Skip to content
Snippets Groups Projects
Commit 3add6bfd authored by Pyry Runko (Puck)'s avatar Pyry Runko (Puck)
Browse files

commit before editing to check ComplexAmplitude sum

parent eb868b1c
No related branches found
No related tags found
No related merge requests found
......@@ -601,13 +601,11 @@ std::complex<double> IPGlasma::SumAmplitudes()
for(int j =0; j < L; j += 1){
int q1[2] = {i,j};
// WilsonLine quark = GetWilsonLine(WilsonLineCoordinate(i, j));
WilsonLine quark = GetWilsonLine(WilsonLineCoordinate(i, j));
// double r1 = DistanceToOrigin(q1);
for(int v =0; v < L; v +=1){
for(int w =0; w < 100; w +=1){
double r = w*lattice_spacing;
if (v>0 || j>0 || i>0) {break;}
// check that the dipole size is not zero
if (i == v && j == w) { continue;}
......@@ -625,21 +623,19 @@ std::complex<double> IPGlasma::SumAmplitudes()
//TODO: is this correct?
//TODO: remove case where r=0
// dipole size for wave function
//double r = sqrt(pow(X(i)-X(v),2) + pow(Y(j)-Y(w),2));
double r = sqrt(pow(X(i)-X(v),2) + pow(Y(j)-Y(w),2));
//cout << "# r in WF is " << r << endl;
// wave functions
double WF_u = WFGammaqqbarLongitudinal(r, z0, Q, u_quark);
//cout << "# WF_u is " << WF_u << endl;
cout << r << " " << WF_u << endl << std::flush;
continue;
double WF_d = WFGammaqqbarLongitudinal(r, z0, Q, d_quark);
double WF_c = WFGammaqqbarLongitudinal(r, z0, Q, c_quark);
// sum += (WF_u + WF_d + WF_c) * ComplexAmplitude(0, quark, q2)
// * exp( -i * ( std::inner_product(std::begin(vecL), std::end(vecL), std::begin(diffX), 0)
// + std::inner_product(std::begin(vecDelta), std::end(vecDelta), std::begin(sumZX), 0) ) );
sum += (WF_u + WF_d + WF_c) * ComplexAmplitude(0, quark, q2)
* exp( -i * ( std::inner_product(std::begin(vecL), std::end(vecL), std::begin(diffX), 0)
+ std::inner_product(std::begin(vecDelta), std::end(vecDelta), std::begin(sumZX), 0) ) );
//TODO: Finish eq (12) from notes
// Finish amplitude
......@@ -672,8 +668,7 @@ double IPGlasma::WFGammaqqbarLongitudinal(double r, double z, double Q, const qu
// fine structure constant set to 1/137
double e = sqrt(4.*M_PI/137.036);
//double epsilon = sqrt(z*(1-z)*Q*Q + pow(quark.mass, 2));
double epsilon = 1;
double epsilon = sqrt(z*(1-z)*Q*Q + pow(quark.mass, 2));
return (quark.e_charge * e * sqrt(NC) * 2*Q*Q*z*(1-z) * std::cyl_bessel_k(0, epsilon*r)*0.5/M_PI);
}
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment