c
c
c =====================================================
subroutine rpn3eu(ixyz,maxm,meqn,mwaves,mbc,mx,ql,qr,
& maux,auxl,auxr,wave,s,fl,fr)
c =====================================================
c
c # Riemann solver for the 3D Euler equations
c # The waves are computed using the Roe approximation.
c
c # This is quite a bit slower than the Roe solver,
c # but may give more accurate solutions for some problems.
c
c # On input, ql contains the state vector at the left edge of each cell
c # qr contains the state vector at the right edge of each cell
c # This data is along a slice in the x-direction if ixyz=1
c # the y-direction if ixyz=2.
c # the z-direction if ixyz=3.
c
c # On output, wave contains the waves, s the speeds
c # and fl, fr the positive and negative Godunov flux.
c
c # Note that the i'th Riemann problem has left state qr(i-1,:)
c # and right state ql(i,:)
c # From the basic routines, this routine is called with ql = qr
c
c Author: Randall J. LeVeque
c
implicit double precision (a-h,o-z)
dimension wave(1-mbc:maxm+mbc, meqn, mwaves)
dimension s(1-mbc:maxm+mbc, mwaves)
dimension ql(1-mbc:maxm+mbc, meqn)
dimension qr(1-mbc:maxm+mbc, meqn)
dimension auxl(1-mbc:maxm+mbc, maux, 3)
dimension auxr(1-mbc:maxm+mbc, maux, 3)
dimension fr(1-mbc:maxm+mbc, meqn)
dimension fl(1-mbc:maxm+mbc, meqn)
c
c local arrays -- common block comroe is passed to rpt3eu
c ------------
parameter (maxmrp = 1005) !# assumes atmost max(mx,my,mz) = 1000 with mbc=5
parameter (minmrp = -4) !# assumes at most mbc=5
dimension delta(5)
dimension sl(2),sr(2)
common /param/ gamma,gamma1
common /comroe/ u2v2w2(minmrp:maxmrp),
& u(minmrp:maxmrp),v(minmrp:maxmrp),w(minmrp:maxmrp),
& enth(minmrp:maxmrp),a(minmrp:maxmrp),g1a2(minmrp:maxmrp),
& euv(minmrp:maxmrp)
c
c # Riemann solver returns flux differences
c ------------
common /rpnflx/ mrpnflx
mrpnflx = 1
c
if (minmrp.gt.1-mbc .or. maxmrp .lt. maxm+mbc) then
write(6,*) 'need to increase maxmrp in rpA'
stop
endif
c
c # set mu to point to the component of the system that corresponds
c # to momentum in the direction of this slice, mv and mw to the
c # orthogonal momentum:
c
if(ixyz .eq. 1)then
mu = 2
mv = 3
mw = 4
else if(ixyz .eq. 2)then
mu = 3
mv = 4
mw = 2
else
mu = 4
mv = 2
mw = 3
endif
c
c # note that notation for u,v, and w reflects assumption that the
c # Riemann problems are in the x-direction with u in the normal
c # direction and v and w in the orthogonal directions, but with the
c # above definitions of mu, mv, and mw the routine also works with
c # ixyz=2 and ixyz = 3
c # and returns, for example, f0 as the Godunov flux g0 for the
c # Riemann problems u_t + g(u)_y = 0 in the y-direction.
c
c
c # compute the Roe-averaged variables needed in the Roe solver.
c # These are stored in the common block comroe since they are
c # later used in routine rpt3eu to do the transverse wave splitting.
