KEY TAKEAWAYS
-
The standard laboratory method produces excellent single devices and unacceptable batches, for five specific reasons: an unspecified top surface, bubbles and dust, hand-punched ports, short-lived SU-8 masters, and a process that cannot be staffed.
-
No single tooling route offers fidelity, life and cost at once: SU-8 holds 2 µm for 30–50 cycles; metal lasts over 500 cycles, but at 10 µm. The replica-mold architecture resolves this — lithography once, production tools replicated from that master, over 100 cycles each at SU-8 fidelity.
-
In the closed mold, thickness, planarity and surface finish are process settings: ±5 µm thickness, flatness under 2 µm, molded-in through-holes from Ø 0.30 mm at 0.10 mm alignment, up to 80,000 pieces a month.
-
Teams that conclude PDMS "cannot scale" start a two-to-five-year transfer to injection molding earlier than the product justifies. The limitation was never the material; the process is replaceable without touching the design.
\ FAILURE MODES
Five reasons lab-scale casting fails
The standard laboratory method — mix 10:1, degas in a desiccator, pour over an SU-8 master in a Petri dish or on a bare wafer, cure in an oven, peel, punch, plasma-bond — produces excellent single devices and unacceptable batches. Five specific mechanisms are responsible.
OPEN MOLD — cast
-
Meniscus curls the edges
-
Bubbles trapped on cure
-
Inlets punched by hand
-
Top surface never gauged
CLOSED MOLD — molded
-
Thickness set by the tool
-
Degassed, bubble-free bulk
-
Through-holes molded in
-
Cured under calibrated pressure
Figure 1 — Open-mold casting versus closed-mold molding. In the open mold the free surface is set by gravity and surface tension; in the closed mold it is set by the tool.
01
The open mold leaves the most important dimension unspecified
In an open mold the top of the part is a free surface. Its position is set by how much prepolymer was poured, how level the bench is and where the meniscus climbs the wall — not by a dimension on a drawing. Thickness varies from part to part and across each part, edges curl, and every downstream operation that depends on a flat, parallel top face — bonding, membrane transfer, cartridge assembly, optical path length — inherits that variation.
02
Hand mixing and desiccator degassing leave bubbles and dust
Hand mixing entrains air. A desiccator removes most of it, not reliably all of it, and the vacuum profile is different every time. Curing on an open bench embeds airborne particles and fibers in the part permanently. In a cell-culture device a trapped bubble is a dead zone; in a droplet generator it destroys monodispersity; in an optical device it is scatter.
03
Manual punching destroys the interconnect before the chip is used
biopsy punch have torn edges, variable diameter and variable pitch. Torn edges shed PDMS debris into the channel. Variable pitch means the chip cannot be registered against rigid features in a holder or instrument. And every punched hole is a stress concentration that fails when tubing is pushed into it repeatedly — the mechanism quantified in Part 3.Inlets and outlets made with a biopsy punch have torn edges, variable diameter and variable pitch. Torn edges shed PDMS debris into the channel. Variable pitch means the chip cannot be registered against rigid features in a holder or instrument. And every punched hole is a stress concentration that fails as tubing is pushed into it again and again — the mechanism quantified in Part 3.
04
SU-8 masters have a short and unpredictable casting life
Photoresist is a patterning material, not a structural one. Repeated heating and cooling through casting cycles delaminates the resist from the silicon, and the problem worsens with thicker resist and higher aspect ratios — exactly the features that make a design worth building [6]. In our experience a master supports a few tens of castings before features begin to lift. Re-running lithography to replace it puts run-to-run variation back into a product that is supposed to be frozen.
05
The process cannot be staffed
Every step above depends on the judgment of the person doing it. Output becomes a function of how many trained scientists you can put on a bench — the hardest resource there is to scale — and batch-to-batch consistency a function of who was working that week. No SOP rescues a process whose critical dimensions are set by hand.
What this costs, in program terms
The cost is not just lower yield. Teams conclude that PDMS "cannot scale" and start a transfer to injection molding far earlier than the product justifies. In practice that transfer is a two-to-five-year program: it forces a redesign of the chip geometry, re-validation of reagents and surface chemistry, and often changes to the analyzer itself. Meanwhile the product is not on sale.
THE CORRECT DIAGNOSIS
The limitation was never the material. It was the process — and the process is replaceable without touching the design.
