Ab-initio molecular dynamics (= AIMD ) is too time-consuming, useless, dealing with unreal atoms.

Home page
Molecular dynamics is useless

Ab-initio molecular dynamics (= AIMD ) treating the whole molecules as one-fake-electron DFT model with pseudopotential lacking real atomic shape is useless, unable to predict anything.

(Fig.1) Ab-initio MD just changes coefficients (= c ) of artificially-chosen fake electron's wavefunction lacking real atomic shape.

Molecular dynamics (= MD ) is impractical

Unreal quantum mechanical shapeless atomic model keeps today's only molecular-motion simulation method called molecular dynamics (= MD ) impractical and too time-consuming.

We can easily predict and explain motions of objects (= or molecules ) with known shape and properties, like when designing and building cars, planes and machines from parts with known shape, which is impossible in today's only atomic theory or unreal quantum mechanical shapeless atomic model.

Due to unphysical quantum mechanical wavefunction lacking real atomic shape, today's only molecular simulating method called ( classical ) molecular dynamics (= MD ) is impractical, too time-consuming to calculate each atomic position at short-time intervals many, many times, based on artificially-chosen force field potential (= only potential, No atomic shape ) that cannot predict any molecular behavior ( this-p.27 ).

Ab-initio molecular dynamics (= MD ) is useless.

Ab-initio molecular dynamics (= AIMD ) can only treat the whole molecules as one-fake-electron DFT model lacking real atomic shape, which is meaningless, cannot predict molecular behavior.

The ordinary (= pseudo-classical ) molecular dynamics based only on artificially-chosen force field potential lacking real atomic shape cannot consider electrons' behavior nor bond-breaking ( this-p.6,  this-Problems that classical MD..  this-p.1-right-1st-paragraph ).

In ab-initio molecular dynamics (= MD ), due to the impractical Schrödinger equation, quantum mechanics has to treat the whole atoms as one pseudo-electron DFT (= Kohn-Sham ) model through artificially-chosen exchange pseudo-potential (= functional ) that is fake ab-initio (= empirical ), unable to predict any physical values ( this-p.2-2nd-paragraph,  this-p.2-left-last-paragraph~p.2-right,  this-p.23-lower ).

Ab-initio-MD in DFT cannot predict molecules.

Ab-initio molecular dynamics (= AI MD ) is based on one-fake-electron DFT model with artificially-chosen exchange pseudo-potential functional that is meaningless, cannot predict any molecular behavior.

This-p.1-middle-last~right says  -- one-fake-electron-DFT
"In modern AIMD (= ab-initio molecular dynamics ), the electronic structure method most commonly used is the Kohn–Sham (KS) formulation of density functional theory (= DFT ), wherein the total energy is expressed as a functional of n mutually orthonormal single-particle electron orbitals... the form of the exchange–correlation energy, Exc, is unknown and, therefore, must be approximated (= so AIMD with DFT cannot predict physical values,  this-p.2-left-2. )"

This-p.1-right-upper says  -- DFT failed
"widely used DFT approximations (e.g. B3LYP ) can fail badly.. DFT-D models can deliver satisfactory results only if carefully parametrized by fitting a few adjustable parameters"  ← DFT relying on fitting parameters cannot predict anything.

This-p.2-left-right say  -- Ab-initio MD cannot predict
"within an efficient ab initio molecular dynamics (AIMD) based on DFT"
"The DFT results appear to be substantially influenced not only by the choice of the functional, but also by other details of the electronic calculations such as the pseudo-potential and the basis set (= one-fake-electron DFT wavefunctions )"  ← Ab-initio MD with DFT relying on choice of various free parameters and potentials cannot predict molecular behavior ( this-p.1-left-right ).

This-p.6-left-last says  -- No solution
"Unfortunately, the density functional has No closed-form solution but many approximations are known"

See also this-p.19-p.20

Ab-initio MD is more impractical.

Ab-initio molecular dynamics (= MD ) is much more time-consuming and more impractical than ordinary (impractical) MD.

Ab-initio or first-principle molecular dynamics (= AIMD, FPMD, CPMD ) changing fictitious electron's wavefunction (= instead of moving real electrons or atoms with shape ) based on one-fake-electron-DFT pseudo-potential is much more impractical and more time-consuming than the impractical classical MD ( this-p.2-left-1st-paragraph,  this-p.2-I-introduction,  this-p.19(or p.13)-3rd-paragraph ).

