CMS 3D CMS Logo

TtFullLepKinSolver.cc
Go to the documentation of this file.
2 #include "TF2.h"
3 
5  : topmass_begin(0), topmass_end(0), topmass_step(0), mw(80.4), mb(4.8), pxmiss_(0), pymiss_(0) {
6  // That crude parametrisation has been obtained from a fit of O(1000) pythia events.
7  // It is normalized to 1.
8  EventShape_ = new TF2("landau2D", "[0]*TMath::Landau(x,[1],[2],0)*TMath::Landau(y,[3],[4],0)", 0, 500, 0, 500);
9  EventShape_->SetParameters(30.7137, 56.2880, 23.0744, 59.1015, 24.9145);
10 }
11 
13  const double b, const double e, const double s, const std::vector<double>& nupars, const double mW, const double mB)
14  : topmass_begin(b), topmass_end(e), topmass_step(s), mw(mW), mb(mB), pxmiss_(0), pymiss_(0) {
15  EventShape_ = new TF2("landau2D", "[0]*TMath::Landau(x,[1],[2],0)*TMath::Landau(y,[3],[4],0)", 0, 500, 0, 500);
16  EventShape_->SetParameters(nupars[0], nupars[1], nupars[2], nupars[3], nupars[4]);
17 }
18 
19 //
20 // destructor
21 //
23 
25  TtDilepEvtSolution fitsol(*asol);
26 
27  //antilepton and lepton
28  TLorentzVector LV_e, LV_e_;
29  //b and bbar quark
30  TLorentzVector LV_b, LV_b_;
31 
32  bool hasMCinfo = true;
33  if (fitsol.getGenN()) { // protect against non-dilept genevents
34  genLV_n = TLorentzVector(
35  fitsol.getGenN()->px(), fitsol.getGenN()->py(), fitsol.getGenN()->pz(), fitsol.getGenN()->energy());
36  } else
37  hasMCinfo = false;
38 
39  if (fitsol.getGenNbar()) { // protect against non-dilept genevents
40  genLV_n_ = TLorentzVector(
41  fitsol.getGenNbar()->px(), fitsol.getGenNbar()->py(), fitsol.getGenNbar()->pz(), fitsol.getGenNbar()->energy());
42  } else
43  hasMCinfo = false;
44  // if MC is to be used to select the best top mass and is not available,
45  // then nothing can be done. Stop here.
46  if (useMCforBest_ && !hasMCinfo)
47  return fitsol;
48 
49  // first lepton
50  if (fitsol.getWpDecay() == "muon") {
51  LV_e = TLorentzVector(
52  fitsol.getMuonp().px(), fitsol.getMuonp().py(), fitsol.getMuonp().pz(), fitsol.getMuonp().energy());
53  } else if (fitsol.getWpDecay() == "electron") {
54  LV_e = TLorentzVector(fitsol.getElectronp().px(),
55  fitsol.getElectronp().py(),
56  fitsol.getElectronp().pz(),
57  fitsol.getElectronp().energy());
58  } else if (fitsol.getWpDecay() == "tau") {
59  LV_e =
60  TLorentzVector(fitsol.getTaup().px(), fitsol.getTaup().py(), fitsol.getTaup().pz(), fitsol.getTaup().energy());
61  }
62 
63  // second lepton
64  if (fitsol.getWmDecay() == "muon") {
65  LV_e_ = TLorentzVector(
66  fitsol.getMuonm().px(), fitsol.getMuonm().py(), fitsol.getMuonm().pz(), fitsol.getMuonm().energy());
67  } else if (fitsol.getWmDecay() == "electron") {
68  LV_e_ = TLorentzVector(fitsol.getElectronm().px(),
69  fitsol.getElectronm().py(),
70  fitsol.getElectronm().pz(),
71  fitsol.getElectronm().energy());
72  } else if (fitsol.getWmDecay() == "tau") {
73  LV_e_ =
74  TLorentzVector(fitsol.getTaum().px(), fitsol.getTaum().py(), fitsol.getTaum().pz(), fitsol.getTaum().energy());
75  }
76 
77  // first jet
78  LV_b = TLorentzVector(
79  fitsol.getCalJetB().px(), fitsol.getCalJetB().py(), fitsol.getCalJetB().pz(), fitsol.getCalJetB().energy());
80 
81  // second jet
