diff --git a/examples/SedovBlast_2D/makeIC.py b/examples/SedovBlast_2D/makeIC.py
index dc2c574bb76387ef3bebd714423e4612e5838e0d..a30e727b4dd42c1f058827959cf12e3b4f152181 100644
--- a/examples/SedovBlast_2D/makeIC.py
+++ b/examples/SedovBlast_2D/makeIC.py
@@ -31,8 +31,6 @@ E0= 1.             # Energy of the explosion
 N_inject = 21      # Number of particles in which to inject energy
 fileName = "sedov.hdf5" 
 
-#L = 101
-
 #---------------------------------------------------
 glass = h5py.File("glassPlane_128.hdf5", "r")
 
@@ -50,10 +48,9 @@ m = zeros(numPart)
 u = zeros(numPart)
 r = zeros(numPart)
 
-for i in range(numPart):
-        r[i] = sqrt((pos[i,0] - 0.5)**2 + (pos[i,1] - 0.5)**2)
-        m[i] = rho0 * vol / numPart    
-        u[i] = P0 / (rho0 * (gamma - 1))
+r = sqrt((pos[:,0] - 0.5)**2 + (pos[:,1] - 0.5)**2)
+m[:] = rho0 * vol / numPart    
+u[:] = P0 / (rho0 * (gamma - 1))
 
 # Make the central particle detonate
 index = argsort(r)
diff --git a/examples/SedovBlast_2D/plotSolution.py b/examples/SedovBlast_2D/plotSolution.py
index af7c3b525dc25b2b10399f9cf27553999b28ab92..69e4e1232dd5c14b06e8a705f4add391f1f597f0 100644
--- a/examples/SedovBlast_2D/plotSolution.py
+++ b/examples/SedovBlast_2D/plotSolution.py
@@ -18,7 +18,7 @@
  # 
  ##############################################################################
 
-# Computes the analytical solution of the 3D Sedov blast wave.
+# Computes the analytical solution of the 2D Sedov blast wave.
 # The script works for a given initial box and dumped energy and computes the solution at a later time t.
 
 # Parameters
@@ -85,6 +85,8 @@ P = sim["/PartType0/Pressure"][:]
 rho = sim["/PartType0/Density"][:]
 
 
+# Now, work our the solution....
+
 from scipy.special import gamma as Gamma
 from numpy import *
 
@@ -190,6 +192,8 @@ def sedov(t, E0, rho0, g, n=1000, nu=3):
     rho *= rho0
     return r, p, rho, u, r_s, p_s, rho_s, u_s, shock_speed
 
+
+# The main properties of the solution
 r_s, P_s, rho_s, v_s, r_shock, _, _, _, _ = sedov(time, E_0, rho_0, gas_gamma, 1000, 2)
 
 # Append points for after the shock
@@ -202,6 +206,8 @@ v_s = np.insert(v_s, np.size(v_s), [0, 0])
 u_s = P_s / (rho_s * (gas_gamma - 1.))  #internal energy
 s_s = P_s / rho_s**gas_gamma # entropic function
 
+
+
 # Plot the interesting quantities
 figure()
 