c
do 10 i = 2-mbc, mx+mbc
rhsqrtl = dsqrt(qr(i-1,1))
rhsqrtr = dsqrt(ql(i,1))
pl = gamma1*(qr(i-1,5) - 0.5d0*(qr(i-1,mu)**2 +
& qr(i-1,mv)**2 + qr(i-1,mw)**2)/qr(i-1,1))
pr = gamma1*(ql(i,5) - 0.5d0*(ql(i,mu)**2 +
& ql(i,mv)**2 + ql(i,mw)**2)/ql(i,1))
rhsq2 = rhsqrtl + rhsqrtr
u(i) = (qr(i-1,mu)/rhsqrtl + ql(i,mu)/rhsqrtr) / rhsq2
v(i) = (qr(i-1,mv)/rhsqrtl + ql(i,mv)/rhsqrtr) / rhsq2
w(i) = (qr(i-1,mw)/rhsqrtl + ql(i,mw)/rhsqrtr) / rhsq2
enth(i) = (((qr(i-1,5)+pl)/rhsqrtl
& + (ql(i,5)+pr)/rhsqrtr)) / rhsq2
u2v2w2(i) = u(i)**2 + v(i)**2 + w(i)**2
a2 = gamma1*(enth(i) - .5d0*u2v2w2(i))
a(i) = dsqrt(a2)
g1a2(i) = gamma1 / a2
euv(i) = enth(i) - u2v2w2(i)
10 continue
c
c
c # now split the jump in q at each interface into waves
c
c # find a1 thru a5, the coefficients of the 5 eigenvectors:
do 20 i = 2-mbc, mx+mbc
delta(1) = ql(i,1) - qr(i-1,1)
delta(2) = ql(i,mu) - qr(i-1,mu)
delta(3) = ql(i,mv) - qr(i-1,mv)
delta(4) = ql(i,mw) - qr(i-1,mw)
delta(5) = ql(i,5) - qr(i-1,5)
a4 = g1a2(i) * (euv(i)*delta(1)
& + u(i)*delta(2) + v(i)*delta(3) + w(i)*delta(4)
& - delta(5))
a2 = delta(3) - v(i)*delta(1)
a3 = delta(4) - w(i)*delta(1)
a5 = (delta(2) + (a(i)-u(i))*delta(1) - a(i)*a4) / (2.d0*a(i))
a1 = delta(1) - a4 - a5
c
c # Compute the waves.
c # Note that the 2-wave, 3-wave and 4-wave travel at the same speed
c # and are lumped together in wave(.,.,2). The 5-wave is then stored
c # in wave(.,.,3).
c
wave(i,1,1) = a1
wave(i,mu,1) = a1*(u(i)-a(i))
wave(i,mv,1) = a1*v(i)
wave(i,mw,1) = a1*w(i)
wave(i,5,1) = a1*(enth(i) - u(i)*a(i))
s(i,1) = u(i)-a(i)
c
wave(i,1,2) = a4
wave(i,mu,2) = a4*u(i)
wave(i,mv,2) = a4*v(i) + a2
wave(i,mw,2) = a4*w(i) + a3
wave(i,5,2) = a4*0.5d0*u2v2w2(i) + a2*v(i) + a3*w(i)
s(i,2) = u(i)
c
wave(i,1,3) = a5
wave(i,mu,3) = a5*(u(i)+a(i))
wave(i,mv,3) = a5*v(i)
wave(i,mw,3) = a5*w(i)
wave(i,5,3) = a5*(enth(i)+u(i)*a(i))
s(i,3) = u(i)+a(i)
20 continue
c
c
c # compute Godunov flux f0 at each interface.
c # Uses exact Riemann solver
c
c
do 200 i = 2-mbc, mx+mbc
rhol = qr(i-1,1)
rhor = ql(i,1)
ul = qr(i-1,mu)/rhol
ur = ql(i ,mu)/rhor
vl = qr(i-1,mv)/rhol
vr = ql(i ,mv)/rhor
wl = qr(i-1,mw)/rhol
wr = ql(i ,mw)/rhor
pl = gamma1*(qr(i-1,5)-0.5*(qr(i-1,mu)*ul + qr(i-1,mv)*vl +
& qr(i-1,mw)*wl ))
pr = gamma1*(ql(i ,5)-0.5*(ql(i ,mu)*ur + ql(i ,mv)*vr +
& ql(i ,mw)*wr))
c
c # iterate to find pstar, ustar:
c
alpha = 1.
pstar = 0.5*(pl+pr)
wsr = dsqrt(pr*rhor) * phi(pstar/pr)
wsl = dsqrt(pl*rhol) * phi(pstar/pl)
c if (pl.eq.pr .and. rhol.eq.rhor) go to 60
c
40 do 50 iter=1,1000
p1 = (ul-ur+pr/wsr+pl/wsl) / (1./wsr + 1./wsl)
pstar = dmax1(p1,1d-6)*alpha + (1.-alpha)*pstar
wr1 = wsr
wl1 = wsl
wsr = dsqrt(pr*rhor) * phi(pstar/pr)
wsl = dsqrt(pl*rhol) * phi(pstar/pl)
if (dmax1(abs(wr1-wsr),dabs(wl1-wsl)) .lt. 1d-6)
& go to 60
50 continue
c
c # nonconvergence:
alpha = alpha/2.
