Single-Photon Avalanche Diodes (SPAD) in CMOS 0.35 µm technology D Pellion, K Jradi, Nicolas Brochard, D Prêle, Dominique Ginhac To cite this version: D Pellion, K Jradi, Nicolas Brochard, D Prêle, Dominique Ginhac. Single-Photon Avalanche Diodes (SPAD) in CMOS 0.35 µm technology. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, Elsevier, 2015, 787, pp.380-385. ฀10.1016/j.nima.2015.01.100฀. ฀hal-01196570฀ HAL Id: hal-01196570 https://hal-univ-bourgogne.archives-ouvertes.fr/hal-01196570 Submitted on 10 Sep 2015 HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés. 1   Single-­‐Photon   Avalanche   Diodes   (SPAD)   in   2   CMOS  0.35µm  technology   3   D.  Pellion1,  K.  Jradi1,  N.  Brochard1,  D.  Prêle2,  D.  Ginhac1   4     5   1:  Le2i  -­‐  CNRS/Univ.  de  Bourgogne,  Dijon,  France   6   2:  APC  -­‐  CNRS/Univ.  Paris  Diderot,  Paris,  France   7     8   Abstract:   9   Some  decades  ago  single  photon  detection  used  to  be  the  terrain  of  photomultiplier  tube   10   (PMT),  thanks  to  its  characteristics  of  sensitivity  and  speed.  However,  PMT  has  several   11   disadvantages   such   as   low   quantum   efficiency,   overall   dimensions,   and   cost,   making   12   them  unsuitable  for  compact  design  of  integrated  systems.  So,  the  past  decade  has  seen  a   13   dramatic   increase   in   interest   in   new   integrated   single-­‐photon   detectors   called   Single-­‐ 14   Photon  Avalanche  Diodes  (SPAD)  or  Geiger-­‐mode  APD.  SPAD  are  working  in  avalanche   15   mode   above   the   breakdown   level.   When   an   incident   photon   is   captured,   a   very   fast   16   avalanche  is  triggered,  generating  an  easily  detectable  current  pulse.     17   This   paper   discusses   SPAD   detectors   fabricated   in   a   standard   CMOS   technology   18   featuring   both   single-­‐photon   sensitivity,   and   excellent   timing   resolution,   while   19   guaranteeing   a   high   integration.   In   this   work,   we   investigate   the   design   of   SPAD   20   detectors   using   the   AMS   0.35µm   CMOS   Opto   technology.   Indeed,   such   standard   CMOS   21   technology   allows   producing   large   surface   (few   mm2)   of   single   photon   sensitive   22   detectors.   Moreover,   SPAD   in   CMOS   technologies   could   be   associated   to   electronic   23   readout  such  as  active  quenching,  digital  to  analog  converter,  memories  and  any  specific   24   processing   required   to   build   efficient   calorimeters1  (Silicon   PhotoMultiplier   -­‐   SiPM)   or   25   high  resolution  imagers  (SPAD  imager).  The  present  work  investigates  SPAD  geometry.   26   MOS   transistor   has   been   used   instead   of   resistor   to   adjust   the   quenching   resistance   and   27   find  optimum  value.  From  this  first  set  of  results,  a  detailed  study  of  the  Dark  Count  Rate                                                                                                                   1  SiPM   is   often   used   to   measure   the   number   of   photons,   proportional   to   the   particle   energy,   which   interacts   with   a   scintillator.   At   the   opposite,   an   imager   gives   both   the   number  of  hit  pixels  and  there  position.  So,  in  particle  physics,  calorimetry  corresponds   to  the  energy  measurement  of  particles  even  if  any  temperature  measurement  is  done.   28   (DCR)  has  been  conducted.  Our  results  show  a  dark  count  rate  increase  with  the  size  of   29   the   photodiodes   and   the   temperature   (at   T=22.5°C,   the   DCR   of   a   10   µm-­‐photodiode   is   30   2020  count.s-­‐1  while  it  is  270  count.s-­‐1  at  T=-­‐40°C  for  a  overvoltage  of  800  mV).  A  small   31   pixel  size  is  desirable,  because  the  DCR  per  unit  area  decreases  with  the  pixel  size.  We   32   also   found   that   the   adjustment   of   overvoltage   is   very   sensitive   and   depends   on   the   33   temperature.  