Earth–abundant  water–splitting  catalysts  coupled  to   silicon  solar  cells  for  solar–to–fuels  conversion.     A  dissertation  presented     by       Casandra  R.  Cox     to       The  Department  of  Chemistry  and  Chemical  Biology     in  partial  fulfillment  of  the  requirements   for  the  degree  of     Doctor  of  Philosophy   in  the  subject  of     Chemistry     Harvard  University   Cambridge,  Massachusetts   September  2014                                               ©  2014  by  Casandra  R.  Cox     All  rights  reserved.         Casandra  R.  Cox     Daniel  G.  Nocera   Earth–abundant  water–splitting  catalysts  coupled  to  silicon   solar  cells  for  solar–to–fuels  conversion.       Abstract     Direct  solar–to–fuels  conversion  can  be  achieved  by  coupling   semiconductors  with  water–splitting  catalysts.  A  10%  or  higher  solar  to  fuels   conversion  is  minimally  necessary  for  the  realization  of  a  robust  future  technology.   Many  water–splitting  devices  have  been  proposed  but  due  to  expensive  designs   and/or  materials,  none  have  demonstrated  the  necessary  efficiency  at  low–cost  that   is  a  requisite  for  large–scale  implementation.  In  this  thesis,  a  modular  approach  is   used  to  couple  water–splitting  catalysts  with  crystalline  silicon  (c–Si)  photovoltaics,   with  ultimate  goal  of  demonstrating  a  stand–alone  and  direct  solar-­‐to-­‐fuels  water– splitting  device  comprising  all  non–precious,  technology  ready,  materials.     Since  the  oxygen  evolution  reaction  is  the  key  efficiency–limiting  step  for   water–splitting,  we  first  focus  on  directly  interfacing  oxygen  evolution  catalysts   with  c–Si  photovoltaics.  Due  to  the  instability  of  silicon  under  oxidizing  conditions,  a   protective  interface  between  the  PV  and  OER  catalyst  is  required.  This  coupling  of   catalyst  to  Si  semiconductor  thus  requires  optimization  of  two  interfaces:  the   silicon|protective  layer  interface;  and,  the  protective  layer|catalyst  interface.  A   modular  approach  allows  for  the  independent  optimization  and  analysis  of  these   two  interfaces.   iii   A  stand–alone  water–splitting  device  based  on  c–Si  is  created  by  connecting   multiple  single  junction  c-­‐Si  solar  cells  in  series.  Steady–state  equivalent  circuit   analysis  allows  for  a  targeted  solar–to–fuels  efficiency  to  be  designed  within  a   predictive  framework  for  a  series–connected  c–Si  solar  cells  and  earth–abundant   water–splitting  catalysts  operating  at  neutral  pH.  Guided  by  simulation  and   modeling,  a  completely  modular,  stand–alone  water–splitting  device  possessing  a   10%  SFE  is  demonstrated.  Importantly,  the  modular  approach  enables  facile   characterization  and  trouble–shooting  for  each  component  of  the  solar  water– splitting  device.  Finally,  as  direct  solar  water–splitting  is  far  from  a  mature   technology,  alternative  concepts  are  presented  for  the  future  design  and  integration   of  solar  water–splitting  devices  based  on  all  earth–abundant  materials.         iv     Table  of  Contents       Title  page   Copyright  page   Abstract   Table  of  Contents   List  of  Figures   List  of  Tables   List  of  Abbreviations   Acknowledgments   i   ii   iii   v   viii   xiii   xiv   xvi–xix                     1. Chapter  1–Introduction   1.1. The  need  for  clean  energy   1.2. Renewable  energy   1.3. Capture  of  solar  power  and  conversion  to  electrical  power   1.4. Conversion  of  electrical  power  into  fuels   1.5. Photoelectrochemical  water–splitting   1.5.1. Buried–junction  PEC  requirements   1.5.2. Buried–junction  PEC  devices   1.6. Crystalline  Silicon   1.7. Earth–abundant  water–splitting  catalysts   1.8. Overview   1.9. References     v       1   2   3   4   5   8   9   11   11   13   14   17   2. Chapter  2–Interfaces  between  crystalline  silicon     solar  cells  and  water–oxidation  catalysts   2.1. Introduction   2.2. Results   2.2.1. Optimization  of  OER–catalyst  functionalized     silicon  solar  cells   2.3. Discussion   2.4. Conclusion   2.5. Experimental   2.6. References   3. Chapter  3–Modeling  a  coupled  photovoltaic  electrochemical     devices  using  steady–state  equivalent  circuit  analysis   3.1. Introduction   3.2. Efficiency  considerations   3.3. Steady–state  equivalent  circuit  analysis   3.4. Results  and  Discussion   3.4.1. Impact  of  ηPV  on  SFE   3.4.2. Impact  of  ηEC  efficiency  on  SFE   3.5. Model  validation   3.6. Conclusion   3.7. Experimental   3.8. References   50   51   52   57   60   61   63   65   67   68   70   32   37   39   40   46   24   25   27     vi       4. Chapter  4–10%  solar–to–fuels  efficiency  with   non–precious  materials   4.1. Introduction   4.2. Results   4.2.1. Device  integration   4.3. Discussion   4.4. Conclusion   4.5. Experimental   4.6. References   5. Chapter  5–Future  directions   5.1. Introduction   5.2. Alternative  PV  materials   5.3. Alternative  catalyst  deposition  methods   5.4. Cell  design   5.5. Conclusion   5.6. Experimental   5.7. References         73   74   75   80   84   90   91   95   98   99   99   101   106   108   108   110       vii       List  of  Figures     Figure   1.1   Schematic   showing   (1)   solar   capture   of   solar   energy   by   a   photovoltaic   device,  (2)  conversion  of  solar  photons  into  a  wireless  current,  and  (3)  storage  via   breaking  the  bonds  of  H2O  to  make  H2  which  can  be  used  as  a  fuel.  Adapted  from  ref.   5.   3     Figure   1.2   Solar   irradiance   at   the   surface   of   the   Earth.   The   band–gap   of   silicon   is   overlaid   as   an   example   showing   that   photons   absorbed   at   the   band–gap   can   be   converted  and  those  absorbed  above  the  band–gap  are  wasted  as  heat.     4       Figure   1.3   Qualitative   schematic   of   an   n–type   semiconductor/electrolyte   junction   for  photoelectrochemical  water–splitting.     9   Figure   1.4  Qualitative  schematic  of  a  buried–junction  photovoltaic  interfaced  with   water–splitting  catalysts  via  Ohmic  contacts  for  solar–water–splitting.   10     Figure  1.5  Chinese  c–Si  PV  module  prices  since  2006.  The  data  was  recreated  from   ref.  43.   12       Figure   1.6   Depictions   of   the   molecular   structure   of   our   Mn,   Co,   and   Ni   water– oxidation  catalysts.  Reprinted  with  permission  from  Mike  Huynh.     14   Figure   2.1   Schematic   of   the   OER–catalyst   functionalized   silicon   solar   cell   used   in   these   studies.   For   electrochemical   measurements,   an   external   voltage   may   be   applied   to   the   contacts   at   either   side   of   the   cell   with   or   without   illumination.   A.   The   solar  cell  is  operating  under  reverse  bias  conditions  with  voltage  applied  to  the  front   metal   contacts   on   the   n–side   in   the   dark.   B.   The   voltage   can   be   applied   directly   to   the   protective–layer,   in   which   case   the   PV   is   bypassed   and   the   current–voltage   characteristics   are   those   of   the   OER–catalyst   on   an   electrode.   C.   The   solar–cell   is   illuminated  with  AM  1.5  illumination  and  the  current–voltage  behavior  reflects  the   activity  of  the  OER–catalyst  functionalized  solar  cell.   27     Figure   2.2  CV  curves  of  (top)  npSi|ITO|CoPi  and  (bottom)  npp+Si|ITO|CoPi  in  0.1  M   KPi   electrolyte   at   pH   7   in   the   dark   with   Vappl   through   the   n–side   of   the   cell   (▬▬,   black),   with   Vappl   thought   the   n–side   under   1   sun   AM   1.5   illumination   (▬▬,   green)   ,   and  in  the  dark  with  Vappl  through  the  ITO  layer  bypassing  the  PV  (▬▬,  blue).  Taken   from  ref.  20.   28       viii       Figure   2.3   Schematic   showing   the   band-­‐diagrams   at   the   p-­‐Si|ITO   interface.   A.   Before  contact.  B.  After  equilibration  of  Fermi  levels  after  interfacing  p–Si  with  ITO.   C.  After  equilibration  of  the  Fermi  levels  after  interfacing  p+–Si  with  ITO.     30     Figure   2.4  Tafel  plots  for  npp+Si|ITO|CoPi  with  potential  applied  to  the  metal  front   contact   for   measurements   in   dark   (black   squares,   n),   at   100   mW   cm–2   (green   squares,   n),   and   1000   mW   cm–2   (orange   red   squares,   n)   illumination.   The   blue   triangles  (▲)  correspond  to  a  measurement  in  dark  where  the  potential  was  applied   through  the  ITO  film  at  the  back  of  the  sample.  Figure  taken  from  ref  20.     31     Figure  2.5  Representative  J–V  curve  for  generation  2  npp+–Si  solar  cells  used  in  this   study  in  the  dark  (▬▬,  black)  and  under  AM  1.5  illumination  (▬▬,  blue).     33     Figure  2.6  Plane  view  SEM  images  of  OER–catalyts  deposited  on  surface–protected   npp+Si|electrodes.  From  left  to  right  A  npp+|FTO|CoBi  and  B  npp+Si|Ni|CoBi.     34     Figure   2.7   O2   production   measured   by   a   fluorescent   sensor   (▬▬,   red)   and   the   amount   produced   based   on   current   passed   assuming   100%   Faradaic   efficiency   (▬▬,  green)  for  (left)  npp+–Si|FTO|CoBi  and  (right)  npp+–Si|Ni|CoBi.     35     Figure   2.8   Tafel   plots   of   (a)   npp+Si|ITO|CoBi   (b)   npp+Si|FTO|CoBi   and   (c)   npp+Si|Ni|CoBi.   With   the   potential   applied   to   the   metal   front   contact   for   measurements  in  the  dark  (●),  under  1  sun  AM  1.5  illumination  (●),  and  in  the  dark   with  the  potential  applied  through  the  protective  coating  at  the  back  of  the  sample     (●).   36     Figure  2.9  Graph  showing  the  variability  in  Tafel  slope  for  various  combinations  of   OER–catalyst  functionalized  c–Si  solar  cells.    The  red  lines  indicate  the  value  based   on  previously  reported  Tafel  analysis.   38     Figure   3.1   Schematic   of   a   wired   and   wireless   PV–EC   based   on   silicon   solar   cells.   Regardless   of   the   mode   of   coupling   between   the   two,   the   equivalent   circuit   is   identical.     55         Figure   3.2   Block  diagram  for  a  photovoltaic  (PV)  powered  electrochemical  cell  (EC),   where  direct  electrical  connection  constrains  JPV  =  JEC  and  VPV  =  VEC.   56   ix       Figure  3.3   The   generalized   current   density–voltage   (J–V)   diagram   of   a   coupled   PV– EC   system   where   the   point   of   intersection   of   the   PV–curve   (▬▬,   blue)   and   EC– curve   (▬▬,   red)   represents   the   operational   point   and   SFE   of   the   coupled   PV–EC   device.  The  SFE  is  maximized  when  the  operating  point  is  equal  to  PMAX.     57         Figure  3.4  Impact  on  the  J–V  curve  for  a  PV  due  to  shunt  (▬▬, dark  red)  or  series   (▬▬, dark  green)  resistance  compared  to  an  ideal  J–V  curve  (▬▬, dark  blue).   58   Figure  3.5  Steady–state  equivalent  circuit  of  a  PV–EC  system.  An  applied  voltage  is   incorporated  to  illustrate  analysis  of  an  externally  assisted  system.     60   Figure   3.6   Impact   on   SFE   via   improvement   in   PV   efficiency   compared   to   the   baseline  ηPV  =  20%  (–––––,  grey  dash).  Given  optimal  coupling  between  the  PV  and   EC   components   (Top)   a   higher   relative   SFE   can   be   obtained   by   improving   the   JSC   (▬▬,   green)   as   opposed   to   the   VOC   (–––––,   dashed   green).   Given   poor   coupling   between   the   baseline   PV   and   EC   (bottom),   only   minor   improvements   in   the   SFE   can   be  obtained.     62     Figure  3.7  J–V  curves  of  multiple  series  connected  solar  cells  with  ηPV  =  20%  (▬▬,   grey)  and  EC  curves  (▬▬,  dark  blue).  The  number  of  solar  cells  required  changes   based  on  choice  of  catalyst  which  causes  the  EC  curve  to  shift  left  or  right  and   resistive  losses  due  to  RSOL  cause  the  EC  curve  to  tilt  down.   63     Figure  3.8  Impact  of  solution  resistance  and  EC  parameters  on  SFE  given  ηPV  =  20%.   Case  I  EC  parameters  (▬▬,  green)  are  based  on  utilizing  the  Co–OEC  and  Case  II  EC   (▬▬,  navy)  are  based  on  utilizing  the  Ni–OEC.   65     Figure   3.9   Graphical   demonstration   of   how   the   predictive   analysis   works   for   PV– assisted   reactions,   where   the   PV–curve   (▬▬,   blue)   is   based   on   the   J–V   characteristics  of  an  in–house  built  single  junction  c–Si  PV  and  the  EC–curve  (▬▬,   red)  is  based  on  the  CoBi  water–oxidation  catalyst  operating  in  pH  9.2  solution.     66     Figure   3.10   Predicted   Tafel   behavior   of   a   PV–assisted   water   oxidation   system   similar   to   the   experiments   described   in   Chapter   2.   The   electrical   properties   of   the   PV   (shown   in   Fig.   3.8)   and   EC   systems   were   measured   independently   (● ,   black   dots)  and  used  to  predict  the  coupled  behavior  (▬▬,  black).  The  Tafel  analysis  of   the  PV–assisted  photoanode  (● ,  red  dots)  and  predicted  behavior  match  to  within   10  mV.     67       x       Figure  4.1  Schematic  of  a  PV–EC  device  based  on  series–connected  single–junction   c–Si  solar  cells  and  water–splitting  catalyst.  In  this  configuration  the  OER–catalyst  is   directly  deposited  on  the  back  of  the  last  solar  cell  in  the  stack.     76     Figure   4.2   Schematic   of   a   PV–EC   device   used   in   these   studies.   In   this   modular   configuration  each  component  can  be  easily  evaluated  and  replaced  independently.   77     Figure  4.3  J–V  curves  of  the  individually  measure  PV  and  EC  components  making  up   the   PV–EC   device.   The   grey   curves   represent   the   J–V   curves   for   the   PV   modules   composed   of   either   three   (––––,   grey–dashed)   or   four   (▬▬,   grey–solid)   single– junction   c–Si   solar   cells   measure   under   AM   1.5   illumination.   The   red   curves   represent   electrochemical   load   J–V  curves   using   NiBi   and   NiMoZn   catalysts,   where   the   ideal   EC   curve   (––––,   red–dashed)   is   based   on   previously   reported   Tafel   analysis  and  the  actual  EC  curve  (▬▬,  red)  measured  in  a  2–electrode  experiment   (0.5M   KBi   /   0.5M   K2SO4,   pH   9.2).   The   point   of   intersection   represents   the   JOP   (● ,   orange  circles)  and  the  SFE  of  the  coupled  system.   78   Figure  4.4.  Steady–state  current  voltage  behavior  for  the  NiBi  operating  in  0.5M  KBi   /   0.5   M   K2SO4   pH   9.2   in   H2   saturated   solution   (● )   and   in   Ar   saturated   solution   (▲).   Since  the  voltage  required  to  achieve  a  given  current  density  under  both  conditions   is   almost   identical   indicates   that   the   contribution   of   H2   oxidation   at   the   anode   is   negligible.     82     Figure   4.5   Current   under   chopped   illumination   representing   JOP   for   the   PV–EC   device   in   0.5M   KBi   /   0.5M   K2SO4   pH9.2.   The   chopped   illumination   illustrates   the   recovery  in  SFE  and  reproducibility  in  measuring  JOP  through  the  PV-­‐EC  device     83     Figure  4.6   Decay   of   the   open–circuit   voltage   of   the   4–cell   PV   mini–module   over   the   course  of  ~15  min.  The  initial  VOC  at  2.42  V  decays  to  a  steady–state  of  2.27  V  after   the   first   10   min   (▬▬,   orange),   which   contributes   to   the   initial   decline   in   the   SFE   of   the   coupled   PV–EC   device.   After   overnight   illumination,   the   Voc   was   measured   (▬▬,  blue)  and  shows  a  slight  recovery  to  2.31  V,  which  corresponds  to  the  initial   increase  in  SFE  of  the  PV–EC  device  during  the  first  24  h.     84     Figure   4.7  Specific  conductance  measurements  for  various  electrolytes  considered   to  minimize  RSOL.  KOH  (∎,  red  squares)  is  the  most  conductive  electrolyte;  in  order   to  operate  in  pH  near  neutral  regimes  0.5M  KBi  was  used  with  additional  supporting   electrolyte,  such  as  KNO3  (●,  green  circles)  or  K2SO4  (●,  black  circles).     85     Figure   4.8   Gas   quantification   for   NiMoZn   cathode   operating   in   (left)   0.5   M   KBi   /   K2SO4  and  (right)  0.5  M  KBi  /  KNO3  both  at  pH  9.2.  The  black  line  represents  100%   Faradaic   efficiency   based   on   the   charge   passes   during   electrolysis.   The   green   circles     xi       represent   H2   measured   by   gas   chromatography.   The   red   arrow   indicates   when   electrolysis   was   stopped.   GC   analysis   was   conducted   until   the   moles   of   gas   measured  in  the  headspace  reached  a  steady–state.  The  lag  period  (▬▬,  black)  in   gas  generation  is  due  to  the  buildup  of  gases  in  the  headspace  of  the  EC  cell.   86     Figure   4.9  Gas  quantification  for  NiBi  cathode  operating  in  (left)  0.5  M  KBi  /  K2SO4   and   at   pH   9.2.   The   black   line   represents   100%   Faradaic   efficiency   based   on   the   charge   passes   during   electrolysis.   The   green   circles   represent   O2   measured   by   gas   chromatography.   The   red   arrow   indicates   when   electrolysis   was   stopped.   GC   analysis   was   conducted   until   the   moles   of   gas   measured   in   the   headspace   reached   a   steady–state.  The  lag  period  (▬▬,  black)  in  gas  generation  is  due  to  the  buildup  of   gases  in  the  headspace  of  the  EC  cell.   87     Figure   4.10   Current   under   chopped   illumination   representing   JOP   for   a   PV–EC   device   composed   of   a   3–cell   PV–module,   a   NiBi   anode,   and   NiMoZn   cathode   operating   in   1M   KOH.   Because   KOH   is   a   more   conductive   electrolyte,   a   12%   or   greater  SFE  can  be  obtain  with  a  3–cell  PV  module  as  opposed  to  a  4–cell  module.   The  initial  drop  in  SFE  is  due  to  the  decrease  in  PV  efficiency,  due  to  heating  of  the   PV–module.  The  chopped  illumination  represents  the  recovery  in  SFE.     88       Figure   4.11  J–V  curves  of  the  individually  measure  PV  and  EC  components  making   up  the  PV–EC  device  operating  in  1M  KOH.  The  grey  curves  represent  the  J–V  curves   for  the  PV  modules  composed  of  either  three  (–––––,  grey–dashed)  or  four   (▬▬,   grey–solid)  single–junction  c–Si  solar  cells  measure  under  AM  1.5  illumination.  The   blue   curves   represent   electrochemical   load   J–V   curves   using   NiBi   and   NiMoZn   catalysts,   where   the   ideal   EC   curve   (–––––,   blue–dashed)   is   based   on   previously   reported  Tafel  analysis  and  the  actual  EC  curve  (▬▬,  blue–solid)  measured  in  a  2– electrode  experiment.  The  point  of  intersection  represents  the  JOP  (●,  orange  circles)   and  the  SFE  of  the  coupled  system.     89     Figure  4.12  SFE  inferred  from  JOP  for  the  PV–EC  device  operating  in  0.5M  KBi  /  0.5M   K2SO4   pH   9.2   measured   for   over   7   days   of   operation   showing   no   decrease   in   SFE   over   operation   time.   Spikes   are   due   to   the   addition   of   solution   to   maintain   the   solution  level  and  pH.     90     Figure   5.1   Tafel   plot   of   a   sputtered   NiFeO   OER   catalyst   operating   in   0.5   M   KBi   /   1.5M  KNO3  pH  9.2.  A  Tafel  slope  of  45  mV  decade–1  is  observed  for  a  50  nm  (∎),  100   nm  (●) and  200  nm  (▲)  thick  NiFeO  film.  Inset:  SEM  image  of  a  NiFeO  shows  a  very   dense,  compact  film.   102       xii       Figure   5.2   Tafel   plots   of   200   nm   thick   NiFeO   (81%   mol   Ni,   19%   mol   Fe)   on   Ni-­‐ coated  glass  operated  in  (▲)  0.2  M  KPi,  pH  7.0,  92  mV  decade–1  slope;  (■)  0.2  M  KBi,   pH  9.3,  61  mV  decade–1  slope;  (●)  1.0  M  KOH,  pH  13.9,  45  mV  decade–1  slope.   103       Figure   5.3   Tafel   analysis   of   Co–OEC   films   formed   and   operated   in   KBi   (●)   as   opposed  to  KPi  (●)  solution.    The  films  formed  from  KBi  exhibit  a  lower  Tafel  slope   and  therefore  demonstrate  higher  activity  than  those  formed  in  KPi.     104     Figure   5.4  Tafel  analysis  of  Co–OEC’s  formed  from  anodizing  metallic  cobalt  in  KBi   solution.  In  all  cases  the  Co–OEC  exhibits  a  Tafel  slope  of  60  mV  decade–1,  however   starting   with   thicker   metallic   films   produces   Co–OEC’s   with   higher   activity   than   thinner  films.     105     Figure   5.5   The   current   density   traces   show   that   recirculating   streams   allow   the   device   to   function   stably   and   continuously   (purple   trace),   while   without   recirculation   the   device   performance   deteriorates   as   concentration   gradients   form   across   the   cell   and   ionic   species   are   depleted   in   the   oxygen-­‐evolution   side   (red   trace).   The   inset   in   the   graph   corresponds   to   a   schematic   representation   of   the   parallel-­‐plate  solar-­‐hydrogen  generator.  Reprinted  with  permission  from  reference   32.   107         xiii       List  of  Tables     Table  2.1.  Summary  of  Faradaic  efficiency  for  npp+–Si | interface| catalyst  films        39   Table  3.1.  Solar  cell  parameters  for  the  modeling.   Table  3.2  Electrochemical  parameters  for  the  modeling.   Table  4.1.  PV  characteristics  for  the  3  and  4–cell  c–Si  mini–modules.       61   61   78     xiv       List  of  Abbreviations     ALD     AM     a–Si     b     BJ     BOS     Bi     cb     CIGS     CoBi     CoPi     c–Si     CV     CVD     D     E-­‐beam   EC     EF     F     FF     FTO     HEC     HER     ITO     J     J0     JSC     kb     MPP     n     NHE     NiBi     NiFeO     OEC     OER     P     PCET     PEC     PSII     PSII     Pi     PV     q       atomic–layer  deposition   air  mass   amorphous  silicon   Tafel  slope   buried  junction   balance  of  systems   borate  buffer   bulk  concentration   copper  indium  gallium  diselenide     cobalt–based  catalysts  deposited  from  borate  electrolyte   cobalt–based  catalysts  deposited  from  phosphate  electrolyte   crystalline  silicon   cyclic  voltammogram  or  cyclic  voltammetry   chemical  vapor  deposition   diffusion  coefficient   electron  beam  evaporation   electrochemical   Fermi  energy  or  level   Faraday’s  constant   fill  factor     fluorine  doped  tin  oxide   hydrogen–evolution  catalyst   hydrogen–evolution  reaction   tin–doped  indium  oxide   current–density   exchange  current  density     short–circuit  current   Boltzmann’s  constant   maximum  power–point   ideality  factor   normal  hydrogen  electrode   nickel  based  catalyst  deposited  from  borate  electrolyte   nickel  iron  oxide   oxygen  evolution  catalyst   oxygen  evolution  reaction   power   proton–coupled  electron  tranfer     photoelectrochemical     photosystem  I   photosystem  II   phosphate  buffer   photovoltaic   charge   xv       R   s   SEI   SEM   SFE   sh   SJ   sol   T   TCO   th   VAppl   VOC   WP   η   ηC   ηEC   ηPV   δ                                             resistance   series     semiconductor  electrolyte  interface   scanning  electron  micrograph   solar–to–fuels  efficiency     shunt     solution  junction   solution   temperature   transparent  conductive  oxide   thermodynamic   potential  applied  to  the  electrode   open  circuit  voltage   watt  at  peak  power   overpotential   coupling  efficiency   electrochemical  efficiency     PV  efficiency   Nernst  diffusion  layer       xvi       Acknowledgments     During  the  past  six  years  at  MIT  and  then  at  Harvard,  I  have  had  the  pleasure  of   meeting  some  amazing  people  and  scientists  during  this  time  and  I  would  like  to   thank  them  for  having  an  impact  on  me  and  my  decisions.       From  a  scientific  perspective  I  first  have  to  thank  my  PhD  advisor  Dan  Nocera.  I  have   learned  a  lot  from  him.  He  has  taught  me  many  technical  skills  including  how  to   write  a  paper,  how  to  make  my  research  accessible  and  interesting  to  others,  and   how  to  make  pretty  figures.  His  unique  advising  style  of  always  pushing  you  when   you  need  it  has  taught  me  more  than  anything  else  I’ve  encountered  in  graduate   school.  He  seems  to  always  be  so  intuitive  to  what  his  students  need  to  be  successful   and  no  matter  how  many  times  we  mess  up  he  never  gives  up  on  us.       I  would  also  like  to  acknowledge  my  long–time  collaborator  Tonio  Buonassisi  for   our  many  thought  provoking  meetings.       From  my  undergraduate  academic  experience  I  like  to  thank  my  under–graduate   research  advisor  Dr.  Stephen  Mezyk  for  sparking  my  interest  in  doing  research  and   pursuing  a  graduate  degree.     I  thank  Dr.  James  Kiddle  for  being  a  great  collaborator  and  kindred  spirit.  We  have   been  great  friends  and  have  had  so  much  fun  together.       Now  for  the  part  most  people  skip  to  acknowledging  all  of  the  lab  mates  and   colleagues  that  have  inspired,  influenced,  and/or  have  just  been  great  friends  over   the  years:     I  would  like  to  thank  Dr.  Liz  Young  for  telling  me  that  I  was  not  alone  in  feeling  like  I   was  the  only  person  in  my  class  who  didn’t  understand  everything  and  felt  way  too   behind  to  keep  on  going.  