This-site ↓

p.9 says  -- Quantum mechanical shapeless atom
"In quantum mechanics, instead, particles are described by a wave-function... No longer have defined positions"

p.14-last says  -- Impractical Schrodinger equation
"the many-particle wave function (= of Schrodinger equation ) is still too complicated for practical use, so we make further simplifications"

p.22 says  -- Nonsense first-principle
"“First-principles (= ab-initio ) results may be worthless nonsense"

p.23 says  -- Slower ab-initio MD
"Even with efficient DFT codes, ( ab-initio MD is ) still several orders of magnitude slower than classical MD"

Ab-initio MD is far slower than classical MD.

Ab-initio molecular dynamics (= AI MD ) is far slower and more impractical than classical MD based on empirical force field potential, even with the help of AI or machine-learning.

Ab-initio molecular dynamics (= AIMD ) based on one-fake-electron DFT model with artificially-chosen exchange pseudo-potential functionals (= which cannot predict any molecular behavior ) is far slower and more impractical than the ordinary (pseudo-)classical molecular dynamics (= MD = this is also impractical, too time-consuming ) with force field pseudo-potentials whose parameters are empirically fitted to experiments.

This site (2026) ↓

p.1-left-last-paragraph says  -- Slower ab-initio MD
"Classical MD with empirical force fields, such as AMBER, CHARMM and GROMACS, scales efficiently on large-scale systems. However, they cannot reliably capture bond breaking/formation, charge transfer, many body polarization, and related quantum-mechanical effects. In contrast, ab-initio MD (AIMD) based on Density Functional Theory (DFT) computes interatomic forces and achieves near quantum-mechanical fidelity. Yet, its O(N³) scaling severely limits accessible system sizes and simulation timescales"

p.6-left-last-paragraph says  -- Impractical ab-initio MD, AI
"The classical MD achieves a throughput between 200 and 400 ns/day (= just simulating 400ns protein motion simulation takes a day, so impractical, too time-consuming MD ).. although ML (= machine-learning ) inference can reduce the high computational cost from O(N3 ) in DFT, the AIMD (= ab-initio molecular dynamics ) models are still computationally more expensive compared to empirical MD simulations"  ← Ab-intio MD is more time-consuming than classical MD even relying on machine-learning or AI.

This-p.3-right-2nd-paragraph says  -- Impractical machine learning
"Among various ML (= machine learning ) methods (= combining quantum mechanical and experimental data ), artificial neural networks (ANNs) and, especially, deep NNs (DNNs) are the most widely used.... Nonetheless, their computational cost typically remains one to two orders of magnitude higher than that of the simplest empirical FFs (= force fields of classical MD )."

↑ Force fields created by machine-learning or AI still takes much more time (= so useless ) than the ordinary classical molecular dynamics with empirical force fields.

Ab-initio MD cannot move real electrons.

Ab-initio MD takes too much time in DFT adjusting coefficients of artificially-chosen one-fake-electron-DFT wavefunction instead of moving real electrons or atoms with shape.

In this impractical ab-initio MD, only nuclei (= positive ions ) are treated like classical particles, moved gradually by differentiating the total fake energy E with respect to each nuclear coordinate R_I (= force = space derivative of total potential energy, this-p.6 ) at short-time intervals.

In Born-Oppenheimer (ab-initio) molecular dynamics (= BOMD ), they have to artificially choose fake one-fake-electron DFT wavefunction ( this-p.14-(40) ), and update their coefficients (= c ) by the extremely time-consuming self-consistent field (= SCF ) method at short-time intervals ( this-p.23,p.27,p.32 ), which is completely impractical ( this-p.35,  this-p.2 ), instead of moving real electrons or atoms with shape.

This-site ↓

p.34 says  -- One fake electron DFT
"DFT is self-consistent single particle (= Kohn-Sham ) equation with effective (= fictitious) potential"

p.39-last says  -- DFT fails
"Limitations of ab-initio-MD based on DFT are – Weak interactions (vdW) are poorly described  – Lack of a systematic improvability"

p.40-last says  -- Time-consuming MD
" the direct BO-MD (= Born-Oppenheimer ab-initio molecular dynamics) involves a SCF solution of the Kohn-Sham equations at each step
computationally very demanding"

Ab-inito MD can change only coefficients of one-fake-electron wavefunction instead of moving atoms.

