Time dependence of charge losses at the Si-SiO2 interface in p+n-silicon strip sensors
Thomas Poehlsen, Eckhart Fretwurst, Robert Klanner, Joern Schwandt, Jiaguo Zhang
TTime dependence of charge losses at the Si-SiO interface in p + n -silicon strip sensors Thomas Poehlsen ∗ , Eckhart Fretwurst, Robert Klanner, Joern Schwandt, and Jiaguo Zhang Institute for Experimental Physics, University of Hamburg, Luruper Chaussee 149, 22761 Hamburg, Germany
Abstract
The collection of charge carriers generated in p + n -strip sensors close to the Si-SiO interfacebefore and after 1 MGy of X-ray irradiation has been investigated using the transient currenttechnique with sub-nanosecond focused light pulses of 660 nm wavelength, which has an absorptionlength of 3.5 µ m in silicon at room temperature. The paper describes the measurement and analysistechniques used to determine the number of electrons and holes collected. Depending on biasinghistory, humidity and irradiation, incomplete collection of either electrons or holes is observed.The charge losses change with time. The time constants are different for electrons and holes andincrease by two orders of magnitude when reducing the relative humidity from about 80 % to lessthan 1 %. An attempt to interpret these results is presented. Keywords: silicon sensors, charge losses, X-ray-radiation damage, humidity, biasing history
1. Introduction
In this paper the time dependence of charge losses close to the Si-SiO interface in p + n -siliconstrip sensors after changing the bias voltage in environments of different humidities is investigated.Measurements were performed on a non-irradiated sensor and on a sensor irradiated by ∼
12 keVX-rays to a dose of 1 MGy. The work is part of the study of X-ray-radiation damage of segmentedsilicon sensors for the AGIPD detector [1] at XFEL, the European X-ray Free-Electron Laser [2],where X-ray doses of up to 1 GGy are expected.In a previous publication [3] it has been shown that time resolved measurements (TCT -Transient Current Technique) for electron-hole pairs generated by a focussed laser close to theSi-SiO interface, allow investigating the electric fields and the properties of possible accumulationlayers. For a non-irradiated sensor it has been observed that, after biasing the sensor, the Si-SiO -interface region is not in steady-state conditions and that the time to reach steady-state conditionsdepends on humidity: It is about two orders of magnitude shorter for high compared to low relativehumidity. In this paper we extend these measurements to a sensor irradiated to an X-ray dose of1 MGy in order to investigate, how the time development of the electric field in the region close tothe sensor surface depends on X-ray-radiation damage.The cause of the time and humidity dependence for the non-irradiated sensor was explained bythe time dependence of the electric boundary conditions on the surface of the sensor: Changing thebias voltage results in surface fields which cause a redistribution of surface charges until a uniformpotential is reached. The strong dependence of the surface resistance on humidity [4] is responsiblefor the big difference in time constants.For given densities of oxide charges and charged interface traps, the electric boundary conditionson the sensor surface strongly influence the formation of accumulation layers and of the electricalfields close to the Si-SiO interface. As the surface boundary conditions can influence the breakdownvoltage [5, 6], their knowledge is highly relevant for the simulation of segmented silicon sensors [7],in particular for high X-ray doses and operation in vacuum. ∗ Corresponding author. Email address: [email protected]. Telephone: +49 40 8998 4725.
Preprint submitted to Elsevier October 9, 2018 a r X i v : . [ phy s i c s . i n s - d e t ] M a y INVESTIGATED SENSOR AND MEASUREMENT TECHNIQUE Figure 1: Schematic layout of the strip region of the DC-coupled Hamamatsu p + n sensor, and coordinate definition.The drawing is not to scale.