I  also  admire  Liz’s  no  nonsense  attitude  and  the  ability  to   always  stand  up  for  herself  and  others  without  being  shy  or  afraid  what  others   might  think.       I  will  also  have  to  thank  Dr.  Matt  Kanan  for  always  being  so  inspiring  and  kind  even   to  a  lowly  first  year  graduate  student.  I  was  always  so  impressed  seeing  him  at  the   Miracle  of  Science  every  Saturday  with  a  new  scientific  paper  to  read  along  side  a   beer  and  burger.  I  also  appreciate  the  friendship  we  have  maintained  over  the  years   and  how  he  always  makes  time  to  meet  me  for  a  drink  when  he  is  town.       Additionally  I  would  like  to  thank  Dr.  Steve  Reece.  Although  we  didn’t  overlap,  being   the  great  mentor  that  he  is  really  helped  me  a  lot  during  my  first  few  years  in   graduate  school.         xvii       Dr.  Mark  Winkler  was  a  great  colleague  and  collaborator.  For  having  so  many   helpful  meetings  and  pep  talks  now  and  then.       Dr.  Joep  Pijpers  got  me  started  on  my  project  and  was  a  great  mentor  during  the   short  time  period  we  worked  together.       Dr.  Dino  Villagran  is  one  of  the  nicest  people  but  somehow  has  made  every  female  in   lab  cry  over  some  ridiculous  thing.       Dr.  Alex  Radosevich  for  teaching  every  one  how  to  bootie  bomb.       Dr.  Bob  McGuire  for  being  such  a  fun  and  nice  person.     Dr.  Dilek  Doğutan  and  I  joined  the  Nocera  lab  around  the  same  time.  It  has  been  nice   seeing  her  progress  from  a  post–doc  to  her  current  position  where  she  has  so  much   leadership  responsibility.  She  really  helps  facilitate  the  research  in  our  lab  on  a  daily   basis.       Dr.  David  Powers  for  giving  great  pep–talks  during  this  last  month  and  being  a  good   friend.       Dr.  Eric  Bloch  has  been  someone  I  have  only  known  a  short  time  but  has  been  a   really  fun  and  kind  person.       Dr.  Tom  Kempa  for  reading  over  portions  of  my  thesis  and  for  all  of  our  long  talks   about  science.       Dr.  Chris  Gagliardi  for  being  such  a  nice  and  funny  person.       Dr.  Emily  McClaurin  for  being  such  a  good  friend  to  me  during  my  first  few  years.   She  always  put  up  with  my  crisis  (which  were  fairly  often).  She  taught  me  a  lot  of   things  about  how  to  handle  myself  in  lab  and  our  “CHEMREF”  sessions  were  always   helpful.       Dr.  Changhoon  Lee  for  being  someone  who  I  could  never  hear  speak  but  I  knew  he   was  a  kind  person.       Dr.  Yogesh  Suredranath  was  the  person  I  was  most  scared  to  present  in  front  of  at   group  meeting  so  I  would  go  through  my  slides  with  him  beforehand.  He  always   gave  time  and  attention  to  people  who  asked  for  it.       Dr.  Matt  Chambers  for  being  the  eternal  optimist  and  always  playing  devil’s   advocate.  We  had  a  lot  of  fun  times  especially  at  the  Bleacher  Bar.  Let’s  go  Buffalo!     Dr.  Arturo  Pizano  we  had  a  lot  of  crazy  times  and  I  could  always  count  on  him  to   enable  a  “f*  it  day.”     xviii       In  my  first  year  I  started  out  as  one  of  four  and  am  the  only  one  who  made  it   through.  I  would  specifically  like  to  thank  Pete  Curtain  for  being  such  a  smart  and   happy  person.  He  is  someone  I  have  always  missed,  especially  in  times  where  I  just   wanted  a  person  who  could  be  a  partner  in  crime  during  the  various  phases  of   graduate  school.  My  last  memory  of  a  big  hung  before  telling  him  good  luck  before   he  left  is  still  one  my  favorite  memories.       Kwabena  Bediako  for  being  so  knowledgeable  and  helpful.  However,  sharing   frustrations  with  science  and  graduate  school  with  someone  who  makes  it  all  seem   so  easy  made  me  feel  not  so  alone.  I  also  always  appreciate  the  pep  talks  walking   home  after  a  long  day  in  lab.       Chris  Lemon  for  being  a  great  friend.  Chris  was  always  there  when  I  needed  him  and   was  always  ready  to  grab  a  beer  and  hang  out  after  a  long  day  in  lab.  He  is  one  of  the   hardest  workers  in  lab  and  never  seems  frustrated.  I  love  our  “gay–tes”  at   Cambridge  Common.       Andrew  Ullman  for  being  so  quirky.  I  have  loved  seeing  the  transformation  from   hippie  to  clean–cut  and  dad–like  (thanks  Anne  Marie).  His  love  for  reading  old   textbooks  is  hilarious  and  he  has  the  best  smile  out  of  anyone  in  lab.       Mike  Huynh  for  being  the  smartest,  hardest–working,  and  kindest  person  in  lab  all   of  which  comes  completely  naturally.  I  think  I  had  the  best  person  to  give  group   meeting  with  and  enjoyed  his  delicious  home–cooked  treats  he  would  surprise  me   with.       Bon  Jun  Koo  I  don’t  even  know  where  to  start.  You  have  been  a  great  friend  and   have  always  been  there  for  me.  Obviously  my  favorite  thing  about  you  is  your   confusion  with  the  English  language  and  American  culture,  which  has  made  me   laugh  countless  times.  I  also  love  your  no  nonsense  attitude  especially  during  long   group  meetings.       Nancy  Li  has  been  like  a  little  sister  to  me  over  the  past  year.  I  love  our  talks  about   all  things  shopping  and  being  terrible  influences  on  one  another  when  it  comes  to   purchasing  things  we  don’t  need.  She  is  so  thoughtful  and  such  a  hard  worker.  I   think  she  will  have  a  very  successful  PhD  experience.       Dan  Graham  all  I  can  say  is  thank  you  for  always  being  the  scape–goat.       Bryce  Anderson  and  Andrew  Maher  are  both  fun,  sincere,  and  kind  people  and  made   sharing  an  office  with  no  windows  seem  not  so  bad.       Evan  Jones  for  always  seeming  to  be  in  the  wrong  place  at  the  wrong  time,  which   makes  me  laugh.         xix       Seung–Jun  Hwang  for  putting  up  with  all  of  our  questions  on  the  Korean  language   after  Bon  Jun  confuses  us.       There  have  been  many  people  I  didn’t  get  to  know  very  well  to  all  of  you  I  wish  the   best  of  luck.     On  a  more  personal  note:       I  would  like  to  thank  my  amazing  husband  Eric  Hontz.  In  the  last  two  years  he  has   helped  me  in  every  aspect  of  life.  We  have  so  much  fun  together  and  I  am  so  excited   about  our  future  together.       I  would  like  to  thank  my  dad  for  always  visiting  me  in  every  place  I’ve  lived  and   being  proud  of  me.       I  would  like  to  thank  my  mother  for  being  the  strongest  person  I  know.  She  has  been   so  encouraging  and  helpful  and  I  love  her  very  much.                               xx       Chapter  1–  Introduction:       1    1.1  The  need  for  clean–energy     One  of  the  greatest  challenges  facing  the  world  today  is  the  need  for  clean– renewable  energy  resources  to  supply  the  needs  of  a  quickly  growing  world– population.  Current  world  energy  consumption  is  524  quadrillion  BTU  (5.5  x  1020   Joules  or  17.5  TW  per  year).1  Due  to  an  increase  in  world  population  to  3  billion   people  by  2050,  the  world  energy  consumption  is  expected  to  increase  by  56%  and   double  by  the  end  of  the  century.  1–3  Most  of  this  population  growth  is  occurring  in   the  developing  world,  which  presently  does  not  have  the  infrastructure  or  wealth  to   keep  up  with  this  demand.4     Presently  86%  of  the  current  world–energy  is  supplied  by  fossil  fuels  and  it  is   projected  that  even  with  increase  world  population,  fossil  fuels  can  continue  to   power  the  planet  for  many  years  to  come.5,6  However,  increasing  levels  of  CO2  in  the   atmosphere  have  been  rising  since  the  industrial  revolution  when  the  world   population  was  seven  times  less  than  today.  Given  that  human  activity  led  to   increased  concentrations  of  CO2  in  the  atmosphere  with  a  considerably  smaller   population,  the  impact  of  today’s  rapidly  growing  world  population  could  lead  to   much  more  severe  results.  The  common  goal  amongst  scientists  and  policy  makers   is  to  prevent  the  concentrations  of  CO2  in  the  atmosphere  from  reaching  levels  such   that  the  change  in  global  temperature  is  more  than  2oC.7  While  it  remains  unclear   what  impact  the  increased  global  temperature  will  have,  it  seems  unwise  to  perform   an  uncontrolled  experiment  on  the  environment.       2   This  quandary  necessitates  new  technologies  to  produce  and  store   renewable  energy  that  minimizes  the  environmental  consequences  associated  with   burning  fossil  fuels.     1.2  Renewable  Energy   Due  to  the  inefficiency  of  photosynthesis  (1%)8  and  the  spatial  limitations  of   wind  power,9  neither  biomass  nor  wind  is  a  viable  option  to  fully  meet  the  world   energy  needs.  The  sun  is  by  far  the  most  abundant  source  of  energy  as  more  energy   from  the  sun  strikes  the  earth  in  just  one  hour  than  is  presently  consumed  in  one   year.  Impressively,  covering  0.1%  of  the  Earth’s  surface  with  solar  cells  with  an   efficiency  of  10%  would  satisfy  present  energy  needs.10,11  Unfortunately  due  to  the   intermittent  and  diurnal  nature  of  sunlight,  in  order  to  make  solar–energy  as  a   viable  resource  requires  capture,  conversion,  and  storage.     Figure   1.1  Schematic   showing   (1)   solar   capture   of   solar   energy  by   a   photovoltaic   device,  (2)  conversion  of  solar  photons  into  a  wireless  current,  and  (3)  storage  via   breaking  the  bonds  of  H2O  to  make  H2  which  can  be  used  as  a  fuel.  Adapted  from  ref.   5.       3   1.3  Capture  of  solar  power  and  conversion  to  electrical  power     An  elegant  technological  approach  to  directly  convert  sunlight  into  electricity   without  moving  parts  or  environmental  emissions  is  to  utilize  semiconductors.   Semiconductors  take  advantage  of  the  fact  that  photons  with  energy  equal  to  the   optical  band–gap  (similar  to  HOMO–LUMO  transition  for  molecules)  can  create  an   electron–hole  pair  that  can  be  separated  between  two  different  materials,  thus   effectively  establishing  a  potential  difference  across  the  interface.  However,  since   semiconductors  are  transparent  to  photons  below  the  band–gap  and  photons  having   energies  much  higher  than  the  band  gap  rapidly  release  heat  to  the  lattice  of  the   solid  the  upper  bound  conversion  efficiency  of  solar  power  input  to  electric  power   output  of  a  single–absorber  is  32%  based  on  a  semiconductor  with  a  band–gap  of   1.4  eV.  12     Figure   1.2  Solar   irradiance   at   the   surface   of   the  Earth.   The  band–gap   of   silicon   is   overlaid   as   an   example   showing   that   photons   absorbed   at   the   band–gap   can   be   converted  and  those  absorbed  above  the  band–gap  are  wasted  as  heat.         4   After  photogenerated  electrons  and  holes  are  created,  an  electric  field  is   required  to  separate  charges  such  that  they  can  be  transferred  to  an  external  load.   An  electric  field  can  be  established  by  interfacing  a  semiconductor  with  another   material  containing  a  different  work  function  (also  called  Fermi  level,  electron   affinity).  This  can  include  a  metal,  another  semiconductor,  doping  two  sides  of  the   same  semiconductor,  or  an  electrolyte  containing  a  redox  couple.  Once  interfaced,   charge  transfer  between  the  two  materials  occurs  until  equilibrium  is  established.   This  produces  a  region  in  each  material  that  is  depleted  of  majority  charge  carriers   (electrons  for  an  n–type  semiconductor  and  holes  for  a  p–type  semiconductor),   which  is  depicted  as  band–bending  within  the  semiconductor  (upward  for  n–type,   downward  for  p–type).  This  translates  to  a  built  in  potential  due  the  electric  field   formed  at  the  junction.  Upon  illumination,  a  non–equilibrium  concentration  of   photogenerated  electrons  and  holes  disturb  the  previously  established  equilibrium   formed  at  the  interface  and  the  electric  field  serves  to  separate  the  photogenerated   electrons  and  holes  such  that  they  can  be  extracted  to  do  electrical  work.  The   electrical  power  generated  could  be  used  directly.  However  due  to  the  intermittent   nature  of  sunlight,  it  is  also  important  to  store  the  electrical  power  generated  in  a   fuel.       1.4  Conversion  of  electrical  power  into  fuels     The  best–known  example  of  converting  solar  energy  and  storing  it  as   chemical  energy  can  be  found  in  nature.  Photosynthetic  organisms  capture  sunlight   and  convert  water  and  carbon  dioxide  into  oxygen  and  reduced  organic  species,   which  can  be  used  as  fuels.  Fuels  are  a  particularly  attractive  modality  for  storage     5   due  to  the  high  energy  density  the  chemical  bond.  The  primary  steps  in   photosynthesis  are  absorption  of  solar  energy  by  chlorophyll  and  other  pigments,   after  which  the  photogenerated  electrons  and  holes  are  separated  in  the   Photosystem  II  (PSII)  reaction  center.  The  oxidative  power  of  the  photogenerated   holes  in  PSII  are  transferred  to  the  oxygen  evolving  complex  to  split  water,   producing  molecular  oxygen  which  is  released  into  the  atmosphere,  as  well  as   protons  and  electrons  which  are  transferred  and  consumed  in  Photosystem  I  (PSI)   to  reduce  NADP+  into  NADPH  (natures  form  of  hydrogen),  which  is  ultimately  used   to  reduce  CO2  to  carbohydrates.  Since  products  from  the  water–splitting  reaction   are  subsumed  in  subsequent  photosynthetic  processes,  water–splitting  is  the  most   critical  step  in  photosynthesis.6,13,14     The  thermodynamics  of  water–splitting  can  be  described  by  the  following  oxygen   evolution  and  hydrogen  evolution  electrochemical  half  reactions  (OER  and  HER,   respectively):     2!   →   ! + 4! + 4 !     4! + 4 !   → 2!         combining  equations  (1)  and  (2)  indicates  that  a  total  voltage  of  1.23  V  is  required   to  drive  the  uphill  water–splitting  reaction.  However,  additional  voltage  is  necessary   to  drive  the  reaction  kinetics  or  rate  of  the  reaction  for  a  given  current  density  (JEC)   making  the  overall  voltage  for  water–splitting  (VEC):       Eoanode  =  1.23V  −  0.059(pH)  vs.  NHE                                  (1.1)   Eocathode  =  0V  −  0.059(pH)  vs.  NHE                                    (1.2)     6       !" !" =   !! +   !"# !" +   !"# !" +   ! (!" )   (1.3)   where,  ηOER  and  ηHER  are  the  anodic  and  cathodic  overpotentials,  respectively,  that   arise  from  the  intrinsic  activation  barrier  for  the  electrochemical  half–reaction   occurring  at  the  electrode–solution  interface  and  ηR  accounts  for  resistive  losses   which  can  arise  from  resistance  through  the  electrodes,  contacts,  or  mass  transport   limitations.  Water–splitting  catalysts  can  minimize  ηOER  and  ηHER.  While  the  impact   of  ηR  can  be  minimized  through  optimal  cell  designs,15–17  the  activation   overpotentials  are  intrinsic  properties  of  the  catalysts  utilized  at  the  anode  and   cathode.  This  overpotential,  which  is  also  a  metric  for  catalyst  activity,  is  typically   reported  in  units  of  mV  decade–1,  and  is  logarithmically  related  to  the  current   density  (J)  as  given  by  the  Tafel  law18:           where    b  is  the  Tafel  slope  and  J0  is  the  exchange  current–density  that  characterizes   the  intrinsic  activity  of  the  electrode  under  equilibrium  conditions.  In  order  to   optimize  the  efficiency  for  water–splitting,  that  is  the  ratio  of  the  thermodynamic   potential  for  water  splitting  to  the  thermodynamic  potential,  catalysts  exhibiting   high  J0  and  a  low  Tafel  slopes  are  necessary.     =  log     ! (1.4)     7   1.5  Photoelectrochemical  water–splitting   The  concept  of  a  photoelectrochemical  (PEC)  device  was  first  popularized  by   the  1976  paper  of  Fujishima  and  Honda.  19    They  described  immersing  a  TiO2   semiconductor  in  solution,  illuminating  it  with  UV  light,  and  observing  upon   application  of  a  potential  bias  the  evolution  of  both  hydrogen  and  oxygen.  Since  this   study,  hundreds  of  device–constructs  have  been  investigated  as  PECS.  They  can   broadly  be  classified  as  those  that  either  employ  a  solution  junction  (SJ)  or  buried   junction  (BJ)  for  charge  separation.20,21,22  While  the  physical  principles  underlying   the  operation  of  the  methods  are  quite  similar,21  the  position  of  charge  separating   interfaces  relative  to  interfaces  injecting  charge  into  water  redox  couples  has   important  consequences  for  implementation  of  either  method.  SJ–PEC  operates  on   the  principle  that  upon  submerging  a  semiconductor  in  a  solution  containing  a   redox  couple,  charge  transfer  at  the  interface  will  occur  provided  appropriate   alignment  between  the  semiconductor  Fermi  level  (EF)  and  the  Nernst  potential  of   redox  species.  The  depletion  region  formed  due  to  band–bending  within  the   semiconductor  allows  for  charge  injection  into  the  solution.  For  the  case  of  water   splitting  by  a  SJ–PEC,  the  quasi–Fermi  level  for  photogenerated  electrons  or  holes   must  straddle  the  thermodynamic  potential  for  the  water–splitting  reaction  (i.e.   1.23  V).23  Due  to  the  previously  mentioned  kinetic  overpotentials  the  actual  voltage   required  for  water–splitting  lies  between  1.6–2  V.  Since  the  photovoltage  generated   from  a  semiconductors  is  typically  at  least  0.4  V  less  than  its  band–gap,24  this   requires  the  semiconductor  to  have  a  band–gap  in  excess  of  2  V.  Therefore,  even   with  proper  band–alignment,  only  a  small  fraction  of  the  solar  spectrum  can  be     8   utilized  limiting  the  efficiency  to  7%.23,25  Furthermore,  semiconductors  are  rarely   good  water–splitting  catalysts.26  This  limitation  may  be  addressed  by  depositing   water–splitting  catalysts  on  the  semi–conductor  surface.  But  surface  modification   often  affects  the  efficiency  of  light  absorption  and  charge  separation  through  the   semiconductor–electrolyte  interface  (SEI).2728  Since  charge  separation  and  catalysis   are  intimately  tied  together,  optimization  of  optimization  of  the  individual   components  of  such  a  device  is  challenging  and  such  devices  have  only   demonstrated  solar–to–fuel  efficiencies  (SFE)  of  less  than  1%.29       Figure   1.3   Qualitative   schematic   of   an   n–type   semiconductor/electrolyte   junction   for  photoelectrochemical  water–splitting.     1.5.1  BJ–PEC  requirements     Many  of  the  aforementioned  challenges  with  SJ–PEC  can  be  overcome  by   relying  on  a  solid–state  semiconductor–semiconductor  junction  (also  referred  to  as   a  buried  junction)  to  perform  charge  separation.  In  the  BJ–PEC  configuration  a     9   solid–state  junction  is  formed  either  between  two  semiconductors  or  by  doping  two   sides  of  the  same  semiconductor.  By  controlling  the  doping–levels,  the  width  of  the   depletion  region  can  be  optimized  for  maximum  charge  separation.30  Thirty  years  of   ongoing  research  in  the  photovoltaic  (PV)  community  has  led  to  doping  as  a  mature   technology  and  optimal  charge  separation  and  photovoltage  characteristics  has   been  achieved.31  The  buried–junction  can  be  connected  to  relevant  interfaces  (e.g.   for  charge  injection  to  catalysts)  through  Ohmic  contacts,  which  can  be  either  thin– metal  films  or  conductive  oxides  deposited  on  the  surface  of  the  semiconductor.   Since  the  semiconductor  surface  is  completely  protected  from  the  aqueous   environment,  semiconductor  stability  no  longer  poses  a  problem.  Ultimately,  the   only  requirement  is  of  the  BJ–PEC  is  that  an  appropriate  voltage  is  supplied  to  drive   the  HER  and  OER  conversions.29     Figure  1.4  Qualitative  schematic  of  a  buried  semiconductor/electrolyte  junction  for   photoelectrochemical  water–splitting.       10   Furthermore,  decoupling  the  absorption  and  charge  rectification  properties  from   the  water–splitting  catalysis  enables  independent  optimization  of  all  the  required   components.  The  Ohmic  contacts  can  be  optimized  by  choosing  highly  conductive   materials  with  proper  band–alignment  to  allow  facile  charge  transport.32,33  Water– splitting  catalysts  can  be  independently  evaluated  and  interfaced.       1.5.2  BJ–PEC  devices     Many  buried  junction  BJ–PEC  devices  have  been  demonstrated  in  the  last  30   years.  To  date  the  highest  solar–to–fuels  efficiency  (SFE)  devices  utilized  either   expensive  multi–junction  III–V  solar  cells,34,35  low  efficiency  amorphous  silicon  (a– Si)  solar  cells,35–38  and  most  recently  copper  indium  gallium  diselenide  (CIGS)  solar   cells.39,40  In  all  cases  the  integrated  BJ–PEC  device  suffered  from  either  low  SFE,36–38   and/or  were  composed  of  expensive  PV  materials,  expensive  catalysts,  and  operated   in  strongly  acidic  or  basic  electrolytes  hindering  long–term  stability.34,35,39,40  For   these  reasons,  none  of  these  devices  were  realistic  for  economic  viability.  In  order  to   make  this  technology  realistic  from  both  a  cost  and  stability  perspective,  low–cost   high  efficiency  PV  materials  and  high  efficiency  earth–abundant  catalyst  that   operate  in  benign  aqueous  environments  are  necessary.       1.6  Crystalline  Silicon     Silicon  is  prime  candidate  material  for  buried–junction  devices  owing  to  its   almost  optimal  band–gap  of  1.1  eV  which  absorbs  a  large  fraction  of  the  solar   spectrum  and  it  is  the  second  most  abundant  material  on  the  planet.  Additionally     11   silicon  solar  cells  and  modules  are  one  of  the  most  mature  technologies  developed   for  solar  capture  and  conversion.31  Currently,  the  record  solar  conversion  efficiency   for  c–Si  solar  cells  has  hit  25%,  which  is  quite  impressive  considering  the   thermodynamic  limit  of  29%.31,41,42  Traditionally  silicon  PV’s  have  been  thought  to   be  too  expensive.11,13,43,44  However,  after  30  years  of  optimization  the  price  of  silicon   solar  cells  has  declined  and  the  conversion  efficiency  has  improved.  31,41  ,45    From   2004–2008  crystalline  silicon  (c–Si)  PV  modules  remained  steady  at  $3.5–$4  per   peak  watt  (WP–1).  However,  due  to  the  price  decrease  in  polycrystalline  silicon,   which  is  used  a  feedstock  material  for  c–Si,  in  2008  the  price  decreased  by  half  and   in  2011  fell  below  $1  Wp–1.41–47  In  order  to  be  cost  competitive  with  current  the   baseload  fossil  fuel  electrical  utility  plants  in  the  US  without  subsidies  the  price     Figure   1.5  Chinese   c–Si  PV   module  prices  since  2006.  The  data  was  adapted   from   ref.  43.     needs  to  further  decrease  to  $0.5–0.75  WP–1.  Modeling  and  outlined  pathways  show     12   that  this  goal  should  be  achievable  by  the  year  2020.47,48    However,  even  the  current   status  of  c–Si  PV’s  has  made  them  a  cost–competitive  technology  with  the  current   resources  used  in  developing  nations  such  as  Africa,  the  Persian  Gulf,  and  India.45       1.7  Earth–abundant  water–splitting  catalysts   Traditionally  catalysts  for  water–splitting  include  rare  earth  elements  of   noble  metals  including  Pt,  Ir,  Ru.49–51    Our  labs  changed  the  paradigm  by  discovering   active  catalysts  composed  of  Earth–abundant  materials.  Oxidation  of  Co2+  salts  in   buffered  solutions  yield  a  cobalt–oxide  water–oxidation  catalyst  self–assembles   onto  conductive  substrates.52,53  This  technique  has  been  extended  to  other  earth– abundant  metals  such  as  Ni  and  Mn.54–56  These  catalyst  are  stable  by  virtue  of  a  self– healing  mechanism,57–60  and  they  operate  under  a  variety  of  pH  ranges,55,56,61,62    and   in  the  presence  of  impurities.  61,62  Additionally,  it  has  been  shown  that  these   catalysts  can  be  easily  interfaced  with  semiconductors63–68  and  specifically  with   buried–junction  silicon  PV’s.36,69–71  Since  these  OER  catalysts  operate  under  a  variety   of  pH  neutral  conditions,  the  choice  of  catalyst  for  the  hydrogen  evolution  reaction   (HER)  has  not  required  platinum.36,72    Specifically,  NiMo(Zn)  alloys  for  hydrogen   evolution,  which  also  self–assemble  onto  conductive  substrates  from  an  aqueous   solution  containing  Ni2+,  sodium  molybdate  and  anhydrous  zinc  chloride  in  the   presence  of  pyrophosphate,  bicarbonate,  and  hydrazine.  