SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf ·...

44
SUSY at The Compression Frontier Tom LeCompte Argonne Na*onal Laboratory

Transcript of SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf ·...

Page 1: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

SUSY at The Compression Frontier Tom  LeCompte  

Argonne  Na*onal  Laboratory  

Page 2: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

2  

Persuasion

Unusually,  I  am  not  here  to  persuade  you  of  anything  –  other  than  that  some  ideas  are  maybe  worth  thinking  about.  

Page 3: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

3  

Outline: Three Areas of Compression

  Compression  from  the  Higgs  Searches  

  Compressed  SUSY  

  The  future  –  and  where  our  compuDng  model  is  squeezing  the  physics  

I  am  a  member  of  the  ATLAS  collaboraDon,  but  am  not  represenDng  them  today.    I  am  here  as  a  “phenomenolgist”.  

I  would  much  rather  this  be  a  discussion  than  a  “lecture”.    Interrupt  me  as  much  as  you  like.  

Page 4: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

4  

Part  I:  Compression  from  the  Higgs  

Page 5: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

5  

CERN Seminar in December:

These  plots  were  shown  at  the  CERN    “jamboree”*  in  December.  

   The  reacDon  of  some  of  the  community  was…  

*  Derived  from  a  word  meaning  “a  big  headache.”  

Page 6: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

6  

A 125 GeV Higgs is too heavy for SUSY!

Page 7: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

7  

There is No Need To Panic:

  First,  the  experiments  are  not  saying  that  there  is  a  125  GeV  Higgs.      –  ATLAS:  there  is  a  window  between  115.5-­‐131  GeV  

–  CMS:  there  is  a  window  between  114  (LEP)-­‐129  GeV  (now  127  GeV)  

–  A  light  Higgs  can  be  anywhere  in  that  window.  

  Second,  trying  to  line  up  the  points  of  poorest  exclusion  between  ATLAS  and  CMS  implicitly  asks  the  wrong  quesDon.  –  Even  if  there  is  a  parDcle  present,  the  point  of  poorest  exclusion  is  not  in  general  at  the  

center  of  mass  of  the  peak.  

At  1σ,  a  SM  Higgs  can  be  almost  anywhere  in  the  non-­‐excluded  region.    At  2σ,  it  could  preby  much  be  anywhere.  

Page 8: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

8  

More Reason Not To Panic…

  Third,  the  ATLAS  H    γγ  search  excess  is  1.5-­‐2.0x  what  one  expects  from  the  SM.    If  one  wants  to  interpret  it  as  a  Higgs,  it  leads  to  one  of  two  possibiliDes:  –  There  is  a  component  of  this  excess  that  is  an  upward  staDsDcal  fluctuaDon.    In  this  case,  

the  inferred  mass  is  altered  by  this  fluctuaDon.    -­‐OR-­‐  –  This  is  a  non-­‐SM  Higgs.  

–  In  either  case,  one  cannot  properly  apply  the  constraints  of  a  125  GeV  SM  Higgs  on  SUSY.  

  Finally  (and  this  will  lead  me  into  my  next  slide),  these  constraints  apply  to  the  MSSM,  and  there’s  more  to  SUSY  than  the  MSSSM.      

We’re  going  to  have  to  wait  a  few  more  months  to  get  something  more  definite.      (Also  a  message  from  the  December  seminar)  

Page 9: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

9  

More Reason Not To Panic…

  Third,  the  ATLAS  H    γγ  search  excess  is  1.5-­‐2.0x  what  one  expects  from  the  SM.    If  one  wants  to  interpret  it  as  a  Higgs,  it  leads  to  one  of  two  possibiliDes:  –  There  is  a  component  of  this  excess  that  is  an  upward  staDsDcal  fluctuaDon.    In  this  case,  

the  inferred  mass  is  altered  by  this  fluctuaDon.    -­‐OR-­‐  –  This  is  a  non-­‐SM  Higgs.  

–  In  either  case,  one  cannot  properly  apply  the  constraints  of  a  125  GeV  SM  Higgs  on  SUSY.  

  Finally  (and  this  will  lead  me  into  my  next  slide),  these  constraints  apply  to  the  MSSM,  and  there’s  more  to  SUSY  than  the  MSSSM.      