82  LV_b_ = TLorentzVector(fitsol.getCalJetBbar().px(),
83  fitsol.getCalJetBbar().py(),
84  fitsol.getCalJetBbar().pz(),
85  fitsol.getCalJetBbar().energy());
86 
87  //loop on top mass parameter
88  double weightmax = -1e30;
89  double mtmax = 0;
90  for (double mt = topmass_begin; mt < topmass_end + 0.5 * topmass_step; mt += topmass_step) {
91  //cout << "mt = " << mt << endl;
92  double q_coeff[5], q_sol[4];
93  FindCoeff(LV_e, LV_e_, LV_b, LV_b_, mt, mt, pxmiss_, pymiss_, q_coeff);
94  int NSol = quartic(q_coeff, q_sol);
95 
96  //loop on all solutions
97  for (int isol = 0; isol < NSol; isol++) {
98  TopRec(LV_e, LV_e_, LV_b, LV_b_, q_sol[isol]);
100  if (weight > weightmax) {
101  weightmax = weight;
102  mtmax = mt;
103  }
104  }
105 
106  //for (int i=0;i<5;i++) cout << " q_coeff["<<i<< "]= " << q_coeff[i];
107  //cout << endl;
108 
109  //for (int i=0;i<4;i++) cout << " q_sol["<<i<< "]= " << q_sol[i];
110  //cout << endl;
111  //cout << "NSol_" << NSol << endl;
112  }
113 
114  fitsol.setRecTopMass(mtmax);
115  fitsol.setRecWeightMax(weightmax);
116 
117  return fitsol;
118 }
119 
120 void TtFullLepKinSolver::SetConstraints(const double xx, const double yy) {
121  pxmiss_ = xx;
122  pymiss_ = yy;
123 }
124 
126  const TLorentzVector& LV_l_,
127  const TLorentzVector& LV_b,
128  const TLorentzVector& LV_b_) {
129  math::XYZTLorentzVector maxLV_n = math::XYZTLorentzVector(0, 0, 0, 0);
130  math::XYZTLorentzVector maxLV_n_ = math::XYZTLorentzVector(0, 0, 0, 0);
131 
132  //loop on top mass parameter
133  double weightmax = -1;
134  for (double mt = topmass_begin; mt < topmass_end + 0.5 * topmass_step; mt += topmass_step) {
135  double q_coeff[5], q_sol[4];
136  FindCoeff(LV_l, LV_l_, LV_b, LV_b_, mt, mt, pxmiss_, pymiss_, q_coeff);
137  int NSol = quartic(q_coeff, q_sol);
138 
139  //loop on all solutions
140  for (int isol = 0; isol < NSol; isol++) {
141  TopRec(LV_l, LV_l_, LV_b, LV_b_, q_sol[isol]);
142  double weight = WeightSolfromShape();
143  if (weight > weightmax) {
144  weightmax = weight;
145  maxLV_n.SetPxPyPzE(LV_n.Px(), LV_n.Py(), LV_n.Pz(), LV_n.E());
146  maxLV_n_.SetPxPyPzE(LV_n_.Px(), LV_n_.Py(), LV_n_.Pz(), LV_n_.E());
147  }
148  }
149  }
151  nuSol.neutrino = reco::LeafCandidate(0, maxLV_n);
152  nuSol.neutrinoBar = reco::LeafCandidate(0, maxLV_n_);
153  nuSol.weight = weightmax;
154  return nuSol;
155 }
156 
157 void TtFullLepKinSolver::FindCoeff(const TLorentzVector& al,
158  const TLorentzVector& l,
159  const TLorentzVector& b_al,
160  const TLorentzVector& b_l,
161  const double mt,
162  const double mat,
163  const double px_miss,
164  const double py_miss,
165  double* koeficienty) {
166  double E, apom1, apom2, apom3;
167  double k11, k21, k31, k41, cpom1, cpom2, cpom3, l11, l21, l31, l41, l51, l61, k1, k2, k3, k4, k5, k6;
168  double l1, l2, l3, l4, l5, l6, k15, k25, k35, k45;
169 
170  C = -al.Px() - b_al.Px() - l.Px() - b_l.Px() + px_miss;
171  D = -al.Py() - b_al.Py() - l.Py() - b_l.Py() + py_miss;
172 
173  // right side of first two linear equations - missing pT
174 
175  E = (sqr(mt) - sqr(mw) - sqr(mb)) / (2 * b_al.E()) - sqr(mw) / (2 * al.E()) - al.E() +
176  al.Px() * b_al.Px() / b_al.E() + al.Py() * b_al.Py() / b_al.E() + al.Pz() * b_al.Pz() / b_al.E();
