Start with the smallest system that still has the parameters you will actually vary. For ultrasonic spray coating that means three things: more than one liquid channel, a substrate heater that reaches the temperatures your chemistry needs, and the same atomizer technology used in the larger systems, so results transfer when you scale up.

What a first system genuinely needs

Multiple liquid channels. Single-channel systems look like a saving until the first multilayer experiment, or the first time you want to compare two formulations without flushing the line between runs. Three channels is the practical number.

A hot plate that is actually hot. Substrate temperature is one of the strongest levers on film morphology, because it sets how fast the solvent leaves after the droplet lands. A plate limited to modest temperatures quietly restricts which materials you can develop.

Controllable flow across a wide range. Thin films need very low flow rates; thicker functional layers need much higher ones. A system that covers from nanolitres to tens of millilitres per minute will not need replacing when the project changes.

The same atomizer as the bigger system. This is the one people miss. If your desktop unit atomises differently from the production system, your process parameters do not transfer and the development work has to be repeated.

What you can do without at the start

Automated Z-axis motion is convenient but rarely decisive: nozzle height is set during development and then left alone, so manual adjustment costs you minutes, not results. A large working area matters only when substrates get large. A vacuum plate is genuinely useful for thin or flexible substrates, but it can usually be added later.

Buy the parameters you will vary. Skip the conveniences you will use twice.

The first month, practically

Expect to spend it on process development rather than results, and plan for that openly with whoever funds the instrument. A sensible sequence: establish a stable spray with solvent only and confirm the mist pattern; move to your actual solution at a mid-range flow and find the nozzle height that gives a wet but not pooling film; then vary substrate temperature, which usually produces the largest change in morphology; and only then optimise scan speed and path spacing for uniformity. Changing one parameter at a time sounds obvious and is routinely ignored, which is why so many early spray films are inexplicable.

Questions worth asking a supplier

• Will you help establish a first recipe for my material, or does the system arrive as a box?
• What training is included, and does it happen on site?
• If I outgrow this system, do the parameters transfer to your larger one?
• What is the lead time for consumables and spare parts in my country?

That last question is worth more than it looks. An instrument waiting three months for a part is not an instrument.

The UltraSprayer-Essential we represent is built for this stage: three channels, a plate to 500 °C, flow from 17.9 nl/min to 84 ml/min, and the same IdaMist atomizer used in the automated Pro system. We handle installation, training and local support. Write to info@rexerlab.com.

Spin coating is simpler and hard to beat on a small flat substrate. Ultrasonic spray coating wins on material efficiency, coating area, textured or porous surfaces, and the path to production. Most laboratories that switch do so for one of those four reasons rather than for film quality alone.

Material waste is usually the first argument

Spin coating discards most of what you dispense; the great majority of the solution is flung off the substrate. With a routine solvent that is an annoyance. With a custom-synthesised polymer, a catalyst ink loaded with platinum, or a perovskite precursor made over three days, it is the dominant cost of the experiment. Ultrasonic spray deposits a much larger fraction of what is pumped, because the mist is directed at the substrate rather than thrown outward.

Where spin coating simply stops working

• Large substrates. Spin coating uniformity degrades as radius grows, and the mechanics become awkward. Spray systems scale by moving the nozzle over a larger area.
• Non-flat and porous surfaces. Gas diffusion layers, membranes, textiles and structured devices cannot be spun usefully. A low-velocity mist follows the surface.
• Non-circular geometry. Spin coating assumes rotational symmetry. Rectangular cells and irregular parts do not have it.
• Multilayer and graded films. Multi-channel spray systems can vary composition between or during passes.

What ultrasonic atomization actually contributes

The word ultrasonic is doing real work here. In conventional pressure spraying, the liquid is atomised by forcing it through a nozzle at high velocity, which produces a broad droplet size distribution and enough momentum to disturb delicate surfaces. Ultrasonic nozzles vibrate the liquid until it breaks into droplets, so droplet size is set by frequency rather than by pressure, the distribution is narrow, and the mist leaves the nozzle slowly. A separate low-pressure air shaper then guides the mist onto the substrate. Narrow droplet distribution is what makes the film uniform; low velocity is what lets you coat something fragile.

Spin coating optimises one sample. Spray coating optimises a process you intend to repeat.

The honest trade-offs

Spray coating has more parameters: flow rate, nozzle height, scan speed, path spacing, substrate temperature, air shaper pressure. That is more control, and also more to establish before you get a reproducible film. Expect a development period that spin coating does not require. Very thin films, in the low tens of nanometres, are also more demanding by spray, and solvent selection interacts with drying behaviour more strongly.

Choosing between systems

Once you have decided on spray, the practical questions are working area, whether the Z axis is automated or manual, how hot the substrate plate gets, and how many liquid channels you need. Substrate temperature deserves particular attention: for many materials the drying dynamics on the hot plate determine morphology, and a plate that reaches 200 °C serves a different set of chemistries than one reaching 500 °C.

