Understanding replications to go from triclinic to orthogonal

Hi

I am working with a triclinic unit cell of albite (mp-721988: NaAlSi3O8 (Triclinic, P-1, 2)) and am hoping to replicate in such a fashion that I can form an orthogonal unit cell which I can then look at different free surfaces.

I have seen in previous discussions the idea of certain replications followed by removing the tilt:

atom_style full
read_data al2o3
comm_modify cutoff 10.0
replicate 1 2 1
change_box all xy final 0.0
change_box all ortho
set group all image 0 0 0
write_data o-al2o3.data

However I am thinking these cases only work for specific tilt angles. Is there a way to know how best replicate it before removing the tilt? I don’t want to just blindly do it as I am concerned about poor physics, overlap etc. Below is the albite data file:

Thanks

Liam

LAMMPS data file. CGCMM style. atom_style full generated by VMD/TopoTools v1.7 on Mon Apr 05 10:16:16 -0230 2021
26 atoms
4 atom types

-1.841500 5.417500 xlo xhi
-1.889768 5.521768 ylo yhi
-0.000055 6.462055 zlo zhi
-1.374985 -2.308739 -3.779208 xy xz yz

Pair Coeffs

1 AL

2 NA

3 O

4 SI

Bond Coeffs

1

Masses

1 26.981539 # AL
2 22.989771 # NA
3 15.999400 # O
4 28.085501 # SI

Atoms # full

1 1 2 0.000000 -2.262000 4.409000 4.810000 # NA UNL
2 2 2 0.000000 5.838000 -0.777000 1.652000 # NA UNL
3 1 1 0.000000 4.686000 -0.570000 5.298000 # AL UNL
4 2 1 0.000000 -1.111000 4.202000 1.164000 # AL UNL
5 2 4 0.000000 2.989000 4.092000 1.126000 # SI UNL
6 1 4 0.000000 0.586000 -0.460000 5.337000 # SI UNL
7 3 4 0.000000 2.789000 0.913000 1.255000 # SI UNL
8 4 4 0.000000 0.786000 2.719000 5.207000 # SI UNL
9 2 4 0.000000 -0.553000 4.037000 2.793000 # SI UNL
10 1 4 0.000000 4.128000 -0.405000 3.669000 # SI UNL
11 1 3 0.000000 -1.139000 -1.305000 5.587000 # O UNL
12 2 3 0.000000 4.714000 4.937000 0.876000 # O UNL
13 2 3 0.000000 -0.040000 5.853000 2.622000 # O UNL
14 1 3 0.000000 3.615000 -2.221000 3.840000 # O UNL
15 3 3 0.000000 4.851000 1.400000 0.474000 # O UNL
16 4 3 0.000000 -1.276000 2.232000 5.989000 # O UNL
17 2 3 0.000000 1.302000 3.983000 2.107000 # O UNL
18 1 3 0.000000 2.273000 -0.351000 4.356000 # O UNL
19 1 3 0.000000 -2.036000 1.944000 4.388000 # O UNL
20 2 3 0.000000 5.611000 1.688000 2.074000 # O UNL
21 3 3 0.000000 2.345000 1.818000 1.809000 # O UNL
22 4 3 0.000000 1.230000 1.814000 4.654000 # O UNL
23 1 3 0.000000 4.833000 -0.977000 4.432000 # O UNL
24 2 3 0.000000 -1.258000 4.609000 2.030000 # O UNL
25 4 3 0.000000 1.289000 1.895000 6.123000 # O UNL
26 3 3 0.000000 2.287000 1.737000 0.339000 # O UNL

An open source external tool that can also easily handle conversion from arbitrary unit cell to orthogonal is Atomsk, which can generate Lammps topology, provided that you do not have permanent bonds (you can add that manually later by yourself).
But as you suspect it is not simple to find a corresponding orthogonal cell, especially for trclinic cells. The generated cell will probably be fairly large and thus impractical. As such, I do not see a good reason to replicate the unit cell before conversion to orthogonal cell.
Best,
Anze