177  F = (sqr(mat) - sqr(mw) - sqr(mb)) / (2 * b_l.E()) - sqr(mw) / (2 * l.E()) - l.E() + l.Px() * b_l.Px() / b_l.E() +
178  l.Py() * b_l.Py() / b_l.E() + l.Pz() * b_l.Pz() / b_l.E();
179 
180  m1 = al.Px() / al.E() - b_al.Px() / b_al.E();
181  m2 = al.Py() / al.E() - b_al.Py() / b_al.E();
182  m3 = al.Pz() / al.E() - b_al.Pz() / b_al.E();
183 
184  n1 = l.Px() / l.E() - b_l.Px() / b_l.E();
185  n2 = l.Py() / l.E() - b_l.Py() / b_l.E();
186  n3 = l.Pz() / l.E() - b_l.Pz() / b_l.E();
187 
188  pom = E - m1 * C - m2 * D;
189  apom1 = sqr(al.Px()) - sqr(al.E());
190  apom2 = sqr(al.Py()) - sqr(al.E());
191  apom3 = sqr(al.Pz()) - sqr(al.E());
192 
193  k11 = 1 / sqr(al.E()) *
194  (pow(mw, 4) / 4 + sqr(C) * apom1 + sqr(D) * apom2 + apom3 * sqr(pom) / sqr(m3) +
195  sqr(mw) * (al.Px() * C + al.Py() * D + al.Pz() * pom / m3) + 2 * al.Px() * al.Py() * C * D +
196  2 * al.Px() * al.Pz() * C * pom / m3 + 2 * al.Py() * al.Pz() * D * pom / m3);
197  k21 = 1 / sqr(al.E()) *
198  (-2 * C * m3 * n3 * apom1 + 2 * apom3 * n3 * m1 * pom / m3 - sqr(mw) * m3 * n3 * al.Px() +
199  sqr(mw) * m1 * n3 * al.Pz() - 2 * al.Px() * al.Py() * D * m3 * n3 + 2 * al.Px() * al.Pz() * C * m1 * n3 -
200  2 * al.Px() * al.Pz() * n3 * pom + 2 * al.Py() * al.Pz() * D * m1 * n3);
201  k31 = 1 / sqr(al.E()) *
202  (-2 * D * m3 * n3 * apom2 + 2 * apom3 * n3 * m2 * pom / m3 - sqr(mw) * m3 * n3 * al.Py() +
203  sqr(mw) * m2 * n3 * al.Pz() - 2 * al.Px() * al.Py() * C * m3 * n3 + 2 * al.Px() * al.Pz() * C * m2 * n3 -
204  2 * al.Py() * al.Pz() * n3 * pom + 2 * al.Py() * al.Pz() * D * m2 * n3);
205  k41 = 1 / sqr(al.E()) *
206  (2 * apom3 * m1 * m2 * sqr(n3) + 2 * al.Px() * al.Py() * sqr(m3) * sqr(n3) -
207  2 * al.Px() * al.Pz() * m2 * m3 * sqr(n3) - 2 * al.Py() * al.Pz() * m1 * m3 * sqr(n3));
208  k51 = 1 / sqr(al.E()) *
209  (apom1 * sqr(m3) * sqr(n3) + apom3 * sqr(m1) * sqr(n3) - 2 * al.Px() * al.Pz() * m1 * m3 * sqr(n3));
210  k61 = 1 / sqr(al.E()) *
211  (apom2 * sqr(m3) * sqr(n3) + apom3 * sqr(m2) * sqr(n3) - 2 * al.Py() * al.Pz() * m2 * m3 * sqr(n3));
212 
213  cpom1 = sqr(l.Px()) - sqr(l.E());
214  cpom2 = sqr(l.Py()) - sqr(l.E());
215  cpom3 = sqr(l.Pz()) - sqr(l.E());
216 
217  l11 = 1 / sqr(l.E()) * (pow(mw, 4) / 4 + cpom3 * sqr(F) / sqr(n3) + sqr(mw) * l.Pz() * F / n3);
218  l21 =
219  1 / sqr(l.E()) *
220  (-2 * cpom3 * F * m3 * n1 / n3 + sqr(mw) * (l.Px() * m3 * n3 - l.Pz() * n1 * m3) + 2 * l.Px() * l.Pz() * F * m3);
221  l31 =
222  1 / sqr(l.E()) *
223  (-2 * cpom3 * F * m3 * n2 / n3 + sqr(mw) * (l.Py() * m3 * n3 - l.Pz() * n2 * m3) + 2 * l.Py() * l.Pz() * F * m3);
224  l41 = 1 / sqr(l.E()) *
225  (2 * cpom3 * n1 * n2 * sqr(m3) + 2 * l.Px() * l.Py() * sqr(m3) * sqr(n3) -
226  2 * l.Px() * l.Pz() * n2 * n3 * sqr(m3) - 2 * l.Py() * l.Pz() * n1 * n3 * sqr(m3));
227  l51 = 1 / sqr(l.E()) *
228  (cpom1 * sqr(m3) * sqr(n3) + cpom3 * sqr(n1) * sqr(m3) - 2 * l.Px() * l.Pz() * n1 * n3 * sqr(m3));
229  l61 = 1 / sqr(l.E()) *
230  (cpom2 * sqr(m3) * sqr(n3) + cpom3 * sqr(n2) * sqr(m3) - 2 * l.Py() * l.Pz() * n2 * n3 * sqr(m3));
231 
232  k1 = k11 * k61;
233  k2 = k61 * k21 / k51;
234  k3 = k31;
235  k4 = k41 / k51;
236  k5 = k61 / k51;
237  k6 = 1;
238 
239  l1 = l11 * k61;
