What each molding process holds on a reagent bottle, in real numbers

A 250 mL lot came back with a 0.45 mm spread in body diameter across 12 cavities, and the filling line rejected 3.1% of units at the guide rail. The drawing said ±0.30 mm. The supplier’s inspection report said “within tolerance.” Both were true, because the supplier was measuring a single-cavity sample and the line was seeing the full mold. That is the gap this article is about: reagent bottle dimensional tolerance capability is not a drawing value, it is a measured distribution across every cavity, every shift, and every dimension that touches your equipment.

I have spent enough years on both sides of the purchase order to say plainly that most tolerance disputes are not quality failures. They are capability questions that nobody asked at the sampling plan stage. If you want to specify, audit, and accept molded reagent bottles on numbers rather than assurances, you need to know what each process can hold, how to measure it honestly, and which dimensions will fight you no matter who molds the part.

There are four processes you will encounter in this category, and they do not share a capability envelope. I have run capability studies on all four. The numbers below are what I would put in a purchase specification, not what a brochure claims.

Read that table as a floor, not a target. Those are the tolerances a well-run tool can hold at Cpk ≥ 1.33 on the body diameter over a 30-lot run. A supplier quoting ±0.10 mm on an extrusion blow molded 500 mL HDPE bottle is either quoting a single-cavity best case or has not run the study. Ask which. The answer tells you whether you are dealing with a molder or a trader.

Why the process, not the polymer, sets the floor

The polymer matters less than people think for dimensional spread. What matters is how the parison or preform is formed and how the mold closes. In EBM, the parison is extruded as a free-hanging tube, and its wall thickness varies with die swell, melt temperature drift, and gravity sag along the length. When the mold closes, that variation is baked into the bottle. You cannot hold ±0.15 mm on a body diameter when the parison wall is moving ±0.15 mm before the mold even touches it.

IBM and ISBM form the preform in a rigid cavity first. The preform has a known, controlled wall. The blow stage then inflates it against a temperature-controlled mold. The dimensional variation comes from preform temperature uniformity and blow pressure stability, both of which are tightly controllable. That is why IBM and ISBM hold roughly half the spread of EBM on the same nominal part.

If your filling line needs a body diameter tighter than ±0.25 mm, you are not buying EBM. You are buying IBM or injection molded, and you are paying for the tool. That is the trade, and it is worth naming explicitly in the RFQ.

How capability is actually measured on a molded bottle

A tolerance is a promise. Capability is the evidence. The two get conflated constantly in supplier documentation, and the conflation is where procurement gets burned.

Cpk on the dimension that touches your equipment

Capability index Cpk compares the distance from the process mean to the nearest specification limit against three standard deviations of the process. A Cpk of 1.33 means the process spread fits inside the tolerance band with a small cushion. A Cpk of 1.0 means you will see roughly 2,700 defects per million on that dimension. For a bottle that has to pass a filling guide rail at 200 units per minute, 1.0 is not acceptable. I specify Cpk ≥ 1.33 on the two or three dimensions that touch the line, and I do not bother with Cpk on cosmetic dimensions.

The study has to be structured correctly or the number is meaningless. Minimum requirements I put in writing:

• 30 consecutive lots minimum, not 30 parts from one lot. Between-lot variation is where EBM lives.
• Every cavity sampled at least 10 times across the run. A 12-cavity tool with one bad cavity shows a Cpk of 1.5 if you average across cavities and 0.8 if you track per cavity.
• Measurement system analysis first. If your caliper or vision system has a repeatability of 0.05 mm on a ±0.15 mm tolerance, you are measuring noise. Gauge R&R should be under 10% of tolerance.
• Stated temperature and humidity during measurement. Polypropylene moves about 0.15% per °C near room temperature. A 5 °C swing in the inspection room shifts a 60 mm diameter by 0.045 mm.

The per-cavity trap

Here is the failure I see most often. A supplier runs a capability study on a 16-cavity mold, pools all 480 measurements, and reports Cpk 1.6. The pooled number hides the fact that cavity 7 runs 0.18 mm high and cavity 12 runs 0.15 mm low. Each cavity is individually capable. The mold as a whole is not, because the spread between cavity means is wider than the tolerance band allows once you stack it on top of within-cavity variation.

The fix is a per-cavity mean chart. Plot the mean of each cavity over the run. If the range of cavity means exceeds about 40% of the tolerance band, the tool needs work, regardless of what the pooled Cpk says. On the 250 mL lot I opened with, the cavity means ranged 0.31 mm, and the tolerance was 0.60 mm total. That is 52% of the band consumed by cavity-to-cavity offset before any random variation. The line felt it.