We’re  going  to  have  to  wait  a  few  more  months  to  get  something  more  definite.      (Also  a  message  from  the  December  seminar)  

Page 10: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

10  

Part  II:  Compressed  SUSY  

Work  done  with  Steve  MarDn,  Northern  Illinois  University  Phys.Rev.D84:015004,2011  (hbp://arxiv.org/abs/1105.4304)  

And  follow-­‐up  (hbp://arxiv.org/abs/1111.6897)  

Page 11: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

11  

Some possibly unpopular thoughts

  SUSY  may  not  be  mSUGRA/CMSSM  

  SUSY  may  not  be  MSSM  

  SUSY  may  not  even  be  SUSY!  –  To  me,  “SUSY”  represents  an  experimental  search  for  a  family  of  topologies  that  are    

predicted  in  supersymmetric  theories,  but  may    have  nothing  to  do  with  them.  –  A  common  characterisDc  is  missing  energy  from  undetected  weakly  or  non-­‐interacDng  

parDcles.  •  These  may  or  may  not  have  something  to  do  with  Dark  Maber  

–  Events  can  be  characterized  by  what’s  going  on  with  the  rest  of  the  event:  leptons,  jets,  etc.  •  These  may  or  may  not  give  you  peaks  in  an  invariant  mass  distribuDon,  but  they  are  

likely  to  have  some  disDnguishing  kinemaDcs.  

Page 12: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

12  

Compressed SUSY

  mSUGRA  has  great  big  gaps  between  the  masses.  

  This  gives  you  energeDc  leptons,  jets  and  usually  lots  of  missing  ET.  

  It’s  almost  as  if  this  was  designed  to  be  easy  to  find  at  colliders.  

  By  compressing  the  spectrum  you  usually  (more  on  this  later)  get  less  energeDc  objects,  less  missing  ET  etc.  

  The  signal  looks  more  like  the  background.  

  Hard  QCD  radiaDon  can  be  important.  

Page 13: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

13  

The “Easy” Thing To Do

  The  LHC  experiments  have  already  published  limits  on  mSUGRA.      

  Can  these  limits  be  adapted  to  more  compressed  spectra?  

  Step  #1  is  to  come  up  with  explicit  compressed  models:  

–  c  =  0  is  mSUGRA-­‐like  

–  c=  1  is  total  compression  (gauginos  degenerate)  

Here  “easy”  means  “without  a  dedicated  search”.  

There  is  nothing  magical  about  these  choices;  our  goal  was  to  come  up  with  a  spectrum  that  was  very  different  than  mSUGRA.  

Page 14: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

14  

Getting acceptances

  Step  #2  is  to  calculate  acceptances  for  these  models.  

  We  used  MadGraph/MadEvent  (generaDon)  +  Pythia  (fragmentaDon)  +  PGS  (simulaDon  of  an  ATLAS-­‐like  detector)    –  ATLAS  made  public  a  number  of  their  own  points  that  we  could  tune  and  compare  to.    

This  was  extraordinarily  useful,  and  I’d  like  to  encourage  this  for  every  analysis.  –  Typically,  we  were  good  to  ~15%  or  beber,  and  the  limitaDon  seemed  to  be  the  staDsDcs  

on  the  ATLAS  side  (more  later)  –  Cross-­‐secDons  were  normalized  to  Prospino  cross-­‐secDons  

  We  ran  four  producDon  models:  –  Gluino/gluino,  gluino/squark,  squark/squark,  and  events  with  stops  and/or  sboboms  

–   allowing  up  to  1  extra  jet  (at  the  matrix  element  level)  

  And  we  got  our  first  surprise:  

Page 15: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

15  

The Surprise!

  The  acceptance  for  moderate  compression  (say  c  <  0.5)  is  comparable  to  (and  someDmes  even  a  lible  bigger  than)  the  mSUGRA-­‐like  case.  

  This  contradicts  our  intuiDon  that  increasing  compression  means  reduced  acceptance.  

  That  reducDon  in  acceptance  doesn’t  seem  to  “switch  on”  unDl  c  >  0.5  or  so.      

These  are  the  three  of  the  four  signal  regions  in  the  35  pb-­‐1  ATLAS  0-­‐lepton  paper  plus  the  1-­‐lepton  paper  signal  region.  

