Montevideo, 27-29 de setiembre de 2006

IBERSENSOR 2006

Voltage controlled color detector with optical readout P. Louro1*, Y. Vygranenko1, J. Martins1, M. Fernandes1, M. Vieira1 1

Electronics Telecommunications and Computer Dept, ISEL, Lisbon, Portugal *Corresponding author: Paula Louro, +351218317287 , [email protected]

Abstract In this work, we report on an amorphous silicon based image sensor with a bias voltage controllable spectral response characteristics. This multilayered device is composed by two stacked p-i-n-i-p structures produced by PE-CVD on a glass substrate and sandwiched between two transparent conductive oxide electrodes with a patterned molybdenum layer between the sensing and switching structures. Optical readout technique is used for image readout. Device performances have been evaluated from the current–voltage characteristics and spectral response measurements performed for the p-i-n-i-p test structures and stacked device. It is demonstrated that the device is sensitive to blue-green or red light depending on polarity of the bias voltage enabling the detection of color images. Device design is discussed and supported by a SPICE model. Keywords: image sensor, spectral response, multilayered structures image sensors based on a-Si:H p-i-n structures using an optical readout technique has been reported in literature [6, 7]. In this work we report an image sensor comprising stacked p-i-n-i-p structures, and discuss the possibility of extracting color images with such device.

1. Introduction Hydrogenated amorphous silicon (a-Si:H) is attractive for imaging/detection applications due to its high absorption efficiency in the visible spectral range and large area processing capability [1]. Besides, the possibility of changing its absorption spectral profile through adding controlled amounts of carbon or germanium during the deposition process resulting in alloys with different band gaps, contributed to the development of a-Si:H based optoelectronics devices [2, 3]. Furthermore, the spectral response of multilayered sensors such as a p-i-n-i-p structures can be controlled by the applied bias [4, 5]. This approach allows to avoid the use of color filters in image sensors, since the red, green and blue components of the incident light can be detected selectively and processed as RGB signal for the reconstruction of color image. Large area

2. Experimental details 2.1 Structure Figure 1 shows the schematic cross section of the color sensor. Device comprises two p-i-n-i-p structures deposited on the glass substrate. Each pi-n-i-p structure consists of two back-to-back diodes composed of a-SiC:H p-layers, a-Si:H ilayers and a common a-SiC:H n-layer.

ITO p type (a-SiC:H)

ITO ITO Switching photodiode

thick

Mo

100 nm

p type (a-SiC:H)

20 nm

i type (a-Si:H)

400 nm

n type (a-SiC:H)

20 nm

i type (a-Si:H)

100 nm

p type (a-SiC:H)

20 nm

switching photodiode

Mo

Mo

thin Sensing photodiode

Sensing photodiode

ITO

a)

thin

thick

i type (a-Si:H)

200 nm

n type (a-SiC:H)

20 nm

i type (a-Si:H)

400 nm

p type (a-SiC:H)

20 nm

ITO

ITO

ITO

Glass

Glass

Glass

b)

20 nm

c)

Figure 1 Schematic cross section of the (a) stacked device, (b) sensing and (c) switching devices. ISBN: 9974-0-0337-7

1/6

Montevideo, 27-29 de setiembre de 2006

IBERSENSOR 2006

The bottom p-i-n-i-p structure is a sensor on which the image will be projected. The top p-i-n-i-p structure is used as a switcher, which will be controlled by the scanning laser beam to readout the image [7]. To reduce the optical coupling between the sensing and the switching diodes, the top and bottom structures are separated by a patterned 120 nm molybdenum layer working as an array of floating nodes. The pixel pitch is 250 µm and the size of the Mo electrodes is 220x220 µm. The bottom and top ITO electrodes are used for the charge collection. 2.2 Fabrication The top p-i-n-i-p structure is used as a switcher, which will be controlled by the scanning laser beam to readout the image [7]. To reduce the optical coupling between the sensing and the switching diodes, the top and bottom structures are separated by a patterned 120 nm molybdenum layer working as an array of floating nodes. The pixel pitch is 250 µm and the size of the Mo electrodes is 220x220 µm. The bottom and top ITO electrodes are used for the charge collection. The fabrication sequence includes the sputtering of the first ITO layer on the Corning glass substrate, deposition of the sensing p-i-n-i-p stack, sputtering and patterning Mo layer, deposition of the switching p-i-n-i-p stack, and finally sputtering and pattering the top ITO layer. Besides, the stacked structure, the simplified test structures have been also deposited during the same deposition processes. The cross sections of the sensing and switching test structures are shown in Figures 1 b) and 1 c). The sensing test structure is identical to the one in the color sensor. The switching test structure is also similar to the original design, except that the Mo contact was replaced by a layer of ITO. The p-i-n-i-p structures were deposited using a parallel-plate PECVD reactor. Deposition conditions such as the partial pressure, RF power and gas flow rates are summarized in Table I. Table 1 Partial flux of pure deposition gases used in the fabrication of the single films of the stacked p-i-n-i-p device. Type