2. Investigated sensor and measurement technique
A DC coupled p + n strip sensor produced by Hamamatsu Photonics [8] was investigated.Previous studies of the sensor are reported in [3, 9]. The cross section of the sensor is shown inFigure 1 and relevant sensor parameters are listed in Table 1. Measurements were performed beforeand after irradiation to a dose of 1 MGy (SiO ) by ∼
12 keV X-rays. All measurements were takenat 200 V bias voltage, well above the depletion voltage of 155 V.Pulsed focussed red laser light with a wavelength of 660 nm was used to generate electron-hole( eh ) pairs in the sensor at the p + -strip side, just below the SiO separating the read-out strips. Alaser of this wavelength has a penetration depth in silicon of about 3.5 µ m at room temperature.Hence, the charges were generated close to the Si-SiO interface. The light was focussed to aGaussian with σ ∼ µ m. The pulses had a duration of 100 ps at FWHM, the number of generated eh pairs was about 100, and the repetition rate 1 kHz.The current signals induced in the read-out strip were read out by DC-coupled Femto HSA-X-2-40 current amplifiers ( ∼
50 Ω input impedance, 180 ps rise time) connected to a Tektronixdigital oscilloscope with 2 . Q i induced on read-out strip i was calculated off-line by integrating thecurrent signal over 16 ns.The humidity during the measurements was varied between ≤ ∼
85 %, and thetemperature was ∼ ◦ C. Humidity and temperature were logged with a TFD128 temperatureproducer Hamamatsucoupling DCpitch 50 µ mdepletion voltage ∼
155 Vdoping concentration ∼ cm − single strip capacitance 1.4 pFgap between p + implants 39 µ mwidth p + implant window 11 µ mdepth p + implant unknownaluminium overhang 2 µ mnumber of strips 128strip length 7.956 mmsensor thickness 450 µ mthickness SiO
700 nmpassivation layer unknowncrystal orientation (cid:104) (cid:105)
Table 1: Parameters of the Hamamatsu sensor.
DETERMINATION OF CHARGE LOSSES
3. Determination of charge losses
This chapter presents the method used to determine the charge losses and shows evidence thatin the p + n -strip sensor investigated, depending on the biasing history, humidity and X-ray dose,situations without charge losses, with electron losses and with hole losses close to the Si-SiO interface have been observed.The electrons and holes generated by the laser drift in the electric field in the sensor therebyinducing currents in the electrodes. The current induced in electrode i by a charge q moving atposition (cid:126)x with drift velocity (cid:126)v dr can be calculated using the weighting potential φ iw : I i = q · (cid:126)E iw ( (cid:126)x ) · (cid:126)v dr ( (cid:126)x ) , (cid:126)E iw ( (cid:126)x ) = (cid:126) ∇ φ iw ( (cid:126)x ) . (1)The weighting potential φ iw ( (cid:126)x ) is a measure of the capacitive coupling of a unit charge atposition (cid:126)x to electrode i . It is one on electrode i , zero on all other electrodes, and between zeroand one in the sensor volume. When a positive charge q moves from position (cid:126)x to electrode i ,it induces a positive current with total charge Q i = q · (1 − φ iw ( (cid:126)x )) in electrode i and negativecurrents with total charges Q j = − q · φ jw ( (cid:126)x ) in all other electrodes j (cid:54) = i .For a p + n -strip sensor electrons drift to the n + -rear contact and holes to the p + strips. Electronscollected at the rear contact induce positive signals in all p + strips. Holes however, induce a positivesignal in the p + strip on which they are collected and negative signals in all other p + strips. If N q eh pairs are generated anywhere in the sensor, and all holes are collected by strip i and all electronsby the rear contact, both electron as well as hole currents induced in strip i are positive, resultingin an induced charge Q i = N q · q , where q is the elementary charge. For strips j the currentsinduced by the electrons are positive and the currents induced by the holes negative, resulting in abipolar signal of total charge Q j = 0.If not all charges are collected by the electrodes but some get stuck at position (cid:126)x , the chargeinduced in a read-out strip RO which does not collect holes is: Q RO = (cid:90) I RO dt = ( N h · q − N e · q ) · (0 − φ ROw ( (cid:126)x )) = ( N e − N h ) · q · φ ROw ( (cid:126)x ) , (2)where N e is the number of electrons collected by the rear contact and N h the number of holescollected by