Low energy response of silicon pn-junction detector

Nuclear Instruments and Methods in Physics Research A 377 (1996) 191-196

NUCLEAR INSTRUMENTS & METHODS IN P H Y S I C S RESEARCH Sechon A

ELSEVIER

Low energy response of silicon pn-junction detector R. H a r t m a n n

a:l<

' , D. H a u f f a, P. L e c h n e r

a

•

, R. R i c h t e r ,

a

..

a

L. Struder , J. K e m m e r b, S. Krisch b,

F. Scholze c, G. Ulm c "MPI Halbleiterlabor, Paul-Gerhardt-Allee 42, D-81245 Miinchen, Germany hKETEK GmbH, Am lsarbach 30, D-85764 OberschleiJ3heim, Germany ~Physikalisch-Technische Bundesanstalt, Abbestr. 2-12, D-10587 Berlin, Germany

Abstract The response function of implanted silicon detectors in the soft X-ray region (150 eV-6 keV) has been measured. To reduce signal charge loss in the highly doped p+-region just beneath the detector surface, different techniques of producing shallow doping profiles and enhancing the electric field at the pn-junction are presented. The spectroscopic resolution could be improved significantly. On {100) detector material, a peak to valley ratio of 5700:1 for the mangan K line was achieved. The measured pulse-height distributions were fitted by a detector model, taking the doping profile of the entrance window into account. The results of the fit were in excellent agreement with the measurement data over the entire energy range.

1. Introduction The lineshapes of implanted silicon X-ray detectors still exhibit an inferior behaviour in the soft X-ray region when compared with surface barrier detectors. Since they are both made of the same material, similar effects are expected. Following the photoabsorption process, a charge cloud of generated electrons is repelled by the electric field from the entrance electrode. Differences might occur in charge loss mechanisms close to the detector entrance side due to a) an insufficient field strength as a result of the doping profile, b) an accumulation layer of holes at the highly doped p~-contact, c) fixed positive oxide charges built in at the S i - S I O 2 interface, and d) crystal lattice defects, caused by the ion implantation. In this work, we focused our interest in optimizing the doping profile of the detector entrance side in order to tailor the electric field tor a minimized charge loss. Taking doping profiles into account as obtained by spreading resistance measurements, we developed a rather simple geometrical detector model in order to describe the structure and behaviour of response functions at various photon energies. This model does not primarily assume any " d e a d " layer, but a charge collection efficiency as a function of depth and detector parameter, to which the charge cloud extension of thermalized electrons is superimposed [ 1]. For an evaluation of the radiation entrance window in the soft X-ray region, different detector structures have been used, depending on the measurement setup. However * Corresponding author.

the fundamental functioning as radiation detectors is identical. It consists of a fully depleted n-type bulk with a p +-contact on its back, which acts as the radiation entrance window. In section 2 we give a description of our technology to produce shallower p+-implants at the radiation entrance window. A further increase of the detector charge collection efficiency was achieved by an additional n-implant, which will be described in Section 3. Calculated response functions are compared with experimental results in Section 4.

2. Entrance windows with a shallow p+-implant 2.1. Technology

The performance of implanted silicon detectors for Xray spectroscopy is limited at higher energies by an increase of the X-ray absorption length in silicon [2]. Due to the deep sensitive volume of 280 txm, the quantum efficiency of our detectors remains higher than 90% for 10 keV photons. For low energy photons, limitations arise due to a reduced collection efficiency of generated signal charges at the radiation entrance window. E.g. for a 277 eV carbon K s photon, the I / e absorption length is about 120 rim. Fig. l a shows the net doping profile and resulting electric field of our standard back p+-implantation. The pn-junction is placed at a depth of approximately 300 nm. With a short penetration of the depletion zone into the implanted p +-region, the result is a highly doped, field free region between the detector surface and the pn-junction. Signal charges generated within this area will reach the

0168-9002/96/$15.00 Copyright © 1996 Published by Elsevier Science B.V. All rights reserved P l l S0168-9002(96 )00254-9

I. DEVICE PHYSICS

192

R. Hartmann et al• / Nucl. Instr• and Meth. in Phys. Res. A 377 (1996) 191-196

'~' 1020 .............................................. E u "'''"

Q 0

%,

depletion• Due to the lower dose of implanted ions, crystal lattice defects are reduced as well.