The  temperature  will  be  adjusted  for  the  subsequent  experiments.     34     35     36   1 37   A  Single-­‐Photon  Avalanche  Diode  (SPAD)  is  a  semiconductor  photon  sensor  operated  in   38   Geiger-­‐mode  where  bias  voltage  is  above  the  diode  breakdown  voltage  (typical  Vbr=10   39   to   100   V)   and   associated   to   a   quenching   circuit   [Ref   1].   A   Silicon   PhotoMultiplier   (SiPM)   40   is   composed   of   hundreds   of   SPAD   (about   10x10µm2   up   to   100x100   µm2)   realised   on   the   41   same  substrate  and  interconnected  together  to  sum  the  photo-­‐current  coming  from  each   42   of   them.   The   typical   density   of   SPAD   is   100-­‐10000   per   mm2.   The   first   development   43   started   about   10   years   ago   in   Russia   [Ref   2].   Hamamatsu   Photonics   produces   44   commercially   SiPM-­‐based   circuits   named   Multi-­‐Pixel   Photon   Counter   (MPPC)   since   45   2008.   Currently,   several   technologies   have   also   been   developed   by   other   companies   46   such   as   Sensl,   or   Ketek.   Micro-­‐electronic   CMOS   technologies   can   also   be   used   to   develop   47   specific  SPAD  and  SiPM  sensors  with  good  performance  [Ref  3][Ref  4].  We  introduce  in   48   this  paper  our  development  of  SPAD  arrays  using  the  CMOS-­‐Opto  “C35B4O1"  technology   49   proposed   by   CMP   (Circuit   Multi-­‐Projects)   in   Grenoble   and   manufactured   by   AMS.   This   50   microelectronic   technology   has   been   chosen   to   design   large   arrays   of   high-­‐resolution   51   SPAD   imagers   for   optical   ultra   low   flux   applications   (for   example,   medical   application   52   [Ref   9]   or   high-­‐energy   astrophysics   [Ref   8]).   CMOS   technologies   allow   integrating   into   53   the   same   substrate   the   SPADs   and   their   specific   readout   electronic.   In   this   paper,   we   54   present   both   the   investigations   on   the   SPAD   design   and   resulting   performance   of   the   55   fabricated  chips.   56     57     Introduction   58   2 The   Technology   "CMOS-­‐Opto   C35B4O1",   and   breakdown   voltage   simulation   59   60   a) Characteristics  of  the  AMS  technology     61   The   "AMS   CMOS-­‐Opto   C35B4O1"   process   is   made   with   a   P   epi-­‐layer   (thickness   ≈14   µm)   62   on a P type substrate. This  0.35µm  CMOS-­‐Opto  process  offers  4  metallization  layers  and  2   63   polysilicon  layers.  Figure  1  shows  the  cross-­‐section.   64   AMS  gives  the  value  of  45  pA/cm2  for  the  Dark  current.  P-­‐epi  wafers  allow  lower  current   65   leakage   in   the   diode,   then   a   lower   dark   current   for   a   better   sensitivity.   This   current   is   66   very  low,  which  is  ideal  for  the  Geiger  mode.     67   AMS   gives   the   saturation   current   for   PMOS:   240   μA/μm   for   L=0.35   µm   and   W=0.4   µm   68   where   L   is   the   PMOS   channel   length   and   W   the   channel   width.   Depending   on   the   69   electrical  simulations  of  the  transistor,  we  selected  W=6  µm  and  L=0,7  µm.  The  aim  is  to   70   have  a  resistive  mode  for  Vds  =  0  to  1  V  (Vds  is  the  drain-­‐source  voltage  of  the  PMOS).  The   71   resistive  mode  range  is  from  10  kΩ  to  100  kΩ  depending  on  Vgs  adjustment  (Vgs  is  the   72   gate-­‐source  voltage  of  the  PMOS).     73   This   technology   is   normally   sensitive   in   the   range   400-­‐1000   nm   [Ref   5]   with   optimal   74   responsivity  of  290  mA/W  for  a  550  nm  wavelength  and  330  mA/W  for  850  nm.   75   b) The  SIMS  results   76   The  first  step  was  to  study  the  different  layers.  There  are  2  n_type  layers,  and  2  p_type   77   layers   of   different   doping   levels   to   modify   the   field   distribution   across   the   structure.   