the strips. For full charge collection as well as for eh recombination N e = N h and Q RO = 0. If only electrons are lost Q RO is positive and negative if only holes are lost.In [3] it has been shown that the charge distribution on the sensor surface at the strip sideinfluences the electric field in the sensor close to the Si-SiO interface and that, depending on thebiasing history and X-ray dose, situations with losses of electrons, losses of holes and no lossescan be realised. This is demonstrated in Figure 2 (taken from [3]) which shows three examples ofcurrents measured on the read-out strip RO , positioned at x = 0. For the coordinate system werefer to Figure 1. The sensor was biased to 200 V, well above the depletion voltage of 155 V, andthe electron-hole pairs were generated by the laser at x = 30 µ m just below the Si-SiO interface.As this position is closer to the neighbouring strip N than to the readout strip RO , holes will onlyreach RO by diffusion and their number will be small. • The solid (blue) line corresponds to the situation e = h , where both electrons and holesare collected (non-irradiated sensor in steady state at 200 V). Holes are collected quickly( ∼ N inducing a negative signal on the read-out strip RO . Electrons, whichhave to pass the entire sensor to reach the rear electrode, are collected more slowly ( ∼
10 ns)inducing a positive signal in the read-out strip. Thus, the signal is bipolar and its integral isapproximately zero. • The dashed (black) line corresponds to the situation e (cid:28) h , where mainly holes are collected(sensor irradiated to 1 MGy by ∼
12 keV X-rays measured after ramping the voltage from 0
TIME DEPENDENCE OF CHARGE LOSSES Figure 2: Induced current in the read-out strip at x = 0 for electron-hole pairs generated at x = 30 µ m. Complete(solid line) and incomplete (dotted and dashed lines) charge collection are shown. For details see text. to 200 V in a dry atmosphere). The positive signal of the electrons is missing and only ashort negative signal induced by the holes collected at the neighbouring strip N is observed. • The dotted (red) line corresponds to the situation e (cid:29) h , where mainly electrons are collected(non-irradiated sensor brought into a steady-state at 500 V and then biased at 200 V in a dryatmosphere). The negative signal of the holes is missing and only a positive current inducedby electrons drifting to the rear side is observed.The following method is used for determining the charge losses: Light is injected sufficiently farfrom the read-out strip RO to assure that no holes are collected by RO . For the measurementspresented in the next chapter we choose RO at x = 0 and inject focussed light at x = 75 µ m,in the middle between two n + strips. To check the consistency of the results data are also takenfor the light injected at x = 40 µ m. In addition, the signal from the rear contact is recorded.Assuming that only one type of charge carrier is lost and eh recombination can be ignored, the“lost charge” is Q lost = ( N e − N h ) · q = Q RO /φ ROw ( (cid:126)x ) . (3) Q lost is positive for hole losses and negative for electron losses. Given the quality of the bulksilicon (lifetime of the charge carriers much longer than the charge-collection time) and the factthat the sensor is operated well above the depletion voltage, charges can only be lost at the Si-SiO interface or in a close-by accumulation layer. Thus the weighting potential at the Si-SiO interfaceat the position where the eh pairs are produced is used for the analysis. Compatible with themeasurements presented in [3], we use: φ ROw = 0 .
05 at x = 75 µ m and 0.35 at x = 40 µ m.
4. Time dependence of charge losses
The measurements of the charge losses were performed for different humidities and biasinghistories. For the measurements the p + strips were grounded, either via the DC-coupled amplifiersor via 50 Ω resistors. The bias voltage was applied at the n + rear electrode. In [3] it has beenobserved for a non-irradiated sensor that, after changing the bias voltage, the distribution of surfacecharges is in a non-steady state, as evidenced by time dependent charge losses. It was found thatthe time to reach the steady state has a strong dependence on humidity. Here, we present furtherresults on the time dependence of charge losses for the non-irradiated sensor and for the sensorirradiated to a dose of 1 MGy.The measurement procedure was the following: By biasing the sensor either to 0 V or to 500 V(400 V for the irradiated sensor) in a humid atmosphere ( >