,80

Cl,)

1018

60 7 2.2. Experimentation

1016

40 ~

"

o

g

~ 1014

20 I2

"~ 1012 Q

1O0

,,-r~,,,~ 200 ,500

Range Into]

1 0 2 0 ! ....... ' . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.o. E

o

1018

. '.,~ ",.k,

p 1016

........ 0 400 500

' ..........

b.)

80 60 "~" 40>~'~

< lO0~.k \ < 1 1 1 >

8

.E C3

E o

1014 . 1012 •~ - . 0

-~

'., ~" \

1 O0

20!.2 i ............................. i~." 0 200 300 400 500 Range [nrn]

1020 ............. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

c)

c 1018 ." / ~

80 60'~'

0

1016

40 >~" .-~. "o

8

1014"

\

"~1012 r.,,

n-[mplont

20E

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

0

1O0

200 300 400 500 Range [nm] Fig. 1. Doping profiles (dotted lines) and electric fields (solid lines) for different entrance window implantations: (a) standard 12.5 keV boron into silicon (profile obtained by spreading resistance measurement), (b) same energy and dose of implanted boron atoms through a SiO, layer (for different wafer material and tilt angle during implantation), and (c) like before, plus an additional phosphorus implantation to increase the electric field. collection anode only after diffusion towards the space charge region. To increase the quantum efficiency for low energetic X-rays, the width of this partially inactive layer has to be reduced. By an implantation of boron ions through a SiO 2 layer, a shallower doping profile is achieveable at the entrance window without consequences on other detector structures. For an implantation through a SiO 2 layer with identical implantation parameters as above, the simulated doping profile and its corresponding electric field distribution are shown in Fig. lb. The pn-junction is situated at a depth of approximately 40nm, while the integrated hole concentration adjacent to the detector surface is still high enough to guarantee the functionality of the detector under full

For our purpose of low energy X-ray spectroscopy, an optimized detector structure was used. It consists of a back illuminated diode with 150 ~m in diameter and an integrated FET to reduce stray capacitances [3]. Following is an AMPTEK A 250 preamplifier, placed near the detector chip. The signals from the preamplifier inside the cryostate are connected to a shaping amplifier and fed to a 12 bit multichannel analyser. At an operating temperature of 140 K and a shaping time of ~-= 3 ~s, the total noise contribution of the detector-amplifier system was 8.2 e for (100) and 10. I e for (111) detector material. To obtain monoenergetic photons in the range between 150 eV and 1 keV, a soft X-ray beamline of the Physikaliscb Bundesanstalt (PTB) at the electron storage ring BESSY in Berlin was used. The beamline is equipped with a plane grating monochromator, yielding an energy resolution E~ dE of 3000 at 100eV and 800 at 1.5 keV [4]. Essential for the precise determination of the detector response function is a suppression of stray light and higher order contributions of the monochromator. This was attained by selecting an appropriate filter for each photon energy. The small dimensions of the detector required a careful alignment with respect to the beam in order to prevent charge splitting between the sensitive region of the detector and the surrounding substrate. Two 25 ~m apertures in a distance of approximately one meter were used to collimate the beam to the necessary size. Typical count rates were between 200 and 800 counts per second, depending on the photon energy. Pulse-height distributions for various X-ray energies are shown in Fig. 2. The detector was operated in a single photon counting mode. Despite a slightly higher electronic noise contribution of detectors on (111) silicon material, the differences in the line shapes between both detector materials are obvious. As shown by numerical simulations of the doping profile in Fig. lb, the range of the implanted boron atoms reaches farther into the bulk for (l 11) silicon material. This will result in a higher loss of signal electrons by trapping. A more detailed description of charge collection mechanisms will be given in Section 4.

