This site keeps growing as I run into more things worth building. There's no fixed release schedule, so entries land whenever a tool or guide is ready, roughly newest first below. Have a suggestion? Open an issue on GitHub.
🧪 Isocyanate Cure Conversion: one number, measured four ways
A polyurethane cure has one quantity worth tracking – how much of the isocyanate has reacted – and no instrument measures it directly. The new Isocyanate Cure Conversion page takes what the instruments do report and turns each into a conversion: a DSC residual exotherm, an FTIR band at 2270 cm−1 ratioed against an unchanging reference, or a glass transition through the DiBenedetto equation. Three routes resting on different physics, side by side, so when they agree the number is probably right and when they diverge the pattern says which to distrust.
The part that catches people out is that the formulation decides how far the cure can go before any measurement means anything. Run the isocyanate in excess and the hydroxyl runs out first, capping conversion at 1/r: at index 105 that ceiling is 0.952, so a specification demanding 98 % NCO conversion is asking for something arithmetically impossible rather than merely slow. Water belongs in that arithmetic and it counts twice, because it takes one isocyanate to make an amine and a second to make the urea. In the worked example, 200 ppm of water in the polyol moves the index by more than two points – larger than most people's weighing error.
From the conversions it fits second-order cure kinetics in the general form, ln[(1 − rα)/(1 − α)] = kapp(1 − r)t, rather than the balanced-stoichiometry shortcut; with two cure temperatures it returns an activation energy, and with that a cure time at any other temperature. Past the gel point it counts elastically effective strands by the same Miller–Macosko recursion the Network Formation page uses, so the two cannot drift apart, and reports the rubbery modulus to set against what a DMA actually measures. There is a section on how each method fails too – baseline drift on a nearly cured DSC scan, carbodiimide and CO2 absorbing near the isocyanate band, and vitrification quietly stopping a reaction the kinetic fit still thinks is running.
📱 Four things that made a page scroll sideways on a phone
A page that scrolls left and right on a phone is a small thing that feels broken every time you touch it. I measured instead of scrolling: each of the 46 pages loaded at a fixed phone width, comparing how wide the page actually was against how wide the screen is. Two pages scrolled at 375 pixels, the commonest phone width, and five at 320. None do now.
Two of the four were long chemical names with nowhere to break. This site is made of names like Poly(bisbenzimidazobenzophenanthroline) — one word, no spaces, no hyphens — and a browser will not break a word unless you let it. One of those renders about 295 pixels wide inside a 262 pixel card, so the name simply ran out past the edge. Names that happen to contain a hyphen were always fine, because a browser will break after a hyphen, which is why this only showed on some entries. The other was a note containing acrylate/methacrylate/acrylamide/methacrylamide: 47 characters, and a slash is not a break either.
The third was the two big cards at the top of the home page, which were set to be at least 340 pixels wide. That is exactly what it sounds like, including on a 360 pixel screen, so they hung 4 pixels off the side. Every grid on the site can now shrink below its preferred width rather than insisting on it. The fourth was the capability table on the diagnostics page, where a rule that keeps a column on one line — correct on a desktop — was still in force after the table had been restacked into cards for a narrow screen.
All four are now checked automatically, so they cannot come back quietly.
🎨 Every colour on the site is now readable in both themes
Text needs to stand out from what is behind it by a measurable amount — 4.5 to 1 for ordinary type, which is the standard the accessibility guidelines set. I measured every page instead of trusting my eye: each of the 46 pages loaded in turn, every element carrying its own text compared against the colour actually painted behind it, in light mode and in dark. It came back with 89 failures in light and 36 in dark. There are now none.
Almost all of it came from one habit: a colour picked while looking at one theme and then used in both. The site has a light and a dark palette for text and backgrounds, but a good number of individual colours were written as fixed values, so whichever theme they were chosen against was the only one they suited.
