Revise main website content after complete proofread pass

Author: godotalgorithm

content/_index.md

@@ -1,4 +1,4 @@
-![MOPAC logo](img/logo.png)
+![MOPAC logo](images/logo.png)
 
 # **M**olecular **O**rbital **PAC**kage
 

content/about/accuracy.md

@@ -12,10 +12,10 @@ There is also a [list of discussions](/Discussions/Discussion_topics.html) about
 
 In the broader context of atomistic simulation, the semiempirical quantum mechanics (SQM) models used by MOPAC are a middle ground in cost, accuracy, and transferability.
 Here, "transferability" generically refers to the ability of a model to retain its accuracy beyond its training data.
-First-principles (*ab initio*) quantum mechanics (QM) calculations are typically at least a thousand times more expensive than an equivalent SQM calculation,
+First-principles (i.e. *ab initio*) quantum mechanics (QM) calculations are typically at least a thousand times more expensive than an equivalent SQM calculation,
 but they are usually more accurate and more transferable. For example, density functional theory (DFT) calculations with a double-zeta or triple-zeta Gaussian basis set
 and the B3LYP density functional typically have errors that are two to five times smaller than any SQM model in MOPAC.
-Molecular mechanics (MM) models based on a force field (interatomic potential) are typically at least a thousand times less expensive than an equivalent SQM calculation.
+Molecular mechanics (MM) models based on a force field (i.e. interatomic potential) are typically at least a thousand times less expensive than an equivalent SQM calculation.
 However, the transferability of force fields is usually quite low. They can be more accurate than an SQM calculation, but that accuracy can rapidly degrade
 if a force field is used for systems that are not well represented by its training data. For the application of force fields to long molecular dynamics trajectories,
 it can be challenging to monitor or estimate the overall accuracy of a force field.
@@ -27,7 +27,8 @@ If you plan to explore a restricted family of molecules or materials very extens
 for more efficient MM simulations or fit a custom force field for your application.
 
 Note that recent developments in machine learning (ML) research have greatly improved the tools for fitting new force fields for MM-based applications.
-ML research has also generated very extensive sets of atomistic training data that are being used to improve the transferability of big-data force fields.
+ML research has also generated very extensive sets of atomistic training data such as [OMol25](https://arxiv.org/abs/2505.08762)
+that are being used to improve the transferability of big-data force fields.
 In principle, SQM model development can also benefit from the data and fitting tools being generated by ML research.
 However, there has been far less research activity in SQM than MM for many decades now, so the progress of SQM in this direction has been much slower.
 
@@ -50,28 +51,28 @@ to a non-interacting system of Kohn-Sham electrons using [DFT](https://doi.org/1
 These exact mappings are computationally intractable just like the exact force field, but they can be approximated by SQM models and practical DFT functionals.
 
 An important distinction between SQM models and DFT functionals is that DFT is used to approximate electron correlation effects beyond a mean-field calculation
-in a large basis set while SQM models are additionally used to approximate the complete basis set limit beyond a minimal-basis calculation.
-More is required of an SQM model, but they also tend to have more free parameters than DFT functionals.
+in a large basis set while SQM models use a small basis set and also rely on the model to approximate the complete basis set limit.
+SQM models are thus used to correct for more sources of error than DFT functionals, but they also tend to have more free parameters.
 Although available SQM models are not as accurate as popular DFT functionals, that is more a reflection of relative development effort and not a fundamental limit.
 DFT functional development has been an extremely popular research activity in chemistry, condensed-matter physics, and materials science for several decades,
 while SQM model development has been kept alive by only a handful of people and research groups since its popularity among method developers faded in the 1970s.
 
 An important additional limitation of MNDO-family models is that they do not directly use diatomic model parameters.
 Most parameters are specific to individual chemical elements and describe their average behavior in all chemical environments.
-Interatomic matrix elements between pair of elements come either from a Slater-type orbital approximation of their valence atomic orbitals,
+Interatomic matrix elements between pairs of elements come either from a Slater-type orbital approximation of their valence atomic orbitals,
 or a Klopman-Ohno approximation of their atomic charge distributions. In either case, a single fixed number---a Slater exponent or a charge radius---is
-used to characterize the average electronic "size" of each atom. The NDDO approximation also neglects electronic charges from pairs of atomic orbitals on different atoms.
+used to characterize the average electronic "size" of each atom. The NDDO approximation also neglects electronic charges from overlapping pairs of atomic orbitals on different atoms.
 Some of the later MNDO-family models like PM6 and PM7 added a limited set of diatomic interatomic repulsion terms to fix error outliers
-and also include diatomic parameters indirectly through their use of dispersion models such as Grimme's [D3 model](https://doi.org/10.1063/1.3382344).
+and also included diatomic parameters indirectly through their use of dispersion models such as Grimme's [D3 model](https://doi.org/10.1063/1.3382344).
 While the lack of diatomic parameters limits the accuracy and transferability of MNDO-family models,
 it was an essential limitation that kept the number of model parameters low enough to be fit to the scarce experimental data that was available in the 1980's.
 