240  l2 = l21 * k61 / k51;
241  l3 = l31;
242  l4 = l41 / k51;
243  l5 = l51 * k61 / (sqr(k51));
244  l6 = l61 / k61;
245 
246  k15 = k1 * l5 - l1 * k5;
247  k25 = k2 * l5 - l2 * k5;
248  k35 = k3 * l5 - l3 * k5;
249  k45 = k4 * l5 - l4 * k5;
250 
251  k16 = k1 * l6 - l1 * k6;
252  k26 = k2 * l6 - l2 * k6;
253  k36 = k3 * l6 - l3 * k6;
254  k46 = k4 * l6 - l4 * k6;
255  k56 = k5 * l6 - l5 * k6;
256 
257  koeficienty[0] = k15 * sqr(k36) - k35 * k36 * k16 - k56 * sqr(k16);
258  koeficienty[1] =
259  2 * k15 * k36 * k46 + k25 * sqr(k36) + k35 * (-k46 * k16 - k36 * k26) - k45 * k36 * k16 - 2 * k56 * k26 * k16;
260  koeficienty[2] = k15 * sqr(k46) + 2 * k25 * k36 * k46 + k35 * (-k46 * k26 - k36 * k56) -
261  k56 * (sqr(k26) + 2 * k56 * k16) - k45 * (k46 * k16 + k36 * k26);
262  koeficienty[3] = k25 * sqr(k46) - k35 * k46 * k56 - k45 * (k46 * k26 + k36 * k56) - 2 * sqr(k56) * k26;
263  koeficienty[4] = -k45 * k46 * k56 - pow(k56, 3);
264 
265  // normalization of coefficients
266  int moc = (int(log10(fabs(koeficienty[0]))) + int(log10(fabs(koeficienty[4])))) / 2;
267 
268  koeficienty[0] = koeficienty[0] / TMath::Power(10, moc);
269  koeficienty[1] = koeficienty[1] / TMath::Power(10, moc);
270  koeficienty[2] = koeficienty[2] / TMath::Power(10, moc);
271  koeficienty[3] = koeficienty[3] / TMath::Power(10, moc);
272  koeficienty[4] = koeficienty[4] / TMath::Power(10, moc);
273 }
274 
275 void TtFullLepKinSolver::TopRec(const TLorentzVector& al,
276  const TLorentzVector& l,
277  const TLorentzVector& b_al,
278  const TLorentzVector& b_l,
279  const double sol) {
280  TVector3 t_ttboost;
281  TLorentzVector aux;
282  double pxp, pyp, pzp, pup, pvp, pwp;
283 
284  pxp = sol * (m3 * n3 / k51);
285  pyp = -(m3 * n3 / k61) * (k56 * pow(sol, 2) + k26 * sol + k16) / (k36 + k46 * sol);
286  pzp = -1 / n3 * (n1 * pxp + n2 * pyp - F);
287  pwp = 1 / m3 * (m1 * pxp + m2 * pyp + pom);
288  pup = C - pxp;
289  pvp = D - pyp;
290 
291  LV_n_.SetXYZM(pxp, pyp, pzp, 0.0);
292  LV_n.SetXYZM(pup, pvp, pwp, 0.0);
293 
294  LV_t_ = b_l + l + LV_n_;
295  LV_t = b_al + al + LV_n;
296 
297  aux = (LV_t_ + LV_t);
298  t_ttboost = -aux.BoostVector();
299  LV_tt_t_ = LV_t_;
300  LV_tt_t = LV_t;
301  LV_tt_t_.Boost(t_ttboost);
302  LV_tt_t.Boost(t_ttboost);
303 }
304 
306  double weight = 1;
307  weight = ((LV_n.E() > genLV_n.E()) ? genLV_n.E() / LV_n.E() : LV_n.E() / genLV_n.E()) *
308  ((LV_n_.E() > genLV_n_.E()) ? genLV_n_.E() / LV_n_.E() : LV_n_.E() / genLV_n_.E());
309  return weight;
310 }
311 
312 double TtFullLepKinSolver::WeightSolfromShape() const { return EventShape_->Eval(LV_n.E(), LV_n_.E()); }
313 
314 int TtFullLepKinSolver::quartic(double* koeficienty, double* koreny) const {
315  double w, b0, b1, b2;
316  double c[4];
317  double d0, d1, h, t, z;
318  double* px;
319 
320  if (koeficienty[4] == 0.0)
321  return cubic(koeficienty, koreny);
322  /* quartic problem? */
323  w = koeficienty[3] / (4 * koeficienty[4]);
324  /* offset */
325  b2 = -6 * sqr(w) + koeficienty[2] / koeficienty[4];
326  /* koeficienty. of shifted polynomial */
327  b1 = (8 * sqr(w) - 2 * koeficienty[2] / koeficienty[4]) * w + koeficienty[1] / koeficienty[4];
328  b0 = ((-3 * sqr(w) + koeficienty[2] / koeficienty[4]) * w - koeficienty[1] / koeficienty[4]) * w +