The worst cases were invisible rather than merely faint. Tin, sodium, zinc, aluminium — any element in the structure drawing canvas outside the eleven that have a set colour — were drawn in near-black on the dark card, a ratio of 1.14 to 1. The symbol was there and could not be read. The lives indicator in the chain game was mint green on a light panel at 1.09 to 1. And links in five places, including the glossary definitions and the footer, had never been given a colour at all, so they fell back to the browser default blue: fine on white, 1.76 to 1 on a dark card.
Where a colour carries meaning it was kept and only its lightness changed, separately per theme. Oxygen still reads red and sulfur still yellow in a drawn structure; sulfur was simply 1.92 to 1 on white before, which is not a readable yellow. The coloured stripes and tints on the home page cards are untouched, because they are decoration and they are what tells the cards apart — only the small text took a deeper shade of the same hue. The one place a colour had to move properly is the red arcade button, which is 13% darker: its yellow lettering was 2.97 to 1 while hovered, and no lettering colour at all, white included, clears the bar on the red it used to lighten to.
Several of these had been sitting there since the pages were first written. None of them would show up in testing, because nothing errors and every page still looks broadly right — you have to be in the other theme, on that page, looking at that one label.
⚙️ Five changes to the structure drawing canvas
Ring templates other than benzene were the wrong size. The ring tool stamps any ring from three to eight atoms, and all seven buttons shared a single radius — the one that is correct for a hexagon. A cyclopropane came out with bonds 73% too long and a cyclooctane 23% too short, which is why a three-ring drawn beside a benzene measured exactly the same width across. Every size now uses the bond length benzene has always had. Benzene and cyclohexane are unchanged, so nothing you drew before has moved.
Loading a worked example could drop it off the screen. The examples are laid out around the middle of the canvas, but the view stayed wherever the previous drawing had been zoomed to. Zoom in, click a load example button, and the structure arrived running off all four edges. Loading an example now returns the view to the one it was drawn for.
There is a Fit button. Next to the zoom controls: it brings the whole drawing into view and centres it, counting brackets, reaction arrows and text labels as well as atoms. Until now a drawing that had grown past the window had to be chased around with the arrow keys or the middle mouse button. It will enlarge a small structure a little, but not without limit — fitting a two-atom fragment to the full canvas would give it bonds a quarter of the screen wide.
Element symbols were unreadable in dark mode. Eleven elements have a set colour, the ones a chemist reads as chemistry: oxygen red, nitrogen blue, sulfur yellow. Everything else — tin, sodium, zinc, aluminium, lithium, anything picked out of the periodic table — was drawn in near-black, which on the dark card is a contrast ratio of 1.14 to 1. The symbol was being painted and could not be read.
The set colours had a quieter version of the same problem: one fixed palette was being used on both a white and a dark background, and not one of the eleven met the 4.5:1 standard on both. Bright hues wash out on white and dark ones vanish on the dark card. Sulfur’s yellow measured 1.92:1 on white, which is the default theme. Each theme now gets its own lightness for those colours with the hue left alone, so oxygen still reads red and sulfur still reads yellow, and every element clears 4.5:1 either way. Exported images are unaffected: they force a white background on purpose and take the colours that suit it, whichever theme you export from.
Ring sizes are on the number keys. The number keys did nothing before, so 3 through 8 now pick the ring of that many atoms and Shift+6 picks benzene. The rule is just “N gives an N-membered ring”; benzene is the aromatic form of the same ring, so it sits on a modifier of the same key rather than getting one of its own. Each ring button’s tooltip names its shortcut. Typing in a text box is unaffected.
๐ Six calculators now have shareable links
The Copy results button has always pasted the page address underneath the numbers it copies, which rather implies the link reproduces them. It did not: the address carried none of the inputs, so whoever you sent it to opened the page on its defaults and saw different numbers under your figures.
Six tools now keep their inputs in the address bar and read them back on arrival: hydrogel mesh size, Flory–Huggins, crosslink density, the distribution simulator, copolymer composition and end-group analysis. Only what you changed goes in, so an untouched page keeps a clean address and a shared one carries exactly the numbers you altered. On crosslink density that means both routes travel together, so the disagreement between swelling and modulus goes with the link rather than having to be described.