Ab-initio Car-Parrinello MD changes only coefficients of chosen one-fictitious electron's wavefunction, which is too time-consuming, impractical, instead of moving real electrons or atoms with shape.

So ab-initio MD often uses the slightly simpler Car-Parrinello molecular dynamics (= CPMD ) using fictitious electron's mass μ ( this-p.36,  this-p.9,  this-p.7,p.22 )

↑ In this CPMD, they update each coefficient (= c ) of the artificially- chosen pseudo-electron's wavefunction (= φ = basis set ) by differentiating the total energy E with respect to the coefficient c ( this-p.41,p.44,  this-p.12-(6.10),  this-p.5,p.7 ) at extremely short-time intervals (= 0.1 fs, this-p.12-22,  this-p.3,p.12 ) together with updating nuclear positions.

So ab-initio molecular dynamics just changes fictitious electron (fake) wavefunction's coefficients instead of moving real electrons.

This-p.42-Density functional theory says  -- Fictitious coefficients
"equation of motion for coefficient dynamics fictitious electron mass,"

And this ab-initio MD (= CPMD, BOMD.., old Ehrenfest MD is not used now, this-13-upper ) needing calculations of both nuclear and fictitious electron's states at much shorter time intervals (= 0.1fs ) is more impractical and more time-consuming than classical MD updating only atoms at the intervals of 2fs ( this-p.19-lower,  this-p.2 ).

The 4th-last and last paragraphs of this hyped news (7/8/2025) say
"Throughout this research, ..the team have investigated the vibrational properties of titanium nitride using Raman spectroscopy, a non-destructive chemical analysis technique which provides detailed information about chemical structure."

"The ultimate goal of this project is to gain an atomistic-level understanding of the role played by the atoms"  ← This means the atomistic-level understanding still has Not been achieved.

↑ Raman spectroscopy's resolution (= just measuring light reemitted from the material ) is about 1μm ( this-middle-Raman is used for microscopic,  this-resolution ), which can Not distinguish single atoms of less than 1nm, nor clarify the atomic mechanism.

↑ This research paper ↓

p.1-abstract says "but a fundamental understanding of how their structures evolve during electrocatalysis remains unknown"  ← quantum mechanics failed to clarify the material's atomic mechanism.

p.4-left-computational methods used the impractical ab-initio molecular dynamics based on unreal DFT's pseudo-potential, which can Not clarify real atomic mechanism.

Quantum mechanical mainstream density functional theory (= DFT ) treats the whole material as one pseudo-electron model with artificially-chosen spin pseudo-potential parameters.  ← No quantum mechanical prediction, so No evidence of spin

(Fig.S)  Quantum mechanical DFT just artificially choosing (fictitious) spin parameters cannot predict ferromagnet or antiferromagnet.  → No evidence of spin

Electron spin is unreal, irrelevant to (anti-)ferromagnet.

Ferromagnet or antiferromagnet is said to be caused by (fictitious) electron spin according to quantum mechanics.

But an electron is known to be Not actually spinning.
Furthermore, quantum mechanics just choosing free spin parameter J has No ability to predict ferromagnet or antiferromagnet.

Quantum mechanical Schrodinger equations are unsolvable and unable to predict any multi-electron atoms or molecules.

Quantum mechanics has to rely on the rough approximation called density functional theory (= DFT ) that tries to treat the whole material as one-pseudo-electron or quasiparticle model with artificially-chosen pseudo-potential (= different exchange-correlation functionals give different spin interaction parameter J as shown in this p.7-Table.III ).

Quantum mechanics just choosing free parameters cannot predict ferromagnet or antiferromagnet.

In this mainstream DFT, physicists have to choose various freely-adjustable parameters such as fictitious wavefunctions, exchange-correlation-pseudo-potential (= U, J,  this p.6-right-1st-paragraph,  this-p.1-right-last ), (spin) magnetic moment (= m,   this p.4~p.7,   this p.136(p.128)-2nd~3rd-paragraphs ), and numbers of electrons (= n ) in different spin states ( this p.16-1st-paragraph ).  ← No quantum mechanical prediction

This p.11-3rd-paragraph says
"In practice treated as fitting parameters, i.e., adjusted to reach agreement with experiment: equilibrium volume, magnetic moment, band gap"

↑ Instead of (useless) quantum mechanical prediction, these artificially-chosen magnetic moment parameter (= m ), electron spin number (= n  or spin polarization parameter ζ ) determine whether the material is ferromagnet or antiferromagnet ( this p.53(p.45),p.60-63,  this p.8-p.12. p.21(or p.20) ).

