the j in this line should be a k

Should be faster overall to use automatic differentiation to compute the Jacobian. I can use fazang.

See relevant post here: https://fortran-lang.discourse.group/t/speed-of-autodif-vs-finite-differencing/2887/8

AD can be useful for climate model jacobians.

Moses boundary condition makes sense for gases. But it does not make sense for particles. Particles should be fallout out of the bottom of the model, but they only diffuse out when Moses is used

https://web.stanford.edu/~cantwell/AA284A_Course_Material/AA284A_Resources/Camberos%20and%20Moubry,%20Chemical%20Equilibrium%20Analysis%20with%20the%20Method%20of%20Element%20Potentials%202001-0873.pdf

Need to add error message when climate is turned on in update_vertical_grid

I suggest adding the topics atmosphere, photochemical-model as you did for photochempy. Maybe earth-science too.

I could improve the relation for mean free path:

Instead use kinetic theory of gases (Equation 9.6, Seinfeld and Pandis (2006))

This is related to #14

Would be cool to also include a 1-D climate model. This could be used on its own, but could also coupled in a time-dependent fashion to the photochemical model.

The code lacks good documentation.

Memory will leak here or here if an error occurs before the memory is deallocated.

all_reactions and all_species in photochem_input_read.f90 could be a leak.

Hello,

I'm following the instructions to install photochem, but I ran into an error when I tried to run part of the code:
python -m pip install --no-deps --no-build-isolation

Here's the error shown:
ERROR: You must give at least one requirement to install (see "pip help install")

What can I do to fix this issue? Thanks!

Looks like I'm screwing up molecular diffusion. There isn't conservation of H2 the atmosphere. There is MORE escaping than is being pushed into the atmosphere from the surface.

It has something to do with the ADU and ADL coefficients. I know this because when I neglect molecular diffusion, then there is conservation. Need to carefully step through all of this.

When water condensation is off, then this breaks cases that fix water in troposphere

Right now we are using a first order upwind scheme for particles falling. This is very diffusive. We can do better with a second order finite volume upwind scheme.

See Chapter 6 in LeVeque.

One potential issue is that these schemes depend on the size of the timestep. We can get the size of the timestep from CVODE, in the RHS function. I just worry that this will be an issue.

Two of the callback functions in PhotochemVars is not implemented in EvoAtmosphere:

• photochem/src/photochem_types.f90

  Line 121 in 8377052

  subroutine time_dependent_flux_fcn(tn, nw, photon_flux)

• photochem/src/photochem_types.f90

  Line 129 in 8377052

  subroutine time_dependent_rate_fcn(tn, nz, rate)

The dynamic viscosity of gas could depend on the background gas. See Table 1-2 in White (2006), "Viscous Fluid Flow"

I interpolate in log space here, but never convert back to linear space:

https://github.com/Nicholaswogan/Photochem/blob/0a3886810d7017a157b908752da0c6a062333bdb/Photochem/fortran/wrapper_utils.f90#L19

When there are not any optical properties for a particle, a massive amount of useless data is still read in and held in memory. If we use an array of types, then we can fix this

Right now I'm using center differencing for advection terms. This causes instability for particles falling. We can improve this by using a "upwind" differencing scheme.

When number of reactants is different than number of productions, I think my codes is getting a different reverse rate than Cantera. The issue would be here

Understand why I need to add this term to the jacobian. Seems like it should not be there.

I'm thinking atoms should be specified like this

atoms:
  - {name: H, mass: 1, redox: - 0.5}

Then we can compute redox state of everything and check for redox conservation.