensor. Figure6 compares the QE from a 1.0 µm pixel CIS product to that from a 0.9 µm pixel CIS product, both using the Gen2 stacking 2. Results and Discussion technology described above. Despite the decrease in pixel size, the 0.9 µm pixel has the same high QE The quantum efficiency (QE) from 1 um, 16 MP Gen1 and Gen2 products is compared in Figure 4. The peak green and red QE are increased 10% and 15%, respectively, while the NIR QE (850 nm) is increased by ~50%. The Gen2 angular response improvement was reported in [4]. Figure 5a compares the Gen1 and Gen2 average crosstalk along the chief ray angle (CRA) of the sensor; crosstalk is calculated from QE measurements along a diagonal line from array center to

Sensors 2018, 18, x 3 of 8 Sensors 2018, 18, x 3 of 8 edge, where the angle is determined by the CRA of the modules lens (Figure 5b). The crosstalk edge, where the angle is determined by the CRA of the modules lens (Figure 5b). The crosstalk increasesedge, where at large the angleangles is on determined a Gen1 sensor, by the while CRA it ofremains the modules relatively lens constant (Figure on5b). a Gen2The crosstalk stacked Sensorsincreases2018 ,at18 ,large 667 angles on a Gen1 sensor, while it remains relatively constant on a Gen2 stacked3 of 8 sensor.increases Figure at large 6 compares angles on the a Gen1QE from sensor, a 1.0 while µm pi itxel remains CIS product relatively to that constant from aon 0.9 a Gen2µm pixel stacked CIS sensor. Figure 6 compares the QE from a 1.0 µm pixel CIS product to that from a 0.9 µm pixel CIS product,sensor. Figure both using6 compares the Gen2 the QEstacking from technologya 1.0 µm pi xeldescribed CIS product above. to Despite that from the a decrease0.9 µm pixel in pixel CIS product, both using the Gen2 stacking technology described above. Despite the decrease in pixel size,andproduct, lowthe 0.9 crosstalkboth µm using pixel as the hasthe 1.0 theGen2µ samem pixel,stacking high demonstrating QE technology and low crosstalk thatdescribed the Gen2 as above. the optical 1.0 Despiteµm benefits pixel, the demonstrating can decrease be extended in pixel that to size, the 0.9 µm pixel has the same high QE and low crosstalk as the 1.0 µm pixel, demonstrating that thesubsize, Gen2 1 theµm 0.9 optical pixel µm pitches.pixel benefits has canthe samebe ex tendedhigh QE to and sub low 1 µm crosstalk pixel pitches. as the 1.0 µm pixel, demonstrating that the Gen2 optical benefits can be extended to sub 1 µm pixel pitches.

FigureFigure 4. QEQE comparing comparing Gen1 Gen1 and and Gen2 1 µm,µm, 16 MP technologies. Figure 4. QE comparing Gen1 and Gen2 1 µm, 16 MP technologies.

Figure 5. (a) Crosstalk vs angle, calculated from QE curves measured in a fix ROI along a diagonal as Figure 5. (a) Crosstalk vs angle, calculated from QE curv curveses measured in a fix fix ROI along a diagonal as shownFigure in5. (thea) Crosstalk (b). Angles vs areangle, the calculatedCRA of the from module QE curvlens esat measuredeach ROI. in a fix ROI along a diagonal as shown in the (b). Angles are the CRA of the module lens at each ROI.

Figure 6. The QE from 1 µm and 0.9 µm pixel products, both using the Gen2 BSI-stacking technology. Figure 6. The QE from 1 µm and 0.9 µm pixel products, both using the Gen2 BSI-stacking technology. FigureFigure 6. 6. TheThe QE QE from from 1 1 µmµm and and 0.9 0.9 µmµm pixel products, both using the Gen2 BSI-stacking technology.