Tools whose input is a pasted dataset are deliberately left out. A chromatogram or a set of frequency sweeps is kilobytes of numbers, which belongs in a file rather than a web address.
๐ Search the library by what a repeat unit contains
The reference library stores a drawn structure for 751 polymers, and until now the only structural questions it could answer were about a whole repeat unit: is this the same polymer, how similar is it, where does it sit on the map. It can now be asked what a unit contains. Twenty-three groups are built in, from ester and amide to imide, anhydride, epoxide and difluoromethylene, and each one is a chip on the search page with the count beside it.
Put next to the row above it, which says what the chain is made of, the two together draw a distinction no tag could: 288 polymers contain an ester and 56 have one in the backbone; 160 contain an amide against 65. The gap is every pendant ester on an acrylate and every amide hanging off a polyacrylamide. Nothing here is declared in the data. It is all read off the structure.
Two details that took some care. A linkage sitting where the repeat bracket happens to be drawn is still in the polymer, so matching runs on the closed unit, which is what lets poly(ethylene oxide) be found as an ether. Rings are the exact opposite: closing a short unit manufactures one, and poly(ethylene oxide) closed is a three-membered ring, so ring groups are matched on the drawn unit instead. Getting that backwards had PEO listed as containing an epoxide, which it is merely made from.
๐ง Hydrogel Mesh Size
A hydrogel is specified by its mesh size more often than by its modulus, because that length decides what can diffuse through it. The new page gets one from a swelling measurement you can make on a balance: dry mass and swollen mass give the polymer volume fraction, Peppas and Merrill's relation gives the molar mass between crosslinks, and Canal and Peppas turn that into the mesh. Give it a solute radius and it estimates what fraction of its free-water diffusion rate that solute keeps, or tells you it does not fit at all.
The reason it is not a card on the crosslink density page is the reference state. A gel crosslinked in water is already swollen when its strands are tied together, so the dry network is not unperturbed and the Flory and Rehner form counts the stretch twice. The relaxed-state volume fraction is an input here, and it is the commonest thing to get wrong, because getting it wrong leaves the answer looking perfectly plausible.
The network is drawn to scale beside your solute, so whether it fits stops being arithmetic. No chain constants are typed into the page: the bond length, bonds per repeat and characteristic ratio come from the chain dimensions dataset, which carries a source for each.
๐งช End Group & Pendant Group Analysis
Three different questions get called end-group analysis, and picking the wrong instrument is the usual reason an answer disappoints: what is on the end, how many chains carry it, and how long the chains are. The new guide lays eight techniques out as cards, each saying what it is best at and what it costs you to get a molecular weight out of it, with calculators for Mn by NMR integration, degree of substitution, and titration equivalent weight.
The familiar ceiling around 20 kg/mol is drawn rather than asserted: the same spectrum window at DP 25, 100 and 400, each normalised to its backbone peak, with the end-group signal at forty times vertical expansion in all three so the collapse you are watching is real and not a change of scale. A MALDI figure does the other half, where the peak spacing names the monomer and the offset names the ends. The titration card draws the functionality trap as what it is: Mn is that many identical blocks, so reading a diol as a triol is a 50% error, not a 5% one.
๐ The GPC trace analyzer now says where its error comes from
Band broadening is usually treated as an instrument constant you measure once. It is, but the constant is set by two numbers on the column datasheet, and you chose them when you bought the columns. The trace analyzer now takes a column series, bank size and length, and reports how much of the dispersity it printed belongs to the columns rather than the polymer.
Two results fall out that are worth knowing before the next purchase order. Range you do not need is not free: a series sold as covering more decades fits them into the same separation volume, which is a steeper calibration, and the wide-range grades are also the low-efficiency ones, so the same injection through a 20 µm mixed bed carries more than twice the broadening of a narrow one. And doubling the bank buys back exactly √2, which is the familiar square-root-of-length result arrived at from the dispersity side, and why the fourth column is a much worse buy than the second.