if (alpha .gt. 0.1) go to 40
write(6,*) 'no convergence',i,wr1,wsr,wl1,wsl
wsr = .5*(wsr+wr1)
wsl = .5*(wsl+wl1)
c
60 continue
ustar = (pl-pr+wsr*ur+wsl*ul) / (wsr+wsl)
c
c
c # left wave:
c ============
c
if (pstar .gt. pl) then
c
c # shock:
sl(1) = ul - wsl/rhol
sr(1) = sl(1)
rho1 = wsl/(ustar-sl(1))
c
else
c
c # rarefaction:
cl = dsqrt(gamma*pl/rhol)
cstar = cl + 0.5*gamma1*(ul-ustar)
sl(1) = ul-cl
sr(1) = ustar-cstar
rho1 = (pstar/pl)**(1./gamma) * rhol
endif
c
c
c
c # right wave:
c =============
c
if (pstar .ge. pr) then
c
c # shock
sl(2) = ur + wsr/rhor
sr(2) = sl(2)
rho2 = wsr/(sl(2)-ustar)
c
else
c
c # rarefaction:
cr = dsqrt(gamma*pr/rhor)
cstar = cr + 0.5*gamma1*(ustar-ur)
sr(2) = ur+cr
sl(2) = ustar+cstar
rho2 = (pstar/pr)**(1./gamma)*rhor
endif
c
c
c # compute flux:
c ===============
c
c # compute state (rhos,us,ps) at x/t = 0:
c
if (sl(1).gt.0) then
rhos = rhol
us = ul
vs = vl
ws = wl
ps = pl
else if (sr(1).le.0. .and. ustar.ge. 0.) then
rhos = rho1
us = ustar
vs = vl
ws = wl
ps = pstar
else if (ustar.lt.0. .and. sl(2).ge. 0.) then
rhos = rho2
us = ustar
vs = vr
ws = wr
ps = pstar
else if (sr(2).lt.0) then
rhos = rhor
us = ur
vs = vr
ws = wr
ps = pr
else if (sl(1).le.0. .and. sr(1).ge.0.) then
c # transonic 1-rarefaction
us = (gamma1*ul + 2.*cl)/(gamma+1.)
e0 = pl/(rhol**gamma)
rhos = (us**2/(gamma*e0))**(1./gamma1)
ps = e0*rhos**gamma
vs = vl
ws = wl
else if (sl(2).le.0. .and. sr(2).ge.0.) then
c # transonic 3-rarefaction
us = (gamma1*ur - 2.*cr)/(gamma+1.)
e0 = pr/(rhor**gamma)
rhos = (us**2/(gamma*e0))**(1./gamma1)
ps = e0*rhos**gamma
vs = vr
ws = wr
endif
c
fl(i,1) = rhos*us
fl(i,mu) = rhos*us**2 + ps
fl(i,mv) = rhos*us*vs
fl(i,mw) = rhos*us*ws
fl(i,5) = us*(gamma*ps/gamma1 + 0.5*rhos*(us**2+vs**2+ws**2))
200 continue
c
do 220 m=1,6
do 220 i = 2-mbc, mx+mbc
fr(i,m) = -fl(i,m)
220 continue
c
return
end
c
c
c
double precision function phi(w)
implicit double precision (a-h,o-z)
common/param/ gamma,gamma1
c
sqg = dsqrt(gamma)
if (w .gt. 1.) then
phi = dsqrt(w*(gamma+1.)/2. + gamma1/2.)
else if (w .gt. 0.99999) then
phi = sqg
else if (w .gt. .999) then
phi = sqg + (2*gamma**2 - 3.*gamma + 1)
& *(w-1.) / (4.*sqg)
else
phi = gamma1*(1.-w) / (2.*sqg*(1.-w**(gamma1/(2.*gamma))))
endif
return
end