More  compression  

Page 16: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

16  

What Is Happening

At  low  compression  (mSUGRA-­‐like)  we  have  large  gaps  between  parDcle  masses.  

At  moderate  to  high  compression,  the  gaps  become  smaller,  but  the  average  masses  shix  upwards.    This  provides  sufficient  Q  for  the  decay  to  be  visible,  provided  c  is  not  too  large.  

This  is  a  statement  about  acceptance,  not  sensiDvity.      In  general,  cross-­‐secDons  will  fall  as  the  mass  spectrum  shixs  upwards.    However,  for  this  model,  since  the  bulk  of  the  cross-­‐secDon  is  through  parDcles  whose  masses  we  fix,  that  doesn’t  happen.      

Page 17: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

17  

What Else Is Happening

  With  increasing  compression,  both  Meff  and  MET  get  small    However,  Meff  gets  smaller  faster  –  so  MET/Meff  can  actually  grow  

–  More  events  can  pass  this  cut,  so  the  acceptance  can  grow…  –  …provided  c  does  not  get  too  big.  

  At  very  high  compression,  both  become  sox  and  the  signal  melts  into  the  background.  

Page 18: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

18  

Setting Limits

  Step  #3:  Take  the  published  ATLAS  fiducial  cross-­‐secDon  limits  and  Step  #2,  and  convert  to  a  limit  in  the  m(gluino-­‐c)  plane.  –  We  slightly  changed  the  axes  

for  ease  of  interpretaDon.  –  We  could  just  have  easily  used  

CMS  limits;  we  used  ATLAS  simply  because  it  was  familiar  to  us.  

–  This  limit-­‐sezng  was  possible  because  the  experiments  published  fiducial  cross-­‐secDon  limits.    This  should  be  strongly  encouraged.   I’ve  switched  to  the  1  {-­‐1  0-­‐lepton  paper  limits,  

which  are  more  stringent  than  the  35  pb-­‐1    limits.    

Page 19: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

19  

Features of Interest For  most  model  points,  Region  A  (two  jets)  has  the  most  sensiDvity  

At  high  compression,  other  search  regions  do  a  lible  beber  than  Region  A.  

Region  C  (four  jets,  low  MET  and  Meff)  does  poorly,  largely  because  it  has  a  large  background  relaDve  to  the  other  points.  

Many  surviving  models  have  m(gluino)  >  1050  GeV  and/or  m(LSP)  >  600  GeV.  

Even  at  high  compression,    m(gluino)  >  600  GeV.  

This  is  not  a  model  independent  statement,  simply  a  shorthand  descripDon  of  the  limits.  

Page 20: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

20  

Another Model – Heavy Winos

  Here  we  move  the  wino  mass  up  to  several  TeV,  effecDvely  decoupling  them.  

  The  decay  energy  is  now  shared  amongst  few  parDcles  –  acceptances  go  up.      

  In  this  family  of  models  gluino  decays  to  stop  +  top  are  kinemaDcally  forbidden.  

  Decays  are    –  Leptonic  decays  are  rare.  

Page 21: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

21  

Limits in Heavy Wino Model Many  surviving  models  have  m(gluino)  >  1100  GeV  and/or  m(LSP)  >  600  GeV.  

Even  at  high  compression,    m(gluino)  >  600  GeV.  

This  is  not  a  model  independent  statement,  simply  a  shorthand  descripDon  of  the  limits.  

Limits  tend  to  be  a  lible  more  stringent  than  in  the  baseline  case,  due  to  parDDoning  the  energy  amongst  fewer  jets.  

At  very  high  mass,  the  cross-­‐secDons  are  all  small,  and  ulDmately  one  hits  the  wall.  

Page 22: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

22  

Yet Another Model – Heavy Squarks

  Here  we  move  the  squark  mass  up  to  several  TeV,  effecDvely  decoupling  them.  

  The  decay  energy  is  now  shared  amongst  many  parDcles  –    –  Decays  go  through  many  

intermediate  jets    –  acceptances  go  down.    

  Leptons  can  again  be  present.    

Page 23: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

23  

Acceptances Get More Complex   When  energy  is  parDDoned  among  many  jets  with  

many  m(jj)  constraints,  things  can  get  complicated.  