60 % relative humidity) for at least 2hours (typically a day and longer) steady-state conditions of the charge distribution on the surfacewere assumed to have been reached. Then, for the measurements under dry conditions the relativehumidity was reduced. After at least 2 hours under dry conditions, the bias voltage was changed
TIME DEPENDENCE OF CHARGE LOSSES Figure 3: Lost charge Q lost , in units of elementary charges, as a function of time after the change of the bias voltagefor the non-irradiated sensor in a dry (red rectangulars) and humid (black dots) atmosphere for light injection at x = 75 µ m. Left: Increase of the bias voltage from 0 V to 200 V. The green curves show control measurements forlight injection at x = 40 µ m. Right: Decrease of the bias voltage from 500 V to 200 V. Note that the time scale for“humid” is on the top and for “dry” on the bottom.Figure 4: Lost charge Q lost , in units of elementary charges, as a function of time after the change of the bias voltagefor the irradiated sensor in a dry (red rectangulars) and humid (black dots) atmosphere. Left: Increase of the biasvoltage from 0 V to 200 V. Right: Decrease of the bias voltage from 400 V to 200 V. Note that the time scale for“humid” is on the top and for “dry” on the bottom. to 200 V, the charge-loss measurements started approximately 30 seconds after 200 V had beenreached, and the time dependence of the lost charges determined from measurements of Q RO for eh pairs generated by the laser at position x .The results for the non-irradiated sensor, obtained according to Equation (3), are shown inFigure 3. When changing the bias voltage from 0 to 200 V (Figure 3 left) initially ∼
40 000electrons are lost for both dry and humid conditions for ∼
140 000 eh pairs generated by the laser.In steady-state conditions ∼
10 000 holes are lost. For dry conditions (relative humidity < ∼ ∼
70 %) in ∼
10 minutes. For comparison, also measurements for light injected at x = 40 µ m areshown. They agree with the x = 75 µ m measurements. As the weighting field at x = 40 µ mis 0.35 compared to 0.05 for x = 75 µ m, the fluctuations due to noise are about an order ofmagnitude smaller. However, it cannot be excluded that some of the holes reach the read-outelectrode RO by diffusion. Therefore we only present results for x = 75 µ m in the following.After changing the voltage from 500 V to 200 V (Figure 3 right), initially ∼
65 000 holes arelost for both dry and humid conditions for ∼
100 000 eh pairs generated. In steady-state conditionsthe hole losses reduce to a value between 0 and 10 000. In humid conditions it takes ∼
50 minutesto reach this state, whereas in dry conditions it takes more than ∼
100 hours.
DISCUSSION ∼
30. For hole losses a factor ∼
120 is found, however, as the 78 % humidity hadan uncertainty of ±
10 % for this particular measurement, no strong conclusions can be drawn.Figure 4 shows the results for the sensor irradiated to 1 MGy, again obtained according toEquation (3). After changing the voltage from 0 to 200 V electron losses are observed as forthe non-irradiated sensor. The number of lost electrons however, is different for dry and humidconditions: ∼
65 000 for “dry” and ∼
15 000 for “humid”, for ∼
100 000 generated eh pairs. For“humid” a steady state of approximately zero electron losses is reached after about 1 hour. For “dry”the electron losses first decrease to ∼
45 000 within approximately 10 hours. Then a much slowerdecrease is observed, reaching ∼
35 000 after 160 hours when the measurements were stopped. Thecurrent collection ring of the irradiated sensor showed a large current for voltages above 400 V.Therefore, 400 V was chosen as maximum bias voltage. The initial charge losses after reducingthe voltage from 400 to 200 V are approximately zero for “humid”, whereas 35 000 holes are lostfor “dry”, for 140 000 generated eh pairs. As function of time the charge losses remain zero for“humid”, whereas for “dry” the hole losses steadily decrease, reaching ∼
10 000 after 50 hours.We also note that the charge losses for the irradiated sensor behave quite differently comparedto the irradiated sensor. For the irradiated sensor the initial losses differ between dry and humidconditions, and the time dependencies do not scale.