n i p

Pressure mTorr

400 400 600

RF power W

2 2 2

ISBN: 9974-0-0337-7

Gas flow, sccm

SiH4

1% TMB + 99% H2

2% PH3 + 98% H2

15 20 15

⎯ ⎯ 15

5 ⎯ ⎯

CH4

15 ⎯ 15

The sensing p-i-n-i-p stack was deposited at substrate temperature of 260 ºC , while the switching p-i-n-i-p stack was deposited at lower temperature of 200ºC to minimize the degradation of sensing diodes.

3 Electric model 3.1 p-i-n-i-p structure Taking into account the geometry of the p-i-n-i-p stack both p-i-n and n-i-p photodiodes are series connected. Thus the current that flows through the whole device (I) is obviously the same that crosses each of the diodes. Thus:

I = I 1p −i − n = − I 2n −i − p

(1)

At the same time the voltage at each diode adds to the voltage at the terminals of the p-i-n-i-p device:

V = V1 − V2

(2)

Assuming for each diode the exponential current dependence on the voltage given by:

Ii

p −i − n

⎛ ηVVi ⎞ = I 0,i × ⎜ e T − 1⎟ − I ph,i ⎜ ⎟ ⎝ ⎠

(3)

Where Vi is the voltage drop across each diode and V the external applied voltage, I0,i and Iph,i are, respectively, the leakage current and the photocurrent of the photodiodes, VT is the thermal voltage and η the ideality factor (assumed to be equal for all photodiodes). Neglecting the series and shunt resistances and assuming: I i = I 0, i + I ph , i the solution of this equations system gives for the current (I) through the device the following expression:

⎛ V I 2 × I 0,1 exp⎜⎜ ⎝ ηVT I= I 0, 2

⎞ ⎟⎟ − I 1 × I 0, 2 ⎠ ⎛ V ⎞ ⎟⎟ + I 0,1 × exp⎜⎜ η V ⎝ T⎠

(4)

and I2 = I0,2 + Iph,2, where I1 = I0,1 + Iph,1 corresponding respectively, to diodes D1 and D2. From equation (4) it is observed that for large bias, either positive of negative, the current magnitude approaches either the current of the reverse biased device. Under illumination it will be controlled by the saturation photocurrent while under dark by the leakage currents. The voltage in each diode is given by:

2/6

Montevideo, 27-29 de setiembre de 2006

IBERSENSOR 2006

(5)

3.2 double p-i-n-i-p stack In a double p-i-n-i-p stack device all the photodiodes are also electrically in series connection. The current that flows through the whole device (I) is given by equation (1). Also, the voltage at each diode adds to the voltage at the terminals of the stacked device:

V = V1 − V2 + V3 − V4 (6) Assuming for each photodiode the exponential current dependence on the voltage as in the single p-i-n-i-p stack, the solution of the equations system for the double p-i-n-i-p stacked device is 2

2

I=

∑ (K × I i =1

2i

2 2 ⎛ ⎞ ⎞ ⎛ 2 − I 2i −1 ) + ⎜ ∑ (I 2i −1 − K × I 2i )⎟ + 4 ⋅ (1 − K ) ⋅ ⎜⎜ K × ∏ I 2i − ∏ I 2i −1 + ⎟⎟ ⎠ ⎝ i =1 i =1 i =1 ⎝ ⎠ 2 × (1 − K )

(7)

structure shows a level of the dark current higher than thick switching p-i-n-i-p structure. It is evident, that for the diodes in back-to-back connection the current is limited by the reverse biased diode with a lowest level of the leakage current. However, for stacked device the dark current is about one order of magnitude large than that for the switching p-i-n-i-p structure. 1E-5

Y#12 1E-6

Dark Current Density (A/cm2)

⎛ I + I1 ⎞ ⎟⎟ V1 = η ⋅ VT × ln⎜⎜ I ⎝ 01 ⎠ V2 = V1 − V

1E-7

1E-8

1E-9

1E-10

sensing diode switching diode Stacked device

Where V

K =e

η ⋅VT

1E-11

2

I 0, 2i −1

i =1

I 0, 2i

×∏

(8)

