Summary for abstract: Measurements of Charge Transfer Inefficiencies in Highly Irradiated CCDs with High-Speed Column Parallel Readout

Please download to get full document.

View again

of 6
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Information Report
Category:

Slides

Published:

Views: 92 | Pages: 6

Extension: PDF | Download: 0

Share
Description
Summary for abstract: Measurements of Charge Transfer Inefficiencies in Highly Irradiated CCDs with High-Speed Column Parallel Readout
Tags
Transcript
  Summary for abstract: Measurements of ChargeTransfer Inefficiencies in Highly Irradiated CCDswith High-Speed Column Parallel Readout Salim Aoulmit, Khaled Bekhouche, Lakhdar Dehimi, Dahmane Djendaoui, Nouredine Sengouga, Andr´e Sopczak   Abstract —The Nobel Prize-winning invention of an imagingsemiconductor circuit (the CCD sensor) has important applica-tions for particle physics detectors. The Charge Coupled Devices(CCDs) have been successfully used in several high energyphysics experiments over the past two decades. Their high spatialresolution and thin sensitive layers make them an excellent toolfor studying short-lived particles. The Linear Collider FlavourIdentification (LCFI) Collaboration has been developing Column-Parallel CCDs for the vertex detector of a future Linear Colliderwhich can be read out many times faster than standard CCDs.The most recent studies are of devices designed to reduce boththe CCD’s intergate capacitance and the clock voltages necessaryto drive it. A method has been developed to measure the ChargeTransfer Inefficiencies. CCD prototypes have been irradiated inseveral steps and the resulting damages have been investigated.The performance of the irradiated devices have been studies fora range of operating temperatures and readout frequencies. TheCCD prototype continues operation with an irradiation of 164krad. I. I NTRODUCTION The invention of an imaging semiconductor circuit (theCCD sensor) [1], [2] has important applications for parti-cle physics detectors. Figure 1 illustrates a vertex detectorgeometry with five CCD layers (ladder 1-5). The study of radiation hardness is crucial for these applications [3]–[5]. TheLCFI collaboration has been developing and testing new CCDdetectors for about 10 years [3]–[5]. Previous experimentalresults on CCD radiation hardness were reported for examplein [6]–[10]. Several theoretical models have increased theunderstanding of radiation damage effects in CCDs [11]–[15].Simulation and modeling of CCD radiation hardness effectsfor a CCD prototype with sequential readout was reported atIEEE2005; comparing full TCAD simulations with analyticmodels was reported at IEEE2006; simulation and modelingof a CCD prototype with column parallel readout (CPCCD)was reported at IEEE2007 and in [14]. Figure 2 shows thecharge transfer inefficiency (CTI) determined with an analyticmodel at different frequencies for temperatures between 100Kand 550K. Experimental measurements using a method todetermine the CTI were performed with a CPCCD prototypeCPC-1 at a test stand at Liverpool University [16]. This work focuses on a new CPCCD prototype, CPC-T, at a test standat Oxford University. The high radiation environment nearthe interaction point at a future Linear Collider damages theCCD material which leads to defects acting as electron trapsin the silicon. The radiation level at a Linear Collider isestimated to be  5 × 10 11 e/cm 2 and  10 10 neutrons/cm 2 peryear at the inner vertex detector layer (14 mm radius) [17],[18]. The mechanism of creating traps has been discussed inthe literature [19]–[21]. These traps result in charge transferinefficiency. Fig. 1. Illustration of a vertex detector concept with five CCD layers withradii 15, 26, 37, 48 and 60 mm. The column parallel technology is in development to copewith the required readout rate. The CPC-T used is a 4-phasevariant of the CPCCD technology capable of 50 MHz readoutfrequency. Experimental work at Liverpool University on anun-irradiated CPC-1 led to CTI values compatible with zerobut with rather large uncertainties [16]. A method to determinethe CTI value, aiming for small CTI uncertainties, was devel-oped and tested with an un-irradiated CPC-T prototype [22].Currently measurements with an irradiated CCD are ongoingwhere the irradiation level is up to about 164 krad [23].Figure 3 [23] shows that the CCD continues to operate afteraccumulated doses of 44, 84, 124 and 164 krad.II. T HEORY Soft X-ray photons (0.1 to 10 keV) interact with siliconatoms within the depleted layer. The depletion layer thicknessis a parameter that determines the quantum efficiency at ener-gies above 4 keV [24]. The absorbed energy generates multiplee-h pairs. For a 5.9 keV X-ray source, one event (photon)generates a cloud of