3. Detectors with an enhanced field entrance window 3.1, Basic concept

A further reduction of signal carrier diffusion towards the backside detector surface was achieved by increasing the electric field at the pn-junction. Within the space charge region, the electric field is given by Poisson's equation as

193

R. Hartmann et al. / Nucl. Instr• and Meth• in Phys. Res. A 3 7 7 (1996) 1 9 1 - 1 9 6

1.0

900

eV

g o.8

8

o.6

"~ 0.4

. . . . . . . . . . . . . . . . . . .

o

1.0

'/:,,i:.,

g 0.8 0

o

0,6

~ 0.4 0

~ 0.2 0.0

~- 0.2 0.0 0

200

400 600 800 Energy [eV]

1000

1200

1.0

200

e~

400 Energy [eV]

600

1.0

800

eV

g o.8

0.8

8 0.6

8 o.6!

.-~ 0.4

•.~ 0.4

~: 0.2 0.0

~. 0.2 0.0

O

100

200 300 400 Energy [eV]

500

1.0

600

1O0

1.0

200 eV

200 Energy [eV] •

"

500

400

' 150eV

=~ 0.8

0.8

0

8 0.6

o

g o.4

0

~- 0.2 0.0

0.6 0.4 0.2 0.0

1O0

200 Energy [eV]

300

400

0

50

O0 150 200 Energy [eV]

250

300

Fig. 2. Comparison of the normalized response functions of detectors produced on both (100) (straight lines) and (l I l) (dotted lines) high ohmic silicon material for photon energies varying from 150 eV to 900 eV. For X-ray energies below 350 eV, measurements of detectors on (1 l 1) material did not reveal a separable signal peak.

O E / O x = q / e ' ( p ( x ) - n(x) + N D ( X ) -- NA(X)) dx.

(1)

Moving the position of the pn-junction closer to the detector surface hardly effects the maximum field strength, since the integrated concentration of ionized donors will change merely by about 0.1% under full depletion. Thus placing an additional n-implant at the falling tail of a shallow p +-profile will enhance the electric field according to the implanted phosphorus dose. The calculated doping profile of such a field enhanced pn-junction is shown in Fig. lc. An overall increase of the electric field between the detector surface and the pn-junction of a factor of three is achieved. For a direct comparison of the detector response with and without an additional n-implantation, a low-noise, large area detector like a CCD is an ideal instrument. With a sensitive area of 1 0 × 3 0 r a m : and a total of 12800 pixel, the electronic noise contribution of a pn-CCD, designed for the X M M satellite mission, is less than 6 e [6]. Due to the long fabrication cycle for a CCD, we performed a test of the described procedure on an already mounted and bonded device. The maximum temperature step for a device mounted on a ceramic carrier is limited by several factors to not more than 400 °C. The assumed

activation rate for a low dose phosphorous implant is less than 5% at annealing temperatures below 400°C [5]. By placing a mechanical mask in front of the chip during the implantation process, an enhanced field pnjunction was implemented in approximately half of the sensitive area of the CCD. The electronic noise contribution rose to 7.1 e within this area, mainly due to the insufficient annealing process. The pixel structure of the device allowed a clear distinction of the two subareas. It is important to notify that the detector was produced on an (100) silicon wafer. 3.2. X - r a y m e a s u r e m e n t s

For measurements, discrete monoenergetic X-ray lines were provided by an X-ray tube with different anodes. Targets of carbon and aluminium in combination with a filter of the same material in front of the device provided carbon and aluminium K lines. Mangan K, and K~ lines were provided by a radioactive 55Fe-source. No filter was used in that case. All measurements were performed at an operating temperature of 150 K. Spectra of carbon K ( E = 277 eV), taken separately for both subareas of the CCD, are shown in Fig. 3. Even

I. DEVICE PHYSICS

194

R. Hartmann et al. / Nucl. Instr. and Meth. in Phys. Res. A 377 (1996) 191

2000

1500

. . . . . . . . .