Subsequent  leaching  in   base  produces  a  high  surface  area  material.73  Theses  alloys  are  able  to  achieve   current  densities  of  700  mA  cm–2  at  100  mV  overpotential  and,  with  continued     13   leaching,  can  attain  activities  as  high  as  at  1000  mA  cm–2  at  an  overpotential  of  35   mV.  72,74     Figure   1.6   Depictions   of   the   molecular   structure   of   our   Mn,   Co,   and   Ni   water– oxidation  catalysts.  Reprinted  with  permission  from  Mike  Huynh.         1.8  Overview   The  following  chapters  of  the  thesis  will  discuss  the  interfacing  of  water– splitting  catalysts  with  c–Si  photovoltaics  to  produce  BJ–PEC  devices  using  a   completely  modular  approach.  Chapter  2  focuses  on  directly  depositing  OER   catalysts  onto  single–junction  c–Si  PV’s  to  create  a  light–assisted  photoanode.  Of   particular  importance  is  the  ability  to  protect  silicon  from  the  oxidizing  conditions   required  for  water–splitting  with  a  protective  interface.  Fabrication  of  these  silicon   photoanodes  requires  optimization  of  two  interfaces:  a  silicon–protective  layer   interface  and  a  protective  layer  catalyst  interface.  Optimization  of  both  lead  to  a   lower  overpotential  (as  determined  by  Tafel  analysis)  required  for  OER.   Since  a  single–junction  c–Si  solar  cell  does  not  supply  the  voltage  required  to   achieve  water–splitting  without  the  use  of  an  external  potential  bias,  in  order  to   realize  a  stand–alone  water–splitting  device  based  on  c–Si,  multiple  single–junction     14   c–Si  solar  cells  need  to  be  connected  in  series.  Given  that  the  technical  aspects  of   device  integration  can  be  quite  challenging,  it  is  beneficial  to  predict  the  behavior  of   a  coupled  photovoltaic–electrochemical  device  (PV–EC).  In  Chapter  3,  steady–state   equivalent–circuit  analysis  of  a  PV  based  on  a  string  of  single–junction  c–Si  solar   cells  driving  an  electrochemical  load  based  on  the  OER–catalysts  developed  in  our   lab  allows  us  to  predict  the  coupled  behavior  between  the  PV  and  EC  components.   Importantly  this  allows  us  to  observe  the  impact  solar–to–fuel  efficiency  based  on   parameters  such  as  choice  of  catalysts  as  well  as  resistive  losses.     Guided  by  modeling  and  simulation,  a  modular  PV–EC  device  is  presented  in   Chapter  4  that  is  constructed  from  c–Si  and  non–precious  catalysts.  A  10%  solar–to– fuels  efficiency    is  demonstrated.  This  chapter  illustrates  how  a  modular  approach   allows  for  independent  characterization  of  each  component.       The  final  chapter  discusses  future  directions  for  the  improved  design  of   buried–junction  devices  for  solar–to–fuel  conversion.  Concepts  are  presented  for   improving  PV–EC  design  integration  and  minimizing  cell  resistances.  Utilization  of   alternative  photovoltaic  materials  is  presented  such  as  perovskite  solar  cells,  which   in  only  the  last  5  years  have  immerged  as  a  cheap  PV  material  with  efficiencies   competitive  with  c–Si.  75,76  Additionally,  in  order  to  facilitate  the  interfacing  of   water–splitting  catalysts  with  PV  materials,  a  brief  overview  and  preliminary  results   are  presented  that  establish  the  potential  of  vapor–deposition  techniques  to  provide   an  alternate  route  to  deposit  OER-­‐catalysts.         15   1.  9  Conclusion   Direct  solar–to–fuel  conversion  has  been  identified  as  a  “holy  grail”  of   science.13  However,  a  commitment  to  solution–based  PECs  over  the  last  50  years  has   impeded  progress  and  few  water–splitting  devices  demonstrating  the  necessary   efficiencies  have  been  conceived.  The  following  chapters  will  highlight  how  utilizing   a  modular  approach  to  design  buried  junction  devices  allows  for  independent   characterization,  optimization,  and  choice  of  all  components,  with  the  eventual  goal   of  creating  a  highly  efficient  stand–alone  water–splitting  device  composed  of  all   non–precious  technology  ready  materials.       16                                                                                                                     1.10  References.     1.    Energy   Information   Administration   (200514)   Annual   Energy   Outlook   (US   Dept.  of  Energy,  Washington  DC.)       2.     Lee  R  (2011)  The  outlook  for  population  growth.  Science  333,  569–573.     3.     United   Nations   department   of   Economic   and   Social   Affairs   (2013)   World   Population  prospects:  The  2012  Revision.  (United  Nations,  New  York).       4.     Holdren   JP   (2008)   Science   and   Technology   for   Sustainable   Well–Being.   Science  319,  424–434.       5.   Lewis   NS,   Nocera   DG   (2006)   Powering   the   planet:   chemical   challenges   in   solar  energy  utilization.  Proc.  Natl.  Acad.  Sci.  103,  15729–15735.       6.   Thomas   JM   (2014)   Reflections   on   the   topic   of   solar   fuels.   Energy  Environ.  Sci.   7,  19–20.     7.   Peters  GP,  Andrew  RM,  Boden  T,  Canadell  JG,  Cias  P,  Le  Quéré  C,  Marland  G,   Raupach  MR,  Wilson  C  (2013)  The  challenge  to  keep  global  warming  below  2   °C.  Nat.  Clim.  Change  3,  4–6.       8.   Barber   J   (2006)   Photosynthetic   energy   conversion:   natural   and   artificial.   Chem.  Soc.  Rev.  38,  15729–15735.     9.   United   Nations   Development   Program   (2000)   World   Energy   Assessment:   Energy  and  the  Challenge  of  Sustainability.  (United  Nations,  New  York).     10.   Grätzel  M  (2001)  Photoelectrochemical  cells.  Nature  414,  338–344.     11.   Grätzel  M  (2007)  Photovoltaic  and  photoelectrochemical  conversion  of  solar   energy.  Philos.  Trans.  R.  Soc.  Math.  Phys.  Eng.  Sci.  365,  993–1005.       12.   Nozik  AJ,  Miller  J  (2010)  Introduction  to  solar  photon  conversion.  Chem.  Rev.   110,  6443–6445.     13.   Bard  AJ,  Fox  MA  (1995)  Artificial  photosynthesis:  solar  splitting  of  water  to   hydrogen  and  oxygen.  Acc.  Chem.  Res.  28,  141–145.     14.   Blankenship   RE,   Tiede   DM,   Barber   J,   Brudvig   GW,   Ghiradi   M,   Gunner   MR,   Junge   W,   Kramer   DM,   Melis   A,   Moore   TA,   Moser   CC,   Nocera   DG,   Nozik   AJ,   Ort   DR,  Parson  WW,  Prince  RC,  Sayre  RT  (2011)  Comparing  photosynthetic  and     17                                                                                                                                                                                                                                                                                                                                             photovoltaic   efficiencies   and   recognizing   the   potential   for   improvement.   Science  332,  805–809.     15.   Haussener   S,   Xiang   X,   Spurgeon   JM,   Ardo   S,   Lewis   NS,   Weber   AZ   (2012).   Modeling,   simulation,   and   design   criteria   for   photoelectrochemical   water– splitting.  Energy  Environ.  Sci.  5,  9922–9935.       16.     Modestino   MA,   Walczak   KA,   Berger   A,   Evans   CM,   Newman   JA,   Ager   JW,   Segalman   RA   (2013)   Robust   production   of   purified   H2   in   a   stable,   self– regulating,   and   continuously   operating   solar   fuel   generator.  Energy  Environ.   Sci.  7,  297–301.     17.   Newman  J  (2013)  Scaling  with  ohm’s  law;  wired  vs.     wireless  photoelectrochemical  Cells.  J.  Electrochem.  Soc.  160,  F309–F311.     18.   Bard   AJ   (2000)   Electrochemical   Methods:   Fundamentals   and   Applications.   (Wiley,  2000).     19.   Fujishima   A,   Honda   K   (1972)   Electrochemical   photolysis   of   water   at   a   semiconductor  electrode.  Nature  238,  37–38.     20.   Surendranath   Y,   Bediako   DK,   Nocera   DG   (2012)   Interplay   of   oxygen– evolution   kinetics   and   photovoltaic   power   curves   on   the   construction   of   artificial  leaves.  Proc.  Natl.  Acad.  Sci.  109,  15617–15621.     21.   Jacobsson   JT,   Fjällström   V,   Edoff   M,   Edvinsson   T   (2014)   Sustainable   solar   hydrogen  production:  from  photoelectrochemical  cells  to  PV–electrolysis  and   back  again.  Energy  Environ.  Sci.  7,  2056–2070.       22.   Conibeer   GJ,   Richards   BS   (2007)   A   comparison   of   PV/electrolyzer   and   photoelectrolytic   technologies   for   use   in   solar   to   hydrogen   energy   storage   systems  Int.  J.  Hydrog.  Energy  32,  2703–2711.       23.   Weber   MF,   Dignam   MJ   (1986)   Splitting   water   with   semiconducting   photoelectrodes—efficiency   considerations.   Int.   J.   Hydrog.   Energy   11,   225– 232.       24.   Lewis   NS   (2005)   Chemical   control   of   charge   transfer   and   recombination   at   semiconductor  photoelectrode  surfaces.  Inorg.  Chem.  44,  6900–6911.       25.     Weber   MF,   Mamiche–Afara   S,   Dignam   MJ,   Pataki   L,   Venter   RD   (1987)   Sputtered  fuel  cell  electrodes.  J.  Electrochem.  Soc.  134,  1416–1419.       18                                                                                                                                                                                                                                                                                                                                             26.     Gerischer   H   (1977)   On   the   stability   of   semiconductor   electrodes   against   photodecomposition.   J.   Electroanal.   Chem.   Interfacial   Electrochem.   82,   133– 143.     27.   Zhong  DK,  Choi  S,  Gamelin  DR  (2011)  Near–complete  supporession  of  surface   recombination   in   solar   photoelectrolysis   by   “Co–Pi”   catalyst–modified   W:BiO4.  J.  Am.  Chem.  Soc.  133,  18370–18377.       28.     Lin   F,   Boettcher   SW   (2013)   Adaptive   semiconductor   /   electrocatalyst   junctions  in  water–splitting  photoanodes.  Nature  Mat.  13,  81–86.     29.   Walter   MG   et   al.   (2010)   Solar   water   splitting   cells.   Chem   Rev   110,   6446– 6473.     30.     Nelson  J.  The  Physics  of  Solar  Cells.  (Imperial  College  Press,  2003).     31.     Green,   MA   (2009)   The   path   to   25%   silicon   solar   cell   efficiency:   history   of   silicon  cell  evolution.  Prog.  Photovolt.  Res.  Appl.  17,  183–189.     32.     Gordon   RG   (2000)   Criteria   for   choosing   transparent   conductors.   MRS   Bull.   25,  52–57.     33.     Beyer  W,  Hüpkes  J,  Stiebig  H  (2007)  Transparent  conducting  oxide  films  for   thin  film  silicon  photovoltaics.  Thin  Solid  Films  516,  147–154.     34.     Khaselev  O,  Turner  JA  (1998)  A  monolithic  photovoltaic–   photoelectrochemical   device   for   hydrogen   production   via   water–splitting.   Science  280,  425-­‐427.     35.   Khaselev   O,   Bansal   A,   Turner   JA   (2001)   High–efficiency   integrated   multijunction  photovoltaic/electrolysis  systems  for  hydrogen  production.  Int.   J.  Hydrog.  Energy  26,  127–132.       36.     Reece   SY,   Hamel   JA,   Sung   K,   Jarvi   TD,   Esswein   AJ,   Pijpers   JJH,   Nocera   DG   (2011)   Wireless   solar   water   splitting   using   silicon–based   semiconductors   and  earth–abundant  catalysts.  Science  334,  645–648.     37.     Cristino   V   ,   Beradi   S,   Caramori   S,   Argazzi   R,   Carli   S,   Meda   L,   Tacca   A,   Bignozzi   CA   (2013)   Efficient   solar   water   oxidation   using   photovoltaic   devices   functionalized   with   earth–abundant   oxygen   evolving   catalysts.   Phys.   Chem.   Chem.  Phys.  15,  13083–13092.     38.     Rocheleu   RE,   Miller   EL,   Misra   A   (1998)   High–efficiency   photo– electrochemical  hydrogen  production  using  multijunction  amorphous  silicon   photoelectrodes.  Energy  Fuels  12,  3–10.     19                                                                                                                                                                                                                                                                                                                                               39.   Jacobsson   TJ,   Platzer–Björkman   C,   Edoff   M,   Edvinsson   T   (2013)     CuInxGa1– xSe2   as   an   efficient   photocathode   for   solar   hydrogen   generation.   Int.   J.   Hydrog.  Energy  38,  15027–15035.     40.   Jacobsson   TJ,   Fjällström   V,   Sahlberg   M,   Edoff   M,   Edvinsson   TA   (2013)   A   Monolithic   device   for   solar   water   splitting   based   on   series   interconnected   thin  film  absorbers  reaching  over  10%  solar–to–hydrogen  efficiency.  Energy   Environ.  Sci.  6,  3676–3683.     41.   Green  MA  (2003)  Crystalline  and  thin–film  silicon  solar  cells:  state  of  the  art   and  future  potential.  Sol.  Energy  74,  181–192.     42.   Swanson   RM   (2005)   Approaching   the   29%   limit   efficiency   of   silicon   solar   cells.   in   Conf.   Rec.   Thirty–First   IEEE   Photovolt.   Spec.   Conf.   889–894   doi:10.1109/PVSC.2005.1488274     43.     Lewis  NS  (2007)  Toward  cost–effective  solar  energy  use.  Science  315,  798– 801.     44.     Wadia   C,   Alivisatos   AP,   Kammen   DM   (2009)   Materials   availability   expands   the   opportunity   for   large–scale   photovoltaics   deployment.   Environ.   Sci.   Technol.  43,  2072–2077.       45.     Bazilian   M   (2013)   Re–considering   the   economics   of   photovoltaic   power.   Renew.  Energy  53,  329–338.       46.   Chu   S,   Majumdar   A   (2012)   Opportunities   and   challenges   for   sustainable   energy  future.  Nature  488,  294–303.     47.   Swanson  RM  (2012)  Plantary  talk  at  the  DOE  SunShot  grand  challenge:   Summit  and  technology  Forum.  (US  Department  of  Energy).     48.   Powell   DM,   Winkler   MW,   Choi   HJ,   Simmons   CB,   Berney   Needleman   D,   Buonassisi   T   (2012)   Crystalline   silicon   photovoltaics:   a   cost   framework   for   determining   technology   pathways   to   reach   a   baseload   electricity   costs.   Energy  Environ.  Sci.  5,  5874–5883.     49.     Morris  ND,  Mallouk  TE  (2002)  A  high–throughput  optical  screening  method   for   the   optimization   of   colloidal   water   oxidation   catalyst.   J.   Am.   Chem.   Soc.   124,  11114–11121.     50.   McCrory   CCL,   Jung   S,   Peter   JC,   Jaramillo   TF   (2013)   Benchmarking   heterogenous  electrocatalysts  for  the  oxygen  evolution  reaction.  J.  Am.  Chem.   Soc.  45,  16977–16987.     20                                                                                                                                                                                                                                                                                                                                               51.   McKone   JR,   Lewis   NS,   Gray   HB   (2014)   Will   solar–driven   water–splitting   devices  see  the  light  of  day?  Chem.  Of    Mater.  26,  407–414.     52.   Kanan   MW,   Nocera   DG   (2008)   In   situ   formation   of   an   oxygen–evolving   catalyst  in  neutral  water  containing  phosphate  and  Co2+.  Science  321,  1072– 1075.       53.   Kanan   MW,   Surendrantath   Y,   Nocera   DG   (2009)   Cobalt–phosphate   oxygen– evolving  compound.  Chem.  Soc.  Rev.  38,  109–114.       54.   Dincă  M,  Surendranath  Y,  Nocera,  DG  (2010)  Nickel–borate  oxygen–evolving   catalyst   that   functions   under   benign   conditions.   Proc.   Natl.   Acad.   Sci.   107,   10337–10341.       55.   Bediako   DK,   Surendranath   Y,     Nocera   DG   (2013)   Mechanistic   studies   of   the   oxygen   evolution   reaction   mediated   by   a   nickel–borate   thin   film   electrocatalyst.  J.  Am.  Chem.  Soc.  135,  3662–3674.     56.   Huynh   M,   Bediako   DK,   Nocera   DG     (2014)   A   functionally   stable   manganese   oxide  oxygen  evolution  catalyst  in  acid.    J.  Am.  Chem.  Soc.  136,  6002–6010.       57.   Lutterman  DA,  Surendranath  Y,  Nocera  DG  (2009)  A     self–healing  oxygen  evolving  Catalyst.  J.  Am.  Chem.  Soc.  131,  3838–3839.       58.   Surendranath  Y,  Lutterman  DA,  Liu  Y,  Nocera  DG  (2012)  Nucleation,  growth,   and  repair  of  a  cobalt–based  oxygen  evolving  catalyst.  J.   Am.   Chem.   Soc.   134,   6326–6336.       59.   Huynh   M,   Bediako   DK,   Liu   Y,   Nocera   DG   (2014)   Nucleation   and   growth   mechanisms   of   an   electrodeposited   manganese   oxide   oxygen   evolution   catalyst.    J.  Phys.  Chem.  C  doi:10.1021/jp501768n.     60.   Bediako   DK,   Costentin   C,   Jones   EC,   Nocera   DG,   Savéant   JM   (2013)   Proton– electron   transport   and   transfer   in   electrocatalytic   films.   Applications   to   a   cobalt–based  O2–evolution  catalyst.  J.  Am.  Chem.  Soc.    135,  10492–10502.       61.     Surendranath   Y,   Dincǎ   M,   Nocera   DG   (2009)   Electrolyte–dependent   electrosynthesis   and   activity   of  cobalt–based   water   oxidation   catalysts.   J.  Am.   Chem.  Soc.131,  2615–2620.       62.   Esswein  AJ,  Surendranath  Y,  Reece  SY,  Nocera  DG  (2011)  Highly  active  cobalt   phosphate   and   borate   based   oxygen   evolving   catalysts   operating   in   neutral   and  natural  waters.  Energy  Environ.  Sci.  4,  499–504.       21                                                                                                                                                                                                                                                                                                                                             63.    Sun   J,   Zhong   DK,   Gamelin   DR   (2010)   Composite   photoanodes   for   photoelectrochemical   solar   water   splitting.   Energy   Environ.   Sci.   3,   1252– 1261.     64.    Zhong   DK,   Gamelin   DR   (2010)   Photoelectrochemical   water   oxidation   by   cobalt   catalyst   (‘Co−Pi’)/α–Fe2O3   composite   photoanodes:   oxygen   evolution   and  resolution  of  a  kinetic  bottleneck.  J.  Am.  Chem.  Soc.  132,  4202–4207.       65.   Steinmiller   EMP,   Choi   KS   (2009)   Photochemical   deposition   of   cobalt–based   oxygen   evolving   catalyst   on   a   semiconductor   photoanode   for   solar   oxygen   production.  Proc.  Natl.  Acad.  Sci.  106,  20633–20636.     66.   Seabold  JA,  Choi  KS,  (2011)  Effect  of  a  cobalt–based  oxygen  evolution  catalyst   on   the   stability   and   the   selectivity   of   photo–oxidation   reactions   of   a   WO3   photoanode.  Chem.  Mater.  23,  1105–1112.     67.   Abdi  FF,  Han  L,  Smets  AHM,  Zeman  M,  Dam  B,  van  de  Krol  R  (2013)  Efficient   solar  water  splitting  by  enhanced  charge  separation  in  a  bismuth  vanadate– silicon  tandem  photoelectrode.  Nat.  Commun.  4,  1–7.     68.    Abdi   FF,     van   de   Krol   R   (2012)   Nature   and   light   dependence   of   bulk   recombination   in   Co–Pi–catalyzed   BiVO4   photoanodes.   J.  Phys.  Chem.  C   116,   9398–9404.     69.     Pijpers   JJH,   Winkler,   MT,   Surendranath   Y,   Buonassisi   T,   Nocera   DG   (2011)Light–induced   water   oxidation   at   silicon   electrodes   functionalized   with   a   cobalt   oxygen–evolving   catalyst.   Proc.   Natl.   Acad.   Sci.   108,   10056– 10061.       70.   Young  ER,  Nocera  DG,  Bulović  V  (2010)  Direct  formation  of  a  water  oxidation   catalyst  from  thin–film  cobalt.  Energy  Environ.  Sci.  3,  1726–1728.     71.   Cox  CR  ,  WinklerMT,  Pijpers  JJH,  Buonassisi  T,  Nocera  DG  (2012)  Interfaces   between   water   splitting   catalysts   and   buried   silicon   junctions.   Energy   Environ.  Sci.  116,  9398–9404.     72.   Nocera  DG  (2012)  The  artificial  leaf.  Acc.  Chem.  Res.  45,  767–776.     73.   Conway   BE,   Bai   L   (1986)   H2   evolution   kinetics   at   high   activity   Ni–Mo–Cd   electrocoated   cathodes   and   its   relation   to   potential   dependence   of   adsorption  of  H.  Int.  J.  Hydrog.  Energy  11,  533–540.       74.   Reece   SY,   Esswein   AJ,   Sung   K,   Green   Z,     Nocera,   DG     (2013)   Compositions,   electrodes,   methods,   and   systems   for   water   electrolysis   and   other   electrochemical  techniques.  US  Patent  Appl  No.  8361288.     22                                                                                                                                                                                                                                                                                                                                               75.   Hodes   G,   Cahen   D   (2014)   Photovoltaics:   perovskite   cells   roll   forward.   Nat.   Photon.  8,  87-­‐88.       76.   Snaith  HJ  (2013)  Perovskites:  the  emergence  of  a  new  era  for  low–cost,  high– efficiency  solar  cells.  The  Journal  of  Physical  Chemistry  Letters.  4,  3623–3630.     23   Chapter  2–Interfaces  between  crystalline  silicon   solar  cells  and  water–oxidation  catalysts     Portions  of  this  chapter  have  been  published:       Pijpers  JJH,  Winkler  MT,  Surendranath  Y,  Buonassisi  T,  Nocera  DG  (2011)  Light– induced  water  oxidation  at  silicon  electrodes  functionalized  with  a  cobalt  oxygen– evolving  catalyst.  Proc.  Natl.  Acad.  Sci.  U.S.A.  108,  10056–10061.   Cox  CR,  Winkler  MT,  Pijpers  JJH,  Buonassisi  T,  Nocera  DG  (2013)  Interfaces  between   water–splitting  catalyst  and  buried  silicon  junctions.  Energy  Environ.  Sci.  6,  532– 538.         24   2.1  Introduction:   The  general  requirements  for  direct  solar–to–fuels  conversion  as  well  as   device  constructs  were  described  in  Chapter  1.  However,  of  the  two  half–reactions   required  for  water–splitting,  the  complex  nature  of  the  proton  coupled  electron   transfer  (PCET)  chemistry  of  the  water–oxidation  reaction1–5  requires  the  largest   overpotential  (typically  around  250–400mV  overpotential  at  10mA  cm–2)6  and  thus   limits  the  overall  efficiency  for  solar–water–splitting  devices.7,8  For  this  reason  most   of  the  research  and  development  for  creating  solar–water–splitting  devices  has   focused  on  creating  photoanodes  that  are  capable  of  driving  the  water–oxidation   reaction  as  well  as  demonstrating  stability  under  the  highly  oxidizing  conditions.9,10   The  primary  focus  has  been  on  large  band–gap  n–type  metal  oxide   semiconductors.9  Large  band–gap  materials  should  be  able  to  supply  large   photovoltages  needed  as  well  as  remain  stable  under  oxidizing  conditions.11   However,  even  after  decades  of  research,  these  metal–oxide  photoanodes  still  face   numerous  challenges.  First  since  they  possess  large  band–gaps,  they  absorb  very   little  of  the  solar  spectrum.  Second,  they  are  very  rarely  configured  within  a  buried– junction,  so  they  suffer  from  all  the  aforementioned  limitations  of  the  solution– junction  approach.  Additionally,  since  the  water–oxidation  reaction  requires  four   hole  equivalents,  under  operating  conditions,  photogenerated  holes  accumulate  in   the  space  charge  layer  at  the  semiconductor–solution  interface  and  recombination   competes  with  water–oxidation.12–14  Lastly,  since  the  kinetics  of  the  oxygen   evolution  reaction  (OER)  are  rate  limiting,  oxygen  evolution  catalysts  (OECs)  must   be  placed  on  the  electrode  surface.15  Surface  modification  with  catalysts  can  affect     25   the  charge–separation  quality;  since  catalysts  typically  aren’t  transparent  only  thin   monolayers  can  be  used  otherwise  they  can  hinder  light  absorption.16   Creating  a  photoanode  with  a  buried–junction  configuration  can  circumvent   many  of  the  challenges  outlined  above.  Years  of  research  have  led  to  optimization  of   doping  crystalline  semiconductors  with  smaller  band–gaps  that  absorb  a  large   fraction  of  the  solar  spectrum.  The  focus  and  technology  ready  materials  consist  of   group  III–V  semiconductors  (GaAs,  GaP),  silicon  (crystalline  or  amorphous),  or  most   recently  focused  on  copper  indium  gallium  diselenide  (CIGS).  Due  to  the  abundance   and  overwhelming  market  advantage  of  silicon  discussed  in  Chapter  1,  it  seems  a   prime  materials  candidate  to  construct  a  photoanode  for  water–oxidation.     Requirements  to  enable  the  use  of  silicon  as  a  photoanode  material  include   the  need  to  overcome  the  instability  of  silicon  under  highly  oxidizing  conditions17  as   well  as  to  decrease  the  kinetic  limitations  for  the  water–oxidation  reaction  by   interfacing  silicon  with  OECs.  It  has  been  known  that  silicon  surfaces  can  be   protected  with  thin  metal  films  or  conductive  oxides.18–23  The  function  of  the   protective  layer  is  two–fold:  it  must  protect  the  silicon  surface  from  oxidation  and   enable  photogenerated  carriers  to  migrate  freely  from  the  buried  silicon  junction  to   immobilized  OECs.  In  other  words,  an  Ohmic  contact  has  to  be  established  between   the  PV  element  and  the  catalytic  sites,  allowing  for  charge  transport  with  minimum   voltage  drops.  This  requires  optimization  of  two  interfaces,  first  the  silicon– protective  layer  interface  and  second  the  protective  layer/catalyst  interface.  Herein   we  present  a  controlled  study  on  silicon  photoanode  performance  as  it  relates  to     26   interfaces  and  choice  of  both  the  protective  layer  (ITO,  FTO,  Ni)  and  OER  catalysts   (CoPi,  CoBi,  NiBi,  NiFeO).     