Outlet1D conditions

[ischoegl]

Problem description

In Outlet1D::eval, the following conditions are used:

cantera/src/oneD/Boundary1D.cpp

Line 372 in f89efe8

rb[c_offset_U] = xb[c_offset_L];

and

cantera/src/oneD/Boundary1D.cpp

Lines 387 to 390 in f89efe8

	// zero Lambda
	if (m_flow_left->fixed_mdot()) {
	rb[c_offset_U] = xb[c_offset_L];
	}

In order to make lambda zero, it should presumably use the same condition as for Inlet1D::eval, i.e.

cantera/src/oneD/Boundary1D.cpp

Lines 196 to 199 in f89efe8

	// if the flow is a freely-propagating flame, mdot is not specified.
	// Set mdot equal to rho*u, and also set lambda to zero.
	m_mdot = m_flow->density(0)*xb[0];
	rb[c_offset_L] = xb[c_offset_L];

While it is currently not used, there is a comment in OutletRes1D::eval:

cantera/src/oneD/Boundary1D.cpp

Lines 470 to 472 in f89efe8

	// this seems wrong...
	// zero Lambda
	rb[c_offset_U] = xb[c_offset_L];

Setting the residual for c_offset_U based on the current value of c_offset_L to drive the value to zero appears non-intuitive, especially as Inlet1D::eval uses the more intuitive rb[c_offset_L] = xb[c_offset_L]; condition. It makes sense that things converge for axisymmetric/unstrained flows, but it nevertheless appears to be incorrect.

A brute-force replacement with the presumably correct version does lead to convergence issues/failures in BurnerFlame unit tests. The likely resolution is to implement the correct condition for rb[c_offset_U].

System information

• Cantera version: 3.0.0b1 (and going back many Cantera versions; lines were last updated in IonFlow #441, but the underlying issue predates that change)
• OS: Any
• Python/MATLAB/other software versions: N/A