78   Figure  2  shows  the  summary  table  of  SIMS  results  (Secondary  ion  mass  spectrometry).   79   These  doping  values  have  been  found  by  SIMS,  after  components  manufacturing.   80   c) Silvaco  simulation:  results   81   We  use  the  doping  profiles  obtained  by  SIMS  to  determine  the  breakdown  voltages.  We   82   expose   here   the   simulation   results   with   these   profiles   obtained.   The   Figure   3   presents   a   83   first   simulation   of   the   structure   with   the   4   zones   and   the   doping   correctly   adjusted.   The   84   software   "Silvaco"   was   used   for   these   simulations.   The   result   of   these   simulations   at   85   22.5°C  (Figure  4)  gives  us  a  breakdown  voltage  of  11.7  V  and  a  guard ring of 40 V. At this 86   point of our work, we can say that this technology is well suited to Geiger Mode. 87   88     89   3 Experimental  results:  Breakdown  voltage   90   We  present  here  the   experimental  results  obtained  for  several  isolated  photodiodes  of   91   different  diameters.  The  diameter  of  the  photodiodes  is  between  D=200  µm  and  D=2.7   92   µm.  The  size  of  the  guard  ring  is  1.7  µm.  The  structural  dimension  is  shown  in  Figure  5.   93   The  breakdown  voltage  values  have  been  determined  from  the  reverse  current–voltage   94   (I–V)   characteristics,   using   a   Keithley   2636A.   A   breakdown   voltage   of   11.7   V   was   95   measured   at   22.5°C   for   photodiodes   with   a   diameter   greater   than   or   equals   to   10   µm.   96   For   photodiodes   with   a   diameter   lower   than   10   µm   diameter,   we   measured   a   higher   97   breakdown   voltage   (near   of   guard   ring   40   V)   (Figure   6).   Measurements   have   been   98   repeated   on   a   significant   number   of   devices,   showing   a   very   good   uniformity   of   the   99   breakdown  voltage  values  and  confirming  the  reliability  of  the  technology  used  for  the   100   Geiger   mode.   We   measured   on   Figure   7   the   temperature   sensitivity   for   breakdown   101   voltage:  9  mV.°C-­‐1.  It  is  found  that  the  temperature  has  a  strong  influence  on  breakdown   102   voltage  and  therefore  on  the  overvoltage.   103     104     105   4 106   This   is   a   first   positive   result   concerning   the   dark   count   rate   (DCR)   using   only   one   107   isolated   photodiode.   The   behaviour   of   the   quenching   system   is   correct.   At   22.5°C   the   108   dark  count  rate,  for  a  photodiode  of  D=10  µm  diameter,  and  an  800  mV  overvoltage,  is   109   2020   count.s-­‐1   (Figure   8).   At   -­‐40°C,   the   dark   count   rate,   for   a   photodiode   of   D=10   µm   110   diameter,  and  an  800  mV  overvoltage,  is  270  count.s-­‐1  (Figure  9).  These  two  results  are   111   presented   in   Figure   10.   The   Figure   12,   summarises   all   these   results.   With   a   diameter   112   lower  than  10  µm,  the  DCR  does  not  decrease  anymore  which  confirms  that  the  smallest   113   diameter   for   this   technology   is   about   D=10   µm.   The   experimental   set-­‐up   is   presented   in   114   Figure  11.  The  Geiger  pulses  were  measured  with  a  universal  counter  "Hameg  HM  8021-­‐ 115   4"  to  the  terminal  of  a  resistor  (100  Ω).   116     117     118     Experimental  results:  Dark  count  rate   119   5 120   We  introduced  in  the  present  document  an  investigation  of  the  technology  "CMOS-­‐Opto   121   C35B4O1"  proposed  by  CMP  (Circuit  Multi-­‐Projects)  in  Grenoble  and  manufactured  by   122   AMS for the  Geiger  mode.  The  main  part  of  our  work  deals  with  the  characteristics  in  the   123   dark   and   allows   us   to   find   the   size   of   the   photodiode   with   the   smallest   DCR/um2:   10µm.   124   These  values  are  comparable  to  those  reported  in  literature  for  CMOS  SPADs  built  in  a   125   similar   technology   [Ref   6]   [Ref   7].   