2.2  Results     Fig.  2.1  shows  a  schematic  of  a  single–junction  crystalline  silicon  (c–Si)   photoanode  used  for  these  studies.  It  is  worth  noting  that  since  illumination  occurs   at  the  n–side,  the  p–side  can  be  coated  with  a  protective  layer  that  is  either  a   transparent  conductive  oxide  (TCO)  typically  either  tin–doped  indium  oxide  (ITO)   or  fluorine–doped  tin  oxide  (FTO),  or  an  opaque  thin  metal  film.  Additionally,  since     Figure   2.1   Schematic   of   the   OER–catalyst   functionalized   silicon   solar   cell   used   in   these   studies.   For   electrochemical   measurements,   an   external   voltage   may   be   applied   to  the  contacts   at   either   side  of  the  cell   with   or   without   illumination.   A.   The   solar  cell  is  operating  under  reverse  bias  conditions  with  voltage  applied  to  the  front   metal   contacts   on   the   n–side  in  the  dark.  B.  The  voltage  can  be  applied  directly  to   the   protective–layer,   in   which   case   the   PV   is   bypassed   and   the   current–voltage   characteristics   are   those   of   the   OER–catalyst   on   an   electrode.   C.   The   solar–cell   is   illuminated  with  AM  1.5  illumination  and  the  current–voltage  behavior  reflects  the   activity  of  the  OER–catalyst  functionalized  solar  cell.     the  OER  catalyst  is  deposited  on  the  protective  material,  which  is  on  the  p–side  of   the  photoanode,  there  are  no  limitations  as  to  the  thickness  of  the  catalyst  layer     27   since  light  absorption  occurs  at  the  n–side.  Because  the  Co–OECs  and  Ni–OECs   developed  in  our  lab  are  porous  in  nature,  this  allows  us  to  deposit  thicker  films,   which  contain  a  larger  number  of  active  sites.24,25  The  second  more  noteworthy   detail  is  the  addition  of  an  optional  highly  doped  p+–Si  layer,  which  was  previously   demonstrated  to  be  necessary  for  the  Ohmic  behavior  between  the  silicon  and   protective  layer  interface.20       Figure   2.2  CV  curves  of   (top)  npSi|ITO|CoPi  and  (bottom)  npp+Si|ITO|CoPi  in  0.1   M   KPi   electrolyte   at   pH   7   in   the   dark   with   Vappl   through   the   n–side   of   the   cell   (▬▬,   black),   with   Vappl   thought   the  n–side  under  1   sun  AM   1.5  illumination   (▬▬,   green)   ,   and  in  the  dark  with  Vappl  through  the  ITO  layer  bypassing  the  PV  (▬▬,  blue).  Taken   from  reference  20.       28     The  importance  of  the  p+–Si  layer  is  illustrated  Fig.  2.2,  which  shows  a  shows   the  cyclic  voltammagrams  (CV)  for  a  npSi|ITO|CoPi  photoanode  and    npp+Si|ITO|CoPi     operating  in  0.1  M  KPi  electrolyte  at  pH  7.  At  pH  7,  the  thermodynamic  potential  for   water–oxidation  is  0.82  V  vs.  the  normal  hydrogen  electrode  (NHE).  In  the  dark,   measuring  the  anodic  current  by  applying  the  potential  across  the  front  metal   contacts  on  the  n–side  (under  reverse  bias),  the  current  is  negligible  as  it  should  be   for  a  high  quality  PV  device.26  The  anodic  potential  can  also  be  applied  between  the   ITO  and  catalyst  layer.  In  this  case  the  PV  is  bypassed  and  the  current–voltage   characteristics  measured  simply  reflect  the  properties  of  the  CoPi  catalyst  on  an  ITO   electrode.  The  CV  shows  that  the  onset  for  water–oxidation  occurs  at  1.2  V  vs.  NHE,   which  is  in  agreement  with  the  overpotential  required  to  drive  the  water–oxidation     reaction  using  the  CoPi  catalyst  on  commercial  ITO  electrodes.27  When  illuminating   the  photoanode  from  the  n–side  with  a  light  intensity  of  100mW  cm–2  (AM  1.5   illumination),  the  potential  onset  for  water–oxidation  has  decreased  as  compared  to   the  measurement  where  PV–bypassed  indicating  that  some  of  the  photogenerated   holes  generated  in  the  silicon  are  injected  into  the  ITO  layer,  and  go  on  to  participate     in  the  water–oxidation  reaction  with  the  CoPi  catalyst.  However,  the  increase  in   current  as  a  function  of  applied  potential  exhibits  a  modest  slope  after  the  onset   potential.  In  the  absence  of  an  interface  oxide,  which  would  render  the  p–Si|ITO  a     Schottky–type  contact,  this  interface  can  be  represented  with  a  traditional   semiconductor|metal  contact  band–diagram.  As  p–Si  (EF  =  5.0–5.2  eV)28  is  brought   into  contact  with  ITO  (EF  =  4.4–4.7  eV),29,30  which  is  an  n–type  semiconductor   degenerately  doped  to  the  metallic  limit,  electrons  flow  from  ITO  generating  a     29   space–charge  layer  (up  to  1  μm)  in  p–Si  that  is  associated  with  downward  band– bending  (Fig.  2.3A).  The  relatively  low  density  of  acceptors  in  lightly  doped  p–Si   yields  a  large  space–charge  layer,  which  in  conjunction  with  downward  band– bending  provides  a  barrier  for  hole  transport  (Fig.  2.3B).  This  barrier  can  be   mitigated  by  heavily  doping  p–Si  to  create  a  p+–layer,  which  contains  a  higher   density  of  acceptors.  The  resultant  reduction  in  width  of  the  space–charge  layer   provides  a  path  for  photogenerated  holes  to  tunnel  from  p+–Si  into  ITO.    This   interface  can  now  be  considered  an  Ohmic  contact  (Fig.  2.3B).  Furthermore,   addition  of  a  p+–layer  introduces  a  back–surface  field,  which  acts  as  a  barrier  for  the   migration  of  photogenerated  electrons  toward  p+–Si|ITO  interface  that  could  result   in  deleterious  charge  recombination.26     Figure   2.3   Schematic   showing   the   band-­‐diagrams   at   the   p–Si|ITO   interface.   A.   Before   contact.  B.  After  equilibration   of  Fermi   levels   after   interfacing   p–Si   with  ITO.   C.  After  equilibration  of  the  Fermi  levels  after  interfacing  p+–Si  with  ITO.       The  effects  of  the  p+–layer  are  exemplified  by  examining  at  the  CV  of  the   npp+Si|ITO|CoPi  photoanode  under  illumination  in  Fig.  2.2B;  in  contrast  to  the   npSi|ITO|CoPi  the  current  increases  much  faster  with  applied  potential.  Additionally,     30   now  the  difference  in  the  onset  potential  for  water–oxidation  between  the  light  and   the  situation  where  the  PV  is  bypassed,  matches  the  VOC  of  the  solar  cell.       The  steady–state  performance  of  the  catalyst  functionalized  silicon  solar  cells   can  be  evaluated  by  using  Tafel  analysis.  While  Tafel  analysis  is  typically  used  to   determine  mechanistic  behavior,  for  the  experiments  described  here,  mechanistic   information  about  the  OER–catalysts  from  the  Tafel  plot  is  convoluted  by  the   electrical  properties  of  the  PV  and  TCO/Ni  components  of  the  anode.  Accordingly,   the  Tafel  data  is  examined  solely  as  a  measure  of  the  steady–state  activity  of  the   photo–assisted  anode  as  a  function  of  applied  potential.  Fig.  2.4  shows  the  Tafel     analysis  of  the  npp+Si|ITO|CoPi  in  the  dark  shows  a  high  slope  of  285  mV  decade–1,   again  corresponding  to  a  high  quality  junction.  In  the  configuration  where  the  PV  is     Figure   2.4  Tafel  plots  for  npp+Si|ITO|CoPi  with  potential  applied  to  the  metal  front   contact   for   measurements   in   dark   (black   squares,   n),   at   100   mW   cm–2   (green   squares,   n),   and   1000   mW   cm–2   (orange   red   squares,   n)   illumination.   The   blue   triangles  (▲)  correspond  to  a  measurement  in  dark  where  the  potential  was  applied   through  the  ITO  film  at  the  back  of  the  sample.  Figure  taken  from  ref  20.         31   bypassed,  the  slope  is  significantly  lowered  110  mV  decade–1.  The  slope  under   illumination  is  identical  and  the  only  difference  is  that  the  applied  voltage  for  the     water–oxidation  reaction  is  decreased  by  the  VOC  of  the  solar  cell.  This  implies  that   the  addition  of  the  p+–layer  allows  for  optimal  conduction  from  the  PV  to  the   catalyst.    However,  a  slope  of  110  mV  decade–1  is  still  higher  than  the  60  mV  decade– 1  slope  we  would  expect  based  on  studies  of  the  CoP  catalyst  on  commercial  ITO   i electrodes.  This  indicates  that  there  is  hindered  conduction  across  the  ITO  layer  to   the  OER  catalyst.20     2.2.1  Optimization  of  OER–catalyst  functionalized  silicon  solar–cells       The  solar  cells  used  in  the  previous  study  demonstrated  less  than  optimal   performance.  The  J–V  curve  for  generation  1  solar  cells  displayed  JSC  =  26.7mA  cm–2,   VOC  =  0.57  V,  and  fill  factors  of  0.47  giving  PV  efficiencies  of  7.1%.  This  limits  the   electrochemical  experiments  to  ~3  mA  cm–2,  which  is  adequate  for  the  purposes  of   general  electrochemical  characterization,  but  with  the  eventual  goal  would  be  to   implement  these  solar  cells  into  a  stand–alone  water–splitting  device  improvement   is  necessary.20  For  generation  2  solar  cell,  a  Si3N4  layer  was  employed  on  the  n–Si   surface  to  act  as  a  passivation  layer  by  tying  up  dangling  bonds  and  as  an  anti– reflection  coating.  Additionally,  the  front  metal  contact  grid  was  patterned   photolithographically  such  that  both  series  resistance  and  shadowing  were   minimized.  The  resulting  generation  2  solar  cells  possessed  JSC  =  28–34  mA  cm–2,  VOC   =  0.5–0.53  V,  and  fill  factors  ranging  from  0.7–0.77,  resulting  in  PV  efficiencies  of   10–13%.       32     Figure  2.5  Representative  J–V  curve  for   generation   2   npp+–Si  solar  cells  used  in  this   study  in  the  dark  (▬▬,  black)  and  under  AM  1.5  illumination  (▬▬,  blue).       The  previous  results  suggest  that  the  activity  of  the  OER–catalyst   functionalized  silicon  photoanode  depends  on  both  intrinsic  catalyst  activity  as  well   as  the  protective  material  used  to  interface  the  OER–catalyst  with  the  solar  cell.  For   typical  experiments  used  to  isolate  the  activity  of  an  OER–catalyst,  high  quality   commercial  electrodes  and  precisely  controlled  deposition  conditions  are  used  such   that  the  observed  current–voltage  behavior  is  truly  indicative  of  the  catalyst’s   mechanism.  However,  most  fabrication  procedures  that  enable  such  high  quality   electrode  materials  implement  high  temperature  deposition  conditions  that  are   detrimental  to  solar  cells.  In  order  to  achieve  maximal  anode  activity,  it  is  preferable   if  the  current–voltage  behavior  closely  mimics  the  intrinsic  behavior  of  the  catalyst   and  the  protective  material  used  has  little  influence  over  anode  activity.     33   Additionally,  since  ITO  is  a  rather  expensive  material,  it  has  been  known  to  show   exhibit  changes  in  electrochemical  properties  at  anodic  potentials,31  and  poor  long– term  stability19  it  is  preferable  to  substitute  ITO  with  alternative  materials  such  as   FTO  or  thin  metal  films.  For  this  reason,  a  variety  of  silicon  solar  cells  with  various   protective–layer/catalyst  combinations  were  examined  to  establish  design   principles  that  need  to  be  considered  when  interfacing  catalysts  with  silicon  solar   cells.  For  example,  Fig.  2.6  shows  scanning  electron  micrograph  (SEM)  images  of  the   CoBi  catalyst  deposited  on  FTO  and  Ni,  respectively.  In  both  cases  the  morphology  of   the  CoBi  catalyst  is  similar  and  comprises  a  roughened  surface  that  exhibits   spherical  nodules;  the  FTO|CoBi  electrode  shows  larger  aggregates  of  these  nodules   owing  to  a  longer  deposition  time  to  achieve  the  same  amount  of  charge  passed.   Since  Ni  is  prone  to  oxidation  at  the  same  potentials  required  for  deposition,  the   amount  of  charge  passed  over  a  given  time  period  does  not  correspond  to  a   controlled  catalyst  thickness,  as  is  the  case  of  an  ITO  or  FTO  electrode.     Figure  2.6  Plane   view  SEM  images  of  OER–catalyts   deposited  on  surface–protected   npp+Si|electrodes.  From  left  to  right  A  npp+|FTO|CoBi  and  B  npp+Si|Ni|CoBi.       The  oxidation  of  Ni  can  be  further  confirmed  by  measuring  the  Faradaic   efficiency  of  the  OER–functionalized  silicon  electrode,  which  allows  us  to  assess  the   extent  to  which  the  current  produced  from  the  cell  is  consumed  in  the  4e––4H+     34   water–oxidation  process  to  O2.     Figure   2.7   O2   production   measured   by   a   fluorescent   sensor   (▬▬,   red)   and   the   amount   produced   based   on   current   passed   assuming   100%   Faradaic   efficiency   (▬▬,  green)  for  (left)  npp+–Si|FTO|CoBi  and  (right)  npp+–Si|Ni|CoBi.     Fig.  2.7  shows  a  comparison  between  Ni  and  FTO  passivated  solar  cells,  both   functionalized  with  the  CoBi  catalyst.  The  Faradaic  efficiency  of  a  FTO–protected   electrode  is  unity  whereas  that  of  the  Ni  protective  coating  is  ca.  80%.       Fig.  2.8  shows  representative  Tafel  plots  for  a  CoBi–functionalized  solar  cell   with  the  three  different  interfaces  materials  ITO,  FTO,  and  Ni.  Consistent  with   previous  results,  in  all  cases,  the  onset  for  water  oxidation  is  shifted  negatively  by   0.5  V  at  any  applied  overpotential,  thus  indicating  that  the  photovoltage  of  the  solar   cell  is  efficiently  harnessed  for  solar–to–fuel  conversion.  Despite  the  similar  voltage   offset,  different  Tafel  slopes  are  observed  for  the  three  systems.  The  remaining   variation  in  the  slopes  of  the  blue  and  red  dots  can  be  fully  attributed  to  different   electrical  resistances  across  the  protective  layer/catalyst  interface,  since  the   composition  of  the  protective  layer  is  the  only  parameter  that  is  varied  between  Fig.   2.8  a–c.  The  Tafel  slopes  presented  in  Fig.  2.8  embody  the  behavior  of  the  inherent   activity  and  interfacial  resistance.     35     Figure   2.8   Tafel   plots   of   (a)   npp+Si|ITO|CoBi   (b)   npp+Si|FTO|CoBi   and   (c)   npp+Si|Ni|CoBi.   With   the   potential   applied   to   the   metal   front   contact   for   measurements  in  the  dark  (●),  under  1  sun  AM  1.5  illumination  (●),  and  in  the  dark   with  the  potential  applied  through  the  protective  coating  at  the  back  of  the  sample     (●).     Resistive  losses  at  this  interface  will  reduce  the  activity  of  the  electrode,  since   the  resulting  voltage  drop  results  in  an  effective  lower  potential  available  for  OER   catalysis.  Other  OER–catalysts  such  as  NiBi  and  NiFeO  were  also  employed  in  order   to  observe  the  influence  of  catalysts  with  better  activity  than  the  CoBi.    For  example,   The  CoBi  and  NiBi  catalysts  have  intrinsic  Tafel  slopes  of  60  and  30  mV  decade–1,   respectively,  arising  from  a  one–  electron  and  two–electron  pre–equilibrium  prior   to  a  chemical  rate–determining  step.27,32,33  For  the  case  of  NiFeO,  the  Tafel  slope  is   known  to  exhibit  variance  in  the  rate–limiting  step  depending  on  the  Fe  content,  but   is  generally  reported  to  be  45  mV  decade–1.34–36     The  Tafel  slopes  for  the  various  systems  are  summarized  in  shown  in  Graph  2.1   and  in  all  cases  the  overall  Tafel  slope  exceeds  the  intrinsic  Tafel  slopes  of  the   catalysts,  thus  indicating  that  there  is  a  series  resistance  resulting  from  the   interface.  Metal  interfaces  are  known  to  make  very  good  Ohmic  contacts18  whereas   the  electrical  properties  of  the  TCO’s  depends  strongly  on  the  deposition  conditions     36   of  the  TCO.18,37,38  It  should  be  noted  that  for  these  studies  the  electrical  and  quality   of  the  interface  materials  used  were  based  on  ease  of  fabrication  and  deposition   conditions  that  were  non–detrimental  to  solar  cell  operation  as  opposed  utilizing   high  temperature  commercial  deposition  techniques  such  as  atomic–layer   deposition  or  chemical  vapor  deposition  (ALD  and  CVD,  respectively),  which   produce  more  uniform,  high  quality  materials.  For  example,  the  sheet  resistance  of   the  protective  coating  used  varied  for  different  materials.  The  resistance  of   sputtered  Ni  was  measured  to  be  very  low  (13  Ω  sq–1).  But  the  sheet  resistance  of   the  FTO,  as  measured  in  a  four–point  probe  measurement  on  an  FTO  deposited  on   glass,  was  found  to  be  41Ω  sq–1.  For  commercial  FTO  on  glass  (prepared  by  CVD  at   600  oC),  the  sheet  resistance  is  7Ω  sq–1.  Using  profilometry,  we  determined  that  our   sprayed  FTO  had  a  thickness  of  400  nm,  so  the  voltage  drop  due  to  resistive  losses   within  such  a  thin  FTO  and/or  ITO  are  presumably  not  very  large.  However,  we   cannot  exclude  the  presence  of  local  inhomogeneities  in  the  FTO/ITO  that  could   potentially  lead  to  local  voltage  drops.       2.3  Discussion   The  different  Tafel  slopes  shown  in  Fig.  2.8  and  shown  in  Graph  2.1  depend   both  on  catalytic  activity  and  Ohmic  resistances  arising  primarily  between  the   catalyst|Si  interface,  and  the  variability  in  the  data  can  be  explained  by  one  or  a   combination  of  the  above  explanations.  To  isolate  the  latter,  we  examined  the  anode   activity  of  a  variety  of  catalyst–protective  coating  combinations  on  the  Si  buried     37   junction.  For  a  given  OER  catalyst,  the  trend  is  the  same.  The  Tafel  slope  for  a  FTO– coated  electrode  is  lower  than  that  of  ITO,  and  Ni–coated  electrodes  exhibit  high   variability.     Figure  2.9  Graph   showing   the  variability  in   Tafel  slope  for  various  combinations  of   OER–catalyst  functionalized  c–Si  solar  cells.    The  red  lines  indicate  the  value  based   on  previously  reported  Tafel  analysis.       We  believe  that  the  variance  of  the  Ni  interface  arises  from  oxidation  of  the   Ni  to  NiO.39  The  attenuated  Faradaic  efficiency  of  the  Ni–coated  electrode  supports   this  contention  inasmuch  as  some  current  would  be  appropriated  for  the  oxidation   of  Ni  to  NiO.  We  note  that  formation  of  a  NiO  layer  does  not  need  to  be  detrimental   to  the  overall  OER  process  since  NiO  is  known  to  be  conductive  and  hence  will  not   necessarily  impair  hole  transport  from  the  buried  junction.22,40    This  suggests  that   Tafel  slopes  do  not  always  just  reflect  Ohmic  resistances  but  that  the  current  does   not  necessarily  translate  to  water  splitting  as  indicated  by  Fig.  2.7.  It  is  conceivable     38   that  during  oxidation  of  Ni  to  NiO,  the  contact  of  the  protective  Si  surface  changes,   leading  to  inhomogeneous  interfacial  properties  and  changes  in  anode  activity,  as   we  observed.  We  show  in  Table  1  that  the  consequences  of  Ni  oxidation  may  be   mitigated  by  using  a  more  dense  catalyst  such  as  NiFeO  (Faradaic  efficiency  of  92%).   Table  2.1  Summary  of  Faradaic  efficiencies  for     npp+–Si | interface| catalyst  films.   Electrode   npp+Si|ITO|CoBi   npp+Si|ITO|NiBi   npp+Si|FTO|CoBi   npp+Si|FTO|NiBi   npp+|Si|FTO|NiFeO   npp+Si|Ni|CoBi   npp+Si|Ni|NiBi   npp+Si|Ni|NiFeO     Importantly,  recent  results  in  the  literature  have  shown  that  thin  metal  films   such  as  Ni  have  been  deposited  on  silicon  using  (using  Electron–beam  evaporation   or  ALD)  have  demonstrated  that  they  can  be  used  to  serve  as  both  a  protective   coating  and  once  oxidized,  as  the  OER–catalyst.22,41  Given  the  good  Ohmic  contact   between  silicon  and  Ni  or  other  metal  films,  such  an  approach  may  contribute  to   optimization  of  electrode  activity.     2.4  Conclusion     Faradaic   Efficiency  (%)   100   100   100   100   100   86   80   92   The  above  demonstrates  that  the  overpotential  for  water  splitting  by  a   buried–junction  photoelectrochemical  device  (BJ–PEC)  can  be  significantly     39   improved  through  optimization  of  the  interface  between  the  Si  junction  and  OER   catalyst.  Although  the  conductivity  and  quality  of  the  interface  materials  could  be   improved  by  optimizing  deposition  conditions  (ALD,  CVD  etc.),  what  remains  clear   is  that  the  materials  used  to  interface  semiconductors  with  water–splitting  catalysts   influence  photoanode  activity.  Furthermore,  the  ability  to  independently   characterize  and  optimize  each  component  of  the  photoanode  highlights  the   modular  approach  that  can  be  used  with  buried–junction  devices.    While  the  single– junction  solar  cells  used  do  not  supply  enough  voltage  to  split  water  without  the  use   of  an  external  potential  bias,  the  design  principles  reported  here  show  that  the   interface  optimization  between  the  silicon  solar  cell  and  OER  catalyst  results  in  a   higher  activity  photoanode,  which  can  be  used  to  construct  a  stand–alone  water– splitting  device.     2.5  Experimental     Solar  cell  fabrication.  Solar  cell  fabrication  generation  1  and  2:  For  the   studies  included,  two  iterations  of  solar  cell  fabrication  procedures  were  employed.   Boron–doped  p–type  silicon  was  commercially  purchased  (International  Wafer   Service)  and  used  as  a  starting  material  (3’’  diameter,  0.3–0.5  mm  thick).  The   resistivity  was  3  Ωcm–2,  corresponding  to  a  dopant  concentration  of  5×1015  cm–3.   Prior  to  phosphorous  diffusion,  wafers  were  cleaned  using  the  RCA  process  to   remove  organic  and  metallic  contaminants  as  described  in  the  Handbook  of   Semiconductor  Wafer  Cleaning  Technology  (41).  An  np–Si  junction  was  created  via   phosphorus  diffusion  by  heating  the  substrate  in  a  tube  furnace  while  flowing     40   phosphoryl  chloride  (POCl3)  in  a  nitrogen  carrier  gas  at  822  °C  for  20  min,  followed   by  20  min  of  annealing  in  an  O2–N2  atmosphere  at  822  °C.  In  this  configuration,   phosphorus  diffusion  occurs  on  both  sides  of  the  wafer,  resulting  in  an  npn–wafer.   The  processing  yields  an  emitter  with  a  sheet  resistance  between  60–70  Ω  square–1.     As  a  result  of  the  high–temperature  treatment  in  air,  the  surface  of  the  npn–wafer   was  covered  with  a  phosphorus  silicate  glass  surface  layer,  which  was  removed  by   dipping  in  10%  HF  solution.  To  create  the  optional  p+  layer,  a  1μm  film  of   aluminum–doped  silicon  (1%  Si)  was  sputtered  onto  one  side  the  npn–wafer   followed  by  a  rapid  thermal  annealing  (RTA)  step  in  N2  at  900  °C.  During  the  RTA   step,  Al  diffuses  through  the  n–Si  layer  and  converts  it  to  an  npp+–Si  wafer.  For   generation  2  solar  cells,  a  Si3N4  layer  was  employed  on  the  n–Si  surface  as  a   passivation  layer  and  as  an  anti–reflection  coating  layer.  The  Si3N4  layer  was  formed   in  a  Tystar  furnace  by  low–pressure  chemical  vapor  deposition  using  a  3:1  mixture   of  ammonia  and  dichlorosiline  (pressure  =250  mTorr)  at  770  °C  for  20  min,  yielding   a  80  nm–thick  film  with  a  refractive  index  of  2.01.  Additionally,  the  grid  was   optimized  for  both  series  resistance  and  shadowing.  Using  photolithography  seven   equally  spaced  grid  lines  (width  25  mm)  collected  current  from  a  1.10  cm2  device.     A  contact–passivation  layer  was  deposited  atop  the  aluminum  contact  to   isolate  it  from  catalytic  processes.  Films  of  indium–tin–oxide  (ITO)  or  Ni  were   deposited  on  the  p+–side  of  the  npp+–Si  wafer  using  an  AJA  International  sputtering   system  (Orion  5).  The  AJA  system  is  equipped  with  three  300  W  guns,  two  of  which   are  RF  for  either  conductive  or  dielectric  materials,  and  one  of  which  is  DC  for   conductive  materials  only  100  nm  thick  ITO  was  reactively  sputtered  in  an  11:1     41   mixture  of  Ar:O2.  To  improve  conductivity  of  the  ITO  layer37  the  entire  silicon  wafer   was  then  annealed  in  a  N2  atmosphere  at  400  °C  for  45  min.  Nickel  films  100  nm   thick  were  sputtered  in  a  pure  Ar  environment.  In  all  cases,  the  pressure  was  ~4  ×   10–6  bar  and  deposition  rates  of  all  films  were  measured  using  a  quartz  crystal   monitor  (QCM)  in  a  preliminary  ‘conditioning’  run.  FTO  was  deposited  on  the  p+– side  of  the  silicon  wafer  by  spray  pyrolysis.  In  a  typical  spray  deposition,  60  mL  of   ethanol  containing  4.2  g  SnCl4 •5H2O  and  0.7  mL  saturated  NH4F  was  used  to  coat   one  3ˊˊ  diameter  wafer  with  FTO.  Before  the  spray  pyrolysis,  the  silicon  wafer  was   heated  to  400  °C  in  air.  After  spraying  the  FTO  precursor  solution,  a  transparent  film   formed  on  the  wafer.  This  FTO  coated  wafer  was  annealed  at  400  °C  in  air  for  1  h.   Ti/Pd/Ag  metal  contacts  were  deposited  (20  /  20  /  250  nm,  Ti  adjacent  to  Si)  on  the   n–side  of  the  silicon  wafer;  the  residual  Al  from  the  formation  of  the  p+  layer  was   used  to  contact  the  base  of  the  PV  device.  Following  metallization,  the  3ˊˊ  diameter   wafer  was  cut  into  1.5  ×  2.5  cm2  pieces  using  a  1064  nm  YAG  laser  cutter.   Electrochemical  Methods.    