329  koeficienty[0] / koeficienty[4];
330 
331  c[3] = 1.0;
332  /* cubic resolvent */
333  c[2] = b2;
334  c[1] = -4 * b0;
335  c[0] = sqr(b1) - 4 * b0 * b2;
336 
337  cubic(c, koreny);
338  z = koreny[0];
339  //double z1=1.0,z2=2.0,z3=3.0;
340  //TMath::RootsCubic(c,z1,z2,z3);
341  //if (z2 !=0) z = z2;
342  //if (z1 !=0) z = z1;
343  /* only lowermost root needed */
344 
345  int nreal = 0;
346  px = koreny;
347  t = sqrt(0.25 * sqr(z) - b0);
348  for (int i = -1; i <= 1; i += 2) {
349  d0 = -0.5 * z + i * t;
350  /* coeffs. of quadratic factor */
351  d1 = (t != 0.0) ? -i * 0.5 * b1 / t : i * sqrt(-z - b2);
352  h = 0.25 * sqr(d1) - d0;
353  if (h >= 0.0) {
354  h = sqrt(h);
355  nreal += 2;
356  *px++ = -0.5 * d1 - h - w;
357  *px++ = -0.5 * d1 + h - w;
358  }
359  }
360 
361  // if (nreal==4) {
362  /* sort results */
363  // if (koreny[2]<koreny[0]) SWAP(koreny[0], koreny[2]);
364  // if (koreny[3]<koreny[1]) SWAP(koreny[1], koreny[3]);
365  // if (koreny[1]<koreny[0]) SWAP(koreny[0], koreny[1]);
366  // if (koreny[3]<koreny[2]) SWAP(koreny[2], koreny[3]);
367  // if (koreny[2]<koreny[1]) SWAP(koreny[1], koreny[2]);
368  // }
369  return nreal;
370 }
371 
372 int TtFullLepKinSolver::cubic(const double* coeffs, double* koreny) const {
373  unsigned nreal;
374  double w, p, q, dis, h, phi;
375 
376  if (coeffs[3] != 0.0) {
377  /* cubic problem? */
378  w = coeffs[2] / (3 * coeffs[3]);
379  p = sqr(coeffs[1] / (3 * coeffs[3]) - sqr(w)) * (coeffs[1] / (3 * coeffs[3]) - sqr(w));
380  q = -0.5 * (2 * sqr(w) * w - (coeffs[1] * w - coeffs[0]) / coeffs[3]);
381  dis = sqr(q) + p;
382  /* discriminant */
383  if (dis < 0.0) {
384  /* 3 real solutions */
385  h = q / sqrt(-p);
386  if (h > 1.0)
387  h = 1.0;
388  /* confine the argument of */
389  if (h < -1.0)
390  h = -1.0;
391  /* acos to [-1;+1] */
392  phi = acos(h);
393  p = 2 * TMath::Power(-p, 1.0 / 6.0);
394  for (unsigned i = 0; i < 3; i++)
395  koreny[i] = p * cos((phi + 2 * i * TMath::Pi()) / 3.0) - w;
396  if (koreny[1] < koreny[0])
397  SWAP(koreny[0], koreny[1]);
398  /* sort results */
399  if (koreny[2] < koreny[1])
400  SWAP(koreny[1], koreny[2]);
401  if (koreny[1] < koreny[0])
402  SWAP(koreny[0], koreny[1]);
403  nreal = 3;
404  } else {
405  /* only one real solution */
406  dis = sqrt(dis);
407  h = TMath::Power(fabs(q + dis), 1.0 / 3.0);
408  p = TMath::Power(fabs(q - dis), 1.0 / 3.0);
409  koreny[0] = ((q + dis > 0.0) ? h : -h) + ((q - dis > 0.0) ? p : -p) - w;
410  nreal = 1;
411  }
412 
413  /* Perform one step of a Newton iteration in order to minimize
414  round-off errors */
415  for (unsigned i = 0; i < nreal; i++) {
416  h = coeffs[1] + koreny[i] * (2 * coeffs[2] + 3 * koreny[i] * coeffs[3]);
417  if (h != 0.0)
418  koreny[i] -= (coeffs[0] + koreny[i] * (coeffs[1] + koreny[i] * (coeffs[2] + koreny[i] * coeffs[3]))) / h;
419  }
420  }
421 
422  else if (coeffs[2] != 0.0) {
423  /* quadratic problem? */
424  p = 0.5 * coeffs[1] / coeffs[2];
425  dis = sqr(p) - coeffs[0] / coeffs[2];
426  if (dis >= 0.0) {
427  /* two real solutions */
428  dis = sqrt(dis);
429  koreny[0] = -p - dis;
430  koreny[1] = -p + dis;
431  nreal = 2;
432  } else