 A nuance of MNDO-family models that complicates their use is that they were designed to approximate heats of formation rather than ground-state total energies.
-A Hartree-Fock calculation is performed on the MNDO-family many-electron Hamiltonian, and the ground-state total energy of that calculation is assigned to be
-the heat of formation at standard temperature and pressure.
-The simple reason for this design decision was that experiments directly measured heats of formation and not total energies.
+A Hartree-Fock calculation is performed on a MNDO-family many-electron Hamiltonian, and the ground-state total energy of that calculation is assigned to be
+the heat of formation at a standard temperature (25°C) and pressure.
+The simple reason for this design decision was that experiments directly measure heats of formation and not total energies.
 If the model output was a total energy, then additional corrections for atomic vibrations would be needed to approximate the heat of formation.
-The direct modeling of heat is formally awkward because it is a statistical property that only relates to a specific atomic configuration if
+However, the direct modeling of heat is formally awkward because it is a statistical property that only relates to a specific atomic configuration if
 the thermal distribution over atomic configuration is tightly peaked around a single equilibrium configuration.
 The heat of formation for a non-equilibrium atomic configuration doesn't make formal sense without introducing constraints that cause it to be an
 equilibrium configuration (e.g. in an external potential).
@@ -112,7 +113,7 @@ see the [COSMO paper](https://doi.org/10.1039/P29930000799).
 The interactions between a solvent and solute can primarily be categorized as electrostatic and statistical. For solute molecules that have a net charge,
 dipole moment, or highly non-uniform charge distribution, the largest solvation effects are electrostatic in nature.
 The solvent polarizes in response to the solute's charge distribution, and this response is approximated by the COSMO model as an ideal conductor located
-at a solvent accessible surface (SAS) of the solute molecule. The electrostatic energy between the SAS conductor and solute is then modified to approximate
+on a solvent accessible surface (SAS) of the solute molecule. The electrostatic energy between the SAS conductor and solute is then modified to approximate
 the energy associated with a finite dielectric constant. The SAS is formally defined by a union of atom-centered spheres with radii proportional to each atom's
 Van der Waals radius, but the SAS is approximated by tessellation to a triangular mesh for numerical work. The tesselation can be adjusted by keywords in the
 MOPAC input file to test numerical sensitivity.
@@ -131,9 +132,9 @@ Numerical errors in SQM and QM calculations are usually much smaller than the er
 Users should be careful when evaluating finite-difference-based derivatives of numerical outputs from MOPAC. The atomic forces evaluated by MOPAC are only
 partially analytical derivatives and do typically contain terms that are evaluated by finite differences. MOPAC evaluates all second derivatives
 (vibrational matrix elements) as finite-difference derivatives of the atomic forces. While MOPAC has keywords that can be used to adjust many numerical details,
-the finite-difference step size cannot be adjusted without altering the source code. The default convergence tolerances of the self-consistent field (SCF) cycle
+the finite-difference step sizes cannot be adjusted without altering the source code. The default convergence tolerances of the self-consistent field (SCF) cycle
 were chosen to achieve reasonable overall accuracy in finite-difference derivatives, but that source of error can be further reduced by setting tighter SCF tolerances
-using keyword in the MOPAC input file.
+using keywords in the MOPAC input file.
 
 Calculations of periodic systems have additional sources of error. MOPAC does not use Brillouin zone sampling, so the only way to reduce finite-size effects in
 periodic calculations is to expand the unit cell into supercells of increasing size. MOPAC also does not have a proper periodic electrostatics solver (e.g.

content/about/contributing.md

@@ -6,7 +6,7 @@ summary: "how to contribute to the open-source development of MOPAC"
 weight: 5
 ---
 
-As an open-source project, MOPAC is open to contributions of a variety of forms. Prior notice is not required for contributions,
+As an open-source project, MOPAC is open to contributions in a variety of forms. Prior notice is not required for contributions,
 but it may be worthwhile to discuss possible contributions as a GitHub [Issue](https://github.com/openmopac/mopac/issues)
 or [Discussion](https://github.com/openmopac/mopac/discussions) or [by email](mailto:openmopac@gmail.com).
 