๐ Five tools that work on distributions, phases and networks
Every calculator here used to take a few numbers in and give a few numbers back. These five take your actual data, or return a whole curve, and between them they cover the two areas the site had nothing on at all: thermodynamics of mixing and viscoelasticity.
- GPC Trace Analyzer. Paste the raw chromatogram and compute Mn, Mw, Mz and Đ yourself, with what the integration limits were worth shown next to the answer. Clipping always narrows a distribution: a sample that is truly Đ = 1.300 reads 1.244 when integrated to 10% of peak height.
- Distribution Simulator. The whole molecular weight distribution a controlled polymerization should give, the dead-chain tail, and what band broadening does to it. Terminating a tenth of the chains drops Mn 8% and barely moves Mw.
- Flory–Huggins Phase Behaviour. Binodal, spinodal and critical point for a blend or a solution, the cloud point from χ(T), and the order–disorder transition of a diblock.
- Time–Temperature Superposition. Paste isothermal frequency sweeps and get a master curve, shift factors and WLF constants, plus why C₁ and C₂ mean nothing without their reference temperature.
- Network Formation. Gel point, sol fraction and elastic strand density for any mix of functionalities, by the Macosko–Miller recursive method.
Each one loads with data generated from a known answer, so the worked example is a measurement of the tool rather than a claim about it, and every published number is re-derived by the site's automated checks on each change.
๐ Chain Dimensions: how big is one chain?
A new tool for the other half of polymer characterisation: not what a chain is made of, but how it is shaped. Pick a polymer and a molar mass and it returns the persistence length, Kuhn length, contour length, end-to-end distance and radius of gyration, using the worm-like chain in its exact form rather than the Gaussian approximation, so it stays correct through the rod-to-coil crossover instead of only in the coil limit. A hundred base pairs of DNA comes back as a rod at 90% of its contour length; three kilobases comes back as a coil with a radius of gyration of 121 nm.
The point of the database is that persistence length is not a material constant. It belongs to a chain in a particular solvent at a particular temperature, and for a charged chain it moves with ionic strength too, so every one of the 25 measurements is quoted with the conditions it was made under and several polymers carry more than one. Single-stranded DNA is 1 nm in high salt and 2.2 nm when the phosphates stop being screened; cellulose reads 6.5 or 9 nm depending on which aggressive solvent dissolved it. Where the values disagree, the disagreement is the data. The 19 entries span from poly(ethylene oxide) at 0.35 nm to an actin filament at 17 ยตm, close to five orders of magnitude.
Where the backbone is a simple aliphatic one the page also predicts lp from the characteristic ratio by geometry alone, and agrees with measurement to within about 15% for polyethylene, polystyrene, PMMA and polypropylene. It refuses to predict for aromatic backbones, because that route genuinely fails there: bisphenol-A polycarbonate has C∞ = 2.4, which the formula turns into 0.22 nm against a measured 1.0. Open Chain Dimensions.
๐ 27 CAS numbers corrected, and you can search by them
Twenty-seven entries in the Polymer Search library listed the CAS number of the monomer where the polymer's number belongs. Poly(caprolactone) carried 502-44-3, which is epsilon-caprolactone, the ring, rather than 24980-41-4, “2-Oxepanone, homopolymer”. It is an easy substitution to make, since the registries index a polymer under the monomer it came from, and a quietly damaging one on a page people copy identifiers out of and paste into a purchase order. All twenty-seven are now the polymer registry number, each checked against EPA's Substance Registry Services or the CAS index name. PCL, PLA, PGA, PCTFE, PECH, poly(butyl acrylate) and poly(alpha-methylstyrene) were among those affected. The data checker now rejects both the specific monomer numbers that were wrong and, more usefully, anything falling outside the ranges CAS actually assigns to polymers, so the error cannot come back the next time the discovery script seeds an entry from monomer data. The name box also searches CAS numbers now, which it never did, so a number off a result card can be pasted straight back in. Open the Polymer Search.