For a related look at how dimensional decisions cascade into downstream handling, the HPLC autosampler vial world has the same problem in miniature: a 12 × 32 mm vial that is 0.2 mm out on crimp neck diameter will jam a 54-position tray just as reliably as a bottle jams a rail.

The dimensions that fight you, and why

Not all dimensions are equally hard. Some come out of the mold nearly free. Others require active process control or post-mold gauging. Knowing which is which lets you write a specification that a molder can actually meet.

Neck thread and finish: the hardest 15 mm on the bottle

The thread finish is where dimensional tolerance capability matters most and where it is hardest to hold. A GPI 24-400 finish has a specified thread major diameter, a T-dimension (thread root to sealing land), and an E-dimension (the sealing bead). The T-dimension on a 24-400 is nominally 17.5 mm with a tolerance of about ±0.25 mm in the GPI standard, but the sealing surface flatness and the thread start position are the killers.

Thread start position drifts with mold rotation and ejection timing. If the thread start moves 15 degrees from cavity to cavity, a torque-applied cap will cross-thread on some cavities and not others. I have seen a capping head set to 1.4 N·m that ran clean on 14 of 16 cavities and stripped on two, purely because the thread start on those two cavities was 20 degrees off. The bottles passed every dimensional check on the drawing. The drawing did not specify thread start position.

Specify it. Thread start angular position should be held within ±5 degrees across cavities. If the molder cannot do that, the tool needs a thread insert rework, and it is worth the cost.

Body diameter and roundness

Body diameter is easier on IBM and injection molded parts, harder on EBM. Roundness is the sneaky one. A bottle can be dead-on nominal diameter on the major axis and 0.4 mm oval. Ovality shows up as a filling nozzle misalignment or a label wrap failure, not as a diameter rejection. Specify roundness separately: ≤ 0.20 mm total indicator reading on the body for IBM, ≤ 0.35 mm for EBM.

Height and base flatness

Overall height tolerance is loose on most drawings (±0.5 mm is common) because it rarely matters for filling. Base flatness matters more than height on anything that runs on a conveyor. A base that rocks 0.3 mm will wobble on a high-speed line and trip a vision reject. Specify base flatness as a rock dimension: ≤ 0.15 mm on a granite surface plate.

Wall thickness and its hidden effect on weight

Wall thickness tolerance is not usually a fit problem, but it is a weight problem, and weight is a fill-volume problem when you are selling by the bottle. A 500 mL HDPE bottle with a nominal 1.2 mm wall that runs 1.05 to 1.35 mm has a mass spread of about 8 grams. That is enough to shift your per-unit cost and enough to shift the tare weight your gravimetric filling system has to compensate for. Specify wall thickness and weight together, and specify them per cavity.

Two field cases where the measurement found the problem

Case 1: the guide rail reject that was a mold temperature problem

A customer running 1 L EBM HDPE bottles on a 240-per-minute filler saw a 2.8% reject rate at the infeed guide rail. The bottles looked fine. Diameter checks on a sample of 20 read within ±0.28 mm, comfortably inside the ±0.40 mm drawing. The line still rejected them.

We pulled 200 bottles from three consecutive lots and measured every one with a laser micrometer at four heights (25, 60, 100, and 140 mm from the base). The mean body diameter was fine. The problem was a taper: the diameter at 25 mm ran 0.35 mm larger than at 140 mm on about 6% of units. The guide rail gap was set for nominal, and the tapered bottles wedged.

The cause was uneven mold cooling on one side of the tool. A blocked water channel on the non-operator side of cavity 4 and cavity 9 was running the mold surface 14 °C hotter than the setpoint, and the parison on that side stayed soft longer and sagged before the mold closed. The fix was a mold water circuit flush and a thermocouple check on every circuit. Reject rate dropped to 0.3% within two lots. The lesson: measure at multiple heights, not just the nominal gauge line, and log mold surface temperature per zone.

Case 2: the cap torque drift that was a thread root dimension

A pharmaceutical packager reported cap back-off on 500 mL PET ISBM bottles filled with a saline solution. The caps were applied at 1.1 N·m and the removal torque 24 hours later was reading 0.6 to 0.7 N·m, below the 0.8 N·m minimum. The bottles passed dimensional inspection. The caps passed incoming inspection.

We sectioned 30 bottles and measured the T-dimension (thread root to sealing land) with an optical comparator. Nominal was 17.5 mm. The lot ran 17.62 to 17.71 mm, high by up to 0.21 mm. That pushed the sealing land down relative to the thread root, so the cap liner was not compressing against the land at the specified torque. The cap was seating on the thread, not the land.

The cause was preform temperature in the ISBM process. A 6 °C drop in preform reheat zone 3 over a shift changed the blow ratio slightly and pushed the T-dimension high. The fix was tighter reheat zone control (±3 °C) and a T-dimension check every 30 minutes with a go/no-go gauge rather than a full CMM run at end of shift. Removal torque came back to 0.95 to 1.05 N·m.