@@ -15,14 +15,14 @@ Developers interested in contributing to MOPAC's source code should read the
 Note that the majority of the code base is written in Fortran 90, with small amounts of
 C, Python, CMake, and Javascript. Welcome software contributions include new features, performance optimization,
 and expanded test coverage. Reorganization or refactoring of the code base is also welcome, but such contributions
-may be more challenging and slow to accept.
+may be more challenging and slower to be accepted.
 
 If you rely on MOPAC for some regular workflow, then you should consider contributing tests that are representative of that workflow.
 This will greatly reduce the chance that future MOPAC development will interfere with your workflow and allow you to benefit from
-future versions with less worry. Some tests may be used for future performance optimization, so contributing tests increases the chances
+future versions with less worry. Some tests may be used for future performance optimization, so contributing tests also increases the chances
 that your MOPAC-based workflow will be made more efficient in the future.
 
-The manual on this website has been retained from the old Stewart Computational Chemistry website, which is raw HTML that was developed
+The manual on this website has been retained from the old Stewart Computational Chemistry website, which is raw static HTML that was developed
 using Microsoft Frontpage. It would be beneficial to migrate more of that content into Hugo (e.g. it could be indexed by Hugo's search functionality),
 but that would be a lot of work and is only viable as a community effort. Such web development contributions are very welcome
 and don't require knowledge of Fortran or the nuances of MOPAC's code base.

content/about/history/_index.md content/about/history/index.mdcontent/about/history/_index.md renamed to content/about/history/index.md

@@ -47,14 +47,14 @@ in 1992, which included d orbitals. MNDO/d was then incorporated into MOPAC in t
 
 ## Austin era
 
-The development of MOPAC began in 1981 after [James "Jimmy" Stewart](stewart_bio/) joined the Dewar group as a visiting scholar.
+The development of MOPAC began in 1981 after [James "Jimmy" Stewart](../stewart_bio) joined the Dewar group as a visiting scholar.
 Stewart was a professor at the University of Strathclyde in Scotland with a lot of prior software development experience in quantum chemistry.
 He was tasked by Dewar to refactor the research software that his group had produced during the development of the MINDO/3 and MNDO models into
 a unified, user-friendly computer program. The first release of MOPAC was in 1983 as software available by mail order from the [Quantum Chemistry Program 
 Exchange](https://en.wikipedia.org/wiki/Quantum_Chemistry_Program_Exchange) (QCPE). After GAUSSIAN became commercial software and was removed
 from the QCPE catalog, MOPAC became the most popular software on the QCPE.
 
-![James Stewart](jstewart.jpg "James Stewart")
+![James Stewart](/images/jstewart.jpg "James Stewart")
 
 The Dewar group also simultaneously developed the Austin Model 1 (AM1) semiempirical model alongside MOPAC, which included the development of 
 new model parameterization software. Stewart also contributed to the development of AM1 and its parameterization software, which he would eventually release as the PARAM
@@ -92,7 +92,7 @@ the Dewar group founded the company [Semichem Inc.](http://www.semichem.com) in
 As part of a broader investment in atomistic simulation software, Fujitsu hired Stewart as a consultant and acquired the distribution rights to future versions of MOPAC.
 The first Fujitsu commercial release of MOPAC was MOPAC 93 in 1993. Fujitsu later acquired the CAChe software from Oxford Molecular Group in 2000 around the time of that
 company's bankruptcy and liquidation, and MOPAC and CAChe formed the foundation for Fujitsu's SCIGRESS software. MOPAC was adapted into the MO-G simulation engine of
-SCIGRESS and its development continued at FQS Poland subsidiary of Fujitsu. Some Fujitsu-exclusive semiempirical quantum chemistry features include the PM5 model and the
+SCIGRESS, and its development continued at FQS Poland subsidiary of Fujitsu. Some Fujitsu-exclusive semiempirical quantum chemistry features include the PM5 model and the
 [LocalSCF](https://doi.org/10.1063/1.1764496) fast solver algorithm, which is technically distinct from MOPAC's 
 [MOZYME](https://doi.org/10.1002/(SICI)1097-461X(1996)58:2%3C133::AID-QUA2%3E3.0.CO;2-Z) fast solver algorithm.
 