Sensors 2018, 18, 667x 4 of 8

Finite-difference time-domain (FDTD) simulations were used to understand the Gen2 improvementFinite-difference in optical time-domain performance. (FDTD)Figure 7 show simulationss the simulation were used results to comparing understand the theGen2 Gen2 and Gen1improvement optical stacks. in optical In performance.the simulation, Figure a 6307 showsnm, monochromatic the simulation plane results wave comparing (Transverse the Gen2 electric and modeGen1 optical(TM)) is stacks. used Into illuminate the simulation, neighboring a 630 nm, monochromatic in each structure; plane wave the left (Transverse pixel has electrica green modecolor (TM))filter while is used the to right illuminate pixel neighboringhas a red color pixels filter. in each The structure; only difference the left between pixel has the a green pixels color in filtereach whilesimulation the right is the pixel color has afilter red colorindex filter. of refracti The onlyon. differenceThe figure between plots the pixelselectric in field each simulationpropagation is throughthe color the filter micro index lens of refraction. (ML) and Thecolor figure filter. plots The theGen1 electric and Gen2 field propagationstructures are through outlined the by micro dashed lens (ML)lines andand colorlabeled filter. for The clarity. Gen1 The and Gen2main structuresdifference are between outlined structures by dashed is lines the andextra labeled silicon for dioxide clarity. Thebarrier main between difference the betweencolor filters structures in the isGen2 the extra stack. silicon The silicon-dioxide dioxide barrier betweenrefractive the index color (~1.45) filters inis lowerthe Gen2 than stack. that Theof the silicon-dioxide CF index (~1.6–2.0), refractive resulting index (~1.45) in light is reflection lower than at thatthe CF/oxide of the CF indexinterface. (~1.6–2.0), In the caseresulting of Gen1, in light partially reflection filtered at the light CF/oxide can pass interface. between In color the case filters, of Gen1, resulting partially in signal filtered loss light due can to passoptical between crosstalk. color The filters, crosstalk resulting difference in signal is seen loss in due the to Figure optical 7 crosstalk.simulation The results; crosstalk comparing difference the circledis seen areas, in the the Figure electric7 simulation field propagation results; comparing between color the circled filters is areas, eliminated the electric by the field Gen2 propagation CMG. The betterbetween light color confinement filters is eliminated in the CF by by the the Gen2 CMG CMG. results The in better higher light QE confinementand lower optical in the CFcrosstalk, by the whichCMG results is consistent in higher with QE the and QE lower data opticalin Figure crosstalk, 4. which is consistent with the QE data in Figure4.

Figure 7. FDTDFDTD simulation simulation comparing comparing the the light light confinement confinement of of Gen2 Gen2 ( (leftleft)) and and Gen1 Gen1 ( (rightright).). Simulations used a monochromatic plan wave at, 630 nm TE mode, incident on the Green and Red pixel of each structure. Structure features are labeledlabeled andand highlighted.highlighted.

In addition to the optical performance benefits, the Gen2 technology noise, white pixel (WP) In addition to the optical performance benefits, the Gen2 technology noise, white pixel (WP) and and dark current performance has been improved over the Gen1, see Table 1. Figure 8 is a plot of the dark current performance has been improved over the Gen1, see Table1. Figure8 is a plot of the noise noise histogram from a two-frame difference image showing a ~2x improvement in the RTS noise. histogram from a two-frame difference image showing a ~2x improvement in the RTS noise. The low The low level of noise was achieved by maximizing the source follow area [5] and reducing source level of noise was achieved by maximizing the source follow area [5] and reducing source follower follower traps by minimizing etching damage, reducing dangling bonds, and by device channel traps by minimizing etching damage, reducing dangling bonds, and by device channel engineering as engineering as discussed in [6]. The WP defect improvement is shown in Figure 9. A greater than 2x discussed in [6]. The WP defect improvement is shown in Figure9. A greater than 2x reduction in WP reduction in WP is achieved by engineering the photo-diode implant process to reduce any high is achieved by engineering the photo-diode implant process to reduce any high electricity field regions electricity field regions and suppress the photo-diode (PD) depletion regions from interacting with and suppress the photo-diode (PD) depletion regions from interacting with high defect surfaces, high defect surfaces, such as front, back, and BS-DTI surfaces. The back surface passivation, for both such as front, back, and BS-DTI surfaces. The back surface passivation, for both Gen1 and Gen2, Gen1 and Gen2, is achieved by deposition of a negatively charged oxide film [7]. In the Gen2 is achieved by deposition of a negatively charged oxide film [7]. In the Gen2 technology, the negative technology, the negative charge density in this film is increased, resulting in an improved dark charge density in this film is increased, resulting in an improved dark current, as shown in Table1. current, as shown in Table 1.