  Part  of  this  is  the  compeDng  effects  of  raising  the  mass  while  at  the  same  Dme  lowering  the  mass  difference.  

  Part  of  this  is  due  to  having  N  jets  in  an  event,  M  of  which  must  pass  certain  cuts.  –  Including  the  effects  of  jet  merging  

  For  low  compression,  region  E  (4  jets,    inclusive  Meff)  is  best,  but  as  the    compression  increases,  B  (3  jets)  and    then  A  (2  jets)  become  best.  

35  pb-­‐1  paper  

1  {-­‐1  paper  

Page 24: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

24  

Limits in Heavy Squark Model

Here  we  only  get  to  a  gluino  mass  limit  around  450  GeV  for  high  compression.      

There  really  isn’t  an  LSP  limit  to  speak  of  in  this  case.  

Limits  are  much  weaker  than  in  the  baseline  case,  due  to  parDDoning  the  energy  amongst  many  jets.  

The  best  limits  are  provided  by  different  signal  regions  than  in  the  previous  two  cases.  

Page 25: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

25  

One Final Model – Heavy Squarks and Winos

  Here  we  move  both  the  squark  and  the  wino  mass  up  to  several  TeV,  effecDvely  decoupling  them.  

  The  decay  energy  is  now  shared  amongst  an  intermediate  number  of  parDcles:  

  As  expected,  the  outcome  is  intermediate  between  the  baseline  (and  heavy  wino)  case  and  the  heavy  squark  case.  

(for  completeness)  

(You  might  recognize  this  spectrum  as  a  typical  “simplified  model.”)  

Page 26: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

26  

Some Intermediate Conclusions

  The  mSUGRA  searches,  while  not  opDmized  for  Compressed  SUSY,  do  a  preby  (maybe  even  surprisingly)  good  job.  

  No  single  search  region  does  the  best  for  every  model.  –  It’s  probably  impossible  to  develop  a  “compressed  signal  region”,  although  it  looks  

possible  to  develop  a  “signal  region  good  for  some  compressed  models”.  

  One  deadly  cut  is  MET/Meff.  –  One  can  sharpen  this  by  including  more  jets  in  Meff.  –  The  discussion  about  whether  this  cut  is  best  applied  to  the  selected  jets  or  the  event  as  

a  whole  is  ongoing  in  the  experiments  now.      It’s  a  delicate  balance.  

  UlDmately,  the  sensiDvity  will  be  driven  by  when  the  signal  dissolves  into  the  background.  –  Today,  most  backgrounds  are  ulDmately  determined  by  the  staDsDcs  of  control  regions  

(or  regions  to  tune  Monte  Carlo).    The  situaDon  will  improve  with  more  data,  at  least  unDl  we  hit  the  true  systemaDcs  “wall”.  

–  We  haven’t  considered  techniques  like  Razor  that  we  know  improve  S/B.    To  what  degree  will  they  help?  

Page 27: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

27  

Part  III:  The  Future  –  and  CompuDng  Compression  

Page 28: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

28  

The Future

  Like  it  or  not,  the  future  of  SUSY  is  compression  –  Since  mSUGRA  is  about  the  least  compressed  –  most  evenly  spaced  -­‐  a  spectrum  one  

can  imagine,  virtually  everything  else  will  be  more  compressed  by  comparison.  

  mSUGRA  can  be  thought  of  as  a  push  outward  into  m0  –  m½  space:  toward  higher  masses.  

  Compressed  SUSY  is  fighDng  a  two-­‐front  war:  –  Towards  higher  masses  for  models  with  moderate  compression  

–  Towards  higher  backgrounds  for  models  with  more  compression  

  FighDng  a  two-­‐front  war  is  not  for  the  faint  of  heart.  

Page 29: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

29  

Exclusion Planes

  In  mSUGRA,  it’s  natural  to  be  discussing  m0  and  m½,  or  alternaDvely,  m(squark)  and  m(gluino)  –  I  confess  that  I  have  a  preference  for  

expressing  limits  in  terms  of  physical  quanDDes  over  parameters  of  the  theory.  

  In  our  compressed  models,  a  more  natural  descripDon  is  in  terms  of  this  compression  parameter  “c”.  –  m(gluino)-­‐m(LSP)  is  essenDally  the  

same  thing,  expressed  in  terms  of  physical  parameters.  