Electrochemical  experiments  were  performed   using  a  CH  Instruments  760D  potentiostat  and  an  Ag/AgCl  reference  electrode   (BASi,  MF–2052).  All  electrode  potentials  were  converted  to  the  NHE  scale  using   E(NHE)  =  E(Ag/AgCl)  +  0.197.  Platinum  mesh  (Alfa  Aesar)  was  used  as  the  auxiliary   electrode.  Unless  otherwise  stated,  the  electrolyte  was  0.5  M  potassium  borate  (Bi)   at  a  pH  of  9.2  with  1.5  M  KNO3  as  a  supporting  electrolyte.     Catalyst  Film  Preparation.  Catalyst  films  were  deposited  on  completed  PV– devices  that  included  a  contact–passivation  layer.  Catalyst  films  of  CoBi  and  NiBi     42   electrodeposited  in  a  two–compartment  electrochemical  cell  with  a  glass  frit   junction  of  fine  porosity.  The  working  compartment  was  charged  with  ~50  mL  of   solution  of  25  mL  of  0.2  M  Bi  electrolyte  and  25  mL  of  1  mM  Co2+  or  Ni2+  solution.   The  auxiliary  compartment  was  charged  with  ~50  mL  of  0.1  M  Bi  electrolyte.  The   working  electrode  was  a  1.5  ×  2.5  cm2  piece  of  npp+–silicon  solar  cell,  fabricated  as   described  above.  Typically,  a  1  cm2  area  of  the  working  electrode  was  immersed  in   the  solution  and  electrolysis  was  carried  by  applying  0.85  V  vs.  NHE  directly  to  the   TCO/Ni  side  until  26  mC  cm–2  of  charge  had  passed  for  CoBi  deposition,  and  2.6  mC   cm–2  of  charge  had  passed  for  NiBi  deposition.  High–activity  NiBi  is  achieved  by   anodizing  NiBi  films  in  1  M  KBi  by  passage  of  3.5  mA  cm–2  for  1  h  with  stirring.33,42   NiFeO  catalyst  films  were  deposited  by  reactive  sputtering  using  the  AJA   International  sputtering  system,  mentioned  above.  Films  of  100  nm  thickness  were   deposited  by  reactive  sputtering  from  an  iron–doped  nickel  sputtering  target  (19%   Fe)  in  an  Ar:O2  atmosphere  of  3:1.34   Photoelectrochemical  Methods.  Photoelectrochemistry  experiments  were   performed  in  a  one–compartment  quartz  cell.  The  light  source  was  a  Sol  2A  solar   simulator  (Newport  Corp.).   Tafel  Data  Collection.  The  Tafel  behavior  of  surface–passivated,  catalyst– functionalized,  solar  cells  was  measured  in  the  region  of  water  oxidation  over  a  200   mV  range  in  10–30  mV  increments.  Depending  on  the  whether  Pi  or  Bi  analogous  of   the  catalysts  were  used,  measurements  were  conducted  in  a  solution  containing  0.5     43   M  KBi/KPi  and  1.5  KNO3  at  pH  9.2,  using  an  Ag/AgCl  reference  electrode  and  a  Pt   auxiliary  electrode.     O2  Quantification.  An  Ocean  Optics  oxygen  sensor  system  (NeoFox®  Phase   Measurement  System)  was  used  for  the  quantitative  detection  of  O2.  The  experiment   was  performed  in  a  custom  built  two–compartment  gas–tight  electrochemical  cell.   The  electrolyte  was  a  0.5  M  KBi  and  1.5  M  KNO3  solution  at  pH  9.2;  the  solution  was   purged  by  bubbling  with  high  purity  Ar  for  2  h  with  vigorous  stirring  and  it  was   then  transferred  to  the  electrochemical  cell  under  Ar.  The  two–compartment  cell   comprised  a  catalyst–functionalized  silicon  solar  cell  working  electrode,  an  Ag/AgCl   reference  electrode,  and  a  Pt  mesh  counter–electrode.  Measurements  of  the  FOXY   probe  were  recorded  at  10  s  intervals,  and  the  data  was  converted  into  the  partial   pressure  of  O2  in  the  headspace  using  a  calibration  curve  defined  by  air,  20.9%  O2   and  high  purity  N2  (0%  O2).  After  recording  the  partial  pressure  of  O2  for  2  h  in  the   absence  of  an  applied  potential,  the  sample  was  illuminated  with  1  sun  AM  1.5  light   and  electrolysis  was  initiated  at  0.6  V  vs.  NHE.  Due  to  variations  in  the  PV   characteristics  for  each  samples,  a  potential  bias  of  0.6  V  produced  operating   current  densities  ranging  from  1.5–5  mA  cm–2.  Electrolysis  with  O2  sensing  was   continued  for  ~12  h.  Upon  terminating  the  electrolysis,  the  O2  signal  was  recorded   for  an  additional  2  h  to  ensure  that  the  O2  yield  reached  a  plateau.  At  the  conclusion   of  the  experiment,  the  solution  and  the  headspace  volumes  in  the  working   compartment  were  measured  (around  55  mL  and  62  mL,  respectively).  The  total   charge  passed  in  the  electrolysis  was  divided  by  4F  to  get  the  theoretical  O2  yield.   The  partial  pressure  of  O2  was  converted  into  µmols,  and  corrected  for  dissolved  O2     44   in  solution  using  Henry’s  Law.  The  total  O2  yields  from  both  the  theoretical  and  the   experimental  results  were  used  to  determine  the  Faradaic  efficiency.     45                                                                                                                     2.6  References:     1.       Cukier   RI,   Nocera   DG   (1998)   Proton–coupled   electron   transfer.   Annu.   Rev.   Phys.  Chem.  49,  337–369.     2.     Eisenberg   R,   Gray,   HB   (2008)   Preface   on   making   oxygen.   Inorg.   Chem.   47,   1697–1699.       3.     Betley,  TA  et   al.  (2008)  A  ligand  field  chemistry  of  oxygen  generation  by  the   oxygen–evolving   complex   and   synthetic   active   sites.   Philos.   Trans.   R.   Soc.   B   Biol.  Sci.  363,  1293–1303.     4.     Hammes–Schiffer,   S   (2009)   Theory   of   proton–coupled   electron   transfer   in   energy  conversion  processes.  Acc.  Chem.  Res.  42,  1881–1889       5.     Concepcion   JJ   et   al.   (2009)   Making   oxygen   with   ruthenium   complexes.   Acc.   Chem.  Res.  42,  1954–1965.       6.     McCrory   CCL,   Jung   S,   Peters   JC,   Jaramillo   TF   (2013)   Benchmarking   heterogeneous   electrocatalysts   for   the   oxygen   evolution   reaction.   J.   Am.   Chem.  Soc.  135,  16977–16987.       7.      Rocheleau   RE,   Miller   EL   (1997)   Photoelectrochemical   production   of   hydrogen:  engineering  loss  analysis.  Int.  J.  Hydrog.  Energy  22,  771–782.     8.      Surendranath   Y,   Bediako,   DK,   Nocera   DG   (2012)   Interplay   of   oxygen– evolution   kinetics   and   photovoltaic   power   curves   on   the   construction   of   artificial  leaves.  Proc.  Natl.  Acad.  Sci.  109,  15617–15621.     9.       Osterloh   FE   (2008)   Inorganic   materials   as   catalysts   for   photochemical   splitting  of  water.  Chem.  Mater.  20,  35–54.     10.       Sun  J,  Zhong  DK,  Gamelin  DR  (2010)  Composite  photoanodes  for   photoelectrochemical   solar   water   splitting.   Energy   Environ.   Sci.   3,   1252– 1261.     11.       Gerischer   H   (1977)   On   the   stability   of   semiconductor   electrodes   against   photodecomposition.   J.   Electroanal.   Chem.   Interfacial   Electrochem.   82,   133– 143.     12.     Zhong  DK,  Choi  S,  Gamelin  DR  (2011)  Near–complete  suppression  of  surface   recombination   in   solar   photoelectrolysis   by   ‘Co–Pi’   catalyst–modified   W:BiVO4.  J  Am  Chem  Soc  133,  18370–18377.     46                                                                                                                                                                                                                                                                                                                                               13.     Lewerenz  HJ,  Peter  L  (2013)  Photoelectrochemical  Water  Splitting       (The  Royal  Society  of  Chemistry)     Available  at:  http://dx.doi.org/10.1039/9781849737739.     14.     Le   Formal   F   et   al.   (2014)   Back   electron–hole   recombination   in   hematite   photoanodes  for  water  splitting.  J.  Am.  Chem.  Soc.  136,  2564–2574.     15.     Walter  MG  et  al.  (2010)  Solar  water  splitting  cells.  Chem  Rev  110,  6446–6473   (2010).     16.   Fudig   L,   Boettcher   SW   (2014)   Adaptive   semiconductor   /   electrocatalyst   junctions  in  water–splitting  photoanodes.  Nat.  Mater.  13,  81–86.     17.     Zhang   XG   (2001)   Electrochemistry   of   Silicon   and   Its   Oxide.   (Springer,   New   York).     18.     Gordon   RG   (2000)   Criteria   for   choosing   transparent   conductors.   MRS   Bull.   25,  52–57.     19.     Kelly   NA,   Gibson   TL   (2006)   Design   and   characterization   of   a   robust   photoelectrochemical   device   to   generate   hydrogen   using   solar   water   splitting.  Int.  J.  Hydrog.  Energy  31,  1658–1673.     20.     Pijpers   JJH,   Winkler   MT,     Surendranath   Y,   Buonassisi   T,   Nocera   DG   (2011)   Light–induced   water   oxidation   at   silicon   electrodes   functionalized   with   a   cobalt  oxygen–evolving  catalyst.  Proc.  Natl.  Acad.  Sci.  108,  10056–10061.     21.     Young  ER.,  Nocera  DG,  Bulović  V  (2011)  Photo–assisted  water  oxidation  with   cobalt–based   catalyst   formed   from   thin–film   cobalt   metal   on   silicon   photoanodes.  Energy  Environ.  Sci.  3,  1726–1728.     22.     Sun   Ket   al.   (2012)   Nickel   oxide   functionalized   silicon   for   efficient   photo– oxidation  of  water.  Energy  Environ.  Sci.  5,  7872–7877.       23.    Kenney   MJ   et  al.   (2013)   High–performance   silicon   photoanodes   passivated   with  ultrathin  nickel  films  for  water  oxidation.  Science  342,  836–840.     24.     Farrow   CL,   Bediako   DK,     Surendranath   Y,   Nocera   DG,     Billinge   SJL   (2013)   Intermediate–range   structure   of   self–assembled   cobalt–based   oxygen– evolving  catalyst.  J.  Am.  Chem.  Soc.  135,  6403–6406.     25.    Bediako   DK,   CostentinC,   Jones   EC,   Nocera   DG,   Savéant,   JM.   (2013)   Proton– electron   transport   and   transfer   in   electrocatalytic   films.   Application   to   a   cobalt–based  O2–evolution  catalyst.  J.  Am.  Chem.  Soc.  135,  10492–10502.     47                                                                                                                                                                                                                                                                                                                                               26.     Nelson  J  (2003)  The  Physics  of  Solar  Cells.  (Imperial  College  Press,).     27.     Kanan   MW,   Surendranath   Y,   Nocera   DG   (2009)   Cobalt–phosphate   oxygen– evolving  compound.  Chem.  Soc.  Rev.  38,  109.     28.     Novikov   A   (2010)   Experimental   measurement   of   work   function   in   doped   silicon  surfaces.  Solid–State  Electron.  54,  8–13.     29.    Park   Y,     Choong   V,   Gao   Y,     Hsieh   BR,   Tang   CW   (1996)   Work   function   of   indium   tin   oxide   transparent   conductor   measured   by   photoelectron   spectroscopy.  Appl.  Phys.  Lett.  68,  2699–2701.     30.     Ho  JJ,  Chen  CY,  Hsiao  RY,  Ho  OL  (2007)  The  work  function  improvement  on   Indium−Tin−Oxide   epitaxial   layers   by   doping   treatment   for   organic   light– emitting  device  applications.  J.  Phys.  Chem.  C  111,  8372–8376.     31.     Kraft,  A.,  Hennig,  H.,  Herbst,  A.  &  Heckner,  K.–H.  Changes  in  electrochemical   and  photoelectrochemical  properties  of  tin–doped  indium  oxide  layers  after   strong  anodic  polarization.  J.  Electroanal.  Chem.  365,  191–196  (1994).     32.     SurendranathY.,   Kanan   MW,   Nocera   DG   (2010)   Mechanistic   studies   of   the   oxygen  evolution  reaction  by  a  cobalt–phosphate  catalyst  at  neutral  pH.  J.  Am.   Chem.  Soc.  132,  16501–16509       33.     Bediako   DK,   Surendranath   Y,     Nocera   DG   (2013)   Mechanistic   studies   of   the   oxygen   evolution   reaction   mediated   by   a   nickel–borate   thin   film   Electrocatalyst.  J.  Am.  Chem.  Soc.  135,  3662–3674.     34.     Miller   EL,   Rocheleau,   RE   (1997)   Electrochemical   behavior   of   reactively   sputtered  iron-­‐doped  nickel  oxide.  J.  Electrochem.  Soc.  144,  3072–3077.     35.     Rocheleau   RE,   Miller   EL,   Misra   A   (1998)   High–efficiency   photoelectrochemical   hydrogen   production   using   multijunction   amorphous   silicon  photoelectrodes.  Energy  Fuels  12,  3–10.     36.     Corrigan   DA   (1987)   The   catalysis   of   the   oxygen   evolution   reaction   by   iron   impurities  in  thin  film  nickel  oxide  electrodes.  J.   Electrochem.   Soc.  134,  377– 384.     37.     Hennig   H,   Heckner   KH,   Hirsch   D,   Ladwig   H   (1982)   The   Influence   of   the   preparation  on  layer  properties  of  transparent  indium–tin  oxide  electrodes.   Phys.  Status  Solidi  A  74,  133–139.       48                                                                                                                                                                                                                                                                                                                                             38.    Kurtz   SR,   Gordon   RG   (1987)   Transparent   conducting   electrodes   on   silicon.   Sol.  Energy  Mater.  15,  229–236.     39.     Visscher  W,  Barendrecht  E  (1980)  The  anodic  oxidation  of  nickel  in  alkaline   solution.  Electrochimica  Acta  25,  651–655.     40.     Carpenter   MK,   Corrigan   DA   (1989)   Photoelectrochemistry   of   nickel   hydroxide  thin  films.  J.  Electrochem.  Soc.  136,  1022–1026.     41.     Hu   S,   Shaner   MR,   Beardslee   JA,   Licterman   M,   Brunschwig   BS,   Lewis   NS   (2014)  Amorphous  TiO2  coatings  stabilize  Si,  GaAs,  and  GaP  photoanodes  for   efficienct  water–oxidation.  Science  344,  1005–1009.     42.     Bediako   DK   et   al.   (2012)   Structure–activity   correlations   in   a   nickel–borate   oxygen  evolution  catalyst.  J.  Am.  Chem.  Soc.  134,  6801–6809.     49   Chapter  3–  Modeling  a  coupled  photovoltaic   electrochemical  device  using  steady–state   equivalent  circuit  analysis.     Portions  of  this  chapter  have  been  published:   Winkler  MT,  Cox  CR,  Nocera  DG,  Buonassisi  T  (2013)  Modeling  integrated   photovoltaic–electrochemical  devices  using  steady–state  equivalent  circuits.  Proc.   Natl.  Acad.  Sci.  110,  E1076–E1082.         50   3.1  Introduction     Direct  solar–to–fuels  conversion  can  be  achieved  by  integrating  a   photovoltaic  device  with  water–splitting  catalysts.  In  Chapter  2,  single–junction   crystalline  silicon  (c–Si)  solar  cells  were  functionalized  with  oxygen  evolving   catalysts  (OECs)  to  construct  a  photo–assisted  device  for  promoting  the  water– oxidation  reaction.  However,  single–junction  c–Si  (VOC    =  0.5–0.7  V)  does  not  supply   enough  voltage  for  water–splitting  (Vth  =  1.23  V)  without  the  use  of  an  external   potential  bias.  In  order  to  supply  the  full  potential  needed  for  water–splitting  based   on  c–Si  solar  cells,  multiple  single–junction  c–Si  solar  cells  need  to  be  connected  in   series.  Steady–state  equivalent  circuit  of  analysis  of  a  series  connected  string  of   single–junction  c–Si  solar  cells,  coupled  to  the  electrochemical  process  of  water– splitting  allows  us  to  predict  the  coupled  behavior  of  a  PV–EC  device.  Specifically,   for  a  PV  component  consisting  of  a  string  of  series–connected  c–Si  solar  cells,  we  can   predict  the  number  of  solar  cells  needed  to  produce  a  10%  or  higher  steady–state   solar–to–fuels  efficiency  (SFE)  based  on  the  choice  of  water–splitting  catalysts  as   well  as  assess  resistive  losses.  This  is  highly  beneficial  since  one  of  the  primary   challenges  in  creating  a  stand–alone  photovoltaic–electrochemical  (PV–EC)  device  is   design  integration.  The  ability  to  predict  the  coupled  PV–EC  behavior  before   undertaking  the  complicated  materials–related  integration  aspects  allows   considerable  research  struggles  to  be  circumvented.         51   3.2  Efficiency  considerations   The  steady–state  solar–to–fuels  efficiency  (SFE)  of  a  photovoltaic  used  to   drive  and  electrochemical  load  has  been  described  previously1–7  and  is  given  by:           where,  ηPV,  ηEC,  ηC  are  the  efficiencies  of  the  PV  device,  the  EC  components  (including   electrodes  and  wires),  and  the  efficiency  of  the  coupling  between  the  two.3,8     The  maximum  efficiency  of  a  single–absorber  solar  cell  is  limited  by  the  solar   spectrum.  For  maximum  conversion  efficiency  to  be  achieved,  a  single–absorber   solar  cell  must  absorb  a  large  portion  of  the  solar  spectrum  to  generate  a  large   photocurrent  with  a  concomitant  large  phototvoltage.  The  latter  contradicts  the   former  since  in  order  to  generate  a  large  photovoltage  a  large  band–gap  is  required,   which  in  turn  results  in  the  absorption  of  only  a  small  fraction  of  the  solar  spectrum,   thus  leading  to  a  small  photocurrent.  Additionally,  due  to  various  types  of  charge   recombination  mechanisms,  typical  solar  cell  performance  is  usually  300–400  mV   lower  than  the  band–gap  at  room  temperature.4,9     The  actual  efficiency  of  a  given  solar  cell  can  be  determined  by  examining  the   current–voltage  characteristics  of  the  solar  cell  under  illumination  (Fig.  3.1).  The   open–circuit  voltage  (VOC)  and  short–circuit  current  (JSC)  are  the  maximum  voltage   and  current  of  the  cell,  respectively.  However,  since  the  VOC  occurs  when  no  current   is  flowing  and  the  JSC  occurs  where  the  voltage  is  zero,  the  actual  maximum  power   (PMAX)  output  of  the  solar  cell  occurs  where  the  product  of  voltage  and  current  is  the   !"# =   !"  ×  !"  ×  !   (3.1)     52   greatest  (PMAX  =  VMP  ×  JMP)  and  is  referred  to  as  the  maximum  power–point  (MPP).   The  fill–factor  (FF)  describes  the  ratio  of  the  maximum  power  of  the  solar  cell  to  the   product  of  VOC  and  JSC.    Therefore,  the  PV  efficiency  given  at  Air  Mass  (AM)  1.5   spectrum  with  an  incident  power  density  of  100mA  cm–2,  is  then  defined  by:         For  silicon  with  a  band–gap  of  1.1  eV,  the  upper–bound  solar  power  to   electrical  power  efficiency  of  29%.10,11  At  present  c–Si  solar  cells  have  reached  a   fabrication  maturity  such  that  the  a  record  efficiency  of  25%  has  been  achieved12   technology  ready  available  PV  efficiencies  are  in  the  18–20%  range  with  module   efficiencies  of  14–20%.13,14     The  additional  voltage  required  to  compensate  for  reaction  kinetics,  for  a   given  operational  current–density,  dominates  the  efficiency  of  an  electrochemical   process  such  as  water–splitting.  The  efficiency  for  water–splitting  is  determined  by   how  much  voltage  is  needed  to  drive  the  reaction  (VOP)  as  compared  to  the   thermodynamic  potential  (Vth)  and  is  given  by:         For  water–splitting  in  which  the  oxygen–evolution  reaction  occurs  at  the   anode  and  the  hydrogen  evolution  reaction  occurs  at  the  cathode,  VEC  accounts  for   !" =   !!   !" (3.3)   !" =      ×  !"  ×  !"   !"# (3.2)     53   the  thermodynamic  potential  and  the  overpotential  for  each  half–reaction,  as  well  as   cell  resistances  (ηR)  (i.e.  resistance  through  electrodes,  contacts,  and  solution   resistance).  The  total  voltage  required  is:           where  the  ηOER  and  ηHER    are  the  overpotentials  for  the  oxygen–evolution  and   hydrogen–evolution  reactions  (OER  and  HER)  and  for  a  given  current–density  JOP   are  given  by  the  Tafel  law:           where  b  is  the  Tafel  slope  and  Jo  is  the  exchange  current  density.  The  overall   efficiency  for  water–splitting  is  typically  limited  by  complex  nature  of  the  proton– coupled  electron–transfer  (PCET)  chemistry  of  the  water–oxidation  reaction15–20  In   order  to  minimize  the  overpotential  catalysts  are  used  to  effect  the  half  reactions   and  different  catalysts  operate  under  different  mechanisms,  which  determines  the   rate  for  the  reaction.  For  example,  mechanistic  studies  for  the  Co–OEC’s  and  Ni– OEC’s  developed  in  out  lab  exhibit  different  Tafel  slopes  of  30  and  60  mV  decade–1,   respectively,  which  are  indicative  of  a  mechanism  involving  a  one  versus  two– electron  reaction  that  is  proceeded  by  a  rate–limiting  chemical  step  for  oxygen   evolution.21,22,23  Typically,  the  relevant  figure  of  merit  is  the  overpotential  required   !"#,!"# = log !" !"#,!"#   ! (3.5)   !" !" =   !! +   !"# !" +   !"# !" +   ! (!" )   (3.4)     54   to  operate  at  a  current  density  of  10mA  cm–2,  which  for  most  OER  catalysts  requires   an  overpotential  of  250–400mV.24,25  Commercial  alkane  electrolyzes  exhibit   efficiencies  around  60–75%,26,27  and  given  the  best  water–splitting  catalysts   developed  in  our  group  a  comparable  efficiency  may  be  achieved.28     The  efficiency  of  coupling  between  the  PV  and  EC  components  is  not   fundamentally  limited  by  any  physical  constraint,  but  is  entirely  dependent  on  how   well  matched  the  maximum  power  output  of  the  PV  is  to  the  electrochemical  load   for  water–  splitting.  The  simplest  mode  of  coupling  the  PV  and  EC  components  is  to   perform  the  anodic  half–reaction  (OER)  on  the  positive  terminal  and  the  cathodic   half  reaction  (HER)  on  the  negative  terminal  of  the  PV–device.  This  type  of  coupling   Figure   3.1   Schematic   of   a   wired   and   wireless   PV–EC   based   on   silicon   solar   cells.   Regardless   of   the   mode   of   coupling   between   the   two,   the   equivalent   circuit   is   identical.       enforces  the  two–half  reactions  to  be  equipotential  with  the  two  terminals  of  the   PV–device,  which  is  true  regardless  if  the  PV  is  a  buried–junction  stack  or  series– connected  single–junction  solar  cells  wired  to  an  anode  and  cathode  (i.e.  a  wired   and  wireless  configuration,  Fig.  3.1).9,29  Fig.  3.2  illustrates  that  this  direct  electrical       55   connection  is  equivalent  to  constraining  the  currents  and  voltages  of  the  PV–EC   system  to  be  identical:  JPV  =  JEC  and  VPV  =  VEC.  The  point  of  intersection  (as  shown  in   Fig.  3.3)  between  the  PV–curve  and  EC–curve  also  defines  the  operational  point  for  a   Figure   3.2  Block  diagram  for  a  photovoltaic  (PV)  powered  electrochemical  cell  (EC),   where  direct  electrical  connection  constrains  JPV  =  JEC  and  VPV  =  VEC       coupled  PV–EC  device.6,7,30  Perfect  coupling  occurs  when  the  intersection  of  the  PV   and  EC  curves  occurs  at  the  PMAX  of  the  solar  cell  meaning  that  the  maximum  power   output  of  the  PV–component  is  utilized  by  the  EC–component  and  is  defined  by:           Using  the  above  definitions  for  ηPV,  ηEC,  and  ηC    of  the,  equation  3.1  can  be  re–written   as:       !"# =      ×  !"  ×  !" !! !"  ×!" !!  ×  !"  ×    ×   =           !"# !"  ×  !"   ×  !" !"# (3.7)   ! =   !"  ×!"    ×  !"   ×  !" (3.6)     56     However,  at  AM  1.5,  100mW  cm–2  of  solar  irradiance  reduces  the  expression  to   1.23V  ×  JOP.     Figure  3.3   The   generalized   current   density–voltage   (J–V)   diagram   of   a  coupled  PV– EC   system   where   the   point   of   intersection   of   the   PV–curve   (▬▬,   blue)   and   EC– curve   (▬▬,   red)   represents   the   operational   point   and   SFE   of   the   coupled   PV–EC   device.  The  SFE  is  maximized  when  the  operating  point  is  equal  to  PMAX.         3.3  Steady–state  equivalent  circuit  analysis     Solar  cells  and  electrochemical  reactions  do  not  operate  at  thermodynamic   limits.  The  steady–state  electrical  behavior  of  practical  PV  and  EC  systems  can  be   described  using  equivalent  circuits.  Circuit  analysis  enables  accurate  modeling  of     57   PV–EC  devices  and  provides  insight  into  their  realistic  performance  and  efficiency   limitations.  The  equivalent  circuit  of  a  PV  device  is  well  known  and  has  been   described  extensively  in  the  literature  and  textbooks.31  For  the  purposes  of   equivalent  circuit  analysis  of  the  PV  component,  the  relevant  circuit  elements  are     Figure   3.4  Impact  on  the   J–V  curve  for  a  PV  due   to  shunt  (▬▬, dark  red)   or   series   (▬▬, dark  green)  resistance  compared  to  an  ideal  J–V  curve  (▬▬, dark  blue).   discussed.  Briefly,  in  a  PV  device  light  absorption  generates  a  current  of  excited   electrons  and  holes.  Some  are  separated  by  the  internal  junction  and  flow  through   an  external  circuit  with  current  density  JPV  and  some  are  lost  to  recombination   processes  and  contribute  to  the  dark  current  of  the  solar  cell  J0.  Typically,  only  two   recombination  mechanisms  are  considered  to  be  relevant.  The  first  recombination   mechanism  is  dominant  at  high  voltages  when  bulk  and  surface  recombination  are   more  prevalent.  In  this  case  the  ideality  factor  n,  which  is  a  measure  of  how  closely     58   the  solar  cell  follows  the  ideal  diode  equation,  is  close  to  1.  At  lower  voltages,   recombination  within  the  junction  itself  is  the  dominant  recombination  mechanism   and  makes  the  ideality  factor  n=2.  Both  of  these  recombination  mechanisms  are   modeled  by  adding  two  diodes  in  parallel  to  the  light  generated  current  JL.  