,

. . . . . . . . .

i

. . . . . . . . .

n . . . . . . . . .

,

196

10000

. . . . . . . .

~C~30

-

c 1000

100

I0

500

,HIIIII. 0

1O0

200

300

400

500

0

2OO0

Ener~ lev]

Fig. 3. Carbon K,, spectra of a back illuminated CCD on (100) silicon material with a differently treated entrance side. The detector was fabricated with a shallow p+'-unplantatmn" on its entrance side (dotted line spectrum). Approximately half of the entrance window was re-implanted with phosphorus to increase the electric field. A peak shift in the spectrum almost vanished (solid line). The data shown here are not corrected for gain variations and charge transfer losses. A full data analysis of a CCD is given in Ref. [6]. Table l Peak position and width of carbon K line (E = 277eV) with different entrance window implantations. The device was produced on (100) silicon material Implantation

Electronic noise [ e ]

Peak position [eV]

Peak width

p+ through SiO, p~ through SiO, + n

6.5 7.1

232.1 262.8

117.1 101.6

though the electronic noise was slightly higher for the enhanced field area, the obtained peak width of 101.6eV (FWHM) is lower than the one obtained at the other area (117.1 eVFWHM). Apparent is also a shift in the peak position between the two spectra. Table 1 gives a brief summary of the measured results. For an energy to ADCchannel calibration, the aluminium K line (E = 1486 eV) was used, where a peak shift due to entrance window effects was supposed to be negligible. Table 2 Peak to valley ratios of mangan K line for different wafer materials and backside implantations. The background value was calculated as the mean value of counts in an energy interval between 800 eV and 1200 eV. Material

Implantation

Peak to valley ratio

Noise [e ]

(111}/{100) ~ (11 I) (100) (100)

p+ p+ through SiOz p + through SiO 2 p+ through SiO 2 + n

2100:1 3700:1 4300 : 1 5700:1

6.4 5.9 6.5 7.1

"No differences in the spectral resolution were observed between the different silicon materials for a standard p~ implantation.

4000 Ene~g~ [ e v ]

6000

8000

Fig. 4. ~Fe spectrum of a CCD with a field enhanced entrance window implantation at an operating temperature of 150 K. The peak to valley ratio reaches a value of 5700 to 1 (also see Table 2).

For X-ray energies above l keV, an entrance window related shift in the pulse-height spectra vanished. As can be seen in Table 2 though, the peak to valley ratio for higher energy photons improved significantly. Fig. 4 shows the spectrum of a radioactive ~SFe-source, recorded with an enhanced field CCD. Reminding that the additional nimplantation could not be annealed properly, the results might improve for the next production cycle.

4. Modeling the detector In order to explain the measured spectra (Figs. 2 - 4 ) by a calculation, the response function of the detector has to be described by a numerical model. By following the path of an incident X-ray photon, the absorption position is given by dN/ckr =

NJ.Zs~(E.,) exp(-/xs,(E )x),

(2)

where /Xs~(E~) is the absorption coefficient in silicon as a function of X-ray energy [2]. No evidence has been found yet on a dependence of/.ts~(E~) on the crystal orientation of silicon. After the absorption process, the energy of the incident X-rays is transferred to photoelectrons plus Auger electrons and, at higher energies, fluorescence X-rays. These primary electrons will generate a whole cascade of secondary electrons along their trajectories in the silicon, resulting in a charge cloud of thermalized electrons. We did not calculate the electron interaction process with the silicon material in detail, but adopted a rather simple geometrical model [7,8]. Assuming an isotropic angular distribution for generated electron, the density of therrealized charge carries is almost uniform within a sphere of radius R, which is larger than the range of the primary electrons. As a result, we obtain a superposition of

