A novel photodiode array structure for gamma camera applications

September 23, 2017 | Autor: Olivier Evrard | Categoria: Quantum Efficiency, Pixel Detector Design, Chip

Descrição do Produto

Nuclear Instruments and Methods in Physics Research A 504 (2003) 188–195

A novel photodiode array structure for gamma camera applications O. Evrard*, M. Keters, L. Van Buul, P. Burger Canberra Semiconductor NV, 25 Lammerdries, B-2250 Olen, Belgium

Abstract Photodiodes coupled to scintillators are more and more used for gamma camera applications in order to improve the efﬁciency at higher energies. However the coupling of photodiode arrays become difﬁcult when the pixel number is large because of the complicated front and backside connections with read-out electronics. Canberra has developed a photodiode array structure that allows all electrical connections to be taken on one side, while the light entrance side can be directly glued on the scintillator. Using such a photodiode array, the dimensions of the scintillator can be the same as the silicon chip and near 100% of the area will detect light. Furthermore, an array of photodiodes coupled to scintillators can be built without any dead gap in between. To improve the quantum efﬁciency of the photodiodes, we had to develop a thin ohmic entrance window covered by an anti-reﬂective coating. A quantum efﬁciency of more than 80% at 565 nm was achieved. r 2003 Elsevier Science B.V. All rights reserved. Keywords: Photodiode; Pixel detector; Quantum efﬁciency; Anti-reﬂective coating

1. Detector structures The detectors are made on 500 silicon wafers using a double sided process. The double sided process combines photolithography and ion implantation. This technique allows to realise the type of complex electrode geometry needed for a photodiode array. On the back side or pixel side, P-type junctions are implanted through a mask determining the implanted pixel geometry. On the photon entrance side, a homogenous blocking n+ contact is implanted. Some applications require that all contacts can be taken on one side (pixel

*Corresponding author. Tel.: +32-14-221975; fax: +32-14221991. E-mail address: [email protected] (O. Evrard).

side) of the detector chip. We have therefore added an n+ contact on the edge of the silicon chip, around the active area, on the pixel side (see Fig. 1). The collected electrons on the ohmic contact (entrance side) have to be transferred to the edge ohmic contact through a resistor whose value is given by the formula below: rw R¼ ð1Þ A where r equals the material resistivity in ohmcm, w equals the detector thickness in cm and A equals the active area of the edge contact in cm2. Typical values for R are 500–2000 O. For applications where timing or ultimate resolution is important, the resistive nþ contact on the pixel side can deteriorate the detector performances (Fig. 2).

0168-9002/03/$ - see front matter r 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0168-9002(03)00789-7

O. Evrard et al. / Nuclear Instruments and Methods in Physics Research A 504 (2003) 188–195

edge contact

R

pixels or P+ junction

guard

ohmic contact or N+ photon entrance side Fig. 1. The detector structure.

189

spread on the pulse width. By applying a very thin aluminium layer (some 20 nm) the time resolution can be improved from the nanosecond range to values under 200 ps. This improved timing feature can only be applied for the measurement of electrons or light protons. And ﬁnally the electric ﬁeld within the depleted region has to be as high as possible. By implementing a multiple ﬂoating guard ring structure, breakdown voltages up to several times the depletion voltage can be achieved. Other applications where efﬁciency and spectroscopic resolution are very important demand thicker detectors. Currently, we can process detectors with a thickness up to 1.5 mm. Concerning the pixel layout, the pixel pitch is limited by the connection to the read-out electronics. The pixel side has to be connected to a multichannel read-out electronic. Three interconnection technologies are currently used. For pixels with a pitch above 1.5 mm one can use screen-printing. Pixels with a pitch down to 250 mm can be interconnected with standard bump-bond technologies. Pixels with a pitch down to 40 mm request a special and very costly bump technology Ref. [2].

2. Requirements Fig. 2. Examples of realised pixel detectors.

Canberra has developed a technique to shunt this resistive contact and bring it down to a value of 2–4 O. The homogeneous n+ contact on the entrance side which does not require any physical connection can very easily be coupled to a scintillator. The applications will determine the detector choice. For timing applications one can improve the detector timing properties by reducing the detector thickness. The standard thickness is 300 mm. The signal rise time improves with a factor of 2/3 by reducing the thickness to 200 mm Ref. [1]. Another important feature is the surface resistance of the implanted n+ contact on the entrance side. The larger the active area, the wider the

For the successful coupling of photodiodes to scintillators and improved detector characteristics, several requirements have to be met. 2.1. Low leakage currents Low leakage currents are needed in order to have a good spectral resolution. The parallel white noise ﬂowing through the input node of the transistor is given in Eq. (2), where A3 is a constant that has a typical value of 1.73. rﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ 1 A3 2qItshaping : ENC ¼ ð2Þ q 2 In a depleted diode having a thickness w; the leakage current density J is dominated by the bulk generation lifetime tgeneration : This lifetime is