⎛ 2 ⎛ (I + I 2i −1 ) × I 0, 2i ⎞ ⎞ ⎟ ⎟ (9) V = ηVT × ln⎜ ∏ ⎜⎜ ⎜ i =1 (I + I 2i ) × I 0, 2i −1 ⎟ ⎟ ⎠⎠ ⎝ ⎝ And the voltage drop across each photodiode is:

⎛ I + Ii ⎞ ⎟ Vi p −i − n = η ⋅ VT × ln⎜⎜ ⎟ I ⎝ 0 ,i ⎠ ⎛ − I + Ii Vi n −i − p = η ⋅ VT × ln⎜⎜ ⎝ I 0 ,i

⎞ ⎟ ⎟ ⎠

1

2

3

Voltage (V)

The voltage in a four diode staked device is given by:

(10)

4. Results and discussion 4.1 Current-voltage characteristics Figure 2 shows the dark current-voltage characteristics of the stacked device and test structures tested at room temperature. For all devices the level of the dark current is different at positive and negative bias polarity due to an asymmetry of the p-i-n-i-p structure having intrinsic layers of different thicknesses. The thinner p-i-n diode has a larger reverse dark current due to the field-enhanced generation of the carriers in the i-layer. For this reason, the thin sensing p-i-n-i-p ISBN: 9974-0-0337-7

0

Figure 2 Dark current -voltage characteristics of the sensing, switching and stacked devices.

This effect may be related with a leakage induced by Mo pixel electrodes. The increased level of the dark current can be also due to the presence of defective pixels. The fabricated imager has 42x42 pixels (the area of the top ITO electrodes is 1.2x1.2 cm2), and most probably some of the diodes are damaged. 4.2 Current-voltage illumination

characteristics

under

4.2.1 p-i-n-i-p structure Under illumination the current voltage characteristic of the devices depends on the wavelength and intensity of the incident light. In Fig. 3 it is displayed the current-voltage characteristic of the p-i-n-i-p sensing photodiode under dark and monochromatic illumination (650 nm, 550 nm and 450 nm) obtained experimentally and though simulation using equation (4).

3/6

Montevideo, 27-29 de setiembre de 2006

IBERSENSOR 2006

λL=650 nm

-6

10

-7

λL=550 nm

10

I (A)

-8

10

λL=450 nm

-9

10

-10

10

-11

10

Simulation parameters: Iph1=0; Iph2=0 Iph1=0.4µA; Iph2=1.2µA Iph1=0.5µA,Iph2=0.15µA Iph1=0.5µA; Iph2=1nA η=5, I01=3nA, I02=0.1nA

Dark

Dashed lines - Experimental data Solid lines - Simulated data

-12

10

-2

-1

0

1

2

V (V) Figure 3 Experimental and simulated current voltage characteristics of the sensing p-i-n-i-p test device under different steady state illumination conditions.

The parameters of each photodiode (Iph1,2, I01,2, η) were adjusted in order to obtain a good compliance with the experimental data. As expected the magnitude of the measured photocurrent changes with the nature of the light. The evident difference on the level of the current magnitude under either positive or negative bias is related to the asymmetry of the device cause by a thin p-i-n and a thick n-i-p photodiodes. Under dark this factor controls the leakage current, and so I01>I02. Under red illumination the light penetrates uniformly both diodes, and thus the photocurrents are similar, being slightly higher in the thickest photodiode due to a higher generation rate (Iph1Iph2). Under blue illumination the generation mechanism is similar to the one observed under green illumination. However, as for this wavelength the absorption coefficient is higher, the penetration depth is much reduced and thus the generation rate will occur mainly in the thin p-i-n photodiode (Iph1>Iph2). In this simulation both the series or shunt resistances were neglected and even so a good agreement was obtained, which reveals the good quality of the deposited material. 4.2.2 Double p-i-n-i-p stack For the double p-i-n-i-p structure the current voltage characteristic was also simulated using equation (7). The simulation parameters were chosen in order to simulate the penetration of light ISBN: 9974-0-0337-7

only into the sensing photodiode. This assumption was based on the optical de-coupling caused by the molybdenum grid, placed between both p-i-n-i-p structures, that prevents the light from entering into the switching photodiode. Thus this photodiode is assumed to be under dark conditions (Iph3=Iph4=0). The results of the simulation current voltage characteristics are displayed in Fig. 4. Results show that the leakage currents of the switching diode are the limiting factor for the current across the device (arrows in the figure), even under illumination. Under the same conditions the potential drop across the back-to-back switching diodes was also simulated (V3, V4) using equation 10. The simulated data are displayed in Fig. 5. 10