195

R. Hartmann et al. / Nucl. Instr. and Meth. in Phys. Res. A 377 (1996) 191-196

spherical charge clouds for each generated primary electron. To calculate the charge collection efficiency r/(x) as a function of detector parameter, a device simulation program TOSCA was used [9]. This allows to take doping profiles as maintained by SIMS or spreading resistance measurements into account as well as higher recombination rates within the implanted area. We developed a detector model, which made it feasible to count signal electrons, originally positioned at any arbitrary point in the detector, after drift and diffusion at the collecting anode. By repeating this for a singular charge packet over the entire detector width, the charge collection efficiency is directly obtained. Computer simulations of doping profiles with consideration of the crystal orientation show a deeper penetration of boron atoms in our (111) material [10]. During implantation all wafers have been tilted off the plain vector by an angle of 7 ° in order to suppress channeling. In case of (111) material, the effective tilt angle between the ion beam and the (111) crystal axis amounts only approximately 3 ° . These wafers were cut out of the monolithic silicon crystal by an angle of 4 ° in order to enable an epitactical growth of a low resistivity silicon layer on the electronic side of the subsequent detector. A smaller tilt angle leads to a definitely deeper penetration of implanted boron atoms (dashed-dotted line in Fig. lb) due to an enhanced channeling. The rotation angle about an axis perpendicular to the center of the wafer also needs to be taken into account for an optimization of doping profiles. The wafer flat was rotated 22 ° off the tilt axis for all implantations, which leads to identical profiles as a 30 ° rotation. Fig. 5 shows numerical calculated values for the charge collection efficiency, being based upon doping profiles as shown lb. The small variations of the CCE close to the Si-SiO 2 interface between both wafer materials are mostly due to different interface states. The range of the asymptotic rise of the CCE is very much correlated with the

1.0

'0.8

y

..,....,. "" ' " ' "

,g 0.6

_/ .."

range of the implanted boron atoms, drawing the conclusion that the annealing process following the ion implantation was not sufficient to completely heal the crystal. Signal electrons drifting through the damaged region experience a higher probability to be captured by traps. For (111) material, the slope of the CCE is lower than for {100) material, since the integrated density of crystal defects from the detector inside towards the interface is higher. A strong variation of the effective entrance window width on the annealing temperature had been observed with a-particles [11] and will be the aim of further research for our X-ray detectors. The charge collection efficiency r/(x), as defined above, is a photon energy invariant function, describing the detector quality for the collection of signal electrons. By a simple convolution of the generated charge cloud distribution, which indeed is depending on the incident photon energy, with r/(x) and averaging overall possible relaxation pathways, a reduced efficiency ~(x) is obtained: ¢1(x, E) =

~ relaxatilJn pathways

P(E)

~'~ primary ' electrons

rl(x + r)p(r)A(r) d r , R

(3) where P(E) denotes the probability of each relaxation pathway, depending on the photon energy [12], R the mean range of all generated secondary electrons by one primary electron, A(r) dr the volume of a slice with thickness dr, as cut out of a sphere with radius R, and p(r) the charge density within the charge cloud created by one single primary electron. As stated above, we took the charge density as being uniform, so that p ( Q / V ) e I. . . . . . . primary" At the present state of our simulation program, an average over all relaxation pathways was performed. Thus ¢/(x) X Ex_r,, shows the calculated pulse-height of an incident X-ray photon, being absorbed at depth x within the detector. By repeatedly applying the number of photons AN, being absorbed within an interval Ax, over the entire detector width, the detector response function is obtained. Simulated line shapes for several energies and wafer materials are shown in Fig. 6. For each wafer material, one function for a charge collection efficiency explained the measured spectra in the energy range between 900 eV and the lower energy resolution limit, which was 150eV for (100) respectively 350 eV for (111) material. =



o

g 0.4 5. Conclusion 0.2

0.0

0

I O0

200

Ronge [ n m ]

300

400

500

Fig. 5. Calculated charge collection efficiencies for detectors with a shallow p+-implant on (100) (solid line) and {111) (dotted line) silicon material.