190

O. Evrard et al. / Nuclear Instruments and Methods in Physics Research A 504 (2003) 188–195

mainly limited by the processing parameters. The expression of J is given in the following equation: qni w with ni J ¼ 2tgeneration 7000 ¼ 3:93 1016 T 3=2 exp : ð3Þ T Therefore, the integrated expression of the ENC will be given by sﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ ni w ð4Þ ENC ¼ A3 Area tshaping : 2tgeneration The intrinsic carrier concentration ni is strongly temperature dependent. The processing steps have a large inﬂuence on the generation lifetime. A nonoptimal process typically leads to lifetimes in the 0.5–2 ms range while an optimised process can lead to lifetimes in the 5–100 ms range. High lifetimes are easier to obtain on substrates with small thickness. Table 1 shows some individual pin diode leakage currents obtained in our foundry, translated in generation lifetime for different areas and thicknesses. Figs. 3 and 4 show the example of a typical leakage current and lifetime distribution in pA for a hexagonal like pixel structure. The thickness of the detector is 200 mm. The area of one hexagon is 25 mm2. (b) Low leakage currents are needed to insure the proper functioning of the ASIC coupled to pixel arrays. Attention has to be given to the leakage current extremes in the case of pixilated detectors. On some ASICS, the pixel that has the highest leakage

Table 1 Leakage current and lifetime values for different detector thicknesses and areas Area (mm2)

Thickness (mm)

Leakage current (pA)

Lifetime (ms)

25 82 50 50 30 25

200 300 500 700 1000 1500

100 400 600 700 1000 900

40 45 55 36 21 28

Fig. 3. The leakage currents in pA of pixels with a thin ohmic windows on the entrance side.

Fig. 4. The extracted lifetimes in ms of pixels with a thin ohmic windows on the entrance side.

can saturate the whole electronics. Reasons can be numerous: the high leakage current can affect the whole biasing of the detectors. Some elements of the resetting mechanisms can be shared by all channels: too numerous resetting in one channel

O. Evrard et al. / Nuclear Instruments and Methods in Physics Research A 504 (2003) 188–195

can therefore hamper the proper functioning of the other channels.

Table 2 Theoretical reﬂectance’s for different optical conﬁgurations ARC

Optimal ARC thickness for 565 nm

Minimum reﬂectance at 565 nm in air (%)

Minimum reﬂectance at 565 nm in a medium with n ¼ 1:5ð%Þ

No ARC SiO2. Ta2O5

— 96 nm 68 nm

35 8.93 0.22

19.77 19.77 2.93

2.2. The need for optical matching with the scintillator material

In our case the optimum refractive index equals 2.35–2.43. Ta2O5 was chosen for its refractive index (2.07) and because it can be deposited at low temperatures by E-beam or sputtering. The optimum ARC thickness doptimum:ARC is given in the following equation: doptimum:ARC ¼

lincident:photon : 4nARC

40.00 35.00 30.00 Reflectance [%]

Bare silicon has a reﬂectivity of 35%. AntiReﬂective Coatings (ARCs) are therefore deposited or grown on the photodiode in order to reduce this reﬂectivity. (a) Choosing the refractive index of the ARC: The refractive index of the ARC and the optical glue need to be different. For maximum absorption of the light, the ARC should be a thin layer in which destructive interference can occur. Optical glues have typical refractive indexes around 1.4– 1.5. SiO2 should not be used as an ARC together with optical glue because of the proximity of its refractive index (1.45) with one of the glue. The optimum refractive index of the ARC is given by Eq. (5), with the refractive index of the silicon substrate equals 3.94. pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ noptimum ¼ noptical:medium nSi : ð5Þ

191

96 nm of SiO2 68 nm of Ta2O5

25.00 20.00 15.00 10.00 5.00 0.00 400

500

600

700

800

900

1000 1100

Wavelenght [nm] Fig. 5. The theoretical reﬂectance of optimised SiO2 and Ta2O5 ARCs in air.