433  /* no real solution */
434  nreal = 0;
435  }
436 
437  else if (coeffs[1] != 0.0) {
438  /* linear problem? */
439  koreny[0] = -coeffs[0] / coeffs[1];
440  nreal = 1;
441  }
442 
443  else
444  /* no equation */
445  nreal = 0;
446 
447  return nreal;
448 }
449 
450 void TtFullLepKinSolver::SWAP(double& realone, double& realtwo) const {
451  if (realtwo < realone) {
452  double aux = realtwo;
453  realtwo = realone;
454  realone = aux;
455  }
456 }
mps_fire.i
i
Definition: mps_fire.py:428
TtFullLepKinSolver::~TtFullLepKinSolver
~TtFullLepKinSolver()
destructor
Definition: TtFullLepKinSolver.cc:22
TtFullLepKinSolver::LV_n
TLorentzVector LV_n
Definition: TtFullLepKinSolver.h:117
TtFullLepKinSolver::topmass_begin
const double topmass_begin
Definition: TtFullLepKinSolver.h:86
TtFullLepKinSolver::sqr
double sqr(const double x) const
Definition: TtFullLepKinSolver.h:80
TtFullLepKinSolver::m2
double m2
Definition: TtFullLepKinSolver.h:110
TtFullLepKinSolver::k46
double k46
Definition: TtFullLepKinSolver.h:105
TtDilepEvtSolution::setRecTopMass
void setRecTopMass(double mass)
Definition: TtDilepEvtSolution.h:205
TtFullLepKinSolver::mb
const double mb
Definition: TtFullLepKinSolver.h:94
mps_merge.weight
weight
Definition: mps_merge.py:88
AlCaHLTBitMon_ParallelJobs.p
p
Definition: AlCaHLTBitMon_ParallelJobs.py:153
h
FWCore Framework interface EventSetupRecordImplementation h
Helper function to determine trigger accepts.
Definition: L1TUtmAlgorithmRcd.h:4
TtFullLepKinSolver::addKinSolInfo
TtDilepEvtSolution addKinSolInfo(TtDilepEvtSolution *asol)
Definition: TtFullLepKinSolver.cc:24
TtFullLepKinSolver::pymiss_
double pymiss_
Definition: TtFullLepKinSolver.h:96
TtDilepEvtSolution::getElectronm
pat::Electron getElectronm() const
Definition: TtDilepEvtSolution.h:41
TtFullLepKinSolver::k61
double k61
Definition: TtFullLepKinSolver.h:108
TtDilepEvtSolution::getCalJetB
pat::Jet getCalJetB() const
Definition: TtDilepEvtSolution.h:117
TtFullLepKinSolver::SetConstraints
void SetConstraints(const double xx=0, const double yy=0)
Definition: TtFullLepKinSolver.cc:120
TtFullLepKinSolver::C
double C
Definition: TtFullLepKinSolver.h:98
TtFullLepKinSolver::NeutrinoSolution::neutrino
reco::LeafCandidate neutrino
Definition: TtFullLepKinSolver.h:30
TtDilepEvtSolution::getCalJetBbar
pat::Jet getCalJetBbar() const
Definition: TtDilepEvtSolution.h:119
b2
static constexpr float b2
Definition: L1EGammaCrystalsEmulatorProducer.cc:82
TtDilepEvtSolution::getTaum
pat::Tau getTaum() const
Definition: TtDilepEvtSolution.h:45
TtFullLepKinSolver::useMCforBest_
bool useMCforBest_
flag to swith from WeightSolfromMC() to WeightSolfromShape()
Definition: TtFullLepKinSolver.h:122
TtFullLepKinSolver::genLV_n_
TLorentzVector genLV_n_
Definition: TtFullLepKinSolver.h:119
TtFullLepKinSolver::F
double F
Definition: TtFullLepKinSolver.h:100
TtFullLepKinSolver::D
double D
Definition: TtFullLepKinSolver.h:99
TtDilepEvtSolution::getWpDecay
std::string getWpDecay() const
Definition: TtDilepEvtSolution.h:124
b1
static constexpr float b1
Definition: L1EGammaCrystalsEmulatorProducer.cc:82
alignCSCRings.s
s
Definition: alignCSCRings.py:92
TtFullLepKinSolver::k16
double k16
Definition: TtFullLepKinSolver.h:102
