Super-droplet multiplicity after coalescence

[jbarr444]

Hi @slayoo,
Quick question: what happens to super-droplet multiplicities when they collide?
Thanks,
Jason

[slayoo]

Hi Jason,

Following step (5) in section 5.1.3 in Shima et al. 2009 (with multiplicities denoted by ξ):

which is reflected here for CPU:

PySDM/PySDM/backends/impl_numba/methods/collisions_methods.py

Lines 49 to 59 in cd9d730

	new_n = multiplicity[j] - gamma[i] * multiplicity[k]
	if new_n > 0:
	multiplicity[j] = new_n
	for a in range(len(attributes)):
	attributes[a, k] += gamma[i] * attributes[a, j]
	else: # new_n == 0
	multiplicity[j] = multiplicity[k] // 2
	multiplicity[k] = multiplicity[k] - multiplicity[j]
	for a in range(len(attributes)):
	attributes[a, j] = gamma[i] * attributes[a, j] + attributes[a, k]
	attributes[a, k] = attributes[a, j]

and here for GPU backend:

PySDM/PySDM/backends/impl_thrust_rtc/methods/collisions_methods.py

Lines 42 to 57 in cd9d730

	auto new_n = multiplicity[j] - gamma[i] * multiplicity[k];
	if (new_n > 0) {
	multiplicity[j] = new_n;
	for (auto attr = 0; attr < n_attr * n_sd; attr+=n_sd) {
	attributes[attr + k] += gamma[i] * attributes[attr + j];
	}
	}
	else { // new_n == 0
	multiplicity[j] = (int64_t)(multiplicity[k] / 2);
	multiplicity[k] = multiplicity[k] - multiplicity[j];
	for (auto attr = 0; attr < n_attr * n_sd; attr+=n_sd) {
	attributes[attr + j] = gamma[i] * attributes[attr + j] + attributes[attr + k];
	attributes[attr + k] = attributes[attr + j];
	}
	}
	}

HTH,
S.

[slayoo]

since PySDM 2.77 (https://github.com/open-atmos/PySDM/releases/tag/v2.77), the saturation vapour pressure functions are now all-SI unit - so no more Celsius. It should be enough to switch from pvs_Celsius(T) to pvs_water(T+T0) in your code. This issue is not related with the version of Python. HTH, S.

[jbarr444]

Oh, wonderful, that fixes it completely. That's strange though because it doesn't upgrade to 2.77 when I pip install PySDM on python 3.12, but it is 2.77 on python 3.11. Thanks anyways.

[slayoo]

Here's the reason - there was a timeout in the PyPI upload job for 2.77, so it has not yet reached PyPI. I've just trigerred a rerun, so it should be fixed in an hour: https://github.com/open-atmos/PySDM/actions/runs/10916574488
Thanks for notifying about it!

[slayoo]

For the record, the upload rerun on Fri went OK and v2.77 is there on PyPI https://pypi.org/project/PySDM/2.77/

[jbarr444]

Updated, thanks!

[jbarr444]

@slayoo That works !! Thank you :)