We represent both ends of the Idasonic range: the automated UltraSprayer-Pro for pilot-scale work and the desktop UltraSprayer-Essential for laboratories starting out. To talk through your material, write to info@rexerlab.com.

What is actually in your chamber, while it is happening. A vacuum gauge gives you one number: total pressure. An online mass spectrometer breaks that number into its components, so instead of knowing that pressure is higher than it should be, you know it is water, or oil, or nitrogen leaking in through a fitting.

The four questions it answers

Is this a leak or is it outgassing? The most common vacuum question, and the most commonly guessed at. A real air leak shows nitrogen and oxygen in roughly atmospheric ratio. Outgassing shows water dominating, falling slowly as the chamber dries. The spectrum distinguishes them in seconds; pump-down curves alone rarely do.

Where is the leak? Spray helium around suspect joints and watch mass 4. The spectrometer sees helium arrive within moments of it reaching the leak, which turns leak hunting into a systematic procedure rather than a soap-bubble ritual.

Is my process gas what the bottle claims? Purity problems and cross-contamination from shared lines show up directly, which matters when trace contaminants change film properties.

What is happening during the process itself? This is where online analysis earns its name. In ALD or ALE, watching precursor and by-product signals cycle tells you whether the surface reaction actually saturated or whether you are wasting precursor and time.

Why the speed specification matters more than it looks

For static residual gas analysis, sampling speed is a convenience. For process monitoring it is the whole point. An ALD half-cycle can be a fraction of a second; a fast transient during a chamber event is shorter still. If the analyser needs longer to scan the mass range than the event lasts, you get an average of something interesting rather than a picture of it. This is why readout rates per amu, rather than the mass range alone, deserve attention when comparing instruments.

Dynamic range, and why two detectors exist

The gases you care about are frequently a millionth of the ones you do not. A Faraday cup is robust and linear for abundant species; a secondary electron multiplier provides the sensitivity for trace species. Instruments carrying both cover the range that real processes present, which is the practical reason dual-detector configurations are common on process tools.

A pressure gauge tells you something is wrong. A mass spectrometer tells you what.

Integration is what separates a laboratory instrument from a process instrument

An analyser that produces beautiful spectra on a laptop next to the tool solves half the problem. On a production system the data has to reach the tool controller, trigger alarms and be logged alongside process parameters. That is why industrial protocol support such as EtherCAT and Modbus, and an integrated pumping system that does not demand its own bench, matter as much as the analytical specification.

The Hakuto Genius online mass spectrometer we represent is built for that role: quadrupole analysis to 300 amu, up to 1000 readings per second per amu, Faraday and SEM detectors, EtherCAT and Modbus, with an integrated turbomolecular pump. To discuss your process, write to info@rexerlab.com.

When chemistry will not cooperate. Reactive plasma etching works by forming volatile compounds that pump away; if a material has no convenient volatile compound, the process stalls or leaves residue. Ion beam etching sidesteps the problem entirely by removing material physically, with accelerated ions knocking atoms off the surface.

The materials that force the decision

Magnetic materials. Permalloy, cobalt alloys and the multilayer stacks used in spintronics have no clean reactive chemistry. This is the classic case for ion milling, and it is why the technique is standard in magnetic head and MRAM work.

Precious and noble metals. Gold and platinum do not form volatile halides under normal conditions. Physical removal is the practical route.

Alloys and multilayers. A stack of different materials would need a different plasma chemistry per layer. A physical process treats them as one, which means one recipe rather than a sequence of them.

Some oxides and dielectrics. Hard-to-etch dielectrics that resist fluorine chemistry can be milled physically.

What you gain

Beyond material independence, there are three practical advantages. The process gas is argon alone, so no toxic gas infrastructure, abatement or associated permits are required. Directionality is excellent because the ions arrive collimated, which gives steep sidewalls. And because the mechanism is momentum transfer rather than chemistry, the etch rate is predictable from material density and beam parameters rather than from a chemistry that shifts with loading.

What you give up, honestly

Ion beam etching is not free of trade-offs, and any supplier who says otherwise is selling rather than advising.

• Selectivity is low. Physical sputtering removes the mask nearly as readily as the film, so mask strategy matters more than in a chemical process.
• Redeposition. Sputtered material can land back on sidewalls. Stage tilt and rotation manage it, but it has to be managed.
• Heat. The energy goes into the substrate. Sample cooling is not an accessory; for temperature-sensitive stacks it is the whole ballgame.
• Throughput. Physical removal is generally slower than an optimised reactive process on a material where chemistry works.

Use chemistry where chemistry works. Use ions where it does not.

Uniformity is the specification that decides quality

The practical quality metric for ion milling is how evenly the beam removes material across the substrate. Fixed-stage systems struggle; planetary motion, where substrates both orbit and rotate under the beam, averages out the beam profile and is what makes the numbers acceptable across a full wafer. When comparing systems, ask for uniformity figures on your substrate size rather than in the abstract.

The Hakuto ion beam etching systems we represent use Kaufman sources from 8 to 20 cm with planetary stage motion, argon only, in configurations from a research tool through to large-scale production. If you want to discuss whether your stack is a candidate, write to info@rexerlab.com.