diff --git a/examples/SedovBlast_2D/sedov.py b/examples/SedovBlast_2D/sedov.py
deleted file mode 100755
index 2439a7fda91831a7c54350a597db5f026e36fee9..0000000000000000000000000000000000000000
--- a/examples/SedovBlast_2D/sedov.py
+++ /dev/null
@@ -1,212 +0,0 @@
-###############################################################################
- # This file is part of SWIFT.
- # Copyright (c) 2015 Bert Vandenbroucke (bert.vandenbroucke@ugent.be)
- # 
- # This program is free software: you can redistribute it and/or modify
- # it under the terms of the GNU Lesser General Public License as published
- # by the Free Software Foundation, either version 3 of the License, or
- # (at your option) any later version.
- # 
- # This program is distributed in the hope that it will be useful,
- # but WITHOUT ANY WARRANTY; without even the implied warranty of
- # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
- # GNU General Public License for more details.
- # 
- # You should have received a copy of the GNU Lesser General Public License
- # along with this program.  If not, see <http://www.gnu.org/licenses/>.
- # 
- ##############################################################################
-
-import numpy as np
-import scipy.integrate as integrate
-import scipy.optimize as optimize
-import os
-
-# Calculate the analytical solution of the Sedov-Taylor shock wave for a given
-# number of dimensions, gamma and time. We assume dimensionless units and a
-# setup with constant density 1, pressure and velocity 0. An energy 1 is
-# inserted in the center at t 0.
-#
-# The solution is a self-similar shock wave, which was described in detail by
-# Sedov (1959). We follow his notations and solution method.
-#
-# The position of the shock at time t is given by
-#  r2 = (E/rho1)^(1/(2+nu)) * t^(2/(2+nu))
-# the energy E is related to the inserted energy E0 by E0 = alpha*E, with alpha
-# a constant which has to be calculated by imposing energy conservation.
-#
-# The density for a given radius at a certain time is determined by introducing
-# the dimensionless position coordinate lambda = r/r2. The density profile as
-# a function of lambda is constant in time and given by
-#  rho = rho1 * R(V, gamma, nu)
-# and
-#  V = V(lambda, gamma, nu)
-#
-# The function V(lambda, gamma, nu) is found by solving a differential equation
-# described in detail in Sedov (1959) chapter 4, section 5. Alpha is calculated
-# from the integrals in section 11 of the same chapter.
-#
-# Numerically, the complete solution requires the use of 3 quadratures and 1
-# root solver, which are implemented using the GNU Scientific Library (GSL).
-# Since some quadratures call functions that themselves contain quadratures,
-# the problem is highly non-linear and complex and takes a while to converge.
-# Therefore, we tabulate the alpha values and profile values the first time
-# a given set of gamma and nu values is requested and reuse these tabulated
-# values.
-#
-# Reference:
-#  Sedov (1959): Sedov, L., Similitude et dimensions en mecanique (7th ed.;
-#                Moscou: Editions Mir) - french translation of the original
-#                book from 1959.
-
-# dimensionless variable z = gamma*P/R as a function of V (and gamma and nu)
-# R is a dimensionless density, while P is a dimensionless pressure
-# z is hence a sort of dimensionless sound speed
-# The expression below corresponds to eq. 11.9 in Sedov (1959), chapter 4
-def _z(V, gamma, nu):
-    if V == 2./(nu+2.)/gamma:
-        return 0.
-    else:
-        return (gamma-1.)*V*V*(V-2./(2.+nu))/2./(2./(2.+nu)/gamma-V)
-
-# differential equation that needs to be solved to obtain lambda for a given V
-# corresponds to eq. 5.11 in Sedov (1959), chapter 4 (omega = 0)
-def _dlnlambda_dV(V, gamma, nu):
-    nom = _z(V, gamma, nu) - (V-2./(nu+2.))*(V-2./(nu+2.))
-    denom = V*(V-1.)*(V-2./(nu+2.))+nu*(2./(nu+2.)/gamma-V)*_z(V, gamma, nu)
-    return nom/denom
-
-# dimensionless variable lambda = r/r2 as a function of V (and gamma and nu)
-# found by solving differential equation 5.11 in Sedov (1959), chapter 4
-# (omega = 0)
-def _lambda(V, gamma, nu):
-    if V == 2./(nu+2.)/gamma:
-        return 0.
-    else:
-        V0 = 4./(nu+2.)/(gamma+1.)
-        integral, err = integrate.quad(_dlnlambda_dV, V, V0, (gamma, nu),
-                                       limit = 8000)
-        return np.exp(-integral)
-
-# dimensionless variable R = rho/rho1 as a function of V (and gamma and nu)