Ohmic   losses  can  be  modeled  by  introducing  shunt  (Rsh)  and  series  resistances  (Rs).  Shunt   resistance  is  caused  by  manufacturing  defects  and  provides  an  alternate  pathway   for  light  generated  charge–carriers  to  flow  ultimately  decreasing  the  output   current–density  (Fig  3.4).  Series  resistance  is  due  to  either  resistance  through   contact  resistance  or  front  metal  contacts  to  silicon  and  decreases  the  output   voltage  of  the  solar  cell.  The  output  current–density  JPV  is  the  difference  between  the   light  generated  photocurrent  JL  and  the  recombination  currents:           where  q  is  the  electron  charge,  T  is  temperature,  and  kB  is  Boltzmann’s  constant.  If   multiple  solar  cells  are  connected  in  series  then  the  output  voltages  are  additive  and   the  current  density  of  the  system  will  decrease  by  dividing  the  current  by  the  active   area  under  illumination.     Since  the  Tafel  law  can  be  rearranged  to  appear  identical  to  the  diode   equation,  each  electrochemical  half–reaction  can  be  represented  as  a  diode  under   reverse  bias  equal  to  the  thermodynamic  potential  for  water–splitting.  Given  the   constraints  imposed  by  coupling  the  two  PV  and  EC  components,  the  equivalent   !" =   ! −   !,!!! ! !!  !!" !! !! !! ! −   !,!!! ! !!  !!" !! !! !! ! − +   !" !   !! (3.8)     59   circuit  in  Fig.  3.5  can  be  solved  numerically  for  the  point  of  intersection  between  the   two  resulting  in  the  predicted  SFE  (recall  1.23  V  ×  JOP)  for  the  coupled  system.     Figure   3.5  Steady–state  equivalent   circuit  of  a  PV–EC  system.  An  applied  voltage  is   incorporated  to  illustrate  analysis  of  an  externally  assisted  system.       3.4  Results  and  Discussion     The  PV  parameters  used  herein  assume  that  the  system  is  powered  by   multiple  high–performance  c–Si  solar  cells  with  ηPV  =  20%  (Table  3.1).  Since  the  OER   reaction  is  typically  efficiency  limiting,  two  different  sets  of  OER–catalyst   parameters  based  on  the  Co–OEC’s  and  Ni–OEC’s  developed  in  our  lab  were  used  for   comparision  (Table  2).7,  22,23,28,  32–35  For  the  HER  reaction  the  Tafel  behavior  is  based   on  a  NiMoZn  alloy.28,  36     60   Table  3.1  Solar  cell  parameters  for  the  modeling   Solar  cell  parameters   J0   RS     VOC   JSC   Efficiency  (%)     Table  3.2  Electrochemical  parameters  for  the  modeling.   Electrochemical  parameters     Case  I:   OER  kinetics    Ni–OEC   Case  II:      Co–OEC       Tafel  slope   (mV  decade–1)   30   60     30   Exchange  current   density  (A  cm–2)   5  x  10–18   2.1  x  10–12     1  ×  10!!     4  x  10–10mA  cm–2   1.5  Ω  cm2   669  mV   41  mA  cm–2   20   HER  kinetics     3.4.1  Impact  of  ηPV  on  SFE     Fig.  3.6  shows  a  comparison  between  a  comparison  between  two  PV’s  where   one  PV  is  improved  above  the  baseline  ηPV  =  20%  to  ηPV  =  23.3%  via  a  higher  JSC  (47   mA  cm–2)  and  the  other  is  improved  via  a  higher  VOC  (0.75  V).  The  first  realization     61     Figure   3.6   Impact   on   SFE   via   improvement   in   PV   efficiency   compared   to   the   baseline  ηPV  =  20%  (–––––,  grey  dash).  Given  optimal  coupling  between  the  PV  and   EC   components   (top)   a   higher   relative   SFE   can   be   obtained   by   improving   the   JSC   (▬▬,   green)   as   opposed   to   the   VOC   (–––––,   dashed   green).   Given   poor   coupling   between   the   baseline   PV   and   EC  (bottom),   only   minor  improvements   in   the   SFE   can   be  obtained.       is  that  for  Case  I  EC  parameters  in  which  there  is  good  coupling  between  the  PV  and     62   EC  components  (i.e.  the  EC  curve  intersects  to  the  left  of  PMAX),  improving  ηPV  via   increasing  JSC  gives  a  larger  relative  increase  in  SFE  as  compared  to  increasing  the     VOC  (15%  vs.  1%).  Additionally,  the  higher  voltage  PV  hardly  improves  the  SFE   compared  to  the  baseline  ηPV  =  20%.  However,  for  Case  II  EC  parameters,  (i.e.  when   the  EC  curve  occurs  to  the  right  of  PMAX),  the  higher  current  and  higher  voltage  PV’s   show  a  similar  increase  in  SFE.     3.4.2  Impact  of  ηEC  efficiency  on  SFE   The  equivalent–circuit  analysis  also  allows  us  to  determine  the  number  of   solar  cells  needed  to  achieve  a  high  SFE  given  choice  of  catalyst  as  well  as Figure  3.7  J–V  curves  of  multiple  series  connected  solar  cells  with  ηPV  =  20%  (▬▬,   grey)  and  EC  curves  (▬▬,  dark  blue).  The  number  of  solar  cells  required  changes   based  on  choice  of  catalyst  which  causes  the  EC  curve  to  shift  left  or  right  and   resistive  losses  due  to  RSOL  cause  the  EC  curve  to  tilt  down.         63   resistive  losses.  Since  an  ionic  conductivity  is  at  least  four  orders  of  magnitude  less   than  electronic  conductivity  (i.e.  resistance  through  electrodes  or  wires)  we  choose   to  specifically  focus  on  solution  resistance  RSOL.  Figure  3.7  shows  that  choice  of   catalyst  shifts  the  EC  curve  horizontally  and  that  RSOL  causes  the  EC  curve  to  tilt   down  and  both  change  the  number  of  solar  cells  required.     Fig.  3.8  shows  the  SFE  as  a  function  of  solution  resistance  given  the  ηPV  =  20%   parameters  and  two  sets  of  EC  parameters  (Tables  3.1  and  3.2).  In  order  to  maintain   an  SFE  of  10%  or  higher,  the  number  of  solar  cells  changes  and  3–5  solar  cells  are   required.  For  Case  I  EC  parameters  an  SFE  of  10%  based  on  3  solar  cells  is  only   achieved  with  minimal  solution  resistance  RSOL  and  rapidly  drops  below  10%.   However,  by  increasing  to  4  cells,  a  12%  SFE  can  achieved  until  the  RSOL  surpasses   90  Ω  cm2.  Interestingly,  given  the  set  of  PV  parameters,  increasing  to  5  cells   produces  just  under  a  10%  SFE  because  JOP  /  5  is  less  than  the  minimum  current– density  needed  to  give  a  10%  SFE  (i.e.  8.13  mA  cm–2).  The  trends  are  the  same  for   Case  II  EC  parameters.  However  the  lower  Tafel  slope  of  the  NiBi  catalyst  results  in  a   larger  ηEC,  producing  a  14.5%  SFE  using  3–cells  and  maintains  >  10%  until  RSOL   attains  20  Ω  cm2,  in  which  case  4  cells  are  needed.  Again,  utilizing  5  cells  results  in  a   JOP  that  is  too  small  to  produce  a  10%  or  higher  SFE.     64     Figure  3.8  Impact  of  solution  resistance  and  EC  parameters  on  SFE  given  ηPV  =  20%.   Case  I  EC  parameters   (▬▬,  green)  are  based  on  utilizing  the  Co–OEC  and  Case  II  EC   (▬▬,  navy)  are  based  on  utilizing  the  Ni–OEC.       3.5  Model  validation     As  a  case  study,  we  used  this  steady–state  equivalent  circuit  analysis  to   analyze  the  OER  half–reaction  in  an  experiment  identical  to  those  discussed  in   Chapter  2.  In  this  experiment,  the  CoBi  catalyst  was  deposited  on  the  p–terminal  of  a   single–junction  c–Si  solar  cell.  Since  additional  voltage  (Vappl)  was  applied  to  assist  in   the  OER  reaction  equation  3.4  was  modified  as  follows:           where  Vth  is  now  the  Nerstian  potential  for  the  water–oxidation  reaction  which  is   0.68  V  vs.  NHE  at  pH  9.2.  This  is  represented  in  Fig.  3.9,  which  shows  the  J–V   !" +   !""# =   !! +   !"# !"   (3.9)     65   characteristics  of  the  PV  device  used  and  J–V  behavior  for  the  OER  reaction   (obtained  via  steady–state  Tafel  analysis  in  isolation  from  the  solar  cell)  which  is   shifted  to  lower  potentials  by  Vappl  until  the  two–curves  intersect,  resulting  in  the   predicted  behavior  of  the  PV–assisted  OER  reaction.     Figure   3.9   Graphical  demonstration  of  how  the  predictive  analysis  works  for   PV– assisted   reactions,   where   the   PV–curve   (▬▬,   blue)   is   based   on   the   J–V   characteristics  of  an  in–house   built   single  junction  c–Si   PV  and   the  EC–curve  (▬▬,   red)  is  based  on  the  CoBi  water–oxidation  catalyst  operating  in  pH  9.2  solution.         In  Fig.  3.10  the  predicted  behavior  of  the  coupled  PV–EC  system  is  compared   to  the  experimentally  measured  Tafel  analysis  of  a  PV–assisted  photoanode.  By   using  the  J–V  properties  of  the  solar  cell  depicted  in  Fig.  3.9,  and  the  independently   measured  Tafel  analysis  of  the  CoBi  catalyst,  the  coupled  behavior  is  determined;  the   measurement  and  prediction  agree  to  within  <  10mV.  Furthermore,  as  mentioned  in     66   Chapter  2,  the  voltage  offset  between  the  Tafel  slopes  for  the  OER–functionalized  c–   Si  photoanodes  under  the  light  configuration  and  the  configuration  in  which  the  PV   is  bypassed  is  exactly  the  VOC  of  the  solar  cell.  Since  Tafel  analysis  is  conducted  at   very  low  current–densities  of  1  mA  cm–2  or  lower,  this  requires  that  the  system   must  be  coupled  near  the  VOC  of  the  solar  cell.     Figure   3.10   Predicted   Tafel   behavior   of   a   PV–assisted   water   oxidation   system   similar   to   the   experiments   described   in  Chapter   2.  The  electrical   properties   of   the   PV   (shown   in   Fig.   3.8)   and   EC   systems   were   measured   independently   (● ,   black   dots)   and  used   to   predict   the  coupled  behavior  (▬▬,  black).  The  Tafel  analysis   of   the   PV–assisted   photoanode  (● ,  red  dots)  and  predicted  behavior  match  to  within   10  mV.       3.6  Conclusion   A  framework  is  outlines  for  integrating  a  series  of  single–junction  c–Si  solar     67   cells  with  recently  developed  water–splitting  catalysts  for  direct  solar–to–fuels   conversion.  The  steady–state  equivalent  circuit  analysis  gives  a  tool  that  allows  us  to   predict  the  efficiency  for  a  coupled  PV–EC  system.  This  analysis  allows  us  to   determine  the  optimal  number  of  solar  cells  required  for  a  stand–alone  water– splitting  device  based  on  catalyst  choice  and  considering  resistive  losses.  The  model   is  validated  by  correctly  predicting  the  J–V  characteristics  of  a  PV–assisted  OER   photoanode.  These  results  pave  a  path  to  our  goal  of  designing  a  stand–alone  water– splitting  device  based  on  all  terrestrially  ready  materials  exhibiting  a  solar–to–fuels   efficiency  of  10%  or  higher.     3.7  Experimental     Sample  Fabrication.  Crystalline  silicon  solar  cells  were  fabricated  according   to  previously  published  procedures.37    The  CoBi  OER–catalysts  was  deposited  via   bulk  electrolysis  in  a  two–compartment  electrochemical  cell  with  a  glass  frit  of  fine   porosity.  For  the  electrodeposition,  the  working  compartment  was  charged  with   ~50mL  solution  (25  mL  of  0.2M  KBi  electrolyte  and  25  mL  of  1mM  Co2+  solution).   The  auxiliary  compartment  was  charged  with  0.1  M  KBi  electrolyte  at  pH  9.2.  The   working  electrode  was  the  surface  passivated  c–Si  solar  cell.  Typically,  1  cm2  are  of   the  working  electrode  was  immersed  in  the  solution  and  electrolysis  was  carried  out   at  0.85  V  vs.  NHE  until  26  mC  cm–2  of  charge  passed.       Photoelectrochemistry  experiments.  Photoelectrochemistry  experiments   were  performed  in  a  one–compartment  quartz  cell.  The  light  source  was  a  Sol  2A   solar  simulator  (Newport).  The  Tafel  behavior  of  the  surface–passivated,  catalyst–   68   functiaonlized  solar  cells  was  measured  in  the  region  of  water–oxidation  over  a  200   mV  range  in  10  to  30  mV  increments.  The  measurements  were  conducted  in  a   solution  containg  0.5  M  KBi  and  1.5  M  KNO3  at  pH  9.2,  using  an  Ag/AgCl  reference   electrode  and  Pt  auxiliary  electrode.       69                                                                                                                     3.8  References     1.     Bolton   JR,   Hall   DO   (1979)   Photochemical   conversion   and   storage   of   solar   energy.  Annu.  Rev.  Energy  4,  353–401.     2.     Parkinson   B   (1984)   On   the   efficiency   and   stability   of   photoelectrochemical   devices.  Acc.  Chem.  Re.s  17,  431–437.     3.     Bolton   JR,   Strickler   SJ,   Connolly   JS   (1985)   Limiting   and   realizable   efficiencies   of  solar  photolysis  of  water.  Nature  316,495–500.     4.     Weber   MF,   Dignam   MJ   (1984)   Efficiency   of   splitting   water   with   semiconducting  photoelectrodes.  J.  Electrochem.  Soc.  131,1258–1265.     5.     Weber   MF,   Dignam   MJ   (1987)   Splitting   water   with   semiconducting   photoelecrodes–efficiency  considerations.  Int.  J.  Hydrog.  Energy  4,  225–232.     6.     Rocheleau   RE,   Miller   EL   (1997)   Photoelectrochemical   production   of   hydrogen:  Engineering  loss  analysis.  Int.  J.  Hydrog.  Energy  22,771–782.     7.     Surendranath   Y,   Bediako   DK,   Nocera   DG   (2012)   Interplay   of   oxygen– evolution   kinetics   and   photovoltaic   power   curves   on   the   construction   of   artificial  leaves.  Proc.  Natl.  Acad.  Sci.  U.S.A  39,  15617–15621.     8.   Kelly   NA,   Gibson   TL   (2006)   Design   and   characterization   of   a   robust   photoelectrochemical   device   to   generate   hydrogen   using   solar   water   splitting.  Int.  J.  Hydrog.  Energy  31,1658–1673.     9.     Walter   MG   et   al.   (2010)   Solar   water   splitting   cells.   Chem.   Rev.   110,   6446– 6473.     10.     Shockley   W,   Queisser   HJ   (1961)   Detailed   balance   limit   of   efficiency   of   p–n   junction  solar  cells.  J.  Appl.  Phys.  32,  510–519.     11.     Swanson   RM   (2005)   in   Conference   Record   of   the   Thirty–first   IEEE   Photovoltaic  Specialists  Conference,  2005,  pp  889–894.     12.     Green   MA   (2009)   The   path   to   25%   silicon   solar   cell   efficiency:   History   of   silicon  cell  evolution.  Prog.  Photovolt.  Res.  Appl.  17,183–189.     13.     Powell   DM   et   al.   (2012)   Crystalline   silicon   photovoltaics:   a   cost   analysis   framework   for   determining   technology   pathways   to   reach   baseload   electricity  costs.  Energy  Environ.  Sci.  5,  5874–5883.     70                                                                                                                                                                                                                                                                                                                                               14.     Rech   B,   Schmidt   SS,   Schlatmann   R   (2013)   Transition   to   Renewable   Energy   Systems,   eds   Stolten   D,   Scherer   Ing   Viktor   (Wiley–VCH   Verlag   GmbH   &   Co.   KGaA).       15.     Cukier   RI,   Nocera   DG   (1998)   Proton–coupled   electron   transfer.   Annu.   Rev.   Phys.  Chem.  49,  337–369.     16.     Eisenberg   R,   Gray   HB   (2008)   Preface   on   making   oxygen.   Inorg.   Chem.   47,1697–1699.     17.     Betley  TA  et  al.  (2008)  A  ligand  field  chemistry  of  oxygen  generation  by  the   oxygen–evolving   complex   and   synthetic   active   sites.   Philos.   Trans.   R.   Soc.   B   Biol.  Sci.  363,1293–1303.     18.     Hammes–Schiffer   S   (2009)   Theory   of   proton–coupled   electron   transfer   in   energy  conversion  processes.  Acc.  Chem.  Res.  42,  1881–1889.     19.     Concepcion   JJ   et   al.   (2009)   Making   oxygen   with   ruthenium   complexes.   Acc.   Chem.  Res.  42,  1954–1965.     20.     Surendranath   Y,   Nocera   DG   (2011)   Oxygen   evolution   reaction   chemistry   of   oxide–based   electrodes.   Prog.   Inorg.   Chem.   57   (John   Wiley   &   Sons,   INc.,   Hoboken,  NJ,  USA).     21.     Surendranath   Y,   Bediako   DK,   Nocera   DG   (2012)   Interplay   of   oxygen– evolution   kinetics   and   photovoltaic   power   curves   on   the   construction   of   artificial  leaves.  Proc.  Natl.  Acad.  Sci.  109,  15617–15621     22.      Surendranath  Y,  Kanan  MW,  Nocera  DG  (2010)  Mechanistic  Studies  of  the   oxygen  evolution  reaction  by  a  cobalt–phosphate  catalyst  at  neutral  pH.  J.  Am.   Chem.  Soc.  132,16501–16509.     23.    Bediako  DK,  Surendranath  Y,  Nocera  DG  (2013)  Mechanistic  studies  of  the   oxygen  evolution  reaction  mediated  by  a  nickel–borate  thin  film   electrocatalyst.  J.  Am.  Chem.  Soc.  135,  3662–3674.     24.     Hernández–Pagán   EA   et   al.   (2012)   Resistance   and   polarization   losses   in   aqueous   buffer–membrane   electrolytes   for   water–splitting   photoelectrochemical  cells.  Energy  Environ.  Sci.  5,  7582–7589.     25.     McCrory   CCL,   Jung   S,   Peters   JC,   Jaramillo   TF   (2013)   Benchmarking   heterogeneous   electrocatalysts   for   the   oxygen   evolution   reaction.   J.   Am.   Chem.  Soc.  135,  16977–16987.       71                                                                                                                                                                                                                                                                                                                                             26.     Nikolic   VM   et   al.   (2010)   Raising   efficiency   of   hydrogen   generation   from   alkaline  water  electrolysis  –  Energy  saving.  Int.  J.  Hydrog.  Energy  35,  12369– 12373.     27.     Zeng   K,   Zhang   D   (2010)   Recent   progress   in   alkaline   water   electrolysis   for   hydrogen   production   and   applications.   Prog.   Energy   Combust.   Sci.   36,   307– 326.     28.     Reece   SY   et   al.   (2011)   Wireless   Solar   Water   Splitting   Using   Silicon–based   semiconductors  and  earth–abundant  catalysts.  Science  334,  645–648.     29.     Winkler   MT,   Cox   CR,   Nocera   DG,   Buonassisi   T   (2013)   Modeling   integrated   photovoltaic–electrochemical   devices   using   steady–state   equivalent   circuits.   Proc.  Natl.  Acad.  Sci.  110,  E1076–E1082.     30.     Rocheleau   RE,   Miller   EL,   Misra   A   (1998)   High–efficiency   photoelectrochemical   hydrogen   production   using   multijunction   amorphous   silicon  photoelectrodes.  Energy  Fuels  12,  3–10.     31.     Nelson  J  (2003)  The  Physics  of  Solar  Cells  (Imperial  College  Press,  London).     32.     Kanan   MW,   Surendranath   Y,   Nocera   DG   (2009)   Cobalt–phosphate   oxygen– evolving  compound.  Chem.  Soc.  Rev.  38:109.     33.    Dincă  M,  Surendranath  Y,  Nocera  DG  (2010)  Nickel–borate  oxygen–evolving   catalyst   that   functions   under   benign   conditions.   Proc.   Natl.   Acad.   Sci.   107,   10337–10341.     34.     Huynh   M,   Bediako   DK,   Nocera   DG   (2014)   A   Functionally   stable   manganese   oxide  oxygen  evolution  catalyst  in  acid.  J.  Am.  Chem.  Soc.  136,  6002–6010.     35.     Huynh   M,   Bediako   DK,   Liu   Y,   Nocera   DG   (2014)   Nucleation   and   growth   mechanisms   of   an   electrodeposited   manganese   oxide   oxygen   evolution   catalyst.  J.  Phys.  Chem.  C.  Available  at:  http://dx.doi.org/10.1021/jp501768n   [Accessed  June  12,  2014].     36.    Steven   Y.   Reece,   Esswein   AJ,   Sung   K,   Green   Z,   Nocera   DG   (2013)   Compositions,   Electrodes,   Methods,   and   Systems   for   Water   Electrolysis   and   Other  Electrochemical  Techniques.  US  Patent  Appl.  No.  8361288.     37.   Cox  CR,  Winkler  MT,  Pijpers  JJH,  Buonassisi  T,  Nocera  DG  (2013)  Interfaces   between   water   splitting   catalysts   and   buried   silicon   junctions.   Energy   Environ.  Sci.  6,  532–538.     72   Chapter  4–10%  solar–to–fuel  efficiency  with  non– precious  materials         73   4.1  Introduction   In  order  to  make  a  solar–water–splitting  device  economically  viable  the   commonly  accepted  metric  is  that  a  10%  or  higher  solar–to–fuels  efficiency  (SFE)  is   required.1–3  Currently,  the  record–holding  SFE  for  solar  water–splitting  devices  is   between  16–18%.1,4  Both  of  these  were  based  on  expensive  group  III–V  multi– junction  solar  cells  and  precious  water–splitting  catalysts  such  as  Pt  and  Ru.  More   recently,  devices  constructed  utilizing  cheaper  PV  components  such  as  triple– junction  amorphous  silicon  (a–Si)  5–9and  copper  indium  gallium  diselenide  (CIGS)   have  also  been  demonstrated.10,11  Of  these  demonstrations,  few  used  earth– abundant  water–splitting  catalysts  and  many  operated  in  highly  acidic  or  basic   solutions,  which  impaired  long–term  stability.  Currently,  the  record  SFE  for  a  stand– alone  water  splitting  device  composed  of  all  earth  abundant  and  technology  ready   materials  operating  in  benign  solutions  is  4.7%.7,12  This  result  was  based  upon  “the   artificial  leaf,”  which  was  composed  of  a  triple–junction  a–Si  solar  cell,  a  CoBi   oxygen–evolution  catalyst  (OEC)  and  a  NiMoZn  alloy  hydrogen–evolution  catalyst   (HEC).  HOwever,  given  the  efficiency  limitations  of  triple–junction  a–Si  solar   cells13,14  a  10%  or  higher  SFE  cannot  be  achieved  until  the  PV  technology  has   improved.     In  Chapter  3,  equivalent  circuit  analysis  of  a  coupled  PV–EC  system  based  on   a  string  of  single  junction  c–Si  solar  cells  and  earth–abundant  catalysts  predicted  a   SFE  of  10%  could  readily  be  achieved.  Guided  by  steady–state  equivalent  circuit   analysis  described  in  Chapter  3,  here  we  demonstrate  that  an  SFE  >10%  can  be   achieved  with  all  non–precious,  low–cost,  commercially  ready  components  and     74   materials.  Specifically,  we  present  a  rational  systems  design  approach  to  evaluate   each  component  of  a  modular  PV–EC  device  comprised  of  a  c–Si  PV  mini–module   and  non–precious  catalysts  for  the  hydrogen–evolution  and  oxygen–evolution   reactions  (HER  and  OER,  respectively).  Although  this  approach  does  not  result  in  a   monolithic  structure  in  which  catalysts  at  directly  deposited  on  the  PV  device  (a.k.a.   an  artificial  leaf),  as  discussed  in  Chapter  3  the  equivalent  circuit  for  both  constructs   is  identical.15  This  approach  allows  for  modular  independent  optimization,  after   which  the  components  could  be  integrated  into  a  monolithic  design.     4.2  Results   Water–splitting  catalysts  can  be  integrated  with  a  c–Si  PV  module  either   directly  by  depositing  catalysts  onto  silicon  (Fig.  4.1)  or  indirectly  by  wiring  the  PV– module  to  electrodes  (Fig.4.2).  In  order  to  make  the  PV–EC  device  with  minimal   fabrication  and  adopt  a  completely  modular  approach  such  that  every  component   can  be  tested,  characterized,  and  replaced  systematically  we  chose  apply  the   indirect  method.  Such  a  configuration  also  allows  us  to  utilize  electrodes  that  are  not   restricted  to  the  PV  area.  However,  for  all  of  the  measurements  conducted  herein   the  electrode  area  was  kept  proportional  to  the  PV–module  area  such  that  all   photoelectrochemical  (PEC)  measurements  would  be  reflective  of  an  equivalent   monolithic  device.     Since  the  PV–EC  modular  configuration  allows  for  independent  optimization   of  the  PV–component  and  electrochemical  components  (electrodes  and  catalysts),   the  operating  point  for  the  coupled  PV–EC  device  can  be  illustrated  graphically  as     75   the  intersection  point  of  the  independently  measured  current–voltage  (J–V)  curves   for  the  PV  and  EC  for  water–splitting.     Figure   4.1   Schematic  of  a  PV–EC  device  based   on  series–connected  single–junction   c–Si  solar  cells  and  water–splitting  catalyst.  In  this  configuration  the  OER–catalyst  is   directly  deposited  on  the  back  of  the  last  solar  cell  in  the  stack.       76     Figure   4.2   Schematic   of   a   PV–EC   device   used   in   these   studies.   In   this   modular   configuration  each  component  can  be  easily  evaluated  and  replaced  independently.   The  point  of  intersection  give  the  operational  current  density,  JOP,  which  is  related  to   the  SFE  by  multiplying  by  the  thermodynamic  potential  for  water–splitting  and   Faradaic  efficiency,  ηFar:         SFE = 1.23  V ∙ !" ∙   !"#   !"# (mW  cm!! ) (4.1)   For  maximum  SFE,  the  intersection  of  the  PV  and  EC  J–V  curves  occurs  at  a   voltage  above  the  minimum  voltage  required  for  water–splitting  (i.e.   thermodynamic  potential,  plus  additional  kinetic  overpotentials  and  cell   resistances),  but  below  the  voltage  at  the  maximum  power–point  of  the  PV  module,     77   VMPP.  This  later  point  is  supported  by  Eq.  4.1,  whereby  the  efficiency  is  proportional   to  JOP  and  maximized  at  voltages  below  VMPP  as  shown  in  Fig.  4.3  which  shows  the  J– V  curve  for  the  mini–modules  used  for  the  PV–EC  device.  The  modules  were   constructed  from  either  3  or  4  commercial  single–junction  c–Si  solar  cells  connected   in  series.  When  connecting  the  cells  in  series,  the  J–V  properties  show  that  upon   addition  of  solar  cells  the  voltages  are  additive,  while  the  current  density  decreases   as  1/area  (see  Table  4.1  for  PV–module  characteristics).  It  should  be  noted  that  the   overall  PV  efficiency  is  maintained  upon  connecting  N  cells  in  series  because  the     Figure   4.3  J–V  curves  of  the  individually  measure  PV  and  EC  components  making  up   the   PV–EC   device.   The   grey   curves   represent   the   J–V   curves   for   the   PV   modules   composed   of   either   three   (––––,   grey–dashed)   or   four   (▬▬,   grey–solid)   single– junction   c–Si   solar   cells   measure   under   AM   1.5   illumination.   The   red   curves   represent   electrochemical   load   J–V  curves   using   NiBi   and   NiMoZn   catalysts,   where   the   ideal   EC   curve   (––––,   red–dashed)   is   based   on   previously   reported   Tafel   analysis  and  the  actual  EC  curve  (▬▬,  red)  measured  in  a  2–electrode  experiment   (0.5M   KBi   /   0.5M   K2SO4,   pH   9.2).   The   point   of   intersection   represents   the   JOP   (● ,   orange  circles)  and  the  SFE  of  the  coupled  system.     78   module  voltage  will  be  xN  larger  than  the  indivdual  cell,  and  the  current  density  will   be  x1/N  that  of  an  indivdual  cell.     Table  4.1.  PV  characteristics  for  the  3  and  4–cell  c–Si  mini–modules.   PV  module  characteristics   VOC  (V)   JSC  (mA  cm–2)   Active  Area  (cm2)   Fill  Factor   Efficiency  (%)     The  choice  of  water–splitting  catalysts  was  based  upon  independent  studies   of  the  catalysts  developed  in  our  group.  Presently,  the  NiBi  OER  catalyst  is  the  most   active  OEC,  requiring  only  430  mV  of  overpotential  to  achieve  a  current  density  of   10mA  cm–2,  making  it  an  order  of  magnitude  better  than  the  previously  developed   Co–OEC’s  given  the  same  amount  of  material.16  For  the  HER–catalyst  a  NiMoZn  alloy   was  used  which  has  been  previously  shown  to  achieve  current  densities  of  700  mA   cm–2  at  100  mV  overpotential  and,  with  continued  leaching  in  6M  KOH  can  attain   activities  as  high  as  at  1000  mA  cm–2  at  an  overpotential  of  35  mV.7,12,17  Given  our   modular  approach  the  current–voltage  characteristics  of  out  EC–component  can  be   independently  evaluated.     Considering  a  PV–EC  device  based  on  commercially  available  single  junction– Si  solar  cells  and  literature  values  for  the  previously  reported  Tafel  behavior  of  the     79   4–  cell   2.46   51.0   6.0   76.9   16.0   3–cell   1.79   52.2   4.5   76.2   15.8   catalysts  utilized  herein,  equivalent–circuit  modeling  predicts  a  10%  or  higher  SFE   can  be  achieved  using  three  single–junction  c–Si  devices  series  connected  in  a  mini– module  with  a  PV  efficiency  of  15%  or  higher  (red  dashed  curve  Fig.  4.3).  However,   this  is  only  the  case  if  all  resistive  losses  are  negligible;  if  resistive  losses  are   present,  the  operating  point  can  occur  to  the  right  of  VMPP,  reducing  JOP  and  SFE.   Modeling  indicates  that  using  a  4–cell  c–Si  module  overcomes  the  impact  of  resistive   losses  on  SFE.15   To  test  these  predictions,  the  steady–state  current–voltage  characteristics  of   the  NiBi  anode  and  NiMoZn  cathode  were  measured  in  a  two–electrode  setup  in  KBi   buffer  at  pH  9.2.  The  intersection  at  which  the  overlaid  current–voltage   characteristics  of  the  half–reactions  with  the  J–V  curve  of  the  PV  mini–modules   illustrates  JOP  and  the  resulting  SFE  for  the  coupled  PV–EC  device.  Confirming  the   design  considerations  for  resistive  losses,  we  estimate  a  SFE  of  2.8%  for  a  3–cell   module  and  10%  for  a  4–cell  module  (Fig.4.3).       4.2.1  Device  integration     The  simplest  way  to  integrate  the  PV  and  EC  components  and  verify  the   independently  estimated  SFE  is  to  connect  the  PV  module  with  the  NiBi  anode  and   NiMoZn  cathode.  The  photocurrent  through  the  integrated  device  can  be  measured   and  should  match  the  predicted  JOP  obtained  in  Fig.  4.3.  The  key  criteria  used  to   validate  the  PV–EC  device  are  the  reporting  protocols  established  by  Chen  et  al.18   These  protocols  include  measurements  utilizing  a  2–electrode  setup  without  the   influence  of  an  potential  bias,  product  quantification  (i.e.  H2  and  O2),  and  assessment     80   of  the  long–term  stability  of  the  device  under  AM  1.5  illumination.  In  addition  to   product  quantification,  we  wished  to  ensure  that  parasitic  currents  due  to  product   crossover  reactions  do  not  influence  JOP.  The  impact  on  the  SFE  due  to  H2  oxidation   can  be  estimated  by  examining  the  mass–transport  limited  current  density,  which  is   given  by:           where,  n  is  the  number  of  electrons,  F  is  Faraday’s  constant,  cb  is  the  bulk   concentration  of  species  in  solution,  D  is  the  diffusion  coefficient  (5.11  ×  10–5  cm2  s–1   for  H2  in  water),  and  δ  is  the  Nernst  diffusion  layer  thickness.    Assuming  H2   saturation  in  water  cb  =  7.8  ×  10–7  mol  cm–3,  and  a  reasonable  value  for  δ  (given  a   planar  electrode  with  no  artificially  imposed  convection)  is  around  0.05  cm19  this   estimates  that  a  current  density  of  8.13  mA  cm–2  (10%  SFE)  would  have  a  parasitic   current  of  0.15  mA  cm–2  (reducing  10.0%  SFE  to  9.8%  SFE).  However,  the  NiBi  is  a   specific  OER  catalyst  and  Fig.  4.4  shows  that  the  steady–state  current  density  of  the   catalyst  under  Ar  and  H2  is  identical,  indicating  that  this  crossover  reaction  is   negligible.     ! = !   (4.2)     81     Figure  4.4.  Steady-­‐state   current  voltage   behavior  for  the   NiBi   operating   in  0.5M  KBi   /   0.5   M   K2SO4   pH   9.2   in   H2   saturated   solution   (● )   and   in   Ar   saturated   solution   (▲).   Since   the  voltage  required  to  achieve   a   given  current   density   under  both   conditions   is   almost   identical   indicates   that   the   contribution   of   H2   oxidation   at   the   anode   is   negligible.       Fig.  4.5  shows  the  measured  JOP  of  the  PV–EC  device,  which  initially  starts  at   8.35  mA  cm–2  corresponding  to  an  SFE  of  10.2%.  During  the  first  few  min  of   illumination  JOP  decreases  to  a  steady–state  value  of  7.8  mA  cm–2.  The  initial  decline   in  JOP  is  consistent  with  heating  of  the  PV–module  under  illumination  causing  a   decrease  in  solar  cell  voltage,  which  shifts  the  maximum  power  point  toward  the   origin.     82     Figure   4.5   Current   under   chopped   illumination   representing   JOP   for   the   PV–EC   device   in   0.5M   KBi   /   0.5M   K2SO4   pH9.2.   The   chopped   illumination   illustrates   the   recovery  in  SFE  and  reproducibility  in  measuring  JOP  through  the  PV-­‐EC  device     This  is  confirmed  by  measuring  the  VOC  of  the  mini–module  as  a  function  of  time   showing  ~130  mV  decrease  which  is  consistent  in  a  temperature  change  of  15  °C     (Fig.  4.6).20  In  line  with  PV  module  heating,  turning  the  lamp  off  for  5  min  and  then   turning  it  back  on  causes  the  SFE  to  recover  to  10.2%  (Fig.  4.6).       83     Figure  4.6   Decay   of   the   open–circuit   voltage   of   the   4–cell   PV   mini–module   over   the   course  of  ~15  min.  The  initial  VOC  at  2.42  V  decays  to  a  steady–state  of  2.27  V  after   the   first   10   min  (▬▬,   orange),  which  contributes  to   the   initial   decline   in   the   SFE   of   the   coupled   PV–EC   device.   After   overnight   illumination,   the   Voc   was   measured   (▬▬,  blue)  and  shows  a  slight  recovery  to  2.31  V,  which  corresponds  to  the  initial   increase  in  SFE  of  the  PV–EC  device  during  the  first  24  h.     4.3  Discussion     The  largest  efficiency  losses  for  the  PV–EC  device  result  from  series   resistance  through  the  electrodes  (REL)  and  solution  resistance  (RSOL).  The  former  is   straightforward  to  address  by  using  metal  electrodes  as  substrates  for  the  OER  and   HER  catalysts.  Presumably  the  use  of  metallic  substrates  makes  resistance  through   the  electrodes  as  well  as  contact  resistance  negligible.  Solution  resistance  in   buffered  electrolytes,  as  opposed  to  strong  acids  or  bases,  remains  a  challenge.  The   primary  reason  for  a  less  than  optimal  RSOL  is  the  limited  solubility  of  the  buffer.21  In   the  case  of  borate  buffer  this  is  the  solubility  limit  of  boric  acid,  which  is  around  1  M     84   corresponding  to  a  specific  conductance  of  26  mS  cm–1.  The  specific  conductivity  can   be  improved  by  adding  an  inert  salt  as  a  supporting  electrolyte  (Fig.  4.7).  For   example,  when  utilizing  KNO3,  the  specific  conductance  of  0.5  M  KBi  /  1.5  M  KNO3  is   126  mS  cm–1.     Figure   4.7  Specific  conductance  measurements  for  various  electrolytes  considered   to  minimize  RSOL.  KOH  (∎,  red  squares)  is  the  most  conductive  electrolyte;  in  order   to  operate  in  pH  near  neutral  regimes  0.5M  KBi  was  used  with  additional  supporting   electrolyte,  such  as  KNO3  (●,  green  circles)  or  K2SO4  (●,  black  circles).       The   choice   or   supporting   electrolyte   is   straightforward   in   typical   electrochemical   experiments   where   only   one   half–reaction   at   either   the   anode   or   cathode   is   of   interest.   When   considering   deleterious   side–reactions   for   both   the   anode  and  cathode,  the  supporting  electrolyte  must  be  inert  over  a  wider  potential   range.   Given   our   modular   approach,   the   choice   of   supporting   electrolyte   was   determined   by   measuring   the   Faradaic   efficiency   for   each   electrode/electrolyte     85   configuration   independently   before   being   implemented   into   the   PV–EC   device.   For   example,   Fig.   4.8   shows   when   operating   the   NiMoZn   cathode   for   HER   at   current   densities  of  10  mA  cm–2  analysis  of  hydrogen  via  gas  chromatography  (GC)  analysis   showed   no   hydrogen   production   indicating   that   NO3–   is   preferentially   reduced   as   opposed  to  protons.       Figure   4.8   Gas   quantification   for   NiMoZn   cathode   operating   in   (left)   0.5   M   KBi   /   K2SO4   and  (right)   0.5   M  KBi  /  KNO3  both  at  pH  9.2.  The   black   line  represents   100%   Faradaic   efficiency   based   on   the  charge   passes   during   electrolysis.   The  green   circles   represent   H2   measured   by   gas   chromatography.   The   red   arrow   indicates   when   electrolysis   was   stopped.   GC   analysis   was   conducted   until   the   moles   of   gas   measured   in  the   headspace   reached  a  steady–state.  The  lag  period  (▬▬,  black)   in   gas  generation  is  due  to  the  buildup  of  gases  in  the  headspace  of  the  EC  cell.     Alternatively,  using  K2SO4  as  a  supporting  electrolyte  results  in  a  Faradaic   efficiency  of  100%.  Fig  4.9  shows  that  the  NiBi  anode  operating  in  0.5  M  KBi  /  K2SO4     86   (pH  9.2)  also  demonstrates  a  100%  Faradaic  efficiency.     Figure   4.9  Gas  quantification  for  NiBi  cathode  operating  in  (left)  0.5  M  KBi  /  K2SO4   and   at   pH   9.2.   The   black   line   represents   100%   Faradaic   efficiency   based   on   the   charge   passes   during   electrolysis.   The   green  circles   represent  O2   measured   by   gas   chromatography.   The   red   arrow   indicates   when   electrolysis   was   stopped.   GC   analysis   was   conducted   until   the   moles   of   gas   measured   in   the   headspace   reached   a   steady–state.  The  lag  period  (▬▬,  black)  in  gas   generation  is  due  to  the  buildup  of   gases  in  the  headspace  of  the  EC  cell.       Although  the  following  gas  quantification  measurements  indicate  that  0.5  M   KBi  /  0.5  M  K2SO4  solution  is  a  reasonable  choice,  K2SO4  is  sparingly  soluble  at  0.5  M,   once  again  limiting  the  specific  conductivity  of  our  electrolyte  to  90  mS  cm–1  (Fig.   4.7).     87     Figure   4.10   Current   under   chopped   illumination   representing   JOP   for   a   PV–EC   device   composed   of   a   3–cell   PV–module,   a   NiBi   anode,   and   NiMoZn   cathode   operating   in   1M   KOH.   Because   KOH   is   a   more   conductive   electrolyte,   a   12%   or   greater   SFE  can   be   obtain   with  a  3–cell  PV  module  as  opposed  to  a  4–cell  module.   The  initial  drop  in  SFE  is  due  to  the  decrease  in  PV  efficiency,  due  to  heating  of  the   PV–module.  The  chopped  illumination  represents  the  recovery  in  SFE.       By  moving  to  a  more  conductive  electrolyte,  such  as  1  M  KOH  (pH  14),  a  12%   SFE  can  be  obtained  with  a  3–cell  mini–module  as  opposed  to  a  4–cell  module  (Fig.   4.10).  This  also  shows  how  minimizing  RSOL  shifts  the  EC  curve  closer  the  ideal  curve   obtained  based  on  the  Tafel  analysis  of  the  catalysts  used  herein  (Fig.  4.11).   However,  it  is  preferential  to  avoid  the  deleterious  effect  of  concentrated  base  on  PV   materials  by  maintaining  neutral  and  near–neutral  conditions.  We  thus  prefer  to   minimize  RSOL  by  utilizing  a  flow–cell  design  and  optimized  cell  geometry.22–24     88     Figure   4.11  J–V  curves  of  the  individually  measure  PV  and  EC  components  making   up  the  PV–EC  device  operating  in  1M  KOH.  The  grey  curves  represent  the  J–V  curves   for  the   PV   modules  composed  of  either  three  (–––––,  grey–dashed)  or  four  (▬▬,   grey–solid)  single–junction  c–Si  solar  cells  measure   under  AM  1.5  illumination.  The   blue   curves   represent   electrochemical   load   J–V   curves   using   NiBi   and   NiMoZn   catalysts,   where   the   ideal   EC   curve   (–––––,   blue–dashed)   is   based   on   previously   reported  Tafel   analysis   and  the   actual  EC   curve  (▬▬,  blue–solid)  measured  in  a  2– electrode  experiment.  The  point  of  intersection  represents  the  JOP  (●,  orange  circles)   and  the  SFE  of  the  coupled  system.       The  operational  stability  of  the  coupled  PV–EC  system  showed  no  decline  in   JOP  for  over  a  week  of  operation  in  0.5  M  KBi  pH  9.2  solution  (Fig.  4.12).   Interestingly,  the  SFE,  inferred  from  the  current,  appears  to  slightly  increase  during   the  first  24  h  of  operation.  This  small  recovery  is  attributed  to  a  recovery  cell   voltage  over  the  course  of  24  h  of  illumination  (blue  line  in  Fig.  4.6).  Initially,  the   module  absorbs  heat  from  the  solar  simulator  photon  flux,  causing  the  initial   decrease  in  PV  efficiency.20  Then,  under  constant  illumination  at  higher   temperatures,  the  observation  of  a  gradual  improvement  in  the  current  density  over     89   a  timescale  of  tens  of  hours  is  consistent  with  the  evolution  of  the  “oxygen–boron   defect”  a  well–  studied  phenomenon  in  p–type  Czochralski  silicon.25,26  Importantly,   the  observed  fluctuations  in  JOP  can  be  attributed  to  fluctuations  in  the  PV  module   output  and  are  not  related  to  the  PV–EC  coupling  or  EC  reactions.   Figure  4.12  SFE  inferred   from  JOP  for  the  PV–EC  device  operating  in   0.5M  KBi  /   0.5M   K2SO4   pH   9.2   measured   for   over   7   days   of   operation   showing   no   decrease   in   SFE   over   operation   time.   Spikes   are   due   to   the   addition   of   solution   to   maintain   the   solution  level  and  pH.       4.4  Conclusion   We  demonstrate  that  an  SFE  efficiency  of  10%  can  be  achieved  utilizing  non– precious  materials  and  c–Si.  This  proof  of  concept  capitalizes  on  the  declining  cost  of   high–quality  PV  devices  and  earth–abundant  catalysts  operating  under  near  neutral   pH  conditions.  This  modular  design  of  the  PV  and  EC  components  allows  for  a  wide     90   variety  or  PV  opposed  materials,  catalysts,  and  electrolytes  to  be  implemented   where  no  one  component  is  constrained  by  the  other.  This  methodology  permits   facile  optimization  and  characterization.  As  PV–EC  device  sub–components  reach   technological  maturity,  an  increasing  emphasis  will  be  placed  on  system  design  and   integration.     4.5  Experimental   Materials  and  Methods.  Nickel  (II)  chloride  hexahydrate,  boric  acid,   potassium  hydroxide,  potassium  nitrate,  potassium  sulfate  were  purchased  from   Sigma  Aldrich  and  used  as  received.  Steel  foil  and  nickel  mesh  were  purchased  from   Strem.     Mini–Module  Fabrication  Crystalline  silicon  mini–modules  were  fabricated   using  commercially  available  single–junction  Czochralski  silicon  solar  cells  with   stand–alone  efficiencies  of  18%.  Mini–cells  were  cut  out  of  commercial  size  wafers   by  laser  scribing  with  a  1064  nm  pulsed  laser  and  mechanical  cleaving.  Mini–cells   were  electrically  connected  via  solar  tabbing  wire  and  silver  epoxy.  The  mini– module  was  constructed  by  sequentially  layering  glass,  EVA,  solar  cells,  and  EVA.   The  glass  maintains  structural  integrity  while  EVA  provides  water  protection.  The   mini–module  was  encapsulated  using  a  double  layer  vacuum  press  heated  to  120  °C.   With  the  module  in  the  lower  chamber  of  the  vacuum  press,  both  the  upper  and   lower  chambers  were  held  under  vacuum  for  5  min.  The  upper  chamber  was  vented   to  atmospheric  pressure  for  5  min  to  remove  air  bubbles  through  the  induced   pressure  difference  between  the  two  chambers.  The  lower  chamber  was  vented  and     91   the  mini–module  was  allowed  to  cool  to  room  temperature.  Excess  EVA  was   removed  and  the  mini–module  was  stored  to  protect  against  mechanical  and  water   degradation.   After  laser–cutting  the  commercial  cells,  connecting  four  in  series,  and   encapsulating  them  with  ethylene  vinyl  acetate,  the  mini–module  efficiency  was   16%.  The  equivalent  mini–module  for  a  three–cell  series  is  15.8%.     Electrochemical  methods.  Electrochemical  experiments  were  performed   using  a  CH–Instruments  760D  potentiostat.  For  three–electrode  measurements   potentials  were  measured  against  an  Ag/AgCl  reference  electrode  (BASi)  and   converted  to  NHE  by  adding  0.197  V.  For  two  electrode  experiments  the  working   electrode  lead  of  the  potentiostat  was  connected  to  the  anode  and  the  reference  and   auxiliary  leads  of  the  potentiostat  were  connected  to  the  cathode.     Catalyst  formation.  The  NiBi  anode  was  electrodeposited  in  a  two– compartment  electrochemical  cell  with  a  glass  frit  junction.  The  working   compartment  was  charged  with  ~25  mL  of  0.2  M  Bi  electrolyte  and  25  mL  of  a  1  mM   Ni2+  solution.  The  working  electrode  was  a  steel  substrate,  and  the  NiBi  catalyst  was   deposited  by  applying  a  voltage  of  0.95  V  (vs.  Ag/AgCl)  for  1  h.  To  improve  anode   activity  the  electrodes  were  then  anodized  at  0.9  V  (vs.  Ag/AgCl)  in  1  M  KOH  for  1  h     The  NiMoZn  cathode  was  electrodeposited  from  a  solution  of  nickel(II)   chloride  hexahydrate  (9.51  g  L–1),  sodium  molybdate  dihydrate  (4.84  g  L–1),   anhydrous  zinc  chloride  (0.0409  g  L–1),  tetrabasic  sodium  pyrophosphate  (34.57  g   L–1)  and  sodium  bicarbonate  (74.77  g  L–1;  VWR).  Hydrazine  hydrate  (1.21  mL  L–1)   was  added  immediately  before  plating.  NiMoZn  was  deposited  onto  a  Ni  mesh     92   substrate  that  had  been  pre–treated  at  2  V  vs.  Ag/AgCl  in  0.5  M  H2SO4  for  3  min.  The   NiMoZn  alloy  was  deposited  at  a  voltage  of  1.8  V  (vs.  Ag/AgCl)  for  30  min.  The   deposit  was  left  to  de–alloy  overnight  in  6  M  KOH.7,17   NiBi  product  crossover.  H2  oxidation  at  the  NiBi  anode  was  examined  via   the  steady–state  activity  of  the  anode  in  Ar  saturated  and  H2  saturated  0.5  M  KBi  /   K2SO4  solution  at  pH  9.2.  The  amount  of  voltage  required  to  achieve  a  given  steady– sate  current–density  in  both  cases  was  the  same  indicating  that  the  contribution  JOP   as  a  consequence  of  H2  oxidation  is  negligible.     Photoelectrochemical  measurements.  The  NiBi  anode  and  NiMoZn   cathode  were  connected  in  series  with  the  c–Si  mini–module.  The  light  source  was  a   Sol  2A  solar  simulator  (Newport  Corp.).  The  current  through  the  PV–EC  device  was   measured  by  using  the  potentiostat  as  an  ammeter.  In  all  cases,  the  area  used  to   convert  current  to  current–density  was  the  active–area  of  the  c–Si  mini–module.   Additionally,  the  geometric  area  of  the  anode  and  cathode  was  scaled  to  match  that   of  the  mini–module.  For  long–term  stability  measurements,  fresh  KBi  buffer  solution   was  added  in  order  to  maintain  the  solution  lost  to  evaporation  as  well  as  pH.     Gas  quantification.  The  Faradaic  efficiency  for  each  electrode  was  evaluated   using  gas  chromatography.  The  experiment  was  performed  galvanostatically  using  a   three–electrode  configuration  in  a  custom  built  two–compartment  gas–tight   electrochemical  cell.  The  working  electrode  was  either  NiBi  on  a  steel  substrate  for   O2  quantification,  or  NiMoZn  on  a  nickel  mesh  substrate  for  H2  quantification.  The   working  electrode  operated  at  a  constant  current  density  of  10  mA  cm–2  for  2  h.   During  the  course  of  the  experiment,  samples  of  evolved  gas  were  removed  from  the     93   headspace  and  injected  into  the  GC.  In  order  to  ensure  that  the  evolved  gas  reached   a  steady  state  value  in  the  headspace,  GC  measurements  were  recorded  for  1  h  after   cessation  of  electrolysis.  The  data  was  converted  into  partial  pressure  of  gas  in  the   headspace  using  calibration  curves  defined  from  known  mixtures  of  H2/N2  or  O2/N2.   The  partial  pressure  of  gas  was  converted  to  µmol,  and  corrected  using  Henry’s  law   to  account  for  the  gas  dissolved  in  solution.  The  total  charge  passed  during   electrolysis  was  divided  by  nF  (n  corresponding  to  the  number  of  electrons  in  each   half  reaction)  to  furnish  the  calculated  gas  yield.  The  total  calculated  and   experimental  gas  yields  were  used  to  determine  the  Faradaic  efficiency.       94                                                                                                                     4.6 References 1. Khaselev  O,  Bansal  A,  Turner  JA  (2001)  High–efficiency  integrated   multijunction  photovoltaic/electrolysis  systems  for  hydrogen  production.  Int.  J.   Hydrog.  Energy  26,  127–132.   Office of Energy Efficiency and Renewable Energy (2014) Hydrogen,  fuel  cells,  &   infrastructure  technologies  program  (U.S.  Department  of  Energy,  Washington   DC,  USA)     Grätzel  M  (2007)  Photovoltaic  and  photoelectrochemical  conversion  of  solar   enegy.  Phil.  Trans.  R.  Soc.  A.  365,  993–1005. Licht  S  et  al.  (2000)  Efficient  solar  water  splitting,  exemplified  by  RuO2– catalyzed  AlGaAs/Si  Photoelectrolysis.  J.  Phys.  Chem.  B  104,  8920–8924.   Miller  EL,  Rocheleau  RE,  Deng  XM  (2003)  Design  considerations  for  a  hybrid   amorphous  silicon/photoelectrochemical  multijunction  cell  for  hydrogen   production.  Int.  J.  Hydrog.  Energy  28,  615–623.   Kelly  NA,  Gibson  TL  (2006)  Design  and  characterization  of  a  robus   tphotoelectrochemical  device  to  generate  hydrogen  using  solar  water  splitting.   31,  1658–1673.   Reece  SY  et  al.  (2011)  Wireless  solar  water  splitting  using  silicon–based   semiconductors  and  earth–abundant  catalysts.  Science  334,  645–648.   Cristino  V  et  al.  (2013)  Efficient  solar  water  oxidation  using  photovoltaic   devices  functionalized  with  earth–abundant  oxygen  evolving  catalysts.  Phys.   Chem.  Chem.  Phys.  15,  13083–13092.   Jin  J,  Walczak  K,  Singh  MR,  Karp  C,  Lewis  NS,  Xiang  C  (2014)  An  experimental   and  modeling/simulation  and  operational  performance,  characteristics  of  and   integrated,   membrane–free,   neutral   pH   solar–driven   water–splitting   system.   Energy  Environ.  Sci.  doi:  10.1039/C4EE01824A 2. 3. 4.     5. 6.     7. 8. 9. 10. Jacobsson   TJ,   Platzer–Björkman   C,   Edoff   M,   Edvinsson   T   (2013)   CuInxGa1−xSe2  as  an  efficient  photocathode  for  solar  hydrogen  generation.  Int   J  Hydrog  Energy  38,15027–15035.   11. Jacobsson   TJ,   Fjällström   V,   Sahlberg   M,   Edoff   M,   Edvinsson   T   (2013)   A   monolithic   device   for   solar   water   splitting   based   on   series   interconnected   thin   film   absorbers   reaching   over   10%   solar–to–hydrogen   efficiency.   Energy   Environ.  Sci.  6,  3676–3683.     95                                                                                                                                                                                                                                                                                                                                             12. Nocera  DG  (2012)  The  artificial  leaf.  Acc.  Chem.  Res.  45,  767–776.   13. Green  MA,  Emery  K,  Hishikawa  Y,  Warta  W,  Dunlop  ED  (2012)  Solar  cell   efficiency  tables  (version  39).  Prog  Photovolt  Res  Appl  20,  12–20.   