Sensors 2018, 18, x 5 of 8 Sensors 2018, 18, 667 5 of 8 Table 1. Pixel performance comparing Gen1 and Gen2 BSI-stacking technologies (1 µm, 16 MP products).

Table 1. Pixel performance comparingParameter Gen1 and Gen2 BSI-stacking Units technologies Gen1 (1 Gen2µm, 16 MP products). Array 16 MP 16 MP Parameter Units Gen1 Gen2 Pixel size µm 1.0 1.0 Full Well CapacityArray (ADC range) e− 165000 MP 166000 MP Pixel size µm 1.0 1.0 Sens-G (530 nm) e−/(Lux × s) 3150 3600 Full Well Capacity (ADC range) e− 5000 6000 Sens-GPRNU (530 (average) nm) e−/(Lux × % s) 3150 0.8 3600 0.8 PRNUSNR10 (average) (ΔE = 2.5) % Lux 0.8 90 0.8 80 DarkSNR10 current (∆E = 2.5)(T = 60 C) Lux e−/s 904 802 − Dark currentBlooming (T = 60 C) e /s% 40% 0% 2 Blooming % 0% 0% FPNFPN (RT) (RT) [e] [e] 0.5 0.5 0.2 ReadRead noise noise (16x (16x gain) gain) [e] [e] 2.0 2.0 1.4 RTSRTS (>1 (>1 mV) mV) ppm ppm 500 500 200

2-frame difference distribution (PPM) 1000000

100000

10000 Gen1 1000 PPM Gen2 100

10

1 -40 -30 -20 -10 0 10 20 30 40 [e], 16xgain

FigureFigure 8.8. Two-frameTwo-frame differencedifference image,image, noisenoise histogramhistogram comparingcomparing thethe 11 µµm,m, 1616 MPMP imagesimages sensorssensors thatthat useuse thethe Gen1Gen1 andand Gen2Gen2 technology.technology.

FigureFigure 9.9.Plot Plot of 1-cumulativeof 1-cumulative density density function function (1-CDF) (1-CDF) comparing comparing the dark imagethe dark from image low white-pixel, from low PDwhite-pixel, process. 1PDµm, process. 16 MP 1 (2 µm, fps, 16 60 MP C). (2 fps, 60 C).

Sensors 2018, 18, 667x 6 of 8

The improved angular performance, as discussed in [4] and shown in Figure 5, is due to the BS-DTI.The A improved relatively angular shallow performance, BS-DTI was as discussedintroduced in [in4] andthe Gen1 shown technology in Figure5, is[1] due and to theresulted BS-DTI. in Aimprovement relatively shallow in the BS-DTI crosstalk was versus introduced angle in perfor the Gen1mance. technology This benefit [1] and was resulted extended in improvement in Gen2 by increasingin the crosstalk the DTI versus depth angle by ~3x, performance. resulting in This the elim benefitination was of extended the crosstalk in Gen2 at high by increasingangles (Figure the DTI5). depthIt byis well ~3x, known resulting that in theincreasing elimination the silicon of the crosstalkthickness at increases high angles the (Figurelong-wavelength5). QE [8]. As discussedIt is wellabove, known the thatGen2 increasing thickness theis increased, silicon thickness thereby increases increasing the the long-wavelength NIR QE. Unlike QE the [8]. AsTSV-bonding, discussed above,which theis limited Gen2 thickness by the etching is increased, and filling thereby of increasing deep trenches, the NIR the QE. hybrid Unlike bond the TSV-bonding,technology is more which flexible is limited for byincreasing the etching the andfinal filling silicon of thickness. deep trenches, In addition the hybrid to increasing bond technology the BSI thickness,is more flexible we improve for increasing the NIR the finalQE with silicon a thickness.new technology In addition [9] to to achieve increasing state-of-the-art the BSI thickness, NIR weperformance improve thefor NIR1 µm QE pixels. with Figure a new 10 technology shows the [QE9] to curve achieve from state-of-the-art a 1 µm pixel NIR NIR performance. performance The for 1QEµm at pixels.850 nm Figure and 940 10 showsnm increases the QE curveto 40% from and a20% 1 µm respectively, pixel NIR performance. without any Thesignificant QE at 850 impact nm and on the940 visible nm increases QE. to 40% and 20% respectively, without any significant impact on the visible QE.