๐งฌ Copolymers in the thermal analysis library
The DSC, TGA and DMA charts now draw from a 70-material library, up from 40, and copolymers have their own category for the first time: 15 of them, including the styrenic thermoplastic elastomers (SBS, SIS), the self-assembling diblocks (PS-b-PMMA, PS-b-P2VP, PS-b-PDMS), the amphiphilic carriers (PEG-b-PLA, Pluronics, PS-b-PEO), and the segmented TPEs (PEBAX, Hytrel). Fifteen homopolymers joined too, filling the obvious gaps: PEO, PCL, PVA, PVAc, PAA, PNIPAM, PVP, PBT, PSU, PPS, PEI and PAN among them. Every entry carries its own provenance and states plainly which numbers are measured and which are estimated, because for block copolymers most of them are composition dependent. A phase separated block copolymer has two glass transitions and the charts draw one, so each entry names the second in its notes rather than pretending it is not there. Open Thermal Analysis.
๐งฑ 29 block copolymers in the reference library
The Polymer Search library went from 2 block copolymers to 29: the styrenic thermoplastic elastomers (SBS, SIS, and SIBS, the TAXUS stent coating), the canonical self-assembly diblocks (PS-b-PMMA of DSA lithography, PS-b-P2VP, PS-b-PDMS), the clinical PEG carriers (PEG-b-PLA of Genexol-PM, the ReGel thermogel triblock), polymersome-forming PB-b-PEO, and the acrylic, silicone, and segmented families (Nanostrength MAM, PEBAX, Hytrel). Nearly all deliberately carry no CAS number: vendor numbers for block copolymers routinely resolve to a homopolymer or a random copolymer of the same monomers, and a wrong identifier is worse than none. The “elastomer” tag is restricted to architectures that actually form a physical network. A diblock with one free chain end does not, however rubbery its soft block. Explore tag chips now also narrow (every selected tag must apply) instead of ORing together. Open the Polymer Search.
๐ Thermal charts scale to fit their data
Every chart on the Thermal Analysis page now fits its axes to the curve instead of assuming a fixed range, so nothing is clipped and a linear polymer’s terminal flow stays on the DMA plot. Marker labels stagger and flip near the edges rather than colliding, and the “binder transitions” view now crops below the first melt as well as the first decomposition, so a binder’s glass transition is actually visible next to a filler that melts before it decomposes.
๐ฅ Thermal Analysis guide: DSC, TGA & DMA
A new interactive guide to the three workhorse thermal techniques, each with a live plot you reshape and read out. Drag a DSC thermogram by heating rate, sample, and first-versus-second heat to watch the glass transition, cold crystallization, melting, and percent crystallinity respond; a TGA/DTG curve by composition and atmosphere to split moisture, polymer, carbon black, and mineral filler; and a DMA plot by frequency and crosslink density to see E′, E″, tan δ, and the three temperatures DMA calls Tg. Open Thermal Analysis.
๐ Sharper publication matching
The publications list on a search result now keeps only papers that actually name the polymer, its monomer, or an alias, instead of everything that shares a family word like “methacrylate.” When a niche polymer has no dedicated literature indexed, it now says so plainly and points you to the PubChem, Scholar, and Patents links rather than filling in with tangential hits. Open the Polymer Search.
๐ฟ Bio-based methacrylates in the reference library
Poly(geranyl methacrylate) and its fully hydrogenated counterpart poly(tetrahydrogeranyl methacrylate), both derived from the plant terpene geraniol, joined the Polymer Search library. They match by name or by drawn structure, and carry a bio-based tag so the renewable-feedstock entries can be browsed together.
๐งฌ Microstructure prediction on the Copolymer Composition tool
The Mayo-Lewis calculator now reads out the likely chain microstructure (alternating, statistical, ideal/random, blocky, or strongly drifting/gradient) from the reactivity-ratio product r₁·r₂, alongside the composition and azeotrope it already computed. The reference table was expanded to 14 common monomer pairs, each value cross-checked against a second source, with the tendency computed live. Open the Copolymer Composition tool.