14. Bagnall  DM,  Boreland  M  (2008)  Photovoltaic  technologies.  Energy  Policy  36,   4390–4396.   15. Winkler  MT,  Cox  CR,  Nocera  DG,  Buonassisi  T  (2013)  Modeling  integrated   photovoltaic–electrochemical  devices  using  steady–state  equivalent  circuits.   Proc.  Natl.  Acad.  Sci.  110,  E1076–E1082.   16. Bediako  DK,  Surendranath  Y,  Nocera  DG  (2013)  Mechanistic  studies  of  the   oxygen  evolution  reaction  mediated  by  a  nickel–borate  thin  film  electrocatalyst.   J.  Am.  Chem.  Soc.  135,  3662–3674.   17. Reece  SY,  Esswein  AJ,  Sung  K,  Green  Z,  Nocera  DG  (2013)  Compositions,   electrodes,  methods,  and  mystems  for  water  electrolysis  and  other   electrochemical  techniques.  US Patent Appl. No. 8361288.   Chen   Z   et   al.   (2010)   Accelerating   Materials   Development   for   Photoelectrochemical   Hydrogen   Production:   Standards   for   Methods,   Definitions,  and  Reporting  Protocols.  J.  Mater.  Res.  25,  3–16.   18. 19. Bockris   JO,   Reddy   AKN,   Gamboa–Aldeco   ME   (1998)   Modern   electrochemistry   (New  York:  Plenum  Press,  c1998–c2000). 20. Jones   AD,   Underwood   CP   (2001)   A   thermal   model   for   photovoltaic   systems.   Sol  Energy  70:349–359.   21. Hernández–Pagán   EA   et   al.   (2012)   Resistance   and   polarization   losses   in   aqueous   buffer–membrane   electrolytes   for   water–splitting   photoelectrochemical  cells.  Energy  Environ.  Sci.  5,  7582–7589.   22. Modestino   MA   et   al.   (2013)   Robust   production   of   purified   H2   in   a   stable,   self– regulating,  and  continuously  operating  solar  fuel  generator.  Energy  Environ.  Sci.   7,  297–301.   Haussener   S   et   al.   (2012)   Modeling,   simulation,   and   design   criteria   for   photoelectrochemical   water–splitting   systems.   Energy   Environ.   Sci.   5,   9922– 9935.   23. 24.     Newman   J   (2013)   Scaling   with   Ohm’s   Law;   wired   vs.   wireless   photoelectrochemical  cells.  J.  Electrochem.  Soc.  160,  F309–F311.     96                                                                                                                                                                                                                                                                                                                                             25. Schmidt   J,   Bothe   K   (2004)   Structure   and   transformation   of   the   metastable   boron–  and  oxygen–related  defect  center  in  crystalline  silicon.  Phys.  Rev.  B  69,   024107.   26. Bianca   Lim   KB   (2008)   Deactivation   of   the   boron–oxygen   recombination   center   in   silicon   by   illumination   at   elevated   temperature.   Phys.   Status   Solidi   RRL   –   Rapid  Res.  Lett.  2,  93–95.   97   Chapter  5–Future  Directions         98   5.1  Introduction   The  previous  chapters  present  a  completely  modular  proof–of–concept   approach  for  direct  solar–to–fuels  conversion  using  all  earth–abundant  and   technology  ready  materials.  Stand–alone  photovoltaic–electrochemical  (PV–EC)   devices  have  been  created  with  few  design  constraints  on  each  component.   Continued  advances  may  be  made  with  consideration  of  alternative  materials  and   concepts  for  various  components  of  a  PV–EC  design.  Given  the  versatility  of  utilizing   buried–junction  photovoltaics,  modifications  to  the  PV–EC  design  are   straightforward.  The  following  chapter  discusses  alternative  concepts  and   preliminary  results  for  each  component  of  the  PV–EC  device,  which  may  allow  for   further  improvements  in  efficiency,  cost  reduction,  and  design  integration.       5.2  Alternative  PV  materials   Although  crystalline  silicon  (c–Si)  is  a  high  quality  material  and  is  an   economically  viable  PV  resource,  the  fact  that  the  solar–to–electrical  power   efficiency  is  reaching  its  thermodynamic  limit  sets  a  ceiling  on  drastic   improvements  in  SFE.1,2  Additionally,  a  large  portion  of  the  current  price  for  c–Si  PV   modules  is  encumbered  by  balance  of  systems  costs  (BOS)  as  opposed  to  the  price  of   silicon  itself,  and  it  remains  unclear  if  these  BOS  costs  can  be  significantly  reduced.3,4   Very  recently,  thin–film  PV’s  consisting  of  a  perovskite  absorber,  typically   CH3NH3PbX3  (X  =  Br,  I),  sandwiched  between  a  TiO2  electron  conducting  layer  and   an  organic  hole  transporting  material  (typically  spiro–OMeTAD),  have  emerged  as  a   formidable  alternative  to  c–Si  PV’s.  Perovskite  PV’s  were  first  introduced  in  2009.5     99   Since  that  time,  laboratory  scale  cells  have  shown  a  rapid  improvement  in  solar–to– electrical  power  efficiency  from  10%  in  2012  to  a  record  19.3%%;6,7  a  band–gap  of   ~1.5  eV  with  a  predicted  practical  efficiency  of  20%  speaks  to  their  promise  as  a   future  PV  material.8  Moreover,  perovskite  PVs  can  be  constructed  from  various  low– cost  liquid  phase  chemical  reactions  and  deposition  methods  such  as  spin–coating   or  spray–pyrolysis.9  Although  yet  to  be  implemented  on  commercial  scale,  the   estimated  cost  for  a  perovskite  solar  cell  could  be  as  low  as  $0.30  WP–1.10   Despite  the  swift  emergence  of  high  efficiency  perovskite  PVs,  current   drawbacks  include  scalability,  stability,  and  environmental  safety.    Currently,  most   of  the  high  efficiency  perovskite  PV’s  have  only  been  demonstrated  to  perform  on   small  scales  of  0.1  cm2  or  less.11  Larger  scales  devices  result  in  lower  fill  factors,   decreasing  the  PV  efficiency.  The  reason  for  the  loss  in  fil–factors  is  unclear,  but  it  is   proposed  that  it  due  to  series  resistance  through  the  device.12  Additionally  since   perovskites  are  water–soluble,  there  is  concern  about  the  long–term  operation  time   and  the  possibility  of  Pb  leaking  into  the  environment.  Substitution  of  Sn2+  for  Pb2+   is  promising  from  the  viewpoint  of  toxicity,  but  Sn–based  perovskites  so  far  have   only  been  stable  in  a  nitrogen  environment  and  have  yet  to  reach  efficiencies  close   to  the  Pb–based  devices.13,14   For  solar–water–splitting  applications,  perovskites  should  be  investigated.  In   order  to  use  perovskite  PVs  either  a  series–connected  approach  can  be  adopted  or   alternatively  perovskites  could  be  used  the  top  cell  in  a  tandem  configuration.  For   example  a  perovskite–silicon  tandem  PV  constructed  from  a  17%  efficient   perovskite  PV  and  a  23–24%  c–Si  solar  cell  could  produce  a  29.6%  PV  efficiency.15     100   These  properties  suggest  that  these  materials  combined  in  a  buried  junction   approach  may  be  a  very  promising  future  line  of  inquiry.     5.3  Alternative  catalyst  deposition  methods   So  far  the  research  our  group  has  focused  on  solution–based   electrodeposition  of  Co,  Ni,  and  Mn  oxygen  evolution  catalysts  (OECs)  and  NiMoZn   alloys  for  hydrogen  evolution  catalyst  (HECs).  Years  of  research  has  given  insight  on   film  formation  and  optimization  of  catalyst  activity.16–20  However,  due  to  the  highly   oxidizing  or  reducing  conditions  required  for  electrodeposition  methods,  it  may  not   always  be  the  most  viable  option  for  direct  integration  of  catalysts  with  PV   materials.  An  alternative  approach  for  catalyst  deposition,  vapor  deposition   techniques  such  as  atomic–layer  deposition  (ALD),  chemical  vapor  deposition   (CVD),  electron–beam  deposition  (E–beam),  and  sputtering  may  be  an  interesting   approach  for  catalyst  integration  to  the  PV.  Vapor–deposition  techniques  for  direct   integration  of  catalysts  with  PV  devices  are  superior  in  terms  of  controlling  film   thickness  and  conformity.  Additionally,  vapor–deposition  techniques  and  may  be   better  in  terms  of  high  throughput  manufacturing.  Recently  E–beam  evaporation   and  ALD  have  been  used  to  deposit  ultra–thin  nickel  or  cobalt  films  onto  silicon  and   function  as  both  a  protective  layer  and  upon  oxidation  an  OER  catalyst.21–23   Interestingly,  it  has  also  been  observed  that  the  oxidation  of  such  crystalline  cobalt   and  nickel  films  used  for  OER  catalysts  become  amorphous  over  the  course  of   operation  and  then  resemble  electrodeposited  versions  of  Co  or  Ni–OEC’s  developed   in  our  lab.21,24,25  These  structural  changes  are  important  since  amorphous     101   electrodeposited  OECs  are  known  to  exhibit  a  porous  film  morphology.  Therefore,   while  a  compact  vapor–deposited  film  may  initially  appear  promising  as  a  dual   protective–layer  and  catalyst  material,  over  the  course  of  operation  time  it  may   become  porous  to  expose  the  underlying  PV.  It  is  important  to  determine  if  the   vapor–deposited  films  will  demonstrate  the  same  electrochemical  activity  as  the   electrodeposited  versions  and  if  structural  changes  occur  over  the  course  of   operation  time.     Preliminary  studies  have  evaluated  the  catalytic  activity  via  Tafel  analysis  for   water–oxidation  of  a  sputtered  NiFeO  film.  Tafel  analysis  shows  that  this  catalyst   achieves  a  45  mV  decade–1  Tafel  slope  making  it  a  highly  active  OER  catalyst.  Of   significance,  in  contrast  to  the  electrodeposited  Co,  Ni,  and  Mn–OECs,  which  are     102     Figure  5.1  Tafel  plot  of  a  sputtered  NiFeO  OER  catalyst  operating  in  0.5  M  KBi  /   1.5M  KNO3  pH  9.2.  A  Tafel  slope  of  45  mV  decade–1  is  observed  for  a  50nm  (∎),   100nm  (●) and  200nm  (▲)  thick  NiFeO  film.  Inset:  SEM  image  of  a  NiFeO  shows  a   very  dense,  compact  film.     porous  in  nature,  NiFeO  appears  to  show  little  improvement  in  activity  with   increased  catalyst  thickness  (Fig.  5.1)  implying  that  the  sputtered  NiFeO  catalyst  is  a   very  dense  film  (Fig  5.1  inset).  Additionally  initial  results  also  show  that  the  NiFeO   catalyst  demonstrates  different  catalytic  activity  depending  on  the  operational  pH  of   the  solution  indicating  that  catalyst  activity  and  therefore  likely  the  mechanism   changes  with  pH  (Fig  5.2).  This  unusual  behavior  suggests  that  further  analysis  is     103   required  to  fully  how  this  dense  catalyst  operates.     Figure   5.2   Tafel   plots   of   200   nm   thick   NiFeO   (81%   mol   Ni,   19%   mol   Fe)   on   Ni– coated  glass  operated  in  (▲)  0.2  M   KPi,  pH  7.0,  92  mV  decade–1  slope;  (■)  0.2  M   KBi,   pH  9.3,  61  mV  decade–1  slope;  (●)  1.0  M  KOH,  pH  13.9,  45  mV  decade–1  slope.     Previous work has shown that the CoPi catalyst can be made by first sputtering a thick film of metallic cobalt (800 nm) followed by subsequent electrochemical anodization in phosphate buffer.26–28 However, the catalytic activity of the films formed from metallic cobalt exhibited inferior activity as compared to those made from solution electrodeposition. We ascribe this difference in behavior to slow charge transport through the thick films.  Studies on electrodeposited Co–OEC’s showed that (a) the structure and thickness influences the charge transfer through the films, and (b) films deposited from borate   104   solutions, rather than phosphate, produce films with an extended structure that improves charge transfer through the film.17,20,29,30  Preliminary results show that the same is true when forming the Co–OEC from metallic films. Anodizing the metallic film in KBi as opposed to KPi solution,  lowers the Tafel slope from 100   Figure   5.3   Tafel   analysis   of   Co–OEC   films   formed   and   operated   in   KBi   (●)   as   opposed  to  KPi  (●)  solution.    The  films  formed  from   KBi  exhibit  a   lower  Tafel  slope   and  therefore  demonstrate  higher  activity  than  those  formed  in  KPi.     to 60 mV/decade, which matches the activity of the electrodeposited catalyst (Fig.   105   5.3).     Figure  5.4  Tafel  analysis  of  Co–OEC’s  formed  from  anodizing  metallic  cobalt  in  KBi   solution.  In  all  cases  the  Co–OEC  exhibits  a  Tafel  slope  of  60  mV  decade–1,  however   starting   with   thicker   metallic   films   produces   Co–OEC’s   with   higher   activity   than   thinner  films.       Additionally, the activity of CoBi films made from metallic cobalt demonstrate an increase in catalytic activity with increased film thickness suggesting that the catalyst films do exhibit some porosity as opposed to terminating in a thin–layer on the surface of the metallic film (Fig. 5.4). Future work to define the parameters for precise control over films formed from vapor deposition techniques and their subsequent activity should be undertaken. 5.4 Cell design   106   Since  many  semiconductors  are  unstable  in  strongly  acidic  or  basic  solutions,   working  in  moderate  pH  regimes  relaxes  the  stability  constraint  on  PV  devices.   However,  solution  resistance  (RSOL)  is  low  in  strongly  acidic  or  basic  solutions.  Due   to  the  solubility  limits  of  buffers  when  working  under  moderate  pH  conditions,  the   concentrations  of  protons  and  hydroxide  ions  are  quite  low  and  addition  of  an  inert   salt  (i.e.  a  supporting  electrolyte)  is  needed  to  carry  the  ionic  current.31,32  However,   as  shown  in  Chapter  4,  use  of  a  supporting  electrolyte  still  doesn’t  compete  with   strong  acids  or  bases  in  terms  of  minimizing  solution  resistance.  Recent   developments  have  shown  that  utilization  of  novel  flow–cell  designs  can  circumvent   some  of  the  ion–transport  problems  imposed  by  solution  resistance.32  Moreover,   some  cell  configurations  require  the  use  of  membranes  in  order  to  prevent  mixing  of   of  H2  and  O2  which  could  lead  to  safety  concerns.  Typically  these  membranes  are   either  Nafion  or  anion/cation  exchange  membranes.31,32When  introducing  a   membrane  additional  resistive  losses  are  imposed  due  to  the  added  resistance  of  the   membrane  as  well  as  formation  of  undesirable  concentration  gradients  in  the  anodic   and  cathodic  compartments.  These  additional  components  increase  the  voltage   required  for  water–splitting  as  follows:           where  ηMEM  is  the  membrane  resistance    and  ηpH  is  the  overpotential  caused  by  pH   gradients  in  solution.  It  has  been  modeled  and  demonstrated  that  ηMEM  can  be   !" !" =   !! +   !"# !" +   !"# !" +   ! !" +   !"! !" + !" !"   (5.1)     107   optimized  by  modifying  the  thickness  and  porosity  of  the  membrane.33,34  However,  a   more  elegant  solution  is  required  in  order  to  prevent  pH  gradients  from  forming  at   the  anode  and  cathode.     Figure   5.5   The   current   density   traces   show   that   recirculating   streams   allow   the   device   to   function   stably   and   continuously   (purple   trace),   while   without   recirculation   the  device   performance  deteriorates   as  concentration  gradients  form   across   the   cell   and   ionic   species   are   depleted   in   the   oxygen–evolution   side(red   trace).   The   inset   in   the   graph   corresponds   to   a   schematic   representation   of   the   parallel–plate   solar–hydrogen  generator.  Reprinted   with   permission   from   reference   32.   Recently,  work  by  Modestino  et.  al  has  shown  that  a  flow–cell  design  with  a   controlled  recirculating    stream  can  prevent  pH  changes  between  the  anode  and   cathode  compartments.32  Additionally,  they  incorporated  a  PV–EC  device  consisting   of  a  high–efficiency  triple–junction  GaInP2/GaAs/Ge  solar  cell  connected  to  an   iridium–oxide  anode  for  oxygen–evolution  and  a  platinum  cathode  on  the   hydrogen–evolution,  separated  by  a  thin  Nafion  membrane  and  1M  KBi  as  an     108   electrolyte.  Utilizing  a  controlled  recirculating  stream  between  compartments,  they   showed  that  pH  gradients  didn’t  form  and  demonstrated  a  solar–to–fuel  efficiency  of   6.2%,  that  operated  continuously  for  15  hours  (Fig.  5.5).     5.5  Conclusion     The  future  challenges  in  solar–water–splitting  stem  primarily  from  low–cost   of  PV  and  EC  components  while  integrating  all  the  necessary  components  to   produce  a  high  efficiency  and  robust  system.  Alternative  materials,  designs,  and   concepts  that  can  be  applied  to  create  next  generation  PV–EC  devices.       5.6  Experimental   Electrochemical  Methods.    Electrochemical  experiments  were  performed  in   a  two–compartment  electrochemical  cell  using  a  CH  Instruments  760C  or  760D   potentiostat  and  an  Ag/AgCl  reference  electrode  (BASi,  MF–2052).  All  electrode   potentials  were  converted  to  the  NHE  scale  using  E(NHE)  =  E(Ag/AgCl)  +  0.197  V.   Platinum  mesh  (Alfa  Aesar)  was  used  as  the  auxiliary  electrode.  Unless  otherwise   stated,  the  electrolyte  was  0.5  M  KBi  at  a  pH  of  9.2  with  1.5  M  KNO3  as  a  supporting   electrolyte.     Catalyst  Film  Preparation  NiFeO  catalyst  films  were  deposited  by  reactive   sputtering  using  the  AJA  International  sputtering  system.  Films  were  deposited  by   reactive  sputtering  from  an  iron–doped  nickel  sputtering  target  (19%  Fe)  in  an   Ar:O2  atmosphere  of  3:1.35  Co–OEC’s  were  made  by  sputtering  metallic  cobalt  onto  a   FTO  electrode.  The  metallic  films  were  anodized  in  either  0.2M  Pi  or  0.2M  Bi  for  20     109   hours  at  a  current  density  of  1  mA  cm–2.  The  films  were  then  rinsed  and  subsequent   Tafel  analysis  was  conducted  in  either  1.0M  KPi  or  KBi  electrolyte  at  pH  7  or  9.2,   respectively.           110                                                                                                                     5.7  References     1.   Swanson  RM  (2005)  in  Conference  Record  of  the  Thirty–first  IEEE   Photovoltaic  Specialists  Conference,  2005,  pp  889–894.     2.   Green  MA  (2009)  The  path  to  25%  silicon  solar  cell  efficiency:  history  of   silicon  cell  evolution.  Prog.  Photovolt.  Res.  Appl.  17,  183–189.     3.   Powell  DM  et  al.  (2012)  Crystalline  silicon  photovoltaics:  a  cost  analysis   framework  for  determining  technology  pathways  to  reach  baseload   electricity  costs.  Energy  Environ.  Sci.  5,  5874–5883.     4.   Bazilian  M  et  al.  (2013)  Re–considering  the  economics  of  photovoltaic  power.   Renew.  Energy  53,  329–338.     5.   Kojima  A,  Teshima  K,  Shirai  Y,  Miyasaka  T  (2009)  Organometal  halide   perovskites  as  visible–light  sensitizers  for  photovoltaic  cells.  J.  Am.  Chem.  Soc.   131,  6050–6051.     6.   Wojciechowski  K,  Saliba  M,  Leijtens  T,  Abate  A,  Snaith  HJ  (2014)  Sub–150  °C   processed  meso–superstructured  perovskite  solar  cells  with  enhanced   efficiency.  Energy  Environ.  Sci.  7,  1142–1147.     7.   Zhou  H  et  al.  (2014)  Interface  engineering  of  highly  efficient  perovskite  solar   cells.  Science  345,  542–546.     8.   Park  N–G  (2013)  Organometal  Perovskite  Light  Absorbers  Toward  a  20%   efficiency  low–cost  solid–state  mesoscopic  solar  cell.  J.  Phys.  Chem.  Lett.  4,   2423–2429.     9.   Nazeeruddin  MK,  Gao  P,  Grätzel  M  (2014)  Organohalide  lead  perovskites  for   photovoltaic  applications.  Energy  Environ.  Sci.  Available  at:   http://pubs.rsc.org/en/content/articlelanding/2014/ee/c4ee00942h   [Accessed  May  27,  2014].     10.   Green  MA,  Ho–Baillie  A,  Snaith  HJ  (2014)  The  emergence  of  perovskite  solar   cells.  Nat.  Photonics  8,  506–514.     11.   Hodes  G,  Cahen  D  (2014)  Photovoltaics:  Perovskite  cells  roll  forward.  Nat.   Photonics  8,  87–88.     12.   Sum  TC,  Mathews  N  (2014)  Advancements  in  Perovskite  Solar  Cells:   Photophysics  behind  the  Photovoltaics.  Energy  Environ.  Sci.  Available  at:     111                                                                                                                                                                                                                                                                                                                                             http://pubs.rsc.org/en/content/articlelanding/2014/ee/c4ee00673a   [Accessed  May  7,  2014].     13.   Noel  NK  et  al.  (2014)  Lead–free  organic–inorganic  tin  halide  perovskites  for   photovoltaic  applications.  Energy  Environ.  Sci.  Available  at:   http://pubs.rsc.org/en/content/articlelanding/2014/ee/c4ee01076k   [Accessed  August  5,  2014].     14.   Hao  F,  Stoumpos  CC,  Cao  DH,  Chang  RPH,  Kanatzidis  MG  (2014)  Lead–free   solid–state  organic–inorganic  halide  perovskite  solar  cells.  Nat.  Photonics  8,   489–494.     15.   Snaith  HJ  (2013)  Perovskites:  The  Emergence  of  a  new  era  for  low–cost,   high–efficiency  solar  cells.  J.  Phys.  Chem.  Lett.  4,  3623–3630.     16.   Surendranath  Y,  Kanan  MW,  Nocera  DG  (2010)  Mechanistic  studies  of  the   oxygen  evolution  reaction  by  a  cobalt–phosphate  catalyst  at  neutral  pH.  J.  Am.   Chem.  Soc.  132,  16501–16509.     17.   Surendranath  Y,  Lutterman  DA,  Liu  Y,  Nocera  DG  (2012)  Nucleation,  growth,   and  repair  of  a  cobalt–based  oxygen  evolving  catalyst.  J.  Am.  Chem.  Soc.  134,   6326–6336.     18.   Bediako  DK  et  al.  (2012)  Structure–activity  correlations  in  a  nickel–borate   oxygen  evolution  catalyst.  J.  Am.  Chem.  Soc.  134,  6801–6809.     19.   Bediako  DK,  Surendranath  Y,  Nocera  DG  (2013)  Mechanistic  studies  of  the   oxygen  evolution  reaction  mediated  by  a  nickel–borate  thin  film   electrocatalyst.  J.  Am.  Chem.  Soc.  135,  3662–3674.     20.   Huynh  M,  Bediako  DK,  Liu  Y,  Nocera  DG  (2014)  Nucleation  and  growth   mechanisms  of  an  electrodeposited  manganese  oxide  oxygen  evolution   catalyst.  J.  Phys.  Chem.  C.  Available  at:  http://dx.doi.org/10.1021/jp501768n   [Accessed  June  12,  2014].     21.   Kenney  MJ  et  al.  (2013)  High–performance  silicon  photoanodes  passivated   with  ultrathin  nickel  films  for  water  oxidation.  Science  342,  836–840.     22.   Hu  S  et  al.  (2014)  Amorphous  TiO2  coatings  stabilize  Si,  GaAs,  and  GaP   photoanodes  for  efficient  water  oxidation.  Science  344,1005–1009.     23.   Yang  J  et  al.  (2014)  Efficient  and  sustained  photoelectrochemical  water   oxidation  by  cobalt  oxide/silicon  photoanodes  with  nanotextured  interfaces.   J.  Am.  Chem.  Soc.  136,  6191–6194.       112                                                                                                                                                                                                                                                                                                                                             24.   Lee  SW  et  al.  (2012)  The  nature  of  lithium  battery  materials  under  oxygen   evolution  reaction  conditions.  J.  Am.  Chem.  Soc.  134,  16959–16962.   25.     Risch  M  et  al.  (2013)  Structural  changes  of  cobalt–based  perovskites  upon   water  oxidation  investigated  by  EXAFS.  J.  Phys.  Chem.  C  117,  8628–8635.     26.    Young  ER,  Nocera  DG,  Bulović  V  (2010)  Direct  formation  of  a  water   oxidation  catalyst  from  thin–film  cobalt.  Energy  Environ.  Sci.  3,  1726–1728.     27.   Young  ER,  Costi  R,  Paydavosi  S,  Nocera  DG,  Bulović  V  (2011)  Photo–assisted   water  oxidation  with  cobalt–based  catalyst  formed  from  thin–film  cobalt   metal  on  silicon  photoanodes.  Energy  Env.  Sci.  4,  2058–2061.     28.   Costi  R,  Young  ER,  Bulović  V,  Nocera  DG  (2013)  Stabilized  CdSe–CoPi   Composite  Photoanode  for  Light–Assisted  Water  Oxidation  by   Transformation  of  a  CdSe/Cobalt  Metal  Thin  Film.  ACS  Appl.  Mater.  Interfaces   5:2364–2367.     29.   Farrow  CL,  Bediako  DK,  Surendranath  Y,  Nocera  DG,  Billinge  SJL  (2013)   Intermediate–range  structure  of  self–assembled  cobalt–based  oxygen– evolving  catalyst.  J.  Am.  Chem.  Soc.  135,  6403–6406.     30.   Bediako  DK,  Costentin  C,  Jones  EC,  Nocera  DG,  Savéant  J–M  (2013)  Proton– electron  transport  and  transfer  in  electrocatalytic  films.  Application  to  a   cobalt–based  O2–evolution  catalyst.  J.  Am.  Chem.  Soc.  135,  10492–10502.     31.   Hernández–Pagán  EA  et  al.  (2012)  Resistance  and  polarization  losses  in   aqueous  buffer–membrane  electrolytes  for  water–splitting   photoelectrochemical  cells.  Energy  Environ.  Sci.  5,  7582–7589.     32.   Modestino  MA  et  al.  (2013)  Robust  production  of  purified  H2  in  a  stable,  self– regulating,  and  continuously  operating  solar  fuel  generator.  Energy  Environ.   Sci.  7,  297–301.     33.   Haussener  S  et  al.  (2012)  Modeling,  simulation,  and  design  criteria  for   photoelectrochemical  water–splitting  systems.  Energy  Environ.  Sci.  5,  9922– 9935.     34.   Berger  A,  Segalman  RA,  Newman  J  (2014)  Material  Requirements  for   Membrane  Separators  in  a  Water–Splitting  Photoelectrochemical  Cell.  Energy   Environ  Sci.  Available  at:   http://pubs.rsc.org/en/content/articlelanding/2014/ee/c3ee43807d   [Accessed  February  12,  2014].     35.   Miller  EL,  Rocheleau  RE  (1997)  Electrochemical  behavior  of  reactively   sputtered  iron–doped  nickel  oxide.  J.  Electrochem.  Soc.  144,  3072–3077.     113