90%

80%

70%

60% Gen2+NIR 50%

QE 40%

30%

20% 10% Gen2 0% 380 580 780 980 Wavelength [nm] Figure 10. QE comparison from a 1 µm, 16MP Gen2 image sensor to one with an additional process Figure 10. QE comparison from a 1 µm, 16MP Gen2 image sensor to one with an additional process for forNIR NIR enhancement. enhancement.

3. Conclusions 3. Conclusions The improvements in the Gen2 over the Gen1 technology are highlighted in Table 1. The higher The improvements in the Gen2 over the Gen1 technology are highlighted in Table1. The higher sensitivity and better SNR10 [10] is achieved by increasing the BSI silicon thickness, extending the sensitivity and better SNR10 [10] is achieved by increasing the BSI silicon thickness, extending the BS-DTI deeper, and introducing the CMG. These optical benefits are maintained at sub 1 µm pixels, BS-DTI deeper, and introducing the CMG. These optical benefits are maintained at sub 1 µm pixels, such as 0.9 µm. The lower noise is due to an improved source follower design and process, while the such as 0.9 µm. The lower noise is due to an improved source follower design and process, while the dark current improvement is achieved by photo-diode and surface passivation engineering. The dark current improvement is achieved by photo-diode and surface passivation engineering. The higher higher QE and low crosstalk, as well as the improved noise, results in excellent low-light QE and low crosstalk, as well as the improved noise, results in excellent low-light performance, as performance, as is shown in Figure 11, where Macbeth chart images taken by a Gen2 and Gen1 CIS is shown in Figure 11, where Macbeth chart images taken by a Gen2 and Gen1 CIS sensor at 5 Lux sensor at 5 Lux are overlaid for comparison. Figure 11b plots the simulated and measured are overlaid for comparison. Figure 11b plots the simulated and measured improvement in SNR vs improvement in SNR vs lux level for a 1 µm Gen2 CIS product. The SNR improvement is most lux level for a 1 µm Gen2 CIS product. The SNR improvement is most significant at low lux levels. significant at low lux levels. The Gen2 technology is currently in mass production for 1.1 µm, 1.0 µm, The Gen2 technology is currently in mass production for 1.1 µm, 1.0 µm, and 0.9 µm pixel CIS products. and 0.9 µm pixel CIS products.

Sensors 2018, 18, 667 7 of 8 Sensors 2018, 18, x 7 of 8

FigureFigure 11. 11. (a(a) )Combined Combined Macbeth Macbeth chart chart images images from from 16 16 MP MP,, 1 1 µmµm pixel pixel Gen1 Gen1 and and Gen2 Gen2 products products to to showshow the the Gen2 Gen2 low low light light (5 (5 Lux) Lux) sensitivity sensitivity improvement. improvement. ( (bb)) Gen2 Gen2 SNR SNR improvement improvement vs vs. lux lux level, level, includingincluding measured measured data data and and simulated simulated results results (F2.0, (F2.0, 15 15 fps). fps).