approximately1620 electrons (Fig. 4 [25])contained within a diameter less than one micrometer [26].The charge from a single X-ray photon, generated within thedepletion region of a target pixel, is not transferred completely  Fig. 2. CTI values from an analytic model as a function of temperature in atwo-phase CPCCD for the two traps, 0.17 eV and 0.44 eV with a concentrationof   10 12 cm − 3 and  1%  hit (pixel) occupancy at readout frequencies 10, 25and 50 MHz.Fig. 3. Signal over noise ratio for irradiations completed with accumulated44, 84, 124 and 164 krad from an 50 kV X-ray source with 30 mA and adose rate 19.9 rad/s for  12  µ m thick Si (active thickness). The uncertainty isabout 10% mainly due to manual positioning. to the next pixel due to two main effects: the generation of thermal dark charge within the depletion region and the trap-ping of signal charge within the n-buried channel [27]. Sincethe buried channel is within the depletion layer, the importantmechanisms are the capture of signal from the conduction bandto the trap level and their subsequent emission back to theconduction band [12]. Therefore, the X-ray event exhibits a‘tail’ of deferred charge. Also, the charge generated in thefield-free region diffuses into neighboring pixels and adds tothe ‘tail’ of deferred charge. The size and shape of this tailis a sensitive indicator of charge transfer inefficiency. X-raystimulation is therefore extremely valuable in characterizingthe CTI [25]. Many analyses have been made to simulate theeffect of traps via the emission and capture processes [13],[27], [28]. The following simplified equations, based on earlierwork by Shockley, Read and Hall [29], [30], have been used Fig. 4.  55 Fe X-ray interacting with a CCD. A 5.9 keV photon generates anelectron cloud of approximately 1620  e − . to analyse the CTI: dn t dt  = − n t τ  e + ( N  t − n t ) τ  c τ  c  = 1 σ n v th n e τ  e  = exp( E  t /kT  ) σ n v th N  C  where  n t  is the density of filled traps,  N  t  is the total densityof traps,  E  t  is the trap energy level below the bottom of theconduction band,  τ  e  is the emission time constant,  τ  c  is thecapture time constant,  σ n  is the trapping cross section,  v th  isthe thermal velocity of carriers,  N  C   is the effective density of states in the conduction band and  n e  is the density of electronsin the conduction band. For a detailed analytic model, thefollowing parameters have been taken into account: •  the order of magnitude of the emission and capture timeconstants compared to the shift time (time needed for acharge packet to move from one pixel to another). •  the shape of the electrostatic potential, which can beassumed to be placed in the middle of the well. •  the level of the signal charge (density of free electrons)within the potential well in comparison to the total densityof traps.III. T EST  S TAND FOR  CCD O PERATION A test stand has been set up with readout electronics and afreezer unit as shown in Fig. 5. The temperature range of thefreezer is from room temperature down to about − 60  ◦ C. Finecontrol of the CPC-T temperature is done using a CAL9500Pcontroller (the temperature is kept constant within  0 . 1  ◦ C).A flux of boiled nitrogen is introduced into the motherboardbox to purge water vapour. The CPC-T chips come in 2 mainvariants: inherent 4-phase CCD driven as 2-phase CCD, and‘pedestal’ 2-phase CCD with 2 additional DC-biased gates.The former was used for this measurement. The first andsecond gates of each pixel, P1A and P1B, are driven by  Fig. 5. Picture of the CPC-T readout. The CPC-T mother board is insidethe freezer.Fig. 6. Clocking scheme for the 4-phase variant. This variant is driven as atwo-phase CCD with additional DC voltage (OPV) to the voltage of the firstgate of a pixel. Phase1, and P1A is offset by the DC voltage OPV (Offsetand Pedestal Voltage), as shown in Fig. 6.The CPC-T has  500 × 10  pixels with a pixel size of   20 × 20  µ m 2 . Initial measurements have been performed on an un-irradiated device in standalone mode, where the signals fromfour columns of the CCD were amplified and connected toexternal 14-bit ADCs. A  55 Fe source emitting 5.9 keV X-rayswas attached to a holder at a distance of 1 cm from the CCDto provide the signal charge. The schematic diagram in Fig. 7illustrates the electronics used to drive and read out the CPC-T. The apparatus is controlled by a LabView program throughinterface modules. The BVM2 sequencer receives the masterclock of 1 MHz from the function generator to provide four Fig. 7. Schematic diagram of the CPC-T readout. signals, two for ADC (clock and trigger) and two to trigger thegenerators which produces a CCD clock and reset gate signals.The 2-resets configuration, when one reset is applied beforereading