Page 30: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

30  

Why Exclusion Planes?

For  a  mSUGRA-­‐like  model,  this  is  the  sort  of  limit  plot  that’s  used.  

For  our  compressed  models,  this  is  a  more  suitable  choice.  

This  is  a  third  logical  possibility,  probably  best  for  some  class  of  models.  

We’d  really  like  to  look  at  the  exclusion  surface  in  a  3-­‐dimensional  space.  

Why  not?  

Page 31: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

31  

How Experiments* Do Exclusions

*  And  wanna-­‐be  phenomenologists  

BEFORE  we  have  taken  any  data,  we  generate  a  “grid”  of  points  in  the  exclusion  plane.  

The  acceptance  and  efficiency  for  each  point  is  calculated  using  Monte  Carlo.  

Page 32: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

32  

Exclusions – Part II

AFTER  we  have  taken  the  data,  we  see  which  points  in  the  grid  have  been  excluded.  

X X X X X X X X

Page 33: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

33  

Exclusions – Part III

If  all  four  corners  of  a  square  are  excluded,  the  enDre  square  is  excluded.  

If  no  corners  of  a  square  is  excluded,  no  part  of  the  square  is  excluded.  

If  1,  2  or  3  corners  of  a  square  are  excluded,  the  exclusion  contour  passes  within  that  square.  

X X X X X X X X

Page 34: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

34  

Exclusions – Part IV

Some  sort  of  interpolaDon  is  then  done  to  infer  the  exclusion  curve  –  e.g.  linear  interpolaDon  followed  by  smoothing.  

Remember  –the  only  actual  informaDon  we  have  is  at  the  grid  points.  X X X X

X X X X

Page 35: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

35  

Some facts

  The  models  in  LeCompte  &  MarDn  required  80,000,000  events  to  be  generated  –  516  models  in  4  categories  (I  am  ignoring  the  5th  model  in  the  paper)  

–  Up  to  4  producDon  processes  per  event  •  Gluino/guino,  gluino/squark,  squark/squark  and  stops  and/or  boboms  

–  80,000-­‐240,000  events  per  point  •  Compressed  models  have  low  acceptance  over  much  of  the  compression  range,  so  we  need  a  

lot  of  events.  

–  It  all  adds  up.    But…  

–  on  a  200-­‐core  cluster,  with  modest  contenDon,  this  can  be  done  over  a  (long)  weekend.  

  The  total  number  of  MC  events  generated  by  a  large  collider  experiment  in  one  of  their  campaigns  is  of  order  a  billion.  –  An  unfair  calculaDon:  109  events  /  102  papers  =  107  MC  events  per  paper  

  A  2-­‐D  grid  is  well-­‐matched  to  what  the  experiments  can  do  today.    A  3-­‐D  grid  (10-­‐20x  as  many  points)  is  mathemaDcally  possible,  but  not  pracDcal.  

Page 36: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

36  

Working Smarter Why  generate  events  here,  where  you  have  no  chance  to  exclude  them?  

X X X X X X X X

Why  generate  events  here,  where  they  are  probably  already  excluded?  

Obviously  this  is  risky;  one  doesn’t  want  to  get  too  close  to  the  exclusion  line.      

SDll,  the  benefit  of  doing  this,  plus  pre-­‐scaling  the  number  of  events  in  each  producDon  process  by  the  cross-­‐secDon  can  be  substanDal  –  perhaps  a  factor  of  around  5.    

Page 37: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

37  

Working Smarter II

Once  we  have  idenDfied  the  approximate  region  of  the  exclusion  curve,  we  can  generate  more  points  near  that  curve.  

AddiDonally,  one  can  generate  more  events  at  each  point  (old  and  new).      

If  you  could  do  this,  one  could  start  with  a  wide  but  sparse  grid,  and  eliminate  the  risk  from  the  previous  “working  smarter”  

X X X X X X X X

Page 38: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

38  

Why Don’t We Work Smarter?

  Today,  we  send  the  grid  to  the  Grid.  

  When  we  get  it  back,  we  can  in  principle  iterate  like  I  described,  and  send  out  a  new  set  of  jobs.  