The   first   results   that   we   have   obtained   are   in   good   126   agreement   with   the   challenge   of   the   Geiger   mode.   Other   results   will   be   reported   in   a   127   forthcoming  paper.   128   6 129   130   Ref  1  :  S.  Cova,  M.  Ghioni,  A.  Lacaita,  C.  Samori,  and  F.  Zappa  (1996),  “Avalanche  photodiodes  and  quenching   131   132   Ref  2:  V.  Golovin  and  V.  Saveliev,  “Novel  type  of  avalanche  photodetector  with  Geiger  mode”,  NIMA  518  (2004)   133   134   Ref   3:   Vilà,   A.,   Arbat,   A.,   Vilella,   E.,   &   Dieguez,   A.   “Geiger-­‐Mode   Avalanche   Photodiodes   in   Standard   CMOS   135   136   Ref  4:  Mandai,  S.,  Fishburn,  M.  W.,  Maruyama,  Y.,  &  Charbon,  E.  “A  wide  spectral  range  single-­‐photon  avalanche   137   138   Ref   5:   K.   Jradi,   D.   Pellion,   D.   Ginhac,   “Design,   Characterization   and   Analysis   of   0.35µM   CMOS   Single   Photon   139   140   Ref   6:   S.   Tisa,   F.   Guerrieri,     F.   Zappa,   “Variable-­‐load   quenching   circuit   for   single-­‐photon   avalanche   diodes”   141   142   143   Ref   7:   E.   Vilella,   A.   Comerma,   O.   Alonso,   A.   Diéguez,   “Low-­‐noise   pixel   detectors   based   on   gated   Geiger   mode   144   145   146   Ref   8:   F.   Lebrun;   R.   Terrier;   P.   Laurent;   D.   Prêle;   E.   Bréelle;   J.-­‐P.   Baronick;   C.   Buy;   A.   Noury;   C.   Olivetto;   R.   147   148   Ref  9:  Taiga  Yamaya  et  al  «  A  SiPM-­‐based  isotropic-­‐3D  PET  detector  X'tal  cube  with  a  three-­‐dimensional  array   149     150   Conclusion   References   circuits  for  single-­‐photon  detection”,  Applied  Optics,  Vol.  35,  No.  12,  1956-­‐1976   560-­‐564   Technologies  “   diode  fabricated  in  an  advanced  180  nm  CMOS  technology”  Opt  Express,  20(6),  5849-­‐5857.   Avalanche  Diode",  Sensors  2014,  14,  22773-­‐22784;  doi:10.3390/s141222773   Optics  express,  16(3),  2232-­‐2244.   avalanche   photodiodes”   Electronics   Letters,   Volume   47,   Issue   6,   17   March   2011,   p.   397   –   398   DOI:     10.1049/el.2011.0017   Chipaux,   "The   Gamma   Cube:   a   novel   concept   of   gamma-­‐ray   telescope"   SPIE   9144,   Space   Telescopes   and   Instrumentation  2014,  SPIE  Proceedings  Vol.  9144   of  1  mm3  crystals  »  Phys.  Med.  Biol.  56  (2011)  6793–6807.  doi:10.1088/0031-­‐9155/56/21/003     151   152     Figure  1:  Cross-­‐section  of  the  Photodiodes  design  (SPAD)  for  Geiger  mode  in  CMOS-­‐Opto  C35B4O1  and  circuit.   153     154     155   156   Figure  2:  SIMS  Results.   157     158   159       + - Figure  3:  Cross-­‐section,  simulation  "Silvaco"  of  the  structure: N /P junction and guard ring N layer.   160     161     162     163   164   165     Figure  4:  Breakdown  voltage  of  the  photodiode  (11.7V)  and  breakdown  voltage  of  the  guard  ring  (40V)  ;   166     167     simulation  results  SILVACO  obtained  at  22.5°C   168   169     Figure  5:  Schematic  structure:  Size  of  guard  rings  and  size  of  photodiodes   170     171     172     173   174     Figure  6:  Breakdown  voltage  of  the  photodiodes;  experimental  results  obtained  at  25°C   175     176     177     178   179     Figure  7:  Breakdown  voltage  versus  temperature  for  different  size   180     181     182   183     Figure  8:  Dark  count  rate  versus  photodiode  voltage  at  22.5°C   184   185     Figure  9:  Dark  count  rate  versus  photodiode  voltage  at  -­‐40°C   186     187   188     Figure  10:  Dark  count  rate  versus  Temperature  for  three  size  at  800mV  overvoltage   189     190     191     192   193     Figure  11:  Electrical circuit used for dynamic characterizations, an exterior resistor (100 Ω for 194   read) has been used.   195     196     197     198     199     200     201   202     Figure  12:  Summary  table  of  our  design   203     204