๐งฉ Copolymer recognition in Polymer Search
Draw a copolymer and bracket each repeat unit separately: the structure search identifies every block with the same exact-match and PubChem machinery a homopolymer uses, then reports the copolymer and pulls publications for the monomer combination. Common named copolymers (SBR/SBS, SAN, ABS, nitrile rubber, PLGA, EVA, EPDM, the Pluronic PEO-PPO surfactants, PS-b-PEG) are in the library too, so they resolve by name and surface automatically when their component blocks are drawn. Open the Polymer Search.
๐ฏ Dispersity (ฤ) Predictor
Predict the narrowest ฤ a controlled radical polymerization can reach from its target DP, conversion, and exchange chemistry. One equation covers RAFT (via the transfer constant Cex), ATRP and NMP (via the deactivator level), and the ideal living limit, with preset RAFT-agent/monomer pairs and ATRP control scenarios drawn from the literature. Shows the molecular-weight distribution as a live SEC trace and how ฤ falls across conversion, with the assumptions the model leaves out stated up front. Open the Dispersity Predictor.
๐งฑ Polyurethane tab in the Calculator
Two-step prepolymer formulation built around capping and chain extension: OH-number-based equivalent weights, NCO:OH capping ratio, theoretical %NCO with a measured-value override for the extension step, extender dosing by index, hard segment content, and a full suggested procedure from polyol drying through cure. Lives alongside ATRP/RAFT/ROMP/FRP in the Calculator, with saved recipes and share links.
๐งด Emulsion Polymerization tab in the Calculator
Micellar (Smith-Ewart) nucleation with SDS and KPS: predicted particle number and diameter from surfactant, initiator, and monomer charge, with a CMC check, solids content, and an advanced option to layer a RAFT or ATRP mediator on top for a controlled/living particle interior. Lives in the Calculator alongside the other techniques.
๐ ๏ธ Site polish: SEO, accessibility, and analytics
Search-friendly meta tags and structured data on every page, a custom 404 page, aria-labels on every icon-only button in the structure editor, a sitemap, and privacy-respecting analytics to see which tools people actually use.
๐ค Automated polymer discovery pipeline
A scheduled job checks a watchlist of common monomers against PubChem, derives the repeat unit structure for the straightforward cases (vinyl, acrylate, methacrylate, diene, and some ring-opening chemistries), and opens a pull request for review. Nothing reaches the live database without a manual check first.
๐ฎ Build a Polymer Chain
A small maze game for practicing skeletal formula conventions: collect the right monomer beads and dodge inhibitors and chain transfer agents before they terminate your chain.
๐ Polymer Search
Look up a common polymer by name, or draw its repeat unit in a freeform structure editor (with ring tools, wedge/hash bonds, and formal charges) and search a curated reference library by structure. Explore mode adds tag browsing and a photo-upload search that reads polymer names or CAS numbers straight off a label or textbook page.
๐ GPC Peak Interpretation, expanded
A troubleshooting quick-reference for common chromatogram artifacts, a multi-detector section, an interactive flow-rate drift calibration demo, and an image-upload trace analyzer for reading Mn/Mw off a real chromatogram photo.
โ๏ธ Recipe Scaling
Scale any polymerization recipe up or down by a factor, or between batch sizes, while keeping every component's ratio exact.
๐ก๏ธ Tg Predictor
Estimate a random copolymer or miscible blend's glass transition temperature from its components using the Fox equation.
๐ GPC Calibration Converter
Convert a polystyrene-equivalent molecular weight from a GPC report into a true molecular weight for the actual polymer, using Mark-Houwink parameters.
๐งฌ Copolymer Composition
Predict instantaneous copolymer composition from feed ratio and reactivity ratios with the Mayo-Lewis equation, including a composition diagram and azeotrope detection.
๐ Glossary, mobile pass, and recipe tools
A glossary of common polymer chemistry terms, a full mobile responsiveness pass, an interactive kinetics plot on the conversion monitoring guide, a print/export stylesheet for recipes, save/reload recipe presets, dispersity (ฤ) guidance on results, and a Stock Solutions tab on the calculator.
๐ Light/dark theme
The calculator and every guide page now follow a light or dark theme, with a toggle in the header.