TtFullLepKinSolver.h
TtFullLepKinSolver::NeutrinoSolution::weight
double weight
Definition: TtFullLepKinSolver.h:29
TtFullLepKinSolver::topmass_end
const double topmass_end
Definition: TtFullLepKinSolver.h:88
funct::cos
Cos< T >::type cos(const T &t)
Definition: Cos.h:22
TtDilepEvtSolution::getMuonp
pat::Muon getMuonp() const
Definition: TtDilepEvtSolution.h:42
TtFullLepKinSolver::FindCoeff
void FindCoeff(const TLorentzVector &al, const TLorentzVector &l, const TLorentzVector &b_al, const TLorentzVector &b_l, const double mt, const double mat, const double pxboost, const double pyboost, double *q_coeff)
Definition: TtFullLepKinSolver.cc:157
TtFullLepKinSolver::pxmiss_
double pxmiss_
Definition: TtFullLepKinSolver.h:96
TtFullLepKinSolver::getNuSolution
NeutrinoSolution getNuSolution(const TLorentzVector &LV_l, const TLorentzVector &LV_l_, const TLorentzVector &LV_b, const TLorentzVector &LV_b_)
Definition: TtFullLepKinSolver.cc:125
TtDilepEvtSolution::getGenNbar
const reco::GenParticle * getGenNbar() const
Definition: TtDilepEvtSolution.h:106
reco::LeafCandidate::py
double py() const final
y coordinate of momentum vector
Definition: LeafCandidate.h:142
w
const double w
Definition: UKUtility.cc:23
TtFullLepKinSolver::NeutrinoSolution
Definition: TtFullLepKinSolver.h:28
mathSSE::sqrt
T sqrt(T t)
Definition: SSEVec.h:19
TtFullLepKinSolver::LV_t
TLorentzVector LV_t
Definition: TtFullLepKinSolver.h:117
DDAxes::z
TtFullLepKinSolver::m1
double m1
Definition: TtFullLepKinSolver.h:109
h
TtFullLepKinSolver::mw
const double mw
Definition: TtFullLepKinSolver.h:92
b
double b
Definition: hdecay.h:118
tools.TF2
TF2
Definition: tools.py:24
TtFullLepKinSolver::k56
double k56
Definition: TtFullLepKinSolver.h:106
TtDilepEvtSolution::getElectronp
pat::Electron getElectronp() const
Definition: TtDilepEvtSolution.h:40
TtFullLepKinSolver::WeightSolfromMC
double WeightSolfromMC() const
Definition: TtFullLepKinSolver.cc:305
TtSemiLepHitFitProducer_Electrons_cfi.mW
mW
Definition: TtSemiLepHitFitProducer_Electrons_cfi.py:35
TtDilepEvtSolution::setRecWeightMax
void setRecWeightMax(double wgt)
Definition: TtDilepEvtSolution.h:206
TtFullLepKinSolver::LV_n_
TLorentzVector LV_n_
Definition: TtFullLepKinSolver.h:117
geometryCSVtoXML.yy
yy
Definition: geometryCSVtoXML.py:19
TtFullLepKinSolver::quartic
int quartic(double *q_coeff, double *q_sol) const
Definition: TtFullLepKinSolver.cc:314
TtDilepEvtSolution::getTaup
pat::Tau getTaup() const
Definition: TtDilepEvtSolution.h:44
TtFullLepKinSolver::TtFullLepKinSolver
TtFullLepKinSolver()
default constructor
Definition: TtFullLepKinSolver.cc:4
TtFullLepKinSolver::NeutrinoSolution::neutrinoBar
reco::LeafCandidate neutrinoBar
Definition: TtFullLepKinSolver.h:31
TtFullLepKinSolver::LV_tt_t
TLorentzVector LV_tt_t
Definition: TtFullLepKinSolver.h:117
printConversionInfo.aux
aux
Definition: printConversionInfo.py:19
createfilelist.int
int
Definition: createfilelist.py:10
TtFullLepKinSolver::TopRec
void TopRec(const TLorentzVector &al, const TLorentzVector &l, const TLorentzVector &b_al, const TLorentzVector &b_l, const double sol)
Definition: TtFullLepKinSolver.cc:275
TtDilepEvtSolution::getWmDecay
std::string getWmDecay() const
Definition: TtDilepEvtSolution.h:125