@@ -101,9 +101,9 @@ SCIGRESS and its development continued at FQS Poland subsidiary of Fujitsu. Some
 Stewart stopped working for Fujitsu in 2004 and distributed subsequent versions of MOPAC through Stewart Computational Chemistry
 in partnership with third-party commercial resellers.
 Versions of MOPAC from this era were free for academic use, but it required a commercial license for government and industrial use. MOPAC's model coverage of the periodic table was
-greatly expanded to include the first 83 elements of the periodic table (with the lanthanides being modeled only in the +3 valence as classical "sparkles") in the
-[PM6](https://doi.org/10.1007/s00894-007-0233-4) and [PM7](https://doi.org/10.1007/s00894-012-1667-x) models. MOPAC's support for biomolecular simulation, particularly of protein 
-structures from the [Protein Data Bank](https://www.rcsb.org) (PDB), was also improved.
+greatly expanded to include the first 83 elements of the periodic table in the [PM6](https://doi.org/10.1007/s00894-007-0233-4) and [PM7](https://doi.org/10.1007/s00894-012-1667-x) 
+models (with the lanthanides being modeled only in the +3 valence as classical "sparkles"). MOPAC's support for biomolecular simulation was also improved, particularly of protein 
+structures from the [Protein Data Bank](https://www.rcsb.org) (PDB).
 
 ![NIH](nih.png)
 
@@ -118,7 +118,7 @@ For a summary of Stewart-authored publications related to MOPAC development afte
 ![MolSSI](molssi.png)
 
 In 2019, Stewart began a partnership with the Molecular Science Software Institute (MolSSI) to migrate MOPAC into an open-source software project.
-On the technical side, this process involved improving the portability of MOPAC's code base, reorganizing its to be easier to understand and contribute to,
+On the technical side, this process involved improving the portability of MOPAC's code base, reorganizing it to be easier to understand and contribute to,
 adapting its building, testing, and distribution to the modern open-source ecosystem, fixing some performance issues, and adding new interfaces to improve accessibility.
 On the social side, this involved contacting and coordinating with various stakeholders and collecting essential institutional knowledge.
 This process concluded in 2022 with a legal transfer of MOPAC's intellectual property to Virginia Tech and its first open-source release.

content/about/history/stewart_bio/index.md

@@ -3,9 +3,6 @@ draft: false
 title: 'James "Jimmy" J. P. Stewart'
 description: "Biography of James Stewart"
 weight: 1
-
-_build:
-  list: false
 ---
 
 ![James Stewart](../jstewart.jpg)

content/about/software.md

@@ -7,7 +7,7 @@ weight: 99
 ---
 
 Given MOPAC's age and popularity, there is a lot of software that works with MOPAC in a variety of ways.
-To get the most out of MOPAC, you should consider the use of one or more of these programs.
+To get the most out of MOPAC, you should consider using one or more of these programs.
 Unless otherwise noted as commercial, the software listed here is free to use for all purposes.
 
 
@@ -33,6 +33,10 @@ A commercial GUI focused on simulations of materials that has support for MOPAC.
 
 An atomistic simulation GUI built from open-source visualization components JSME, CH5M3D, and Jmol/JSmol with support for MOPAC.
 
+### [Molecular Operating Environment (MOE)](https://www.chemcomp.com/en/Products.htm) (commercial)
+
+A commercial GUI focused on drug discovery that has support for MOPAC.
+
 ### [Signals ChemDraw](https://revvitysignals.com/products/research/chemdraw) (commercial)
 
 This popular GUI software has gone through multiple names and owners (most recently Chem3D Ultra from PerkinElmer), and it is now developed and sold by Revvity Signals Software.
@@ -69,11 +73,11 @@ For example, a driver can enable molecular dynamics in an engine that doesn't na
 Note that MOPAC executables not distributed with MDI support enabled. To use MOPAC as an MDI engine, you must build it yourself with the CMake command-line option
 `-DMDI=ON`.
 
-### [Mopactools](/mopactools/)
+### [mopactools](/mopactools/)
 
 A Python wrapper for MOPAC, for both command-line usage and its native API.
 
-### [Pymopac](https://pymopac.readthedocs.io)
+### [pymopac](https://pymopac.readthedocs.io)
 
 Another Python wrapper for MOPAC, for both command-line usage and its native API.