\ TOOLING
Mold technology: choosing the right tool
A PDMS part is only as good as the tool it is formed in. The four common tooling routes trade cost, tool life and feature fidelity against one another, and none is good at all three.
Mold type
Cost
Life (cycles)
Lead time
Min. feature
Best used for
SU-8 on silicon
$1k – 3k
Prototyping, and as the master of record
30 – 50
2 – 7 d
2 µm
3D-printed resin
$100 – 1k
Concept checks and complex geometry
10 – 30
1 – 10 d
50 µm
PMMA machined
$50 – 200
Fast, simple structural parts
< 10
1 – 3 d
50 µm
Metal alloy
$500 – 10k
Volume production of coarse features
> 500
5 – 30 d
10 µm
HiComp closed replica mold
—
Volume PDMS at SU-8 fidelity
> 100
—
2 µm
Read the table as a single problem statement: the route with the fidelity you need (SU-8, 2 µm) has the shortest life, and the route with the life you need (metal, > 500 cycles) cannot hold the features. Volume PDMS manufacturing is the work of resolving that contradiction.
The replica-mold architecture
We resolve it by separating the master of record from the tool that touches product. Lithography runs once; everything downstream is replicated from that single master.
SU-8 MASTER
· Layout design
· SU-8 lithography
· Wafer dicing
· Features to 2 µm
MASTER MOLD
· Primary tool design & assembly
· Thickness control
· 1.00 mm ± 0.05 mm
REPLICA MASTER
· Vacuum-degassed resin casting from the master — copy tooling
CLOSED PRODUCTION MOLD
· Molded-in through-holes & reservoirs
· ± 0.10 mm
· Life > 100 cycles
Figure 2 — The replica-mold chain. Lithography is performed once; all production tooling descends from that master.
SU-8 master
Layout design, SU-8 lithography and wafer dicing produce the master, with features down to 2 µm. This part never sees production volume.
Master mold
A primary tool is designed and assembled around the master, establishing the gauged thickness reference — 1.00 ± 0.05 mm on a typical layer.
Replica master
A vacuum-degassed resin casting taken from the master becomes the copy tooling, so surface fidelity survives the copy instead of degrading with it.
Closed production mold
Production tools are built from the replica master, with through-holes molded in. Tolerance ± 0.10 mm, service life above 100 cycles per tool.
Why this matters beyond cost
The architecture buys three things that matter more than tooling cost. Capacity is elastic: adding a line means replicating tools, not re-running lithography, so new capacity never waits on a lithography run. Every chip in every batch traces to one lithographic origin — the traceability a design-control auditor will ask for. And the production tool can carry features the master never had — through-holes, reservoirs, registration datums — because it is designed, not just copied.
WHAT THE CLOSED MOLD ADDS THAT NO OPEN MOLD CAN
-
A specified top surface — thickness and planarity are properties of the tool, not of the pour
-
Molded-in through-holes and reservoirs, eliminating manual punching and its debris entirely
-
Registration features that let the part be located against rigid pillars in a housing
-
A repeatable cure under calibrated pressure, so surface finish is a process setting
\ PROCESS
The closed-mold production process
The distinction that matters is molding versus casting. Casting is an open process with an open outcome; molding is a closed process with a specified one. The production sequence below is run by trained technicians against written SOPs, in an ISO 7 cleanroom, under an ISO 13485 quality system.
1
Metered mixing
.png)
5
Demold (holes formed)
2
Controlled vacuum degas
6
Bond 2‑8 layers
3
Closed‑mold fill
7
100% inspection
__.png)
4
Press cure (calibrated P)
8
Cleanroom pack
Figure 3 — The eight-step PDMS production sequence.
1/ Metered mixing
Base and curing agent are metered and mixed in dedicated equipment, not by hand, which removes operator variation from the most consequential ratio in the process.
3/ Closed-mold fill
The degassed prepolymer is introduced into a closed, integrated mold. There is no free surface, so there is no meniscus.
5/ Demold
The part leaves the tool complete, with through-holes and reservoirs already formed. No punching step exists, so no punching defects exist.
7/ 100% inspection
Every part is inspected under the microscope; particle control keeps critical areas debris-free.
2/ Controlled vacuum degassing
Degassing follows a defined vacuum profile in dedicated equipment: bubble-free bulk PDMS, batch after batch, instead of results that depend on the desiccator.
4/ Press cure under calibrated pressure
Pressure during cure sets thickness, planarity and surface finish. These become process settings with tolerances, not outcomes to be measured after the fact.