Acknowledgements: The authors would like to acknowledge the contributions of Howard Rhodes to this work. HowardAcknowledgments: played a majorThe role authors in OmniVision’s would like to pixel acknowledge technology the for contributions many years, of including Howard the Rhodes development to this work. of theHoward 1µm and played sub a1µm major pixels role discussed in OmniVision’s in this pixelwork. technology for many years, including the development of the 1µm and sub 1µm pixels discussed in this work. Author Contributions: Vincent C. Venezia designed the experiments, analyzed data and wrote the Author Contributions: Vincent C. Venezia designed the experiments, analyzed data and wrote the manuscript. manuscript.Alan Chih-Wei Alan Hsiung Chih-Wei and Hsiung Wu-Zang and Yang WuZang developed Yang processdeveloped modules process and modules integrated and into integrated the process into flow.the processYuying flow. Zhang Yuying and Zhiqiang Zhang and Lin performedZhiqiang Lin pixel performe characterization.d pixel characterization. Cheng Zhao simulated Cheng Zhao the structuresimulated using the structureFDTD. Lindsay using FDTD. A. Grant Lindsay oversaw A. Grant the project oversaw and the provided project technical and provided guidance. technical guidance.

ConflictsConflicts of of Interest: Interest: TheThe authors authors declare declare no no conflict conflict of of interest. interest.