The technologies described in this paper represent important advances in the performance of back illuminated silicon X-ray detectors. However special attention has to be put on the wafer entrance side protection during fabrication and its orientation during ion beam implantation.

I. DEVICE PHYSICS

196

R. Hartmann et al. / Nucl. Instr. and Meth. in Phys. Res. A 377 (1996) 191-196

,, . . . . . . . .

10.00

[,, . . . . . . . . .

, .........

, ........

:.



, .........

, ........

. . . . .

10.000

.

, ,350 eV

.

.

,

.

.

m

1.000

~

O.lOO

.

,

•

.

.

, 900

,

.

.

.

,

-

.

.

eV

1.00 o 0

_o

k~ O

0.10

rv 0.01

0.001 0

1O0

200

300

E n e r g y

400

500

600

400

200

0

600 800 Energy [eV]

[eV]

10.00

1000

1200

1000

1200

I 0.000



, 350



eV

¢o

c

0.010

,

900 eV

1.000

1.00

g L~

-~

=

o.loo

n.

0.010

O.lO

0.01

0.001 0

1O0

200 300 400 Energy [eV]

500

600

, i

0

'

200

400 600 800 Energy [eV]

Fig. 6. Comparison of normalized response functions for photon energies of 350eV and 900eV and different wafer materials with calculations. Measurements are represented by dotted lines. The spectroscopic results of detectors with a back p+ and an additional n-implantation have shown, that an increase of the electric field at the entrance window will improve the signal charge collection significantly. Its implementation into the regular detector fabrication process will allow spectroscopy down to the silicon L-edge.

Acknowledgement The authors are indebted to M. Posselt and B. Schmidt from Forschungszentrum Rossendorf/Dresden for their support of performing numerous doping profile simulations and to H. Kranz from Fraunhoferlnstitut ftir Festkrrpertechnologie/Miinchen for his cooperation during the reimplantation of mounted detector chips. We also want to thank Eurisys Measures for supporting this work.

References [1] R. Hartmann, E Lechner, L. Striider, F. Scholze and G. Ulm, Metrologia 32 (1995), in press.

[2] B.L. Henke, P. Lee, T.J. Tanaka, R.L. Shimabukuro and B.K. Fujikawa, At. Data Nucl. Data Tables 27 (1982) 1. [3] E Lechner et al., Nucl. Instr. and Meth. A 326 (1993) 284. [4] F. Scholze, M. Krumrey, E Miiller and D. Fuchs, Rev. Sci. Instr. 65 (1994) 3229. [5] H. Ryssel and I. Ruge, Ionenimplantation (Teubner, Stuttgart, 1978). [6] H. Soltau et al., these Proceedings (7th Europ. Syrup. on Semiconductor Detectors, Schloss Elmau, Bavaria, Germany, 1995) Nucl. Instr. and Meth. 377 (1996) 340. [7] Y. Inagaki, K. Shima and H. Meazawa, Nucl. Instr. and Meth. B 27 (1987) 353. [8] F. Scholze and G. Ulm, Nucl. Instr. and Meth. A 339 (1994) 49. [9] H. Gajewski, B. Heinemann, H. Langmach, G. Telschow and K. Zacharias, TOSCA Handbuch, Institut fiir Angewandte Analysis und Stochastik, Berlin (1992). [10] M. Posselt, Radiat. Eft. Def. Solids 130/131 (1994) 87. [11] T. Maisch, S. Kalbitzer et al., Nucl. Instr. and Meth. A 288 (1990) 19. [12] G.W. Frazer et al., Nucl. Instr. and Meth. A 350 (1994) 368.