5. Discussion
In [3] the electron and hole losses and their time dependence have been explained with thehelp of detailed simulations, and their relevance for the operation of segmented p + n sensors forthe detection of X-rays and charged particles discussed: Biasing a p + sensor which has been insteady-state conditions at 0 V, results in a transverse electric field E y (see Figure 1) in the sensorclose to the Si-SiO interface pointing into the sensor, and a tangential field E x on the sensorsurface pointing to the p + implants. The transverse field and a possible electron accumulation layercause the loss of electrons, and the surface field causes a movement of charges on the sensor surfaceuntil the steady-state of a uniform potential is reached. For a sensor in steady state under bias,reducing the voltage results in the opposite situation: The electric field in the sensor points towardsthe SiO and therefore holes are lost. The tangential fields initially point away from the strips untilby charge movement the steady-state of a uniform potential is reached. The strong dependenceof the time dependencies on humidity is explained by the decrease of the surface resistivity withhumidity [4]. If the steady state is reached due to surface currents, a scaling of the time dependencewith the surface conductivity is expected.The new measurement shown in Figure 3 for the non-irradiated sensor confirm the picture:Initial electron losses occur when the voltage is ramped up and hole losses when the voltage isramped down. The initial steady-state charge losses are the same for “dry” and for “humid”. Thesame holds for the steady-state losses. However, the time constants for reaching the steady statefor hole losses is about an order of magnitude larger than for electron losses. This could be relatedto the surface charge distribution in steady state. We also note that in steady state a small holeloss of ∼ and theresulting high electric fields in the SiO are the cause. SUMMARY
6. Summary
The collection of charge carriers generated in p + n -strip sensors close to the Si-SiO interfacebefore and after 1 MGy of X-ray irradiation has been investigated using the transient currenttechnique with sub-nanosecond focused light pulses of 660 nm wavelength, which has an absorptionlength of 3.5 µ m in silicon at room temperature. Depending on the applied bias voltage, biasinghistory, humidity and irradiation, incomplete collection of either electrons or holes has been observedwhen generating the charges at the strip side of the sensor.For the non-irradiated sensor electron losses are observed when the voltage is ramped up fromzero to the operating voltage and hole losses when the operating voltage is reached from highervoltages. As a function of time the charge losses decrease until a steady state with hole lossesbetween 0 and 10 % is reached. The shape of the time dependence and the time constant aredifferent for hole and electron losses. At a given humidity the electron losses are about an orderof magnitude faster. However, the time constants for both electron and hole losses increase byabout two orders of magnitude, when the humidity is reduced from about 80 % to below 1 %. Theobservations are explained by changes of the charge distribution on the sensor surface. Initially,after changing the bias voltage the surface-charge distribution is not in a steady state, and the timeconstant to reach the steady state is related to the surface conductivity of positive and negativecharge carriers, which are strong functions of humidity.For the sensor irradiated to 1 MGy the situation is more complex: As for the non-irradiatedsensor, in a dry atmosphere electron losses are observed when the operating voltage is reachedfrom lower, and hole losses when it is reached from higher voltages. The time constant for the holelosses is similar as for the non-irradiated sensor, however, the shape of the dependence is quitedifferent. For electron losses a much longer time constant is found for the irradiated sensor. In ahumid atmosphere already 30 seconds after the bias voltage is reached, the charge losses are small:approximately 15 % electron losses when the voltage is ramped up, and less than 10 % chargelosses, when it is ramped down. We speculate that chemical effects and electric fields inside theSiO and passivation layer due to charged traps are responsible for the differences.The measurements demonstrate the well-known fact that X-ray-radiation damage in the SiO and its impact on sensor performance is highly complex. Acknowledgements
This work was performed within the AGIPD Project which is partially supported by the XFEL-Company. We would like to thank the AGIPD colleagues for the excellent collaboration. Supportwas also provided by the Helmholtz Alliance “Physics at the Terascale” and the German Ministry ofScience, BMBF, through the Forschungsschwerpunkt “Particle Physics with the CMS-Experiment”.J. Zhang is supported by the Marie Curie Initial Training Network “MC-PAD”.
References [1] B. Henrich et al.,
The adaptive gain integrating pixel detector AGIPD, a detector for the European XFEL ,Nucl. Instr. and Meth. A 633 Suppl. 1(2011) S11, and http://photon-science.desy.de/research/technical_groups/detectors/projects/agipd/ .[2] M. Altarelli, et al. (Eds.),
XFEL: The European X-Ray Free-Electron Laser, Technical Design Report , PreprintDESY 2006-097, DESY, Hamburg 2006, and .[3] T. Poehlsen et al.,
Charge losses in segmented silicon sensors at the Si-SiO interface , Nucl. Instr. and Meth. A700 (2013) 22-39[4] A.S. Grove, Physics and Technology of Semiconductor Devices , John Wiley & Sons (1967).[5] A. Longoni, M. Sampietro and L. Str¨uder,
Instability of the behaviour of high resistivity silicon detectors due tothe presence of oxide charges , Nucl. Instr. and Meth. A 288 (1990) 35.[6] F.G. Hartjes,
Moisture sensitivity of AC-coupled silicon strip sensors , Nucl. Instr. and Meth. A 552 (2005) 168.[7] R.H. Richter et al.,
Strip detector design for ATLAS and HERA-B using two-dimensional device simulation ,Nucl. Instr. and Meth. A 377 (1996) 412.[8] .[9] J. Becker,