ð6Þ

The incident wavelength depends upon the scintillator material. Using CsI(Tl), the ARC thickness must be optimised for a wavelength of 565 nm. The optimal thickness and minimal reﬂectance at 565 nm are given in Table 2 for Ta2O5 and SiO2 both in an air medium and in a medium with a refractive index of 1.5. The minimum reﬂectance of Ta2O5 is signiﬁcantly lower than for the SiO2, both in air and in an optical glue medium. The theoretical reﬂectance curves of SiO2 and Ta2O5 are given in air for an optimal thickness in Fig. 5: (b) Optical characterisation of the ARC: 68 nm of Ta2O5 was deposited on the ohmic side of our

substrates. The reﬂectance in air was measured after the Ta2O5 deposition and is given in Fig. 6, overlaid with the theoretical reﬂectance. We see that the theoretical and experimental curves have a good match. The minimal experimental reﬂectance at 565 nm is 3% and is somehow higher than the theoretical value mentioned in Table 2. 2.3. The need for a thin entrance window Great care must be given to the junction or ohmic contacts so that electron–hole pairs created by photons do not recombine in the undepleted

O. Evrard et al. / Nuclear Instruments and Methods in Physics Research A 504 (2003) 188–195

192

Table 3 Minority carrier lifetimes, diffusion lengths and collection potentials for a 0.2 mm deep junction with different doping levels

40.00

Reflectance (%)

35.00 Experimental Theoretical

30.00 25.00 20.00

ND (cm3)

Lifetime (ms)

Diffusion length Lp (mm)

Collection potential exp(xj =Lp )

1.00E+17 1.00E+18 1.00E+19 1.00E+20

1.25E-05 1.04E-06 3.88E-08 5.32E-10

86.91 16.48 2.43 0.26

1.00 0.99 0.92 0.47

15.00 10.00 5.00 0.00 350

450

550

650

750

850

950

Wavelenght (nm)

Fig. 6. The experimental and theoretical reﬂectance versus wavelength curves of 68 nm of Ta2O5 in air.

part of these contacts. Electron–hole pairs generated by photons will indeed recombine: (a) in the bulk of the undepleted layer by Auger radiative recombination if the doping level is too high. (b) at the surface of the undepleted layer by recombination with dangling bonds if the surface is unpassivated. The photodiode that is described in this article has an nþ undepleted layer exposed to the entrance side. The expression of the recombination lifetime tp of holes minority carriers in an n+ layer Ref. [3] is given by tp ¼

7:8

1013

1 : ND þ 1:8 1031 ND2

ð7Þ

This expression can be used to compute the diffusion length Lp of the holes. pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ ð8Þ Lp ¼ Dp tp : It can be stated that 1=e of the generated hole carriers will still be available for collection at a distance of Lp from the generation point. For an undepleted depth xj and a minority carrier diffusion length Lp ; the probability of recombining before reaching the depleted region can be approximated by xj p ¼ 1 exp : ð9Þ Lp Provided that the doping level of the n+ layer is kept to reasonably low values and that the

dangling bonds at surface are passivated with a thermally grown oxide or a deposited PECVD oxide, the undepleted layer can contribute in a useful way to the collection of charges. Electrons and holes will migrate by diffusion in these regions. This is a slower collection mechanism that might not be compatible with high counting applications. Therefore, there is still a need to keep the thickness of the undepleted region as thin as possible. In order to give an idea of the collection potentials of a 0.2 mm deep layer that is homogeneously doped, the following table gives values of lifetimes, diffusion lengths and probabilities of recombination for different doping levels (Table 3). The internal quantum efﬁciency versus incident photon wavelength plot is given in Fig. 7 for different ohmic window thicknesses, with a doping level of 1020/cm3 assuming a rectangular step doping proﬁle and a very good surface passivation. On this plot, it can be seen that high doping levels are only compatible with thin n+ ohmic structures if a good quantum efﬁciency is desired. For a wavelength of 565 nm, the thickness of the layer must be kept under 1 mm in order to have a quantum efﬁciency above 70%. In practice, high quality surface passivations are difﬁcult to implement on very thin layers, so, these results give a better estimate of the actual quantum efﬁciency. The nþ proﬁles realised by technology are Gaussian. Nevertheless, these simple assumptions can be useful for the knowledge of an upper estimate of the quantum efﬁciency for a given ohmic contact. Fig. 8 shows the inﬂuence of the doping level for an undepleted layer depth of

O. Evrard et al. / Nuclear Instruments and Methods in Physics Research A 504 (2003) 188–195

193

1 0.9

Quantum efficiency

0.8 0.7 0.6 0.5 0.4

External QE no ARC

0.3

External QE with Ta2O5

0.2

Internal QE

0.1

Fig. 7. Modelling of the internal quantum efﬁciency for different ohmic window thicknesses.

0 400

500

600

700

800

900

1000

1100

Wavelenght (nm) 1.1

Fig. 9. The external quantum efﬁciencies with and without Ta2O5 ARC, the derived internal quantum efﬁciency of the Ta2O5 ARC coated substrate.

Internal quantum efficiency (%)

1 0.9 0.8 0.7 0.6 0.5

10^18 #/cm3

0.4 0.3 0.2

5.10^19 #/cm3 10^20 #/cm3

0.1 0 300 400 500 600 700 800 900 1000 1100 1200 Wavelenght (nm)

Fig. 8. Modelling of the internal quantum efﬁciency for different doping levels of a 0.2 mm thick ohmic window.