the first pixel of the CCD and one after reading thelast pixel, is used in this measurement. This configurationleadsto low noise since the reset noise is absent. The occupancy isabout 1% for the integration time of 100 ms given the strengthof the X-ray source and the experimental layout. We mademeasurements with two different numbers of frames (1000and 10000) to study the effect on the statistical uncertainties.Our method is based on the typical methods used for seriallyread out CCDs, where the CTI is determined by fitting a lineto the readout charge signal as a function of the pixel number.A linear function can be expected when the CTI is small.IV. S IGNAL  M EASUREMENT AND  CTI D ETERMINATION M ETHOD The fast ADCs convert the signal charge after amplificationwith a wideband preamplifier. The four columns are read outin 4 channels by ADCs. The signal charges of 500 pixelswere acquired in 1000 and 10000 frames per measurement foreach temperature. Using 10000 frames leads to a sufficientstatistical precision (around few times  10 − 6 for the CTImeasurement). The collected data have been analysed usingMATLAB [31]. First, we begin by applying correlated doublesampling, where the difference between the signals of twoconsecutive pixels is taken to be the signal charge collectedby the latter pixel. This reduces some of the noise components(e.g. 1/f, kTC, white noise, etc.) in the CCD signal. As an  example, Fig. 8 shows the pulse-height distribution of ADCcodes for the last 10 adjacent pixels in column 2. −5005010015020025010 0 10 1 10 2 10 3 10 4 ADC codes    C  o  u  n   t  s Distribution of ADC codesFit noiseFit X−rayReadout frequency=1 MHz,T=−27.5 o C, OPV=−0.81 V, CLK voltage=4 V,integration time=100 ms, 10000 frames. (I)(II)(III) Fig. 8. Distribution of ADC codes for channel (column) two. Three regionsare observed: (I) the high peak region which represents the noise, (II) theregion separating the two peaks which represents the charge sharing betweenpixels, and (III) the X-ray peak region which represents the fully collectedcharge in a single pixel. The noise and X-ray peaks are fitted by a Gaussianfunction to determine the centroids. The X-ray signal is the difference betweenthe two centroids. The charge transfer inefficiency (CTI) in one pixel is definedas the ratio of signal lost during transfer (captured by traps) tothe initial signal charge. The CTI is calculated following thesesteps: creation of a histogram with ADC codes of 10 pixels in acolumn. These pixels have nearly the same baseline (defined asthe centroid of the noise distribution). The histogram creationis repeated 50 times to cover all 500 pixels in a column.Fits with Gaussian functions are made to the noise and X-ray peaks. We use  x 0 − nσ  and  x 0  +  nσ  as limits for noisepeak and X-ray peak ( x 0  and  σ  are the centroid and thestandard deviation respectively of the two Gaussian functionsresulting from fitting the noise and X-ray peaks) to determinenoise centroid and X-ray centroid for each pixel. The factor n  is chosen between 1 and 3 depending on the amount of charge sharing. The factor is the same for all pixel groupsin one measurement. It can vary from one measurement toanother depending on the distribution shape. The value  n  = 1 is most often used. The X-ray signal for each pixel is thedifference between the X-ray peak centroid and the noisepeak centroid. Figure 9 shows the distribution of the signalvalues as a function of pixel number. The distribution is fittedwith the first-order polynomial function  P  0  +  P  1  j , where  P  0 corresponds to the charge at the first pixel,  P  1  is the slopeand  j  is the pixel number. The CTI is determined using CTI   = − P  1 /P  0 .V. CTI R ESULTS  P RE -I RRADIATION Figure 10 shows the CTI values as a function of temperaturefor an un-irradiated CPC-T for 1000 and 10000 frames. The 50100150200250300350400450188189190191192193194195196Pixel number    X −  r  a  y  p  e  a   k   (   A   D   C   c  o   d  e   ) X−ray peakLinear fitReadout frequency=1 MHz,T=−27.5 o C, OPV=−0.81 V, CLK voltage=4 V,integration time=100 ms, 10000 frames. Fig. 9. Linear fit to average ADC codes. The X-ray centroid of each pixel iscalculated by averaging ADC codes within the interval  x 0 − nσ  and  x 0 + nσ where 1 ≤ n ≤ 3. CTI has been calculated using a linear fit of signal valuesversus pixel number. Figure 11 shows a comparison with theCPC-1 measurement [16] taken at the test stand in Liverpool.Uncertainties have been reduced mostly by increasing thenumber of frames. For this CCD with 500 pixels per column aCTI value of   10 − 5 means that only  0 . 