ReferencesReferences

1.1. Venezia,Venezia, V.C.; V.C.; Shih, Shih, C.; C.; Yang, Yang, W.Z.; W.-Z.; Zhang, Zhang, B.; B.;Rhodes, Rhodes, H. H.Stacked Stacked Chip Chip Technology: Technology: A New A New Direction Direction for CMOSfor CMOS Imagers. Imagers. In Proceedings In Proceedings of the of theInternational International Image Image Sensor Sensor Work Workshop,shop, Vaals, Vaals, The The Netherlands, Netherlands, 8–118–11 June June 2015; 2015; pp. pp. 36–39. 36–39. 2.2. Sukegawa,Sukegawa, S.; S.; Umebayashi, Umebayashi, T.; T.; Nakajima, Nakajima, T.; T.; Kawanobe, Kawanobe, H.; H.; Koseki, Koseki, K.; K.; Hirota, Hirota, I.; I.; Haruta, Haruta, T.; T.; Kasai, Kasai, M.; M.; 1 Fukumoto,Fukumoto, K.; K.; Wakano, Wakano, T.; T.; et etal. al.A A¼ inch4 inch 8M 8M pixel pixel back-illuminated back-illuminated stacked stacked CMOS CMOS image image sensor. sensor. In ProceedingsIn Proceedings of the of theIEEE IEEE International International Solid-State Solid-State Circ Circuitsuits Conference Conference Digest Digest of ofTechnical Technical Papers Papers (ISSCC), (ISSCC), SanSan Francisco, Francisco, CA, CA, USA, USA, 17–21 17–21 February February 2013. 2013. 3.3. Chen,Chen, S.-H.; S.-H.; Thijs, Thijs, S.; S.; Linten, Linten, D.; D.; Scholz, Scholz, M.; M.; Hellings Hellings,, G.; G.; Groeseneken, Groeseneken, G. G. ESD ESD protection protection devices devices placed placed insideinside keep-out keep-out zone zone (KOZ) (KOZ) of of through through Silicon viavia (TSV)(TSV) inin 3D3D stackedstacked integrated integrated circuits. circuits. In In Proceedings Proceedings of ofthe the Electrical Electrical Overstress/Electrostatic Overstress/Electrostatic Discharge Discharge Symposium Symposium 2012, Tucson, 2012, AZ, Tu USA,cson, 9–14 AZ, September USA, 9–14 2012. 4. SeptemberVenezia, V.C.; 2012. Shih, C.; Yang, W.-Z.; Zang, Y.; Lin, Z.; Grant, L.A.; Rhodes, H. 1.0 µm pixel improvements 4. Venezia,with hybrid V.C.; bond Shih, stacking C.; Yang, technology. W.Z.; Zang, In Y.; Proceedings Lin, Z.; Grant, of the L.A.; 2017 Rhodes, International H. 1.0 µm Image pixel Sensor improvements Workshop, withHiroshima, hybrid bond Japan, stacking 30 May–2 technology. June 2017; In pp. Proceedin 8–11. gs of the 2017 International Image Sensor Workshop, 5. Hiroshima,Abe, K.; Sugawa, Japan, S.;30 Kuroda,May–2 June R.; Watabe, 2017; pp. S.; 8–11. Miyamoto, N.; Teramoto, A.; Ohmi, T.; Kamata, Y.; Shibusawa, K. 5. Abe,Analysis K.; Sugawa, of Source S.; Follower Kuroda, Random R.; Watabe, Telegraph S.; SignalMiyamoto, Using nMOSN.; Teramoto, and pMOS A.; Array Ohmi, TEG. T.;In Kamata, Proceedings Y.; Shibusawa,of the 2007 K. International Analysis of ImageSource Sensor Follower Workshop, Random Ogunquit, Telegraph ME, Signal USA, Using 7–10 nMOS June 2007;and pMOS pp. 62–65. Array TEG. 6. InTakahashi, Proceedings S.; Huang,of the 2007 Y.-M.; Internat Sze, J.-J.;ional Wu, Image T.-T.; Sensor Guo, F.-S.;Workshop, Hsu, W.-C.; Ogunquit, Tseng, ME, T.-H.; USA, Liao, 7–10 K.; June Kuo, 2007; C.-C.; pp.Chen, 62–65. T.-H.; et al. A 45 nm Stacked CMOS Image Sensor Process Technology for Submicron Pixel. Sensors 6. Takahashi,2017, 17, 2816. S.; Huang, [CrossRef Y.-M.;][PubMed Sze, J.-J.;] Wu, T.-T.; Guo, F.-S.; Hsu, W.-C.; Tseng, T.-H.; Liao, K.; Kuo, C.-C.; 7. Chen,Tachibana, T.-H.; T.;et al. Sameshima, A 45 nm Stacked T.; Iwashita, CMOS Y.;Image Kiyota, Sensor Y.; Chikyow,Process Technolo T.; Yoshida,gy for H.; Submicron Arafune, Pixel. K.; Satoh, Sensors S.; 2017Ogura,, 17, 2816. A. Material Research on High-Quality Passivation Layers with Controlled Fixed Charge for 7. Tachibana,Crystalline T.; Silicon Sameshima, Solar Cells. T.; Iwashita,Jpn. J. Appl. Y.; Phys.Kiyota,2011 Y,.;50 Chikyow,, 04DP09. T.; [CrossRef Yoshida,] H.; Arafune, K.; Satoh, S.; Ogura, A. Material Research on High-Quality Passivation Layers with Controlled Fixed Charge for Crystalline Silicon Solar Cells. Jpn. J. Appl. Phys. 2011, 50, 04DP09. 8. Janesick, J.R. Chapter 3.1 QE formulas. In Scientific Charge-Coupled Devices; SPIE Press: Bellingham, WA, USA, 2001; Chatpter 3, pp. 170–178. 9. Available online: http://www.ovt.com/purecel-pixel-tech/nyxel-technology (accessed on 23 October 2017).

Sensors 2018, 18, 667 8 of 8

8. Janesick, J.R. Chapter 3.1 QE formulas. In Scientific Charge-Coupled Devices; SPIE Press: Bellingham, WA, USA, 2001; Chatpter 3; pp. 170–178. 9. Nyxel™ Technology. Available online: http://www.ovt.com/purecel-pixel-tech/nyxel-technology (accessed on 23 October 2017). 10. Alakarhu, J. Image sensors and Image Quality in Mobile Phones. In Proceedings of the International Image Sensor Workshop, Ogunquit, ME, USA, 7–10 June 2007; pp. 1–4.

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).