-# found by inverting eq. 5.12 in Sedov (1959), chapter 4 (omega = 0)
-# the integration constant C is found by inserting the R, V and z values
-# at the shock wave, where lambda is 1. These correspond to eq. 11.8 in Sedov
-# (1959), chapter 4.
-def _R(V, gamma, nu):
-    if V == 2./(nu+2.)/gamma:
-        return 0.
-    else:
-        C = 8.*gamma*(gamma-1.)/(nu+2.)/(nu+2.)/(gamma+1.)/(gamma+1.) \
-                *((gamma-1.)/(gamma+1.))**(gamma-2.) \
-                *(4./(nu+2.)/(gamma+1.)-2./(nu+2.))
-        lambda1 = _lambda(V, gamma, nu)
-        lambda5 = lambda1**(nu+2)
-        return (_z(V, gamma, nu)*(V-2./(nu+2.))*lambda5/C)**(1./(gamma-2.))
-
-# function of which we need to find the zero point to invert lambda(V)
-def _lambda_min_lambda(V, lambdax, gamma, nu):
-    return _lambda(V, gamma, nu) - lambdax
-
-# dimensionless variable V = v*t/r as a function of lambda (and gamma and nu)
-# found by inverting the function lamdba(V) which is found by solving
-# differential equation 5.11 in Sedov (1959), chapter 4 (omega = 0)
-# the invertion is done by searching the zero point of the function
-# lambda_min_lambda defined above
-def _V_inv(lambdax, gamma, nu):
-    if lambdax == 0.:
-        return 2./(2.+nu)/gamma;
-    else:
-        return optimize.brentq(_lambda_min_lambda, 2./(nu+2.)/gamma, 
-                               4./(nu+2.)/(gamma+1.), (lambdax, gamma, nu))
-
-# integrand of the first integral in eq. 11.24 in Sedov (1959), chapter 4
-def _integrandum1(lambdax, gamma, nu):
-    V = _V_inv(lambdax, gamma, nu)
-    if nu == 2:
-        return _R(V, gamma, nu)*V**2*lambdax**3
-    else:
-        return _R(V, gamma, nu)*V**2*lambdax**4
-
-# integrand of the second integral in eq. 11.24 in Sedov (1959), chapter 4
-def _integrandum2(lambdax, gamma, nu):
-    V = _V_inv(lambdax, gamma, nu)
-    if V == 2./(nu+2.)/gamma:
-        P = 0.
-    else:
-        P = _z(V, gamma, nu)*_R(V, gamma, nu)/gamma
-    if nu == 2:
-        return P*lambdax**3
-    else:
-        return P*lambdax**4
-
-# calculate alpha = E0/E
-# this corresponds to eq. 11.24 in Sedov (1959), chapter 4
-def get_alpha(gamma, nu):
-  integral1, err1 = integrate.quad(_integrandum1, 0., 1., (gamma, nu))
-  integral2, err2 = integrate.quad(_integrandum2, 0., 1., (gamma, nu))
-  
-  if nu == 2:
-    return np.pi*integral1+2.*np.pi/(gamma-1.)*integral2
-  else:
-    return 2.*np.pi*integral1+4.*np.pi/(gamma-1.)*integral2
-
-# get the analytical solution for the Sedov-Taylor blastwave given an input
-# energy E, adiabatic index gamma, and number of dimensions nu, at time t and
-# with a maximal outer region radius maxr
-def get_analytical_solution(E, gamma, nu, t, maxr = 1.):
-    # we check for the existance of a datafile with precomputed alpha and
-    # profile values
-    # if it does not exist, we calculate it here and write it out
-    # calculation of alpha and the profile takes a while...
-    lvec = np.zeros(1000)
-    Rvec = np.zeros(1000)
-    fname = "sedov_profile_gamma_{gamma}_nu_{nu}.dat".format(gamma = gamma,
-                                                             nu = nu)
-    if os.path.exists(fname):
-        file = open(fname, "r")
-        lines = file.readlines()
-        alpha = float(lines[0])
-        for i in range(len(lines)-1):
-            data = lines[i+1].split()
-            lvec[i] = float(data[0])
-            Rvec[i] = float(data[1])
-    else:
-        alpha = get_alpha(gamma, nu)
-        for i in range(1000):
-            lvec[i] = (i+1)*0.001
-            V = _V_inv(lvec[i], gamma, nu)
-            Rvec[i] = _R(V, gamma, nu)
-        file = open(fname, "w")
-        file.write("#{alpha}\n".format(alpha = alpha))
-        for i in range(1000):
-            file.write("{l}\t{R}\n".format(l = lvec[i], R = Rvec[i]))
-
-    xvec = np.zeros(1002)
-    rhovec = np.zeros(1002)
-    if nu == 2:
-        r2 = (E/alpha)**0.25*np.sqrt(t)
-    else:
-        r2 = (E/alpha)**0.2*t**0.4
-
-    for i in range(1000):
-        xvec[i] = lvec[i]*r2
-        rhovec[i] = Rvec[i]
-    xvec[1000] = 1.001*r2
-    xvec[1001] = maxr
-    rhovec[1000] = 1.
-    rhovec[1001] = 1.
-    
-    return xvec, rhovec
-
-def main():
-    E = 1.
-    gamma = 1.66667
-    nu = 2
-    t = 0.001
-    x, rho = get_analytical_solution(E, gamma, nu, t)
-    for i in range(len(x)):
-        print x[i], rho[i]
-
-if __name__ == "__main__":
-    main()