5%  of the signal chargeis lost, which is acceptable in normal operation. The apparenttrend of the CTI at high temperatures in the operating rangeused is probably due to the contribution of two effects. First,there is the effect of thermal carrier generation (dark current)which is highly temperature-dependent.The dark current, non-uniform by nature, can have a large effect on the signal chargetransfer for high temperatures, long integration time and largenumber of pixels in the column [32]. Second, there is thepossibility of the presence of low trap density that could havebeen created during the long duration (around two years) of exposure to a soft X-ray source while studying the device.This significant positive value of CTI before irradiating theCCD was observed experimentally and modeled by a simpleanalytic model by including one trap level [4]. Using ouranalytic model [15], the CTI is expressed as CTI   = 2 N  t n s [1 − exp( − t ( 1 τ  c + 2 τ  e ))] × [( τ  s τ  e (1 − exp( −  tτ  s ))(1 − exp( − t (  1 τ  s +  1 τ  e ))))exp( − tτ  e ) − exp( − t w τ  e )] . We have fitted the CTI curve including two deep traps asshown in Fig. 12. Both are electron traps at 0.37 eV and0.44 eV below the bottom of the conduction band and havinga trapping cross-section  σ n  = 3 × 10 − 15 cm 2 . We have consid-ered that there is no interaction between the two traps, so theyinteract independently with the signal charge. Therefore, the  −60−55−50−45−40−35−30−25−20−15−2−1012345678x 10 −5 Temperature ( o C)    C   T   I   (   1   0  −   5    ) 1000 frames10000 framesReadout frequency=1 MHz,integration time=100 ms,OPV=−0.81 V,CLK voltage=4 V. Fig. 10. CTI as a function of temperature for different numbers of frames.The error bars have been significantly reduced by increasing the number of frames. total CTI is the sum of CTIs resulting from the effect of thetraps. The fit is in good agreement with the data and showsthat the 0.44 eV trap is the dominant one in this range of temperature as its density ( N  t 2  = 5 . 22 × 10 10 cm − 3 ) is muchlarger than that of the 0.37 eV trap ( N  t 1  = 2 . 63 × 10 9 cm − 3 ).The accuracy of the CTI calculation can be improved bypositioning the  55 Fe source so that it irradiates uniformlythe CCD, carefully choosing the gain to use the ADC in itsmaximum dynamic range and acquiring data in a large numberof frames. The non-uniformity of the X-ray source coveragehas an effect on the CTI determination. This is well understoodand reproduces the estimate based on a geometry where thesource is placed 1 cm away from the CPC-T. Figure 13shows the non-uniformity of the X-ray source coverage. Thefigure contains two curves, the measured and the estimatedX-ray distributions. The measured distribution is determinedby counting all ADC codes above the noise threshold  x 0 +3 σ .The estimated distribution is determined using the followingformula for the given geometry: f  ( n ) =  h/l  p  2  h/l  p  2 +  n − n 0  2  3 / 2 where  f   is the distribution of X-rays upon the CPC-T,  n  isthe pixel number,  h  is the distance between source and CPC-T,  l  p  is the length of one pixel and  n 0  is the pixel numbercorresponding to the vertex position. In order to avoid theeffect of the non-uniformity, the first and last 50 pixels areexcluded from the fit to the averages for the CTI determination(Fig. 8). −55−50−45−40−35−30−25−2005101520253035Temperature ( o C)    C   T   I  u  n  c  e  r   t  a   i  n   t  y   (   1   0  −   6    ) Liverpool results (2008)Oxford results (2009) Fig. 11. Comparison of the Oxford results with 10000 frames with theLiverpool results [16] where the number of frames was 5000 and a fractionof data was lost because of a sampling inefficiency. −60−55−50−45−40−35−30−25−20−15012345678Temperature ( o C)    C   T   I   (   1   0  −   5    ) DataFit E  C  −E  t1 =0.37 eV,N  t1 =(2.31  ±  0.93)x10  9   cm  −3  ,E  C  −E  t2  =0.44 eV,N  t2  =(5.44   ±  0.64)x10  10   cm  −3  . Fig. 12. Non-linear fit of the measured CTI using our analytic model [15].The model includes two acceptor traps, 0.37 and 0.44 eV below the conductionband. A trapping cross-section  σ n  = 3  ×  10 − 15 cm 2 is used for both traps. VI. C ONCLUSIONS AND  O UTLOOK The application of CCDs as particle detectors lead to thevery interesting field of radiation hardness studies. Specifically,as a measure of radiation damage effects, the charge transferinefficiency (CTI) has to be determined in measurements.These measurements have been discussed for an example of an un-irradiated CPC-T operated in the temperature range − 15  ◦ C to − 60  ◦ C (freezer cooling) with different numbers of frames, 1000 and 10000. A clear X-ray signal is extracted bycalculating the difference between the noise centroid (baseline)and the X-ray centroid. The statistical uncertainties have beenreduced compared to a previous work with CPC-1 [16].
Recommended
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x