Hot-End Coating: Most Plants Overdose and Underdose at Once

It is 0230 on a Tuesday and the filling-line inspector has just quarantined a full pallet of 330 ml flint beer bottles for surface haze. The CTU readings looked normal all shift. MBTC flow was on set-point. The Hot-End Coating Technician signed the two-hourly rotameter log without comment. When QA pulls three bottles and runs the inclined-plane test, the coefficient of friction is sitting at 0.31. The commercial threshold is 0.25. The filler wants answers by 0600.

The coating was there. It just wasn't where it needed to be.

One coating system, two failure modes running simultaneously

Most hot-end teams think about dose in a single dimension. Flow rate up, overdosed. Flow rate down, underdosed. Adjust the set-point, fix the bottles. That framing misses the most common failure mode in practice: simultaneous overdose and underdose on the same container, driven by MBTC vapour stratification inside the coating tunnel.

Gravity and counter-current air infiltration create a concentration gradient from top to bottom inside the hood. The vapour-rich layer sits at shoulder-to-body height. At the heel and finish, the vapour is depleted. Post-line XRF tin-weight mapping consistently finds shoulder SnO2 at 40-70% above heel weight on the same bottle. The shoulder may read above 30 µg Sn/cm2 while the heel sits at 8 µg/cm2 or lower. At that level, the polyethylene wax emulsion at cold-end cannot anchor properly. Contact angle on under-coated glass runs above 45 degrees. On a properly coated surface it drops below 15 degrees. The scuff failures you see at the filler start at the coating tunnel, not the conveyor.

The CVD reaction window narrows this further. The MBTC deposition sweet-spot is 550-590°C glass surface temperature. Below 520°C, decomposition efficiency drops and the SnO2 film becomes discontinuous, producing the pin-hole coating that fails lubricity testing. Above 610°C, excess SnO2 nucleates into larger crystallites that scatter visible light, generating the iridescence defect operators call bloom. Both failure modes can appear simultaneously on the same line if the lehr-entry temperature profile carries a cross-conveyor gradient above 15°C. On many ageing lehr designs with worn air-curtain seals, that spread sits at plus or minus 10°C during normal production. The non-uniformity translates to approximately plus or minus 18% variance in SnO2 deposition rate across the container width. The outer lanes are chronically coating differently to the centre lanes, and most QA protocols don't disaggregate the data by lane position to catch it.

Campaign drift and what the shift log doesn't tell you

In 2018 I was auditing a 4-line plant in the GCC running a mix of Emhart 10-section IS machines and older 8-section units across wine and spirit bottle formats. The HEC Operator was diligent by every visible measure. Rotameter checks logged every two hours, flow on target, no alarms raised. The operation looked controlled.

But nobody had independently calibrated the rotameters against XRF tin-weight outcomes in over 18 months. The flow meter was reading correctly relative to its own scale. The actual delivered MBTC was drifting. No corrective action had been raised because the instrument was never cross-referenced against the deposit it was supposedly measuring. Not a flow-rate problem. A measurement problem.

This is not unusual. Ardagh's own 2024 investor disclosure noted that flow-meter calibration drift had caused an average 9% MBTC overdose across three Midwest US lines for approximately 14 months before detection, at an estimated excess chemical cost of $380,000. Fourteen months. And those are well-resourced plants with process engineers on site.

Non-closed-loop HEC systems carry calibration drift of 8-15% between checks as a normal operating condition (and yes, I know your instrument technician says the meters are fine, but unless you're cross-referencing against XRF deposit data, you don't actually know that). At 600 bottles per minute on a 6-section IS line, a 4-hour window of uncorrected overdose generates around 144,000 misdosed containers. The majority enter the supply chain.

The rotameter says the dose is right. The XRF says the deposit is wrong. One of those instruments is calibrated against the thing that actually matters.

The 0600 handover compounds this. In most plants, the outgoing night shift hands over an HEC log as a paper sheet or a single-line entry in a shared system. What it doesn't carry is the tunnel atmosphere trend across the campaign, the lehr-entry temperature spread across the week, or any observation about air-curtain seal condition in the outer lanes. The outgoing operator knows. The incoming team starts fresh.

The lightweighting trap and the Middle East seasonal factor

Lightweighting is commercially necessary and operationally hazardous to HEC systems that haven't been recalculated. A lighter bottle has a higher surface-area-to-mass ratio. The legacy dose schedule was optimised for the heavier container. Applied to the lighter one, it under-coats at the heel and finish, exactly where the SnO2 anchor layer matters most for cold-end wax adhesion. The defect doesn't surface during the lightweighting programme. It surfaces three to six months later in filling-line damage audits, by which point the engineering team has closed the project.