0.2 mm. It can be seen that low doping levels provide the best quantum efﬁciencies. Again, the assumption of a high quality surface passivation is made. For scintillators producing photon wavelengths of 420 nm, the combination of a thin undepleted layer, low doping levels and a very good surface passivation is needed.

computed by dividing the external quantum efﬁciency by 12RðlÞ; RðlÞ being the substrate reﬂectance: the internal quantum efﬁciency is represented by the upper plain curve. Theoretical ﬁttings and extractions have been derived from these results. Further optimisation on the ohmic contact can be done in order to improve the internal quantum efﬁciency at wavelengths lower than 565 nm such as the wavelength of 420 nm for CsI(Na).

3. Applications The traditional approach, using the pixel detector in combination with a scintillator can be used in 2 ways. *

2.4. Quantum efficiency characterisation External quantum efﬁciency was measured on photodiodes and is represented in Fig. 9. The lower bold and bold dashed curves show the evolution of the external quantum efﬁciency prior to and after the ARC deposition. The quantum efﬁciency rises from 65% to 83% at 565 nm. In order to assess the potentials of our detector technology, the internal quantum efﬁciency was

*

The ﬁrst and more simple method is to use 1 single scintillator coupled to the pixel detector. Position information can be obtained by weighing the different pixels which are hit by one event. This is an easy solution but the resolving capacity of the sensor is less good. When adding the signals from different pixels energy resolution loss occurs. The second method is to use separated scintillator crystals with the same dimensions as the pixels. The set-up is more complicated, but the

194

O. Evrard et al. / Nuclear Instruments and Methods in Physics Research A 504 (2003) 188–195

1 2 3

Nuclear physics also uses these kinds of detector arrays. One example is a pixel detector with very thin entrance window developed to measure 10 keV protons.

4 Detector 1

Detector 2

Fig. 10. Example of 2 detectors coupled to scintillators, where 1 is the read-out electronics, 2 the detector, 3 the optical grease and 4 the scintillators.

resolving character of the sensor is very good. This type of camera is mostly used to examine small tissue or animals. The main advantage is that several detectors can be placed side to side with no dead space inbetween Fig. 10. Another application of the pixel detector is to use it as a hybrid photomultiplier tube HPD Ref. [4]. In a HPD the PM is replaced by a vacuum tube where the anode is a silicon pixel detector. The multiplication of the signal is done in the silicon detector chip, a typical gain is around 3000. The entrance window should be as thin as possible in order to reduce the energy loss of the electron within the silicon. Typically the threshold for these devices lays in-between 1 and 2 keV. In this application a thin aluminium layer (for timing properties) replaces the anti-reﬂective coating Fig. 11. Building further on the HPD, a gamma camera was developed at CERN Ref. [5] using the HPD tube and a pixel detector bump-bonded to binary read-out electronics with the same dimensions as the pixels 50 mm 500 mm. The HPD was coupled to different scintillator crystals. Other applications using pixel detectors involve direct X-ray measurements where detector efﬁciency and therefore detector thickness are dominant features. Detector thicknesses up to 1.5 mm have been processed. Pixel detectors have been developed for lung monitoring applications where the energy range between 8 and 200 keV is studied. A pixel detector for a space application, in the 3– 50 keV range will is currently under development (Fig. 12).

Fig. 11. Proximity HPD for the use in high ﬁeld, distance between cathode and photoelectric material is short in order to avoid a deviation of electron.

Fig. 12. A 1.5 mm thick X-ray detector system with 5 5 mm pixels.

O. Evrard et al. / Nuclear Instruments and Methods in Physics Research A 504 (2003) 188–195

195

4. Conclusions

References

CANBERRA has devised a novel photodiode structure that allows an easy optical coupling with scintillating materials. This structure allows all contacts to be taken on one side. The doping and depth of the ohmic window has been carefully optimised in order to have good quantum efﬁciency and minimise the dead layer. Using an ARC compatible with the optical glue and the scintillator refractive index, CANBERRA photodiodes show an external quantum efﬁciency of 83% at a wavelength of 565 nm.

[1] P.A. Cushman Heering, Nucl. Instr. and Meth. Phys. Res. A 442 (2000) 289. [2] E. Heijne, Nucl. Instr. and Meth. Phys. Res. A 465 (2000) 1. [3] A. Rohatgi, IEEE Trans. Electron. Devices ED-31(5) (1984). [4] R. DeSalvo, Nucl. Instr. and Meth. Phys. Res. A 387 (1997) 92. [5] C. D’Ambrosio, et al., Nucl. Phys. B (Proc. Suppl. ) 78 (1999) 598.

Lihat lebih banyak...

A novel photodiode array structure for gamma camera applications

Descrição do Produto

Comentários