Current (A)

p-i-n-i-p sensing structure

-7

10

-8

10

-9

10

-10

10

-11

Dark : Iph, 1,2=0) Red : Iph1=0.4µA; Iph2 = 1.2µA Green: Iph1=0.5µA; Iph2 = 0.15µA Blue : Iph1=0.5µA; Iph2 = 1nA

pinipinip

-1,0

-0,5

0,0

0,5

1,0

Voltage (V) Figure 4 Simulated current-voltage characteristics of the double p-i-n-i-p stack device under different illumination conditions. 0,5

[D1(pin)D2(nip)][D3(pin)D4(nip)] [sensing] [switching]

0,0

V3,4 (V)

-5

10

-0,5 D3

-1,5

D4 Dark Red Blue Green

-1,0

-1,0

-0,5

0,0

0,5

1,0

V (V)

F

igure 5 Simulated potential drop across the back-to-back switching diodes.

The external bias is shared asymmetrically (equations 9 and 10) among all the diodes, depending on the illumination conditions and applied bias. The molybdenum inter-layer decreases optical coupling between the sensing and 4/6

Montevideo, 27-29 de setiembre de 2006

IBERSENSOR 2006

green/blue light the switching photodiode is also photo excited, however the photocurrents assigned to the switching photodiodes are one or two orders of magnitude less than the ones assigned to the sensing photodiode. Thus, in this sensor the switching photodiode is not working only as a reading element, as most of the light projected onto the sensing photodiode goes through the Molybdenum grid and enters into the switching photodiode. Under these same conditions, where the switching photodiode is assumed not to be under dark, the potential drop of both photodiodes of the switching structure (D3 and D4) was also simulated with equation 10). Results show that, in these conditions, the self biasing effect in still observable, which confers still to the device light to-dark sensitivity and the ability of colour discrimination. Thus, the use of the grid interlayer between both sensing and switching devices reveals not to be so much relevant. This means that the use of a single p-i-n-i-p device as a colour imaging structure, instead of a more complex device, is quite feasible. 1,5 1,0

[D1(pin)D2(nip)][D3(pin)D4(nip)] [sensing] [switching]

0,5 0,0

V3,4 (V)

switching diodes preventing most absorption of the light across the switching photodiode. Consequently, the non irradiated back-to-back switching photodiodes (D3 and D4) self bias in order to get in line with the irradiated sensing ones (D1 and D2) to which they are series connected. Under negative bias, D3 is reverse and D4 is forward biased, while the opposite trend occurs under positive bias. Thus if a probe beam (the laser scanner) is directed to the pixel, the voltage drop across the switching diodes (V3,4) decreases enabling the charge transfer from the sensing diodes to the input amplifier. By changing the polarity of the bias, it is possible to achieve colour discrimination. Negative bias voltage applied to the front ITO electrodes enhances the sensitivity in blue and positive bias to detect red light. Selectivity to green light can be also achieved by adjusting the biasing conditions. Thus the use of this double pi-n-i-p stack device for colour image sensing using the LSP technique is strongly dependent on the self bias effect of the switching photodiode that works, actually, as a reading element, while the sensing photodiode is the effective sensing element. In order to test the self bias effect of the switching photodiode the current voltage characteristics of the double p-i-n-i-p stacked device were measured and simulated under different steady state illumination conditions (Fig. 6).

-0,5 -1,0

1E-5

λL=650 nm

D3

D4 Dark Red Blue Green

-1,5

1E-6

λL=550 nm

Current (A)

1E-7

-2,0 -2

λL=450 nm

-1

0

1

2

V (V)

1E-8

Figure 7 Simulated potential drop across the back-to-back switching diodes.

Dark 1E-9

1E-10

Iph1 Iph2 Iph3 Red 4.0 µA 8µA 1.5 µA Green 3.3 µA 8µA 15 nA Blue 2.5 µA 8µA 1.5 nA I01=I04=10 nA, I03=1 nA, I02=6 nA

Dashed lines - Experimental data Solid lines - Simulated data -2

-1

0

1

Iph4 6µA 0.3 µA 10 nA

2

Voltage (V)

Figure 6 Experimental and simulated current voltage characteristics of the double p-i-n-i-p stacked device under different steady state illumination conditions.

