Computational artifacts

The various atomic charges and bond overlap populations of CH2 = NH are shown below and confirm our expectations based on consideration of sigma conjugative effects. However, an anomaly is noted in the case of the 4—31G calculation of the C—Hb and C—Hc overlap populations. Since the STO—3G optimization leads to a longer C—Hc bond, as predicted, the anomaly represents most likely a computational artifact. 

In a study which is relevant to the mechanism of hydrolysis of phosphonium salts, Glaser and Streitwieser297 studied the ions H4PO- and H3PFO- and their derivatives with Li +, NH4 and HF at the 6-31G level augmented by diffuse functions. They found that the structures of the anions are those of a hydride or fluoride ion solvated by or complexed with phosphine oxide, rather than phosphoranes297. A very important point is that earlier studies with diffuse functions yielded the pentacoordinated phosphoranes which they judged297 to be computational artifacts of the small basis set. 

The isomers of C2H2 are 2Ol(s), acetylene and 2O2(s), vinylidene. Both have singlet ground states. A third isomer was reported to have appeared in a theoretical calculation, but, it turned out to have been a computational artifact [8]. Table 17.1 presents relative self consistent field (SCF) energies, zero point vibrational energies (ZPVE), and dipole moments of the isomers. 

In summary, simulations with finite reservoirs have the distinct advantage that TDDFT algorithms, propagation methods and computer codes are well established for isolated systems and can be used for transport calculations without major changes. Computational artifacts due to the finite reservoirs can and need to be kept under control by a systematic enlargement of the contact sizes. 

The set of linear regression coefficients (a or b), correction terms (if needed), and constant c in Equation 2.10, are adjustable parameters. The % and % values are determined exactly from the hydrogen-suppressed graph of the repeat unit. There is no additive constant term in Equation 2.9. A constant does not change as a function of the amount of material present, so that it is an intensive property which does not belong in correlations for extensive properties. For example, so long as there is some material present, the density (an intensive property) has the same constant positive value. On the other hand, the total volume becomes infinitesimally small (approaches zero) in the limit of an exceedingly small quantity of any material. The omission of the constant in Equation 2.9 is therefore essential to prevent the introduction of a computational artifact into correlations for extensive properties. 

From their QSERR they find solute lipophilicity and steric properties as being responsible for analyte retention (k ) while enantioseparation (a) varied mainly with electronic and steric properties. The main difference between the analytes is that the enantioseparation of the esters is correlated with steric parameters that scale linearly with log a while the sulfoxides scale nonlinearly (parabolic), but this may be due to a computational artifact. The 3D-QSERR derived from field analysis revealed that while superpositioning of field maps for both analytes are not exactly the same, a similar balance of physicochemical forces involved in the chiral recognition process are at play for both sets of analyes. This type of atomistic molecular modeling, then, is a powerful adjunct to the type of modeling described earlier in this chapter and will, no doubt, be used more frequently in future studies. 

Eh or 30,000 K) and is linear with temperature at high temperatures. This behavior has already been explained by Eq. (51). This equation also indicates that whether p increases or decreases with temperature is determined by the numbers of basis functions (n) and electrons (N) and is, therefore, another computational artifact. When N > n, increases with temperature otherwise, p decreases. In physical reality or in a realistic calculation where N, is expected to display the — ln 2n)/p behavior. 

Once the significant components of the system have been chosen, a computational domain is then defined to enclose them. The geometry of the simulation box must define a volume that realistically encloses the physics of the system, with boundary conditions mimicking the effects of the larger, real system being modeled. Within the ion channel framework, only a small fraction of the cellular lipid membrane is simulated thus, the dimension of the computational domain is minimized to reduce the computational burden. Consequently, the boundary conditions must be chosen carefully so that unwanted computational artifacts are not introduced into the simulation results. 

There used to be two realities in the world of physics Experiment and Theory. Now there are three, and the third one is The Computer. In the community of physicists outside of the computer-simulation enclave, there is a good deal of skepticism about computer simulations of many-particle systems. This skepticism is certainly justified at the present time. Nevertheless, in my opinion, computer simulation will affect the progress of physics in a profound way. The many-body problems that we worked on for decades will finally yield to computer simulation. This does not mean that many-body problems will suddenly become simple the complications of these problems will appear in a new form. The question will be, how are computer simulations to be interpreted in terms of the mathematically posed problem, or in terms of physical reality. A computer-generated many-particle process contains an enormous amount of information, and the challenge will be to extract the information we want, leaving computer artifacts behind. In other words, the computer may have the answer, but we will have to figure out the question. 

Geometry optimizations of ArBN at HF/6-31G(d,p) and MP2/6-31G(d,p) levels predict rather short Ar-B distances of 1.858 and 1.900 A, respectively. At the MP2/6-31G(d,p) level, the dissociation of ArBN is calculated to be exothermic by 7.7 kcal/mol a very small barrier (<1 kcal/mol) was calculated at MP2/6-31G(d,p), which is probably a computational artifact. The geometry optimization of ArBN at the MP3/6-31G(d,p) level also yields a rather short (1.847 A) Ar B distance, and at this same level the molecule is predicted to be stable toward dissociation by 8.9 kcal/mol. Single-point energy calculations at MP4/6-311G(d,p) using the MP3/6-31G(d,p) optimized structure predict that ArBN is unstable by 17.2 kcal/mol [1]. 

There are three primary differences between MNDO and AMI. The most important is the addition of Gaussian functions to the description of the core-repulsion function (CRF) in AMI that was not present in MNDO. These Gaussian functions were added to adjust the shape of the CRF and to patch MNDO for certain theoretical inadequacies. As such, they have met with mixed success in this role, improving results in some cases, but producing computational artifacts in others. 

Becke-Johnson damping needs an adjustment of an additional parameter (in the case of DFT-D3) but is more physically justified, as dispersion energy converges to a finite value when -> 0 [41] and allows to avoid some computational artifacts [37]. This approach improves non-covalent bond distances over zero-damping and significantly improves predictions of thermochemical properties. The approach indeed affects the short-range interaction covered by the underlying functionals but possible overcorrelation effects seem to cancel out for chemically relevant model systems [37]. 

---