  A  typical  queue  for  a  job  this  big  can  be  weeks  long  –  maybe  a  month.  –  It’s  not  easy  or  fast  to  generate  a  billion  events  

–  While  this  grid  is  not  a  billion  events,  there  are  many  other  needs  that  need  to  be  serviced  –  one  can’t  think  of  a  single  request  in  isolaDon.  

  It’s  simply  not  true  that  the  grid  is  fast.    The  grid  has  a  lot  of  capacity.      

High  capacity   High  speed  

This  is  a  problem  we  will  need  to  solve  if  we  ever  want  to  go  beyond  a  small  number  of  2D  grids.  

Page 39: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

39  

Why Don’t We Work Smarter II?

  One  reason  why  LeCompte  &  MarDn  could  run  80M  events  over  a  weekend  while  ATLAS  is  struggling  with  a  few  10’s  of  million  is  because  we  used  PGS.  

  PGS  reproduced  ATLAS  simulaDon  to  within  ~15%.  –  Some  of  this  15%  was  due  to  staDsDcal  fluctuaDons  in  the  ATLAS  samples  

–  Using  PGS  to  interpolate  between  two  nearby  fully  simulated  points  is  probably  good  to  a  few  percent.  

–  QCD  uncertainDes  mean  Prospino  cross-­‐secDons  are  probably  good  to  (at  best)  20%  

  The  experiments  are  hesitant  about  opening  the  door  to  PGS  –  Concern  that  PGS  will  be  misused  –  a  concern  that  is  not  

enDrely  unjusDfied.  •  PGS  will  let  us  generate  orders  of  magnitude  more  points  than  

we  could  otherwise.  •  PGS  can  also  facilitate  lazy  and  sloppy  analyses  

This  is  not  a  trivial  problem  to  solve.    Any  soluDon  beginning  with  “Just”  probably  will  not  work.    

Page 40: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

40  

Summary

  mSUGRA  is  about  the  least  compressed  model  imaginable.  

  Whatever  succeeds  it  is  likely  to  be  more  compressed  –  there’s  only  one  direcDon  to  go.  

  “Ordinary”  SUSY  searches  do  surprisingly  well  with  models  with  moderate  compression.  

  Models  with  more  compression  will  require  a  more  dedicated  search.      –  These  searches  oxen  need  to  extend  to  regions  of  higher  background.  

–  There  are  two  upcoming  challenges:  •  Advance  the  analysis  techniques  to  reach  parity  with  the  mSUGRA  searches  •  Solve  the  compuDng  problems  associated  with  mulDdimensional  grids  

  It’s  too  soon  to  worry  about  the  Higgs.  

  Thanks  to  the  organizers  for  inviDng  me!  

  Thanks  to  my  colleague  Steve  MarDn,  my  ATLAS  collaborators  and  my  CMS  friends    

Page 41: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

41  

Page 42: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

42  

ATLAS 1 fb-1 signal regions

Page 43: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

43  

A Word on Backgrounds

  Consider  Z+jets,  followed  by  Z    neutrinos.  

  We  will  never  get  enough  Z    lepton  events  to  fully  constrain  this:  –  BF(Z    vv)  =  2  x  BF(Z    ee)  

•  The  Z  has  only  the  axial  coupling  to  charged  leptons,  but  the  full  vector  coupling  and  the  axial  coupling  to  neutrinos.  

–  Three  neutrino  families  vs.  two  charged  lepton  families  

–  Acceptance  for  neutrinos  is  100%.  

–  All  in  all,  there  is  about  a  factor  of  15x  more  Z    neutrinos  as  Z    leptons.  

  The  best  we  can  do  is  use  the  events  we  have  to  constrain  Monte  Carlos  –  The  staDsDcal  uncertainty  will  be  driven  by  the  number  of  events  in  the  control  region,  

which  means  the  uncertainDes  are  4x  larger  than  we  would  like.  

–  This  sDll  falls  like  1/sqrt(N),  though.  

Page 44: SUSY at The Compression Frontierresearch.kek.jp/people/nojiri/shingakujyutu/LeCompte.pdf · 2012-02-19 · 2 Persuasion Unusually,$Iam$not here$to$persuade$ you$of$anything$– other$than$that

44  

Published ATLAS/CMS Plots