6/ Bonding
Bubble-free PDMS-to-PDMS and PDMS-to-glass bonding at 0.5–0.8 MPa, in stacks of two to eight layers, with a reversible bonding option where the application requires it. Membranes and electrodes can be embedded in the stack.
8/ Cleanroom packaging
Parts are packed in the cleanroom and shipped as bare chips, chips with tubing pre-assembled, or fully integrated cartridge assemblies.
WHY THIS CAN BE STAFFED
-
Every judgment call in the laboratory method has become a machine setting or a tool dimension. That is what lets trained technicians run the line instead of the scientists who designed the chip — and why output is counted in thousands of chips a day rather than tens a week.
The service page for this process: PDMS manufacturing.
\ SPECIFICATIONS
What "industrial grade" means: the production specification
"Industrial grade" is only meaningful if it resolves to numbers. The table below is the current HiComp PDMS process window. Every value is a specification held in production, not a best-case laboratory result.
Geometry
Maximum chip size
≤ 300 × 300 mm
Minimum channel width
≥ 2 µm
Channel depth tolerance
± 5%
Channel aspect ratio
up to 10:1
Minimum through-hole diameter
0.30 mm
Minimum thickness
> 5 µm
Thickness tolerance
± 5 µm
Thickness uniformity
± 10 µm (for t > 100 µm)
Surface flatness
< 2 µm
Alignment accuracy
0.10 mm (tighter available on request)
Bonding, environment and delivery
Multi-layer bonding
2 – 8 layers
Bond strength
0.5 – 0.8 MPa
Embedded materials
Membranes, metals, electrodes
Options
Tinted or colored PDMS; reversible bonding
Manufacturing environment
ISO 7 cleanroom; particle-controlled, debris-free in critical areas
Quality system
ISO 13485; FDA registered
Capacity
Up to 80,000 pieces per month
Lead time
2 – 3 weeks after design approval; expedites available
Minimum order
No strict minimum; pricing improves with batch size
Delivery formats
Bare chips, chips with pre-assembled tubing, or integrated cartridge assemblies
\ COMPARISON
Fab PDMS versus lab PDMS, criterion by criterion
CRITERION
Lab PDMS (open cast)
HiComp Fab PDMS (closed mold)
Mold & forming
Open dish or bare wafer; PDMS poured and left to level under gravity
Closed, integrated mold; cured under calibrated pressure
Thickness & planarity
Meniscus edges; thickness varies part to part and across each part
1.00 ± 0.05 mm at the tool; ± 5 µm on the part; flatness < 2 µm
Bubbles
Hand mixing and desiccator degassing; trapped voids are routine
Metered mixing plus controlled vacuum degassing; bubble-free bulk
Through-holes
Punched by hand — torn edges, variable pitch, scrap at assembly
Molded in — Ø ≥ 0.30 mm, alignment 0.10 mm, no post-processing
Contamination
Open bench; ambient dust and fibers embedded in the cured part
ISO 7 cleanroom; microscopic inspection; debris-free critical areas
Tooling
SU-8 master cast directly; delaminates after tens of cycles
Replica tooling from one master; > 100 cycles per production tool
Bonding & packaging
Manual plasma, variable strength; ad-hoc packing
PDMS/PDMS and PDMS/glass at 0.5–0.8 MPa; 2–8 layers; cleanroom packed
Who runs it, and output
Scientists; tens of chips per week
Trained technicians; 2,000+ chips per day, 80,000 per month
Source on this page: [6] Amadeo et al., Reusable master molds for soft lithography, Micromachines 12, 1392 (2021). Manufacturing specifications and capacity figures are HiComp production specifications. Full list on the guide overview.
\ CONTINUE
Continue reading
The complete paper is available as a PDF (16 pages, 1 MB): Industrializing PDMS — a manufacturing and decision guide.
How an engagement runs
01. Proposal &
Contract
Timeline: 6 - 7 Weeks
From first conversation to purchase order — initial meeting and indicative quote, NDA, a written design evaluation, technical meeting, firm quote, contract.
02. Tooling & First Article
Timeline: 2 - 3 Weeks
Mold design, replication from the master pattern, and first-article inspection against your drawings.
03. Ramp & Mass Production
Timeline: Month 1 – Month 4+
Roughly 1,000 pieces in month one, 5,000 by month two, 20,000 by month four — up to 80,000 a month at full capacity.
