Single-Point Calculations
The most fundamental feature of xtb is computing electronic energies. This calculation is performed automatically every time an xtb run is initiated.
You can explicitly request a single-point calculation using:
xtb struc.xyz --sp
The electronic energy, given in Hartree, is printed at the end of the output as TOTAL ENERGY
, along with the GRADIENT NORM
and the HOMO-LUMO GAP
.
Single-point calculations in xtb can be useful for quickly identifying reasonable conformers. For example, you can determine the lowest-energy gas-phase conformer by comparing electronic energies.
Exercise
As an exercise, identify the lowest-energy conformer for the following two structures using both GFN-FF and GFN2-xTB. Then, compute their energy difference in kcal/mol.
40
Energy =
O 2.8656865 0.5086790 1.3197062
C 2.9608465 1.3781190 0.4782162
O 3.9258365 2.3120790 0.4899462
H 4.4705065 2.1269290 1.2702563
C 2.0089165 1.5718490 -0.6774237
C 1.2218365 2.8726290 -0.5104438
H 0.5452365 2.9727090 -1.3548738
H 0.6374265 2.8387490 0.4080463
H 1.8941065 3.7273190 -0.4842237
H 2.5858265 1.6221090 -1.6024638
N 1.1570565 0.4147690 -0.7077437
H 1.1682765 -0.1851510 0.1004562
C 0.4547365 0.0498590 -1.8004238
O 0.4143565 0.7088490 -2.8348837
C -0.3359535 -1.2532710 -1.6476638
H -0.0269735 -1.8943310 -2.4724237
N -0.0439735 -1.9464710 -0.4179737
H -0.4437335 -1.6251410 0.4556562
C 1.0449065 -2.7567710 -0.3028238
O 1.7205465 -3.1364810 -1.2489437
C 1.3476865 -3.1714810 1.1313163
H 2.4156765 -3.3901810 1.1668263
H 0.8153865 -4.1048210 1.3166462
N 0.8864165 -2.1654010 2.0880563
H 0.7812065 -2.5725610 3.0070963
H 1.5766565 -1.4271210 2.1744762
C -1.8356235 -0.9421110 -1.7606137
H -1.9948435 -0.4470910 -2.7176738
H -2.3823335 -1.8853110 -1.7612237
C -2.2945635 -0.0653810 -0.6306337
C -2.1976335 1.3274490 -0.7299938
H -1.8465635 1.7652190 -1.6561638
C -2.5483335 2.1427590 0.3430863
H -2.4749535 3.2179390 0.2471862
C -2.9994235 1.5773690 1.5353163
H -3.2729535 2.2106090 2.3681563
C -3.1088535 0.1928090 1.6437062
H -3.4691135 -0.2535310 2.5606562
C -2.7612435 -0.6201710 0.5658963
H -2.8620635 -1.6960210 0.6479063
40
Energy =
O -0.7784810 0.9576770 5.7268255
C -0.9737110 0.2611170 4.7601655
O -2.2037210 -0.1090930 4.3828155
H -2.1370810 -0.6604030 3.5681155
C 0.1586390 -0.2846530 3.8806455
C 1.5146990 0.1371970 4.4029555
H 2.3036390 -0.2844730 3.7831355
H 1.5969990 1.2219670 4.4013155
H 1.6411190 -0.2009830 5.4270155
H 0.0646390 -1.3746330 3.8792655
N 0.0108590 0.1662570 2.5005855
H 0.7060590 0.7969570 2.1009255
C -0.8578710 -0.3970130 1.6490855
O -1.7221410 -1.2152430 1.9931455
C -0.6966010 0.0606870 0.2003455
H -0.6989610 1.1548170 0.1998355
N 0.5849790 -0.3575730 -0.3352345
H 0.6392690 -1.0698530 -1.0538745
C 1.7056390 0.3670370 -0.1649545
O 1.8033390 1.3083670 0.6275255
C 2.8738890 -0.0710330 -1.0300845
H 3.5601090 -0.6186530 -0.3843745
H 3.3852490 0.8427970 -1.3386345
N 2.4434290 -0.9468530 -2.1165445
H 3.2259990 -1.4733230 -2.4795245
H 2.0704490 -0.4026530 -2.8854245
C -1.8214910 -0.4657430 -0.6848745
H -1.8713110 -1.5502830 -0.5862745
H -2.7651510 -0.0718430 -0.3085445
C -1.5891810 -0.0664830 -2.1143745
C -1.1455410 -0.9995930 -3.0548145
H -1.0385310 -2.0377230 -2.7632045
C -0.8512410 -0.6142830 -4.3622745
H -0.5161810 -1.3521630 -5.0788045
C -0.9972910 0.7172970 -4.7447145
H -0.7733810 1.0197170 -5.7584145
C -1.4472710 1.6567670 -3.8169645
H -1.5731910 2.6901770 -4.1098845
C -1.7396210 1.2655270 -2.5133545
H -2.0910510 2.0001870 -1.7985545
The predicted energy difference between these two conformers, as calculated using CCSD(T), is that conf1 is 1.4 kcal/mol higher in energy than conf2. Are both results reasonable?
A way to improve the performance of GFN-FF for conformer ranking is ConfRank.
Further Settings
For the energy calculations, xtb tries to converge the atomic charges of the molecule, which is equal to converging a Wavefunction. The behaviour of this self-consistent charge (SCC) convergence can be adjusted, e.g., the --acc <REAL>
flag changes the accuracy and the --iterations <INTEGER>
lets you change the maximum number of SCC cycles. The latter is particularly useful when the wavefunction does not converge within the default 250 cycles, which may occur in very large or electronically complex systems.
However, if converging the Wavefunction fails, it should first be checked manually before increasing the number of iterations, as some systems may not converge at all.
More details on adjusting convergence parameters can be found in the xtb manual.