Orientation. A standard Raman spectrum tells you what a material is; polarized Raman tells you how it is arranged. For anything anisotropic, and that includes most crystals, layered materials, drawn polymers and stressed films, orientation is often the property that actually matters.

The physics in one paragraph

Raman modes have symmetry. How strongly a given mode scatters depends on the angle between the polarization of the incident light and the crystallographic or molecular axes of the sample. Rotate the polarization and the intensity of individual peaks rises and falls in a pattern that is characteristic of the material and its orientation. That angular pattern is the measurement.

What laboratories use it for

Crystal orientation and domain mapping. In 2D materials and thin films, polarized Raman distinguishes grains and domains that look identical in an unpolarized spectrum, which is how uniformity across a wafer gets assessed.

Stress and strain. Peak positions shift under strain, and the shift is direction-dependent. Polarization resolves the direction, not just the magnitude, which is what makes it useful for device reliability work.

Polymer chain alignment. Drawn fibres, films and packaging materials derive their mechanical properties from chain orientation. Polarized Raman measures it without destroying the sample.

Layer stacking. In stacked 2D materials the twist angle between layers changes electronic behaviour, and polarization-dependent modes are a practical way to read it.

Why the measurement is usually done badly

The physics is standard; the execution is where laboratories lose time. Done by hand, an angle-resolved measurement means rotating a polarizer or the sample stage, taking a spectrum, recording the angle, and repeating. A full 360° scan at reasonable resolution is dozens of manual steps for a single point. Multiply that by a map and the measurement stops being practical, so people take four angles instead of thirty-six and hope the interpolation holds.

Most polarization data is sparse not because the science needs it that way, but because collecting it by hand is tedious.

Automating the rotation changes what is realistic. When the instrument scans the full circle under software control, angle-resolved imaging becomes something you run routinely rather than something you plan a week around. Adding a second excitation wavelength on top of that lets you check whether an orientation effect is genuine or an artefact of resonance at one particular wavelength.

Practical questions to ask

• Is the polarization scan automated across the full 360°, and at what angular step?
• Can the scan be run per pixel across a map, or only at single points?
• If the system has two lasers, does the polarization scan run independently for each?
• How repeatable is the laser power between measurements? Intensity comparisons are meaningless if the power drifts.

The MRID system we represent was designed around exactly this: two software-switched lasers, automated 360° angle-resolved polarized imaging for each, and laser power calibrated to under 1% error so intensity comparisons hold. Questions about your samples are welcome at info@rexerlab.com.

Because the wavelength that produces a clean spectrum on one sample can bury the next one in fluorescence. Raman scattering is weak, and the excitation laser decides how much signal you get, how much fluorescence competes with it, and how deep into the sample you are looking. A single-wavelength instrument quietly limits which samples your laboratory can handle.

Three things the wavelength changes

Signal strength. Raman scattering intensity scales steeply with excitation frequency, so shorter wavelengths return far more signal. A blue or green laser can produce a usable spectrum in a fraction of the acquisition time a near-infrared laser would need on the same sample.

Fluorescence. This is the practical reason most laboratories end up wanting a second laser. Many organic materials, biological samples, minerals and polymers fluoresce strongly under visible excitation, and the fluorescence background can be orders of magnitude larger than the Raman signal you are trying to see. Moving to 785 nm or 1064 nm often removes the problem entirely, at the cost of signal.

Resonance and sampling depth. When the excitation approaches an electronic transition of the material, specific vibrational modes are enhanced dramatically. Choosing the wavelength therefore selects which part of the sample you emphasise. Longer wavelengths also penetrate further into scattering materials, which matters for coatings and layered structures.

Why researchers end up wanting three, not two

A shared research instrument rarely serves one sample type. A group working on 2D materials wants a wavelength on resonance with the layer of interest; the collaborator down the corridor arrives with a fluorescent polymer; a third project needs photoluminescence rather than Raman. Each of these is a different optimum. The practical consequence is that the instrument either constrains the science or the science queues behind the instrument.

The part that is easy to underestimate: alignment

Adding lasers to a Raman system is not simply a purchasing decision. Every wavelength needs its own filters and grating position, and unless the system is designed for it, switching sources means realignment, recalibration, and an afternoon gone. That is exactly why multi-laser platforms are built so that all sources land on the same sample spot and the switch happens in software rather than with an Allen key.

The question is not how many lasers an instrument has, but how much work it takes to change between them.

What to check before you buy

• Do all wavelengths reach the same spot on the sample, or do you re-find your region of interest after each switch?
• Is switching automated in software, including filters, grating and calibration?
• Can further wavelengths be added later, including UV or 1064 nm mounted externally?
• Does the system also handle photoluminescence and lifetime measurement, or will that be a second instrument?

The RAMaker platform we represent is built around this problem: up to six excitation wavelengths onto one spot without realignment, with Raman, photoluminescence and time-resolved measurement in the same chamber. If you would like to talk through which wavelengths your samples actually need, write to info@rexerlab.com.