Both cases share a pattern: the drawing tolerance was met, the functional dimension was not on the drawing. If you take one thing from this section, take that. Write the functional dimension into the specification, not just the geometric one.

Writing a tolerance specification a molder can meet and you can audit

Most reagent bottle drawings I see are geometric wish lists. They specify every dimension to ±0.1 mm because that is what the CAD template defaulted to, and the molder quotes to the drawing and then quietly ships to whatever the process holds. Here is the specification structure I use now.

• Tier 1 (functional, Cpk ≥ 1.33): body diameter at the guide-rail contact height, neck finish T-dimension, thread start position, base flatness. Four dimensions, no more. These get per-cavity capability data with the PPAP.
• Tier 2 (fit, Cpk ≥ 1.0): overall height, neck OD, wall thickness at the label panel. These get a 30-piece sample per lot with a variable data chart.
• Tier 3 (reference, no Cpk): everything else. Cosmetic, weight, capacity to the nominal fill line. Inspected by attribute, not variable.

The tiering does two things. It tells the molder where to spend tool maintenance money, and it tells your incoming inspection where to spend time. A 12-cavity tool with four Tier 1 dimensions gets a 48-point check per lot. That is 20 minutes on a vision system and it catches the cavity drift before it reaches the line.

For the standards side, the U.S. EPA SW-846 test methods set container requirements that flow back into dimensional specs, particularly around closure integrity and headspace. If you are molding bottles for environmental sampling, the dimensional spec has to be written against those requirements, not against a generic catalog drawing.

Troubleshooting checklist when a lot drifts out of capability

When a capability study fails or a line starts rejecting, work the checklist in this order. It is roughly ordered by frequency of cause, based on the last decade of investigations I have run.

• Check the measurement first. Gauge R&R, room temperature, fixture wear, operator technique. About one in five “capability failures” is a measurement system failure.
• Check per-cavity means. If one or two cavities are out, it is a tool or cooling circuit problem. If all cavities drifted together, it is a process setpoint or material lot problem.
• Check mold surface temperature per zone. A 5 °C drift on one zone moves the local wall thickness and the local diameter. Log it every shift.
• Check material lot moisture and melt index. A 0.05 g/10 min shift in melt index on HDPE moves the parison die swell and the EBM diameter by about 0.1 mm.
• Check the tool for flash and wear at the parting line. Flash at the parting line on an EBM tool adds 0.1 to 0.2 mm to the body diameter on the flash side. It is easy to miss because it looks like a cosmetic issue.
• Check ejection and cooling time. Short cooling time leaves the part soft at ejection and it deforms under its own weight. That shows up as ovality and base rock, not diameter.
• Check the blow pressure and preform reheat profile on IBM and ISBM. A 0.5 bar drop in blow pressure moves the T-dimension high.

If the checklist clears and the dimension is still out, the tool needs work. That is a capital conversation, not a process conversation, and it is better to have it with data in hand than with a complaint.

What to put in the purchase order

Dimensional tolerance capability is a contract term, not a hope. Put the Cpk requirement, the per-cavity sampling plan, the measurement method, and the gauge R&R limit in the PO. Require a first-article inspection with per-cavity data and a running capability chart for the first three production lots. Require notification if the tool is moved, reworked, or run on a different press.

If you are buying bottles alongside other consumables, the same discipline applies across the reagent bottle line and the sample vial line: specify the functional dimensions, measure per cavity, and tie acceptance to Cpk, not to a single-pass inspection report. For the broader instrument side, the technical resources at Agilent, Thermo Fisher, and Restek are useful for understanding what the downstream system expects from the container, and the chromatography products side of the network covers the consumable interface in more detail.

The decision rule is simple. For any dimension that touches your filling line, your capper, or your customer’s autosampler, specify Cpk ≥ 1.33 with per-cavity data, and reject any lot whose capability study pools cavities. For everything else, specify a generous geometric tolerance and inspect by attribute. Do not pay for precision you will not measure, and do not accept a drawing tolerance as evidence of capability. The evidence is the distribution, and the distribution is per cavity, per shift, per dimension. Ask for it before the PO, not after the reject pile.

References & further reading

Related resources from our lab consumables network:

• Reagent Bottle at reagentbottleforsale.com
• Hplc Autosampler Vial at hplcautosamplervials.com
• Sample Vial at samplevialprice.com
• Chromatography Products at chromatographyselling.com

External references and standards:

• Agilent chromatography consumables and columns
• Thermo Fisher chromatography and sample handling
• U.S. EPA SW-846 test methods for waste and materials
• Restek technical resources for chromatography