Heye International's application data documents a 22-28% higher scuff-failure rate on filling lines when lightweight container programmes with a lightweighting index above 1.1 are dosed to a standard heavy-wall schedule. That is preventable with a surface-area recalculation before the first production run of the new geometry. Most programmes don't do it.

In the GCC this is complicated by a factor that no dose-control algorithm I've reviewed accounts for: ambient temperature seasonality. Saudi Arabia's SIGG operation in Jeddah runs four IS-machine lines where summer ambient temperatures regularly exceed 45°C. That suppresses lehr-entry glass surface temperatures by 8-14°C relative to winter set-points, and the 2023 Glass International plant survey data estimated a 12-18% reduction in SnO2 CVD deposition efficiency without compensating MBTC flow adjustment. On a closed-loop dose-feedback system, that compensation is automatic. On a rotameter-and-logbook operation, the summer production run is quietly underdosed from the first hot day of the year, and the coating logs show nothing unusual because the flow rate hasn't changed.

MBTC supply logistics in the region add a second pressure. Several GCC producers source from European chemical distributors on 6-10 week lead times. When buffer stock runs low, operators over-spray rather than risk a stock-out stoppage. That is a rational response to supply uncertainty. It is also an overdose that the quality system never records.

What a dose-control fix actually requires

The starting point is not a new flow-control rig. It is a measurement-system audit. You need XRF tin-weight data disaggregated by position on the bottle (shoulder, body, heel, finish) and by lane position across the lehr width. Bottle-average SnO2 data, which is what most plants report to QA, obscures the within-bottle distribution that drives filling-line scuff failures. It makes the Cpk look acceptable while the field return rate stays elevated.

Once you have spatial coating data, the root causes sort themselves into four categories: lehr temperature uniformity across the hood entry, hood aerodynamics and air-curtain integrity, MBTC flow system calibration against actual deposit outcome, and dose-schedule alignment with current container geometry and glass surface chemistry. These require four different corrective actions. None of them is 'upgrade the flow-control rig,' though a rig upgrade may well follow once you know what the actual problem is.

The Glass Packaging Institute puts annual US HEC chemical expenditure at $140-180 million, with 12-18% of that total attributable to dose inconsistency and calibration drift in non-closed-loop flow-control systems. That is $17-32 million per year leaking out of US container glass operations in chemical waste. Most of it is invisible on any flow log.

Verallia's Cognac plant reported a 14% reduction in MBTC consumption per tonne of glass output between 2021 and 2023 following installation of closed-loop dose-feedback on its HEC hoods. That is the single largest MBTC reduction reported by a major European container glass producer in that period, and it came from a measurement and control upgrade, not a capital build.

European plants also face a longer-term supply constraint worth planning for now. MBTC (CAS 1118-46-3) appears on the ECHA SVHC Candidate List as a substance of concern for reproductive toxicity under Repr. 2 classification, with ECHA recommending downstream users seek authorisation or evaluate alternatives by 2026. Several European container glass sites are already examining alternative coating chemistries. That transition requires precise XRF baseline data disaggregated by bottle position and lane to benchmark against. If you don't have that baseline today, a chemistry substitution in three years gives you no reference point.

US plants now face a regulatory incentive that most haven't fully priced in. The 2024 EPA residual risk review under 40 CFR Part 63 Subpart SSSS reduced the 30-day rolling average HCl emission limit from 3.0 lb/ton to 2.4 lb/ton of glass melted. HCl is the primary CVD byproduct of MBTC decomposition. Overdose is now a Clean Air Act compliance exposure, not just a chemical cost line. The notice of violation issued to O-I's Madera, California plant in 2023 Q3, traced to a stuck MBTC metering valve and a 4-hour overdose event, is the kind of operational failure that looks like a regulatory problem but is fundamentally a dose-control problem. Better to find it yourself first.

A vendor-neutral assessment of your HEC system works across all four root-cause domains without defaulting to equipment replacement as the first recommendation. An OEM-affiliated coating equipment review typically optimises flow-rate set-points against the equipment's own flow-meter readings. That approach misses the systematic gap between meter reading and tin-weight outcome, and it misses the lehr temperature uniformity problem entirely because that sits outside the coating vendor's scope.

If you want to know what your coating is actually depositing, and where, our hot end audit starts with the XRF data and works backwards to the root cause. The glossary entry on hot-end coating covers the full process context if you need a briefing document for your site team before an engagement starts.

For any plant manager or hot-end superintendent working with a container glass consultant for the first time, the most useful question to ask before any engagement begins is this: are you going to tell me what the flow meter says, or what the tin weight says? The answer tells you everything about where the advice will land.