TtFullLepKinSolver::cubic
int cubic(const double *c_coeff, double *c_sol) const
Definition: TtFullLepKinSolver.cc:372
submitPVResolutionJobs.q
q
Definition: submitPVResolutionJobs.py:84
b0
static constexpr float b0
Definition: L1EGammaCrystalsEmulatorProducer.cc:82
TtFullLepKinSolver::SWAP
void SWAP(double &realone, double &realtwo) const
Definition: TtFullLepKinSolver.cc:450
TtFullLepKinSolver::n3
double n3
Definition: TtFullLepKinSolver.h:114
HltBtagPostValidation_cff.c
c
Definition: HltBtagPostValidation_cff.py:31
TtFullLepKinSolver::n2
double n2
Definition: TtFullLepKinSolver.h:113
cmsLHEtoEOSManager.l
l
Definition: cmsLHEtoEOSManager.py:204
DDAxes::phi
multPhiCorr_741_25nsDY_cfi.px
px
Definition: multPhiCorr_741_25nsDY_cfi.py:10
TtFullLepKinSolver::LV_tt_t_
TLorentzVector LV_tt_t_
Definition: TtFullLepKinSolver.h:117
TtFullLepKinSolver::genLV_n
TLorentzVector genLV_n
provisional
Definition: TtFullLepKinSolver.h:119
math::XYZTLorentzVector
XYZTLorentzVectorD XYZTLorentzVector
Lorentz vector with cylindrical internal representation using pseudorapidity.
Definition: LorentzVector.h:29
TtDilepEvtSolution::getMuonm
pat::Muon getMuonm() const
Definition: TtDilepEvtSolution.h:43
TtFullLepKinSolver::topmass_step
const double topmass_step
Definition: TtFullLepKinSolver.h:90
TtSemiLepEvtBuilder_cfi.mt
mt
Definition: TtSemiLepEvtBuilder_cfi.py:47
TtFullLepKinSolver::LV_t_
TLorentzVector LV_t_
Definition: TtFullLepKinSolver.h:117
TtFullLepKinSolver::k51
double k51
Definition: TtFullLepKinSolver.h:107
reco::LeafCandidate::energy
double energy() const final
energy
Definition: LeafCandidate.h:125
TtFullLepKinSolver::m3
double m3
Definition: TtFullLepKinSolver.h:111
Pi
const double Pi
Definition: CosmicMuonParameters.h:18
funct::pow
Power< A, B >::type pow(const A &a, const B &b)
Definition: Power.h:29
reco::LeafCandidate
Definition: LeafCandidate.h:16
TtFullLepKinSolver::n1
double n1
Definition: TtFullLepKinSolver.h:112
TtFullLepKinSolver::k26
double k26
Definition: TtFullLepKinSolver.h:103
submitPVValidationJobs.t
string t
Definition: submitPVValidationJobs.py:644
d0
static constexpr float d0
Definition: L1EGammaCrystalsEmulatorProducer.cc:84
TtFullLepKinSolver::k36
double k36
Definition: TtFullLepKinSolver.h:104
TtFullLepKinSolver::WeightSolfromShape
double WeightSolfromShape() const
use the parametrized event shape to obtain the solution weight.
Definition: TtFullLepKinSolver.cc:312
reco::LeafCandidate::px
double px() const final
x coordinate of momentum vector
Definition: LeafCandidate.h:140
TtFullLepKinSolver::pom
double pom
Definition: TtFullLepKinSolver.h:101
reco::LeafCandidate::pz
double pz() const final
z coordinate of momentum vector
Definition: LeafCandidate.h:144
TtFullLepKinSolver::EventShape_
TF2 * EventShape_
Event shape.
Definition: TtFullLepKinSolver.h:124
TtDilepEvtSolution::getGenN
const reco::GenParticle * getGenN() const
Definition: TtDilepEvtSolution.h:76
d1
static constexpr float d1
Definition: L1EGammaCrystalsEmulatorProducer.cc:84
geometryCSVtoXML.xx
xx
Definition: geometryCSVtoXML.py:19
weight
Definition: weight.py:1
TtDilepEvtSolution
Definition: TtDilepEvtSolution.h:25
MillePedeFileConverter_cfg.e
e
Definition: MillePedeFileConverter_cfg.py:37