DFT's exchange-potential and pseudo-potential parameters (= U, J ) must be chosen from experimental results in most cases (= instead of being predicted by quantum mechanics,  this p.4-right-3rd-last-paragraph,  this p.2-left-2nd-last-paragraph,  this p.3-1,  this p.3,  this p.22 chose t or J,  this p.6-right-1st-paragraph ).

This p.5-right says
"However, apart from the intentionally fitted LSDA+U result, all functionals overestimate the antiferromagnetic coupling J by at least a factor of two, and perform even worse for the interchain coupling J
 ( this p.16,  this p.8-left-1st-paragraph )"

So quantum mechanics (= unphysical spin model ) relying on various freely-chosen parameters ( this p.10,  this p.21 ) cannot predict whether the material is ferromagnetic or antiferromagnetic ( this p.22(or p.21)-2nd-paragraph ), which means there is No evidence that (anti-)ferromagnetism is caused by electron spin.

Quantum mechanics or DFT relying on artificially-chosen exchange pseudo-potentials can Not predict electron spin or magnetic moments of materials

This research paper ↓

p.30(or p.29)-3rd-paragraph says "The treatment described electrons in a non-interacting, fictitious potential"

p.47(or p.46) says " The most straightforward inclusion of magnetic materials in DFT is to attach two spinors ( χ+ and χ−) to the wave function (= artificially-chosen fake wavefunctions for expressing unreal spin )"

p.95(or p.94)-Table 4.7 exchange (spin) interaction parameters J are different among different chosen exchange-correlation (= XC ) potential functionals such as LDA, PBE, rSCAN.. ( this-p.6,  this-3rd-paragraph ),  ← Spin exchange parameters affected by the artificial choice of exchange potential means the spin magnetic moments are Not quantum mechanical prediction.

p.96(or p.95)-3rd-paragraph says "DFT is known to overestimate the magnitude of the exchange interaction parameter"

p.126(or p.125)-2nd-paragraph says "the magnetism within iron arsenide depends heavily on the choice of XC (= exchange-correlation ) functional"

Stable ferromagnet is caused by the realistic electron's orbits (= instead of spin ) meshed with each other by Coulomb electric interaction.

The 2nd, 5th, 10th, 2nd-last paragraphs of this hyped news (7/24/2025) say
"Understanding the origins of magnetism—deeply rooted in quantum mechanics—enables scientists and engineers to manipulate magnetic properties for further innovative applications and emerging technologies."  ← hype

"With their method, the team successfully mapped out every electronic contribution to magnetic interactions, tracing the intricate exchange pathways through both magnetic atoms"  ← This quantum exchange interaction lacking real exchange force is unreal, which abstract unphysical math ( this-p.9-upper ) clarifies No real magnetic mechanism.

"This breakthrough therefore provides a solid foundation for the rational design and precise engineering of magnetic materials, accelerating the path toward smart, efficient, and tunable technologies"  ← hype

Quantum mechanics just choosing free parameters and artificial models can Not predict nor clarify any real magnetic mechanism of materials.

↑ This research paper ↓

p.1-abstract says nothing about technological application, contrary to the above hyped news.

p.1-inroduction-2nd-paragraph says
"While these consequences of magnetic exchange are well understood, the exact quantitative determination of magnetic exchange between a pair of atoms for a specific material remains challenging"  ← Quantum mechanics fail to clarify material's magnetism nor (unphysical spin) exchange interaction.

p.4-p.6 shows this research with No experiment just artificially chose this unphysical exchange interaction parameters J without quantum mechanical prediction nor clarifying any real mechanism.

This-p.11-left-2nd-paragraph says
"allows us to extract the magnetic exchange interaction,... The method is however tied to the choice of DFT basis (= one fictitious electron model clarifying No real mechanism ), implementation of exchange correlation functional used in DFT and to specific computational parameters (i.e. k-points sampling, Hubbard U corrections within DFT+U, etc.)."

↑ So the present quantum mechanics relying on artificial choice of various free parameters and exchange energy functional of one-pseudo-electron DFT model can Not predict nor clarify any fictional spin exchange interaction, so No technological development, contrary to hypes.

Feel free to link to this site.