The good agreement between the simulated and experimental data was obtained through a careful adjustment of the leakage currents and photocurrents of each photodiode. To obtain such agreement the photocurrents of the switching diode were assumed to be non zero, i.e., the switching photodiode is actually, optically coupled to the sensing photodiode. This effect is particularly relevant for the red light that penetrates almost uniformly through the whole device. Under ISBN: 9974-0-0337-7

4.3 Color discrimination Fig. 8 shows a spectral response of the sensing p-in-i-p test structure measured at a positive and a negative polarity of the bias of 2 V applied to the front ITO electrode. Spectral response data show that color recognition is feasible. The spectral sensitivity in the green/blue range is enhanced under negative bias, while under positive bias, the red sensitivity is improved. A good color separation between blue-green and red is observed. By reverse biasing the diode where most of the photons are absorbed (the front under blue and the back under red irradiations) an enhanced sensitivity is expected. Combining the

5/6

Montevideo, 27-29 de setiembre de 2006

IBERSENSOR 2006

information obtained under positive and negative applied bias color discrimination can be achieved. 1.0 -2 V +2 V

Spectral Responce (a.u.)

0.8

0.6

0.4

0.2

0.0

400

500

600

700

800

Wavelength (nm)

Figure 8 Spectral response of the sensing test structure at two different applied voltages.

4.4 Light-to-dark sensitivity To show the ability of the sensor as an image sensor, the photocurrent generated by the scanner is shown in Figure 9, at ± 1 V. Here the same blue and red optical images (two stripes) were projected, one by one, on the active surface of the sensing diode and acquired through the switching diode using the moving probe beam. The line scan frequency was close to 1 kHz. For readout time of 1 ms a the frame time for a 50 lines image is around 50 ms.

Photocurrent (A)

4x10

-3

3x10

-3

2x10

-3

1x10

-3

λL=650 nm

V=+1 V V=-1 V

0

-1x10

Dark

-3

0

10

20

30

40

50

Dark 60

70

Position (a.u) -3

2x10

-3

1x10

-3

Photocurrent (A)

3x10

a)

λ L=500 nm

V=+1V V=-1V

Defining the light-to-dark sensitivity as the difference between the photocurrent magnitude with (ON state) and without (OFF state) optical bias we observe an enhanced light-to-dark sensitivity at -1 V for the blue images and at +1V for the red ones. 5. Conclusions Stacked p-i-n-i-p image sensor with a voltage controlled spectral responsivity has been fabricated and characterized. Dark current density below 10 nA/cm2 is achieved for the structure having 42x42 pixels. A good color separation between blue-green and red is observed. A SPICE model of the backto-back diodes based device has been developed to support the obtained results. Sensor shows a good color discrimination and light-to-dark sensitivity. To improve color separation and image resolution, and to optimize the optical decoupling between both sensing and switching diodes further work has to be done. Acknowledgment This work has been financially supported by POCTI/ESE/38689/2001 project and IPL 13 project.

References [1] R. A.Street, I. Fujieda, R. Weisfield, P. Nylen, Mater. Res. Soc. Symp. Proc. 258 (1992) 1145. [2] M. Kunii, K. Hasegawa, H. Oka, Y. Nakazawa, T. Takeshita, and H. Kurihara, IEEE Trans. Electron Devices 36, 2877 (1989). [3] L. E. Antonuk, J. Boudry, Y. El-Mohri, W. Huang, J. Siewerdsen, J. Yorkston, and R. A. Street, Proc. SPIE 2163, 118 (1994). [4] C. Chang, C. Y. Chang, Y. K. Fang, and C. Jwo, IEEE Electron Device Lett. 8, 64 (1987). [5] H. Stiebig, J. Giehl, D. Knipp, P. Rieve, and M. Bohm, Mater. Res. Soc. Symp. Proc. 377, 815 (1995). [6] M. Vieira, M. Fernandes, J. Martins, P. Louro, R. Schwarz, M. Schubert, IEEE Sensor Journal, 1, no.2, pp. 158-167 (2001). [7] M. Vieira, A. Fantoni, P. Louro, M. Fernandes, G. Lavareda and C.N. Carvalho , J. Non-Crystalline Solids, in Press.

0

-1x10

-3

0

10

20

Dark 30 40

Position (a.u.)

50

60

70

b)

Figure 9 Photocurrent generated by the scanner at ± 1 V with a red (a) and a blue (b) two stripes image projected onto the device.

ISBN: 9974-0-0337-7

6/6